Considerable physical and psychosocial benefits are associated with participation in sporting activities.1 2 Physical activity has been demonstrated to reduce the risks of coronary heart disease, some cancers, obesity, hypertension, and type 2 diabetes mellitus.2 3 4 5 6 Its obvious merits have prompted several national health programmes, including the health programme in Hong Kong, to promote sports to the general public.7 8 9 However, sports participation carries a risk of injury, especially traumatic brain injury (TBI). It has been estimated that 20% of all TBIs are sports-related.10 In addition, up to 20% of sports-related TBI survivors (usually adolescents or young adults) experience chronic symptoms including headache, fatigue, and cognitive and balance difficulties.11 There is a global trend of increasing sports-related TBI incidence: from 3.5 to 31.5 per 100 000 population in the past decade.12 13 Because many patients with mild TBI do not seek medical attention, these figures likely underestimate the total burden of this condition.13 Population-based studies reviewing sports-related TBI are sparse; most target specific groups of individuals (eg, professional athletes) or rely on self-reporting surveys that lack a uniform definition and comprehensive assessment of brain injury.14 For similar reasons, studies reviewing outcome predictors have also been inadequate, thus hindering the generation of meaningful conclusions to guide governmental policy initiatives.14

Hong Kong is highly urbanised with an established public transport system, such that cycling is mainly regarded as a recreational activity.15 16 In terms of fatality rate, Hong Kong is among the most dangerous areas for cycling, compared with other cities such as New York, the US, or countries such as France.17

This study was performed to document the epidemiology of sports-related TBI among patients who required inpatient care by adopting territory-wide uniform diagnostic coding criteria, clear data definitions, and systematic assessments of radiologic findings. Because cycling is a popular sport in Hong Kong, factors predictive of intracranial haemorrhage (eg, the effect of helmet use) and poor functional outcomes among cyclists hospitalised with TBI were determined.

Patients who required inpatient care at any Hospital Authority institution for sports-related TBI from 1 January 2015 to 31 December 2019 were reviewed. The Hospital Authority is a public health service highly subsidised by the Hong Kong SAR Government; it is responsible for 90% of inpatient bed days in the city. Patients were identified by the International Classification of Diseases, Tenth Revision, Clinical Modification code (ICD-10-CM) designation for TBI 854.0 secondary to a sports-related external cause (E codes: E006-10). Data from clinical records, operation notes, medication prescriptions, and computed tomography (CT) brain scans from a central digital imaging repository were reviewed. In particular, the type of sport played, the clinical presentation of symptoms, the Injury Severity Score (ISS), the need for neurosurgery, length of hospitalisation, and diagnosis of post-concussion syndrome were recorded. Head injury was classified into mild (presenting Glasgow Coma Scale [GCS] score, 14-15); moderate (presenting GCS score, 9-13), and severe (presenting GCS score, 3-8), in accordance with criteria established by the Neurotraumatology Committee of the World Federation of Neurosurgical Societies.18 Post-concussion syndrome was defined in accordance with ICD-10 criteria. This required a 4-week duration of symptoms from at least three categories following a traumatic loss of consciousness. The symptom categories were headache, irritability, concentration impairment, insomnia, and a preoccupation with the aforementioned symptoms. For neurosurgical patients with cycling related–TBI, the mechanism of injury and their experience level (ie, professional athlete or amateur rider) were documented. All CT scans were reviewed by an independent assessor with 6 months of neurosurgical training experience who was blinded to the patients’ clinical characteristics and outcomes. The scans were first evaluated using the Rotterdam CT score, a commonly utilised validated radiological assessment system for the prognosis of patients with TBI. The classification has four distinct elements that require the appraisal of the degree of basal cistern obliteration, degree of midline shift, the presence (or absence) of an epidural mass lesion, and the presence (or absence) of intraventricular or traumatic subarachnoid haemorrhage (Table 1). In addition, the scans were assessed for skull fractures, cerebral contusions, and acute subdural haematomas (ASDHs). The primary outcome of the study was the presence of intracranial haemorrhage on the admitting CT scan. All potential predictors were categorised into patient-related, trauma-related, and radiological factors. The secondary outcome was unfavourable functional performance, defined as a Glasgow Outcome Scale (GOS) score of 3 to 5 on discharge from the hospital (3, severe disability; 4, persistent vegetative state; and 5, death).

Statistical analyses utilised the Chi squared test and Fisher’s exact test were used for categorical data such as patient gender or the use of antiplatelet medication. Independent-samples t test was used for continuous data such as patient age or duration of hospitalisation. Multivariate binary logistic regression was used to identify independent factors for the presence (or absence) of traumatic intracranial haemorrhage. A P value of <0.05 was considered statistically significant. Statistical analysis was conducted using SPSS (Windows version 20.0; IBM Corp, Armonk [NY], US).

In total, 720 consecutive patients were hospitalised with sports-related TBI during the 5-year study period, and 705 (97.9%) of them were admitted under neurosurgical care. This was equivalent to a crude incidence of 1.9 per 100 000 general population. The mean (± standard deviation [SD]) age was 32 ± 19 years; 521 (72.4%) patients were adults (≥18 years) and 568 (78.9%) were male. The most common sport was cycling (59.2%), followed by football (21.3%) and basketball (7.5%) [Fig a]. On admission, most (86.1%) patients were fully conscious. Overall, 658 (91.4%) patients had mild TBI, 41 (5.7%) patients had moderate TBI, and 21 (2.9%) patients had severe TBI. Post-traumatic seizures occurred in 36 (5.0%) patients. Furthermore, 324 (45.0%) patients had a loss of consciousness and 269 (37.4%) patients experienced post-traumatic retrograde amnesia. Only 19 (2.6%) patients were taking either antiplatelet or anticoagulant medication. Extracranial injuries were sustained by 208 (28.9%) patients; among them, injuries were mainly either limb abrasions or contusions (62.1%). The median ISS was 2 (interquartile range=2-8).

Intracranial haemorrhage was noted in 283 (39.3%) patients with TBI; 166 (23.1%) patients exhibited traumatic subarachnoid haemorrhage and 157 (21.8%) exhibited ASDH. Skull fractures were detected in 179 (24.9%) patients, with a median Rotterdam CT score of 2 (interquartile range=2-2). In total, 59 (8.2%) patients required neurosurgical intervention; 32 (54.2%) of them had good recovery with a median GOS score of 5 on discharge from the hospital and at 6 months. The mean (± SD) duration of hospitalisation was 5 ± 28 days. Among 307 (42.6%) patients in whom 6-month GOS scores could be assessed, good recovery was observed in 260 (84.7%). Post-concussion syndrome was diagnosed in 30 (6.2%) of 482 patients who attended scheduled follow-up outpatient consultations.

The crude incidence of recreational cycling-related TBI requiring hospitalisation was 1.1 per 100 000 population. A comparison was performed between cyclists with sports-related TBI and patients who had TBI because of other sports (Table 2). Cyclists were significantly older (P<0.001) [Table 2; Fig b]. Among 426 cyclists, 306 (71.8%) were male and 120 (28.2%) were female. However, the proportion of patients who were female was significantly higher among those who had TBI because of cycling (28.2%) than among those who had TBI because of other sports (10.9%; P<0.001). Cyclists were likely to exhibit more severe TBI with an almost three-fold greater risk of sustaining extracranial injury (odds ratio [OR] 2.8; 95% CI: 1.9-4.0), resulting in a higher ISS (P<0.001). Of cyclists admitted for head injury, 201 (47.2%) had intracranial haemorrhage, which was radiologically more extensive in terms of the Rotterdam CT score, compared with haemorrhage in non-cyclists (P<0.01). As a consequence, a greater proportion of cyclists had a worse GOS score on discharge from hospital (OR 2.8; 95% CI: 1.3-6.2) and at 6 months (OR 4.7; 95% CI: 2.1-10.5). The cause of death for all cyclists with 30-day mortality was severe TBI with medically refractory intracranial hypertension. Although the overall incidence was low, cyclists also had a greater risk of post-concussion syndrome (OR 2.5; 95% CI: 1.2-5.4).

Among the 426 cyclists in this study, 128 (30.0%) experienced bicycle accidents during the weekend; 10 (2.3%) of the cyclists were professional athletes. Most cyclists (273; 64.1%) accidently fell off their bicycle on their own (ie, without colliding into another object) on level ground. Of the injuries, 103 (24.2%) were sustained when the cyclist was traveling downhill; for the 28 patients with records of self-reported velocities, the estimated mean (± SD) velocity at the moment before injury was 40 ± 15 km/h. At the time of injury, 361 (84.7%) cyclists had not been wearing a helmet. Among eight (1.9%) patients who subsequently died, none had been wearing protective head gear.

Risk factors for traumatic intracranial haemorrhage among cyclists are shown in Table 3. Univariate analysis identified the following risk factors: age ≥60 years, use of antiplatelet medication, involvement in a motor vehicle collision, presence of moderate to severe TBI, and skull fracture. In univariate analysis, helmet wearing was protective against intracranial haemorrhage. Multivariate logistic regression identified the following independent risk factors: age ≥60 years, antiplatelet medication intake, moderate or severe TBI, and the presence of a skull fracture. Table 4 shows independent significant predictors for unfavourable GOS score on discharge from hospital: age ≥60 years, antiplatelet intake, severe TBI, intracranial haemorrhage, and the need for neurosurgical operative intervention.

As shown in Table 2, 375 (88.0%) hospitalised cyclists had mild TBI, whereas only 36 (8.5%) cyclists had moderate TBI and 15 (3.5%) had severe TBI. No protective effect of helmet use was noted in terms of reducing TBI severity across these GCS-defined categories (Table 3). However, among cyclists with mild TBI, helmets were significantly protective against intracranial haemorrhage and skull fracture, regardless of age, antiplatelet medication intake, or mechanism of injury (Table 5). Although the median Rotterdam CT score was comparable between cyclists with mild TBI who did or did not wear helmets (P=0.68), significantly fewer patients with head protection had epidural haematoma or ASDH. For patients with mild TBI who had intracranial haemorrhage, this difference in radiological factors led to a significantly shorter mean (± SD) duration of hospitalisation for patients who wore helmets (2.6 ± 2.9 days), compared with patients who did not (7.1 ± 11.6 days, P<0.001). However, there was no difference in the need for neurosurgical intervention among patients with mild TBI who had intracranial haemorrhage according to head protection status (P=0.17). Similarly, unfavourable GOS scores on discharge from hospital (P=0.43) and at 6 months (P=0.71) were comparable among patients with mild TBI who had intracranial haemorrhage, regardless of head protection status (Table 5).

The incidence of sports-related TBI in Hong Kong is 2 per 100 000 general population; this is lower than in other countries (eg, the US, Australia, or Italy), where the incidences range from 4 to 32 per 100 000 population.12 The lower incidence in Hong Kong is consistent with a previous finding that Hong Kong residents (especially children and adolescents) have lower physical activity and fitness levels than in other regions, according to a global evidence-based evaluation of such indicators from 49 countries.19 Considering the health benefits of an active lifestyle, there is a clear need to promote sports engagement in Hong Kong. A survey of 5701 residents performed by the Transport Department of the Hong Kong SAR Government estimated that 10% of households had bicycles available for use; moreover, 69% (4 million) of residents aged 15 years or older knew how to ride one.16 In addition, most survey respondents (73%) cycled for recreational or fitness purposes.15

Previous epidemiological studies of sports-related brain injuries revealed that cycling was one of the most frequent activities involved.10 12 20 21 In 2019, 1738 road traffic accidents involving cyclists were reported to the Hong Kong Transport Department.22 Half of these accidents (50.6%, 879/1738) occurred in recreational areas such as cycling tracks, parks, or playgrounds; eight (0.5%) patients experienced fatal injuries.22 In the past 10 years, the number of cyclist injuries in Hong Kong has increased by 5.2% per year.17 Compared with other regions worldwide, Hong Kong is one of the most dangerous areas for cycling.17 The fatality rate (per billion minutes cycled) in the city was 34, substantially higher than the rates in Stockholm, Sweden (3), France (4), and other metropolitan areas (eg, New York City [18] and Los Angeles [8]).17 These studies included riders primarily involved in commuting and the causes of death were not elucidated, but they indicate a growing need to enhance the safety of vulnerable road users.

To our knowledge, this is the first multicentre study to comprehensively document the outcomes of inpatients with recreational cycling-related TBI using standard assessment criteria. By comparison with patients who had TBI because of other sports, we found that cyclists in Hong Kong exhibited greater risks of more severe injury, intracranial haemorrhage, unfavourable GOS score at discharge from hospital, and post-concussion syndrome. Despite these findings, our results suggest that cycling is generally safe and hospitalised patients had a high (92.7%) likelihood of favourable functional outcomes on discharge from hospital.

Single-centre reviews of cycling-related injuries among various suburban districts in Hong Kong found that limb injuries were the most common form of trauma followed by head injury (10%-39% of patients).23 24 25 26 Among patients with TBI, 16% to 53% exhibited “severe” injury; however, the studies did not provide explicit definitions to qualify this categorisation, and did not describe radiological data regarding the extent of injury or the need for neurosurgical intervention.23 26 In the present study, 12% of hospitalised cyclists with head injury had moderate to severe TBI. There was also a high incidence of intracranial haemorrhage involving almost half of the patients. Both these factors were independent predictors of poor GOS score on discharge from hospital. Our results are consistent with the findings in a previous study where 75% of all cycling-related deaths were caused by severe TBI.27 The lack of an independent association with motor vehicle collisions, which constituted only a minority of injuries in this cohort, suggests that recreational cycling at comparatively low speeds can be fatal. Notably, the mechanism of injury for four (50%) of the eight recreational cyclists who died in this study was a loss of balance, followed by a fall on level ground without colliding into another object, person, or motor vehicle.

Previous studies in Hong Kong, the most recent of which was performed >10 years ago, revealed that recreational cyclists rarely wore protective headgear (eg, frequencies of 0.2% to 2.2% among emergency department attendees).24 25 26 Our findings revealed that significantly more patients (15%) wore helmets at the time of injury. Governmental advocacy initiatives for promoting helmet wearing in recent years may have resulted in heightened public awareness regarding the risks of head trauma.28 There is little doubt that helmets are protective. In the past 30 years, several case-control and epidemiological studies have delivered compelling evidence to support the efficacy of bicycle helmet wearing in reducing the risk of life-threatening TBI.29 30 31 32 33 34 35 36 37 In a case-control prospective multicentre study of over 3000 patients, Thompson et al33 noted that helmets (irrespective of design) conferred up to a 74% reduction in TBI during accidents. A subsequent meta-analysis of five studies observed that helmets provided a 63% to 88% reduction in the risk of head, brain, and severe TBI for all ages of cyclists; this included equal levels of protection for collisions involving motor vehicles and collisions due to other causes.38 In the present study, helmet wearing did not reduce TBI severity according to our broadly predefined categories. However, among hospitalised recreational cyclists with mild head injury, helmets did provide significant protection against intracranial haemorrhage, including potentially life-threatening epidural haematomas and ASDHs, as well as skull fractures. Thus, our findings may have important public health implications with regard to introducing mandatory bicycle helmet wearing legislation in the city.

Whether such laws should exist is a particularly divisive issue among public health experts and interest groups.39 40 41 42 43 In Australia, a nation with all-age helmet wearing safety laws, an overall 46% decline in cyclist fatalities per 1 000 000 population has been reported, compared with the pre-legislation period.44 Similar findings were noted in New Zealand: a 67% decline in severe TBI was recorded after the introduction of helmet laws.45 In the US, a significant reduction in paediatric cyclist fatalities involving motor vehicles was observed in states with such laws.46 A meta-analysis of the effectiveness of bicycle helmet legislation revealed that it increased helmet usage, while significantly reducing head injuries and mortality.47 Several medical associations have expressed support for introducing such legislation; these include the World Health Organization, the British Medical Association, the American Medical Association, and the Royal Australasian College of Surgeons.48 49 50 51 However, critics of such compulsory policies have hypothesised that helmets could encourage risk-compensation behaviour, whereby cyclists may be more willing to engage in potentially injurious risks or for motorists to exercise less caution when encountering them.39 52 53 Other reasons for opposition include infringement on individual liberties; some public health scholars have theorised that such laws could discourage cyclists from participating in gainful physical activity.39 40 41 42 54 From a Hong Kong Transport Department survey (5701 respondents) regarding attitudes towards possible helmet law and enforcement measures, the majority of respondents (78%-90%) were in favour of introducing such legislation, especially when riding on carriageways.16 However, among respondents who knew how to ride a bicycle (3933 respondents), 23% declared they would ride less frequently if mandatory helmet wearing was required.16

An inherent limitation of a retrospective study of this nature was the likely under-reporting of the number of patients with sports-related head injuries. In the only existing population-based study of TBI epidemiology that included community-based injuries, 95% were considered mild and 28% of respondents did not seek medical attention.55 Among professional or university-level athletes, under-reporting is more apparent: questionnaire surveys reveal that 31% to 78% of respondents neglected to pursue medical care despite experiencing a concussion during the preceding 12 months.56 57 At the emergency department level, no territory-wide TBI registry exists in Hong Kong; moreover, diagnostic coding to facilitate data retrieval is typically not performed after consultations. Therefore, we could only identify hospitalised patients with sports-related TBI by means of an administrative database that utilised the ICD-10 coding system. However, the validity of such administrative data for research has been questioned.58 Studies have shown significantly lower TBI rates among young adults, men, and patients with less severe injuries when the ICD system was utilised, compared with thorough medical record review.59 Analysis of a population-based TBI sample showed that only 19% of individuals were assigned a TBI-related diagnostic ICD code.60 In addition, a degree of selection bias may have existed because some non-hospitalised helmet-wearing cyclists with mild head injury may have been discharged from the emergency department, mitigating the protective effects of helmet use. This may explain why no considerable differences in outcomes were detected for patients with moderate or severely injured patients. Despite the low rate of helmet use among recreational cyclists (15%), significant protective effects were detected among mildly injured patients with regard to intracranial haemorrhage and skull fracture. This limited participant identification approach also allowed for a pragmatic review of patients with clinically significant TBI who were hospitalised following evaluation by an emergency care physician. Computed tomography scans are generally performed only for hospitalised patients with head injury in our public healthcare system; this approach offered an opportunity to evaluate imaging data for intracranial haemorrhage. Because the ICD coding system for traumatic intracranial haemorrhage reportedly has high sensitivity and specificity (both >80%),59 we adopted this coding outcome as the study’s primary endpoint. Another important limitation was the definition of mild TBI, which affected most patients in this study. The definitions offered by several authorities range from conventional GCS-based criteria such as the US Centers for Disease Control and Prevention,61 and the American College of Surgeons62 to additional symptoms of confusion, memory impairment, transient loss of consciousness, and irritability proposed by the World Health Organisation and the American Congress of Rehabilitation Medicine.18 63 64 A better delineation of these symptoms would have enhanced the identification of patients with “high-risk” mild TBI; however, because these relevant symptoms were often not systematically documented in most medical records retrieved in our study, we used GCS-based criteria to reduce the overall rate of underdiagnosis. Using GOS score on discharge from hospital as a secondary study endpoint, we found that only 31 (7%) patients had unfavourable outcomes. Although statistically significant predictors for TBI were identified, the wide confidence intervals for these predictors suggest that the sample size was insufficient to draw robust conclusions. Finally, we could only retrospectively assess GOS score as a fundamental measure of functional outcome. More sensitive instruments (eg, the extended GOS or the Sport Concussion Assessment Tool65 66) might have been better for assessing the psychosocial and cognitive aspects of TBI, considering that a large proportion of mildly injured cyclists had intracranial haemorrhage.

The incidence of sports-related TBI in Hong Kong is low and cycling is the most frequently associated activity. Almost half of hospitalised recreational cyclists sustained intracranial haemorrhage. Compared with patients who had head injury because of other sports, cyclists are more likely to experience severe consequences. There is evidence that helmet use offers protection against intracranial haemorrhage and skull fracture among cyclists with mild head injury. Cycling is a safe physical activity, but further legislative measures should be introduced to promote and protect the welfare of individuals enjoying this sport.

Concept or design: PYM Woo, E Cheung.
Acquisition of data: PYM Woo, E Cheung, FWY Lau, NWS Law, CKY Mak, P Tan, B Siu, A Wong.
Analysis or interpretation of data: PYM Woo, E Cheung, CKY Mak.
Drafting of the manuscript: All authors.
Critical revision of the manuscript for important intellectual content: All authors.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

This research has not been presented or published in any form prior to submission.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

This study was approved by the Kowloon Central Cluster/Kowloon East Cluster research ethics committee (Ref KCC/KEC-2020-0331). All patients were treated in accordance with the Declaration of Helsinki. Informed consent was obtained from either the patient, next-of-kin, or their legal guardian.

1. Allender S, Cowburn G, Foster C. Understanding participation in sport and physical activity among children and adults: a review of qualitative studies. Health Educ Res 2006;21:826-35. Crossref

2. Warburton DE, Nicol CW, Bredin SS. Health benefits of physical activity: the evidence. CMAJ 2006;174:801-9. Crossref

3. Batty D, Thune I. Does physical activity prevent cancer? Evidence suggests protection against colon cancer and probably breast cancer. BMJ 2000;321:1424-5. Crossref

4. Batty GD, Lee IM. Physical activity and coronary heart disease. BMJ 2004;328:1089-90. Crossref

5. Aune D, Norat T, Leitzmann M, Tonstad S, Vatten LJ. Physical activity and the risk of type 2 diabetes: a systematic review and dose–response meta-analysis. Eur J Epidemiol 2015;30:529-42. Crossref

6. Liu X, Zhang D, Liu Y, et al. Dose-response association between physical activity and incident hypertension: a systematic review and meta-analysis of cohort studies. Hypertension 2017;69:813-20. Crossref

7. Malm C, Jakobsson J, Isaksson A. Physical activity and sports-real health benefits: a review with insight into the Public Health of Sweden. Sports (Basel) 2019;7:127. Crossref

8. Puckett M, Neri A, Underwood JM, Stewart SL. Nutrition and physical activity strategies for cancer prevention in current national comprehensive cancer control program plans. J Community Health 2016;41:1013-20. Crossref

9. Department of Health, Hong Kong SAR Government. Towards 2025: strategy and action plan to prevent and control non-communicable diseases in Hong Kong. 2018 Available from: https://www.chp.gov.hk/files/pdf/saptowards2025_fullreport_en.pdf. Accessed 2 Aug 2020.

10. Theadom A, Starkey NJ, Dowell T, et al. Sports-related brain injury in the general population: an epidemiological study. J Sci Med Sport 2014;17:591-6. Crossref

11. Manley G, Gardner AJ, Schneider KJ, et al. A systematic review of potential long-term effects of sport-related concussion. Br J Sports Med 2017;51:969-77. Crossref

12. Theadom A, Mahon S, Hume P, et al. Incidence of sports-related traumatic brain injury of all severities: a systematic review. Neuroepidemiology 2020;54:192-9. Crossref

13. Daneshvar DH, Nowinski CJ, McKee AC, Cantu RC. The epidemiology of sport-related concussion. Clin Sports Med 2011;30:1-17. Crossref

14. Selassie AW, Wilson DA, Pickelsimer EE, Voronca DC, Williams NR, Edwards JC. Incidence of sport-related traumatic brain injury and risk factors of severity: a population-based epidemiologic study. Ann Epidemiol 2013;23:750-6. Crossref

15. Transport Department, Hong Kong SAR Government. Cycling study. 2004. Available from: https://www.td.gov.hk/filemanager/en/publication/cyclingstudy.pdf. Accessed 22 Aug 2020.

16. Transport Department. Hong Kong SAR Government. Travel Characteristics Survey 2011 Final Report. Available from: https://www.td.gov.hk/filemanager/en/content_4652/tcs2011_eng.pdf. Accessed 22 Aug 2020.

17. Xu P, Dong N, Wong SC, Huang H. Cyclists injured in traffic crashes in Hong Kong: a call for action. PLoS One 2019;14:e0220785. Crossref

18. Servadei F, Teasdale G, Merry G, Neurotraumatology Committee of the World Federation of Neurosurgical Societies. Defining acute mild head injury in adults: a proposal based on prognostic factors, diagnosis, and management. J Neurotrauma 2001;18:657-64. Crossref

19. Huang WY, Wong SH, Sit CH, et al. Results from the Hong Kong’s 2018 report card on physical activity for children and youth. J Exerc Sci Fit 2019;17:14-9. Crossref

20. Harris AW, Jones CA, Rowe BH, Voaklander DC. A population-based study of sport and recreation-related head injuries treated in a Canadian health region. J Sci Med Sport 2012;15:298-304. Crossref

21. Beck AJ, Kee J. The epidemiology of sports head injuries in a rural population. Br J Sports Med 2011;45:A20. Crossref

22. Transport Department, Hong Kong SAR Government. Road Traffic Accident Statistics. Year 2019. Road traffic accidents involving bicycle by location and severity 2019. Available from: https://www.td.gov.hk/en/road_safety/road_traffic_accident_statistics/2019/index.html. Accessed 9 Aug 2020.

23. Lee LL, Yeung KL, Chan JT, Chen RC. A profile of bicycle-related injuries in Tai Po. Hong Kong J Emerg Med 2003;10:81-7.Crossref

24. Yeung JH, Leung CS, Poon WS, Cheung NK, Graham CA, Rainer TH. Bicycle related injuries presenting to a trauma centre in Hong Kong. Injury 2009;40:555-9. Crossref

25. Ng CP, Siu AY, Chung CH. Bicycle-related injuries: a local scene. Hong Kong J Emerg Med 2001;8:78-83. Crossref

26. Sze NN, Tsui KL, Wong SC, So FL. Bicycle-related crashes in Hong Kong: is it possible to reduce mortality and severe injury in the metropolitan area? Hong Kong J Emerg Med 2011;18:136-43. Crossref

27. Rivara FP, Thompson DC, Thompson RS. Epidemiology of bicycle injuries and risk factors for serious injury. Inj Prev 1997;3:110-4. Crossref

28. Transport Department, Hong Kong SAR Government. Be a smart cyclist. Wear bicycle helmet always. 2020. Available from: https://www.td.gov.hk/filemanager/en/content_4551/221201047_leaflet_a.pdf. Accessed 21 Aug 2020.

29. Attewell RG, Glase K, McFadden M. Bicycle helmet efficacy: a meta-analysis. Accid Anal Prev 2001;33:345-52. Crossref

30. Maimaris C, Summers CL, Browning C, Palmer CR. Injury patterns in cyclists attending an accident and emergency department: a comparison of helmet wearers and non-wearers. BMJ 1994;308:1537-40. Crossref

31. McDermott FT, Lane JC, Brazenor GA, Debney EA. The effectiveness of bicyclist helmets: a study of 1710 casualties. J Trauma 1993;34:834-44. Crossref

32. Thomas S, Acton C, Nixon J, Battistutta D, Pitt WR, Clark R. Effectiveness of bicycle helmets in preventing head injury in children: case-control study. BMJ 1994;308:173-6. Crossref

33. Thompson DC, Rivara FP, Thompson RS. Effectiveness of bicycle safety helmets in preventing head injuries. A case-control study. JAMA 1996;276:1968-73. Crossref

34. Thompson RS, Rivara FP, Thompson DC. A case-control study of the effectiveness of bicycle safety helmets. N Engl J Med 1989;320:1361-7. Crossref

35. Strotmeyer SJ, Behr C, Fabio A, Gaines BA. Bike helmets prevent pediatric head injury in serious bicycle crashes with motor vehicles. Inj Epidemiol 2020;7(Suppl 1):24. Crossref

36. Sethi M, Heidenberg J, Wall SP, et al. Bicycle helmets are highly protective against traumatic brain injury within a dense urban setting. Injury 2015;46:2483-90. Crossref

37. Persaud N, Coleman E, Zwolakowski D, Lauwers B, Cass D. Nonuse of bicycle helmets and risk of fatal head injury: a proportional mortality, case-control study. CMAJ 2012;184:E921-3. Crossref

38. Thompson DC, Rivara FP, Thompson R. Helmets for preventing head and facial injuries in bicyclists. Cochrane Database Syst Rev 2000;1999(2):CD001855. Crossref

39. Robinson DL. No clear evidence from countries that have enforced the wearing of helmets. BMJ 2006;332:722-5. Crossref

40. Hong Kong Cycling Alliance. Helmets. Available from: http://hkcyclingalliance.org/on-the-road/helmets. Accessed 21 Aug 2020.

41. Bateman-House A. Bikes, helmets, and public health: decision-making when goods collide. Am J Public Health 2014;104:986-92. Crossref

42. Goldacre B, Spiegelhalter D. Bicycle helmets and the law. BMJ 2013;346:f3817. Crossref

43. Rivara FP, Thompson DC, Patterson MQ, Thompson RS. Prevention of bicycle-related injuries: helmets, education, and legislation. Annu Rev Public Health 1998;19:293-318. Crossref

44. Olivier J, Boufous S, Grzebieta R. The impact of bicycle helmet legislation on cycling fatalities in Australia. Int J Epidemiol 2019;48:1197-203. Crossref

45. Tin Tin S, Woodward A, Ameratunga S. Injuries to pedal cyclists on New Zealand roads, 1988-2007. BMC Public Health 2010;10:655. Crossref

46. Meehan WP 3rd, Lee LK, Fischer CM, Mannix RC. Bicycle helmet laws are associated with a lower fatality rate from bicycle-motor vehicle collisions. J Pediatr 2013;163:726-9. Crossref

47. Macpherson A, Spinks A. Bicycle helmet legislation for the uptake of helmet use and prevention of head injuries. Cochrane Database Syst Rev 2008;2008(3):CD005401. Crossref

48. World Health Organisation. Why are helmets needed? 2006. Available from: https://www.who.int/roadsafety/projects/manuals/helmet_manual/1-Why.pdf. Accessed 13 Oct 2021.

49. British Medical Association. Healthy transport=healthy lives. 2012. Available from: https://thepep.unece.org/sites/default/files/2017-06/Healthy%20transport%20healthy%20lives%20Britsh%20Medical%20Association.pdf. Accessed 13 Oct 2021.

50. American Medical Association. Helmets for riders of motorized and non-motorized cycles H-10.964. Available from: https://policysearch.ama-assn.org/policyfinder/detail/bicycle helmets?uri=%2FAMADoc%2FHOD.xml-0-3.xml. Accessed 20 Sep 2021.

51. McDermott FT. Helmet efficacy in the prevention of bicyclist head injuries: Royal Australasian College of Surgeons initiatives in the introduction of compulsory safety helmet wearing in Victoria, Australia. World J Surg 1992;16:379-83. Crossref

52. Gamble T, Walker I. Wearing a bicycle helmet can increase risk taking and sensation seeking in adults. Psychol Sci 2016;27:289-94. Crossref

53. Walker I. Drivers overtaking bicyclists: objective data on the effects of riding position, helmet use, vehicle type and apparent gender. Accid Anal Prev 2007;39:417-25. Crossref

54. Hooper C, Spicer J. Liberty or death; don’t tread on me. J Med Ethics 2012;38:338-41. Crossref

55. Theadom A, Barker-Collo S, Feigin VL, et al. The spectrum captured: a methodological approach to studying incidence and outcomes of traumatic brain injury on a population level. Neuroepidemiology 2012;38:18-29. Crossref

56. Delaney JS, Lamfookon C, Bloom GA, Al-Kashmiri A, Correa JA. Why university athletes choose not to reveal their concussion symptoms during a practice or game. Clin J Sport Med 2015;25:113-25. Crossref

57. Meehan WP 3rd, Mannix RC, O’Brien MJ, Collins MW. The prevalence of undiagnosed concussions in athletes. Clin J Sport Med 2013;23:339-42. Crossref

58. van Walraven C, Bennett C, Forster AJ. Administrative database research infrequently used validated diagnostic or procedural codes. J Clin Epidemiol 2011;64:1054-9. Crossref

59. Carroll CP, Cochran JA, Guse CE, Wang MC. Are we underestimating the burden of traumatic brain injury? Surveillance of severe traumatic brain injury using Centers for Disease Control International Classification of Disease, Ninth Revision, Clinical Modification, traumatic brain injury codes. Neurosurgery 2012;71:1064-70. Crossref

60. Barker-Collo S, Theadom A, Jones K, Feigin VL, Kahan M. Accuracy of an International Classification of Diseases Code Surveillance System in the identification of traumatic brain injury. Neuroepidemiology 2016;47:46-52. Crossref

61. Centers for Disease Control and Prevention. Glasgow Coma Scale. Available from: https://www.cdc.gov/masstrauma/resources/gcs.pdf. Accessed 20 Sep 2021.

62. American College of Surgeons. Best practices in the management of traumatic brain injury. Available from: https://www.facs.org/-/media/files/quality-programs/trauma/tqip/tbi_guidelines.ashx. Accessed 20 Sep 2021.

63. Carroll LJ, Cassidy JD, Holm L, Kraus J, Coronado VG, WHO Collaborating Centre Task Force on Mild Traumatic Brain Injury. Methodological issues and research recommendations for mild traumatic brain injury: the WHO Collaborating Centre Task Force on Mild Traumatic Brain Injury. J Rehabil Med 2004;(43 Suppl):113-25. Crossref

64. Kay T, Harrington DE, Adams R, et al. Definition of mild traumatic brain injury. J Head Trauma Rehabil 1993;8:86-7. Crossref

65. Kean J, Malec JF. Towards a better measure of brain injury outcome: new measures or a new metric? Arch Phys Med Rehabil 2014;95:1225-8. Crossref

66. McCrory P, Meeuwisse WH, Aubry M, et al. Consensus statement on concussion in sport: the 4th International Conference on Concussion in Sport, Zurich, November 2012. J Athl Train 2013;48:554-75. Crossref

Klebsiella pneumonia causes various clinically important infections. In 2017, K pneumoniae was the third most common isolate in intensive care units (ICUs) and second most common isolate in all patients in the Hong Kong East Cluster.1

The emergence of multidrug-resistant K pneumoniae infections is an increasing concern.2 There have been outbreaks of extended-spectrum beta-lactamase (ESBL)–producing strains, carbapenem-resistant (CR) and carbapenemase-producing (CP) strains, and hypervirulent K pneumoniae infections both in Hong Kong and worldwide.2 3 4 In Greece and Italy, CP K pneumoniae comprises 68.3% of all K pneumoniae strains.5

The importance of appropriate early empirical antibiotics has been repeatedly emphasised in the management of septic shock by the Surviving Sepsis Campaign 2016.6 Previously, we highlighted the importance of appropriate early antibiotics for successful patient outcomes.4 To the best of our knowledge, risk factors for receiving inappropriate empirical antibiotics have not yet been explored. Therefore, in the present study, we aimed to evaluate the impact of appropriate empirical antibiotics on outcomes in patients with K pneumoniae bacteraemia; we also performed subgroup analysis on ICU patients with K pneumoniae bacteraemia.

We conducted a retrospective analysis of adult patients with K pneumoniae bacteraemia who were admitted to Pamela Youde Nethersole Eastern Hospital within the period from 1 January 2009 to 30 June 2017. Pamela Youde Nethersole Eastern Hospital is a 1700-bed hospital in Hong Kong which provides extensive services except cardiothoracic surgery, transplant surgery, and burns. Patients were excluded if they were aged <18 years or had incomplete information. Patient medical records were reviewed, as were data in clinical management and clinical information systems (IntelliVue Clinical Information Portfolio; Philips Medical, Amsterdam, The Netherlands). The clinical management system is a database that stores patients’ demographics, laboratory results, and drug administration records for all public hospitals in Hong Kong. In the event of missing data, medical records were reviewed manually. Baseline demographics, clinical characteristics, and microbiological data for all included patients were identified from the above databases and medical records.

Disease severity was quantified using the maximum Sequential Organ Failure Assessment (SOFA) score.7 The following clinical outcome data were investigated: use of invasive organ supports (eg, inotropic use, mechanical ventilation, and renal replacement therapies), ICU and hospital lengths of stay (LOSs), ICU ventilator duration, and mortality. The primary outcome was 90-day all-cause mortality; secondary outcomes were ICU and hospital mortalities, ICU and hospital LOSs, and ICU ventilator duration.

Klebsiella pneumoniae bacteraemia was defined as the growth of K pneumoniae in one or more blood cultures. If more than one positive blood culture result was recorded, only the first sample was included. Empirical antibiotic treatment was defined as the antibiotic used within 24 hours after a culture sample was collected. The empirical antibiotic treatment was considered appropriate if at least one of the antibiotic agents was consistent with the in vitro susceptibility results.8 9 10 Community-acquired infection was defined as K pneumoniae identified in patients upon admission or within 48 hours after admission; hospital-acquired infection was defined as K pneumoniae identified in patients at >48 hours after admission.11 Hepatobiliary sepsis comprised liver abscess, cholangitis, and cholecystitis; gastrointestinal sepsis comprised spontaneous bacterial peritonitis, peritonitis caused by bowel perforation, and intra-abdominal abscesses (excluding liver abscess). Medical co-morbidities (eg, diabetes mellitus, cirrhosis, congestive heart failure, chronic renal impairment, haematological malignancy and solid tumour) were defined in accordance with the International Classification of Disease coding. Prior steroid use was defined as oral or intravenous steroid consumption within 30 days before the index positive blood culture result. Chemotherapy use was defined as oral or intravenous infusion of biological agents or chemotherapy administered within 30 days before the index positive blood culture result. Any antibiotics usage within 30 days before the index hospital admission was regarded as prior antibiotics usage.

Blood cultures were incubated and processed. Blood culture results were considered negative if no positive growth occurred after 5 to 7 days. Susceptibility interpretation was based on Clinical and Laboratory Standards Institute interpretive criteria. The ESBL testing was based on Clinical and Laboratory Standards Institute testing criteria12 13 14 15 16 17 18 or the double-disk synergy test described by Jarlier et al.19 Regarding CR K pneumoniae, the isolates were sent to the Public Health Laboratory Centre of Hong Kong if further genetic testing was required to confirm carbapenemase production. Multiplex real-time polymerase chain reaction assays were performed to detect Classes A, B, and D carbapenemase gene targets.

We compared the characteristics and clinical parameters between patients treated with appropriate and inappropriate empirical antibiotics, as well as between 90-day survivors and non-survivors. Results are expressed as the median ± interquartile range (IQR) or as the number (percentage) of patients, as appropriate.

Categorical variables were compared by the Pearson Chi squared test or Fisher’s exact test, as appropriate for univariate analysis. Continuous variables were compared by using the Mann–Whitney U test. Variables with P<0.2 in univariate analysis or with clinical significance from previous studies were included in the multivariate analysis. Independent predictors for 90-day mortality were assessed by Cox regression analysis. Logistic regression analysis was used to assess independent predictors for receiving appropriate and inappropriate empirical antibiotics. Post hoc analysis was performed for patients with chronic renal failure and resistant organisms.

The Statistical Package for Social Sciences (Windows version 24.0; IBM Corp, Armonk [NY], US) was used to perform statistical analyses.

During the 8.5-year study period, we identified 984 patients with K pneumoniae bacteraemia; of these, 686 (69.7%) and 298 (30.3%) received appropriate and inappropriate empirical antibiotics, respectively. Table 1 shows the baseline demographics of patients who received appropriate and inappropriate empirical antibiotics. The median patient age was 75 years (IQR=63-83 years). The most common types of infection were hepatobiliary tract infection (33.1%), urosepsis (24.4%), and respiratory tract infection (18.4%). The overall 90-day mortality was 32.7%, hospital mortality was 22.5%, and median hospital LOS was 10.68 days (5.38-22.81 days, P<0.001).

Univariate analysis (Table 1) revealed that the risk factors for receiving inappropriate empirical antibiotics were age >65 years (P=0.044), chronic renal impairment (P<0.001), respiratory tract infection (P=0.002), mechanical ventilation (P=0.001), CR or ESBL-producing isolates (both P<0.001), and higher total SOFA score (P=0.048). Hepatobiliary sepsis was associated with a higher rate of appropriate empirical antibiotic treatment (P=0.009).

Table 2 demonstrates the logistic regression analysis of the predictors for the appropriateness of empirical antibiotics for all patients and ICU subgroup. These include older patients (P=0.010), chronic kidney disease (P=0.007), mechanical ventilation (P=0.005), respiratory tract infection (P=0.034), and either carbapenem resistance or ESBL production (P<0.001). Table 1 shows that the 90-day and hospital mortalities were significantly higher in patients with inappropriate empirical treatment (both P<0.001). Moreover, the hospital LOS was shorter in patients who received inappropriate empirical antibiotics (P<0.001).

Receipt of inappropriate empirical antibiotics was associated with higher hospital mortality; this finding was consistent in ICU subgroup (56% vs 23%). The absolute risk increases in mortality associated with the receipt of inappropriate empirical antibiotics were 18.3% and 33% in ICU subgroup. The number of inappropriate empirical antibiotics associated with each mortality was five in the ICU subgroup and three in all patients.

Antibiograms showing the proportions of non-susceptible K pneumoniae isolates are described in the Online Supplementary Table 1. Twenty (2.0%) patients had CR K pneumoniae bacteraemia, but molecular tests in the Public Health Laboratory Centre of Hong Kong revealed that none of them had CP strains. Overall, 113 (11.5%) patients had ESBL-producing infections.

Post hoc analysis revealed that patients with chronic renal failure were more likely to have ESBL infections (18.3% vs 10.5%; P=0.011) and CR infections (0.056% vs 0.015%; P=0.003).

The 90-day all-cause mortalities were 32.7% in all patients with K pneumoniae bacteraemia and 34.6% in the ICU subgroup. Univariate analysis (Table 3) showed that 90-day non-survivors were more likely to be aged >65 years (P<0.001), admitted through the Department of Medicine (P<0.001), have septic shock (P=0.005), have a higher total SOFA score (P<0.001), receive inappropriate or no empirical antibiotics (P<0.001 and P<0.001, respectively), have solid tumour (P<0.001), have respiratory tract infection (P<0.001), be mechanically ventilated (P<0.001), have gastrointestinal infections (P<0.001), and require renal replacement therapy (P=0.044). Patients with diabetes (P=0.001), hepatobiliary sepsis (P<0.001), and urosepsis (P<0.001) had lower 90-day mortalities.

Table 4 shows the Cox regression analysis of predictors for 90-day mortality. Independent predictors for increased 90-day mortality were respiratory tract infection (P<0.001), gastrointestinal infection (P<0.001), inappropriate empirical antibiotics (P<0.001), older age (P<0.001), solid tumour (P<0.001), patients admitted through the Department of Medicine (P=0.012), and higher total SOFA score (P<0.001). Patients with diabetes had lower 90-day mortality (P=0.001). The Figure shows the Kaplan–Meier survival plot and according to log rank analysis, the results demonstrated a statistically significant improvement in survival among patients who received appropriate empirical antibiotics (P<0.001).

Online Supplementary Table 2 shows the demographics for inappropriate empirical antibiotics by logistic regression analysis in ICU subgroup. Overall, 205 (20.8%) patients required intensive care; among them, 148 (72.2%) received appropriate empirical antibiotics, while 57 (27.8%) received inappropriate or no empirical antibiotics. The median patient age was 68 years (IQR=58-78 years). The commonest infection was hepatobiliary infections (42.9%), followed by respiratory tract (23.4%) and urosepsis (14.6%). Furthermore, 82.4% of the patients had septic shock, 33.2% received renal replacement therapy, 60% received mechanical ventilation, and 78.5% had vasopressor use. The ICU and overall 90-day mortalities were 18.5% and 34.6%, respectively. The receipt of inappropriate empirical antibiotics was significantly associated with higher 90-day mortality (59.6% vs 25.0%; P<0.001), higher ICU mortality (35.1% vs 12.2%; P<0.001), higher hospital mortality (56.1% vs 23.0%; P<0.001), and longer ventilator duration (2 d vs 1 d, P=0.026).

Cox regression analysis showed that the receipt of inappropriate or no empirical antibiotics (P<0.001; Table 4) was the strongest independent predictor of 90-day mortality in critically ill patients with K pneumoniae bacteraemia. Other independent predictors were congestive heart failure (P=0.02), admitted through the Department of Medicine (P=0.016), and a higher total SOFA score (P<0.001). Ninety-day non-survivors had longer hospital LOS (P<0.001).

Among all patients in this study, 686 (69.7%) received appropriate empirical antibiotics. Furthermore, 148 (72.2%) critically ill patients received appropriate empirical antibiotics. Importantly, 93 (9.5%) patients did not receive any empirical antibiotics. The median hospital LOS for these 93 patients was significantly shorter than the LOS for all patients in the study. We performed event-free survival analysis and found that the LOSs were similar in both groups; these findings suggested that patients who received inappropriate antibiotics had more severe disease and earlier death, leading to a shorter hospital LOS.

Micek et al20 found that prior antibiotic exposure was a risk factor for inappropriate empirical therapy. Lautenbach et al21 described a positive correlation between the total cumulative dose of antibiotics and ESBL K pneumoniae infection. In our cohort, prior antibiotics exposure was not significantly correlated with inappropriate empirical antibiotic treatment; furthermore, prior antibiotics exposure was not associated with ESBL infections. We examined the presence of prior antibiotics use 30 days prior to positive blood culture results, but information regarding the total cumulative antibiotics exposure in terms of dosing and duration were not available. Other information was unavailable regarding antibiotics prescribed outside hospital settings.

Patients with chronic renal failure are reportedly more prone to developing resistant infections.22 They were at greater risk of receiving inappropriate antibiotics. Additionally, hospital-acquired infection has been associated with a higher rate of inappropriate empirical antibiotic treatment and higher 90-day mortality.11

The rates of CR and CP K pneumoniae bacteraemia were much lower than the rates reported in other endemic countries.5 A study from Shanghai reported that approximately 22% of all patients with K pneumoniae bacteraemia had CR infections; moreover, approximately 59% of the isolates were CR infections in patients who required intensive care.23

The prevalence of ESBL K pneumoniae bacteraemia in Hong Kong is low.4 The ESBL infection rate in our cohort was 11.5%, similar to previous local studies.1 24 Another study in China demonstrated a much higher rate of ESBL infections (approximately 39%).25 Both ESBL and CR infections were not associated with increased mortality in our study in either the overall patient population or the ICU subgroup. However, ESBL and CR infections were significantly associated with longer hospital LOS.

The antibiotics sensitivity in our cohorts was comparable with the antibiogram data in the IMPACT guideline.1 Ampicillin-sulbactam or amoxicillin-clavulanate is recommended as the drug of choice for treatment of K pneumomiae infections in local guidelines.1 In our cohort, more than 80% of K pneumomiae isolates were susceptible to amoxicillin-clavulanate, indicating that it is a reasonable option for broad empirical coverage. Add-on therapy with aminoglycoside improves the coverage of this regimen, because more than 90% of the isolates in our cohort were sensitive to either gentamicin or amikacin. While combination therapy improves the chance of successful empirical therapy, routine use of combination therapy remains controversial.6

The 90-day all-cause mortalities in our study were comparable with the findings in previous studies.24 26 Respiratory tract infections and gastrointestinal infections have consistently been associated with a worse outcome and greater mortality.27 28 In contrast, urosepsis and hepatobiliary sepsis have repeatedly associated with better survival outcomes.29 These sources of infections may be amendable to percutaneous, endoscopic, or surgical drainage, allowing more rapid and definitive control of sepsis, which leads to better survival.28

A greater proportion of patients with respiratory tract infection did not receive any empirical antibiotics, which might explain the worse outcomes in these patients. The symptoms and signs of respiratory tract infection may overlap with other diseases (eg, heart failure) and treatment may be delayed while waiting for laboratory results and imaging. Given the greater proportion of patients with respiratory tract infections who did not receive any empirical antibiotics, there is a need for early consideration of empirical antibiotics in patients with signs and symptoms of respiratory tract infections.

Consistent with the findings of previous studies, we found that older age, solid tumour, and admission through the Department of Medicine were factors associated with higher 90-day mortality.30 31 Patients with these factors were more likely to have other pre-existing co-morbidities, worse premorbid functional status, and be institutionalised. They may also have received a more conservative approach to treatment overall.

Diabetes mellitus is well-known to predispose patients to infections, such that affected patients are reportedly 4.4-fold more likely to develop bloodstream infection.32 Similar to the findings by Peralta et al,32 we did not find increased mortality among patients with diabetes. In our cohort, diabetic patients were more likely to had urosepsis than respiratory tract infections. Greater frequency of urosepsis and smaller frequency of respiratory tract infection may have an overall positive effect on survival. Moreover, advances in diabetes care in recent decades (eg, newer generations of medication and integrated multidisciplinary care) have led to neutral effects of diabetes on short-term mortality in patients undergoing major operations and patients with sepsis.34 34 35 Glycaemic status, haemoglobin A1c levels, and diabetes severity were not available in the present study; thus, we could not delineate how diabetes control affected bacteraemia outcomes. Our results are limited to demonstrating that 90-day mortality and diabetes have a greater impact on long-term survival, although this conclusion may not be apparent in the current study.33

In addition to mortality, we demonstrated that the inappropriate use of empirical antibiotics was associated with longer ICU ventilator duration, which leads to greater costs and more extensive use of ICU resources. A large retrospective cohort from the US regarding Enterobacteriaceae infections showed that each additional day without appropriate antibiotics was associated with an increased hospital expenditure of US$750 and an increased risk of 30-day readmission.36

To the best of our knowledge, the appropriateness of empirical antibiotics has consistently been identified as one of the strongest independent predictors of 90-day mortality in all affected patients and in critically ill patients.37 38 39 Kumar et al8 demonstrated a fivefold increase in mortality among patients with sepsis who received inappropriate initial antibiotics. In this study, we demonstrated a twofold increase in mortality in all patients with K pneumoniae bacteraemia, as well as a threefold increase in mortality in critically ill patients with K pneumoniae bacteraemia. Furthermore, receipt of inappropriate initial antibiotics was the strongest independent predictor of 90-day mortality in the ICU subgroup. Zilberberg et al40 found that the detrimental effect of inappropriate empirical antibiotic treatment could not be corrected despite subsequent targeted antibiotic treatments. A meta-analysis revealed a slower rate of bacterial clearance and increased treatment failure rate when patients were administered inappropriate empirical antibiotic therapy.41 Appropriate early antibiotics allow rapid reduction of bacterial load and modulate host defences, thus alleviating some organ dysfunction.42 A more pragmatic approach would be the early administration of broader-spectrum empirical antibiotics and timely de-escalation, according to bacterial sensitivity and the patient’s clinical progression.

The chain of sepsis management begins during the first encounter in the Emergency Department and in general hospital wards, rather than in the ICU. Early administration of antibiotics within 3 hours of hospital admission and aggressive sepsis care (beginning in the Emergency Department) can improve survival.42 Another prospective observational study demonstrated that the adequate use of empirical antibiotics prior to ICU admission was the strongest independent factor associated with survival.29 Extensive efforts are needed to facilitate early, appropriate use of empirical antibiotics, including the use of a current antibiogram, implementation of multidisciplinary sepsis management guidelines, and establishment of protocols among pharmacists, microbiologists, clinicians, and nurses.

To the best of our knowledge, this is the largest study of K pneumoniae bacteraemia in Hong Kong and in the Asia-Pacific region. By including both ICU and general ward patients, we achieved clarity regarding the diverse characteristics of K pneumoniae bacteraemia. We also identified many potential predictors of K pneumoniae bacteraemia–related mortality, based on our extensive literature review and previous publications, then tested these predictors using real-world patient data. By evaluating 90-day mortality, hospital LOS, and ventilator duration, our study more comprehensively evaluated immediate and longer-term complications of bacteraemia; it also provided information for future studies of cost-effectiveness in terms of empirical antibiotics and resource utilisation. Finally, we used the maximum total SOFA score for severity assessment. This score has been repeatedly validated in determining disease severity and predicting mortality in critically ill patients.43 44 45

This retrospective study was subject to potential confounding factors, including selection bias that could not be completely eliminated from the analysis. Notably, the results of this single-centre study may not be generalisable to other countries with higher CR or CP K pneumoniae infections. Furthermore, this study encompassed a long duration, in which the definitions of sepsis or septic shock might have changed.45 The care of patients with sepsis evolved over time, including advances in source control by percutaneous and endoscopic means that potentially improved patient survival. Nonetheless, the role of empirical antibiotics in patients with sepsis remains an essential sepsis consideration.

Antibiotic pharmacodynamics also has a fundamental role in bacteraemia treatment. In this study, we could not collect information regarding the timing of first-dose antibiotics, time to appropriate antibiotics, duration of antibiotics, or time to surgical treatments. Moreover, antibiotic stewardship and therapeutic de-escalation efforts, as well as their impacts on patient outcomes, were not assessed in the present study. Future studies may be needed concerning the prolonged infusion of beta-lactam antibiotics, use of combination therapies, duration of antibiotics, and serum monitoring of antibiotics.

The receipt of inappropriate empirical antibiotics led to twofold greater 90-day mortality in patients with K pneumoniae bacteraemia. In critically ill patients, inappropriate use of empirical antibiotics was the strongest independent predictor of mortality. Early identification of high-risk patients and administration of appropriate empirical antibiotics can improve patient outcomes.

Concept or design: MY Man, HP Shum.
Acquisition of data: MY Man, HP Shum.
Analysis or interpretation of data: MY Man, HP Shum.
Drafting of the manuscript: MY Man.
Critical revision of the manuscript for important intellectual content: WW Yan.

All authors read and approved the final manuscript. All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

This research was presented by KC Li as an abstract at the 31st Annual Congress of the European Society of Intensive Care Medicine (ESICM), 21-24 October 2018, Paris, France.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

This study was approved by the Hong Kong Easter Cluster Ethics Committee of the Hospital Authority (HKECREC-2018-018). The requirement for written informed consent was waived because of the retrospective nature of the study.

1. Ho PL, Wu TC, Chao DV, et al, editors. Reducing bacterial resistance with IMPACT. Interhospital multi-disciplinary programme on antimicrobial chemotherapy. 5th edition. 2017. Available from: https://www.chp.gov.hk/files/pdf/reducing_bacterial_resistance_with_impact.pdf. Accessed 7 Jun 2020.

2. Ben-David D, Kordevani R, Keller N, et al. Outcome of carbapenem resistant Klebsiella pneumoniae bloodstream infections. Clin Microbiol Infect 2012;18:54-60. Crossref

3. Wong MH, Shum HP, Chen JH, et al. Emergence of carbapenem-resistant hypervirulent Klebsiella pneumoniae. Lancet Infect Dis 2018;18:24. Crossref

4. Man MY, Shum HP, Chan YH, et al. Clinical predictors and outcomes of Klebsiella pneumoniae bacteraemia in a regional hospital in Hong Kong. J Hosp Infect 2017;97:35-41.Crossref

5. Alicino C, Giacobbe DR, Orsi A, et al. Trends in the annual incidence of carbapenem-resistant Klebsiella pneumoniae bloodstream infections: a 8-year retrospective study in a large teaching hospital in northern Italy. BMC Infect Dis 2015;15:415. Crossref

6. Rhodes A, Evans LE, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock: 2016. Intensive Care Med 2017;43:304-77.Crossref

7. Ferreira FL, Bota DP, Bross A, Mélot C, Vincent JL. Serial evaluation of the SOFA score to predict outcome in critically ill patients. JAMA 2001;286:1754-8. Crossref

8. Kumar A, Ellis P, Arabi Y, et al. Initiation of inappropriate antimicrobial therapy results in a fivefold reduction of survival in human septic shock. Chest 2009;136:1237-48. Crossref

9. Vazquez-Guillamet C, Scolari M, Zilberberg MD, Shorr AF, Micek ST, Kollef M. Using the number needed to treat to assess appropriate antimicrobial therapy as a determinant of outcome in severe sepsis and septic shock. Crit Care Med 2014;42:2342-9. Crossref

10. Micek ST, Welch EC, Khan J, et al. Empiric combination antibiotic therapy is associated with improved outcome against sepsis due to Gram-negative bacteria: a retrospective analysis. Antimicrob Agents Chemother 2010;54:1742-8. Crossref

11. Juan CH, Chuang C, Chen CH, Li L, Lin YT. Clinical characteristics, antimicrobial resistance and capsular types of community-acquired, healthcare-associated, and nosocomial Klebsiella pneumoniae bacteremia. Antimicrob Resist Infect Control 2019;8:1. Crossref

12. Clinical and Laboratory Standards Institute. M100-S16 Performance Standards for Antimicrobial Susceptibility Testing; Sixteenth Informational Supplement. Clinical and Laboratory Standards Institute, Wayne, PA; January 2006.

13. Clinical and Laboratory Standards Institute. M100-S17 Performance Standards for antimicrobial susceptibility testing; Seventeenth Informational Supplement. Clinical and Laboratory Standards Institute, Wayne, PA; January 2007.

14. Clinical and Laboratory Standards Institute. M100-S19 Performance Standards for antimicrobial susceptibility testing; Nineteenth Informational Supplement. Clinical and Laboratory Standards Institute, Wayne, PA; January 2009.

15. Clinical and Laboratory Standards Institute. M100-S20 Performance Standards for Antimicrobial Susceptibility Testing; Twentieth Informational Supplement. Clinical and Laboratory Standards Institute, Wayne, PA; January 2010.

16. Clinical and Laboratory Standards Institute. M100-S23 Performance Standards for Antimicrobial Susceptibility Testing; Twenty-Third Informational Supplement. Clinical and Laboratory Standards Institute, Wayne, PA; January 2013.

17. Clinical and Laboratory Standards Institute. M100-S25 Performance Standards for Antimicrobial Susceptibility Testing; Twenty-Fifth Informational Supplement. Clinical and Laboratory Standards Institute, Wayne, PA; January 2015.

18. Clinical and Laboratory Standards Institute. M100-S26 Performance Standards for Antimicrobial Susceptibility Testing; Twenty-Sixth Informational Supplement. Clinical and Laboratory Standards Institute, Wayne, PA; January 2016.

19. Jarlier V, Nicolas MH, Fournier G, Philippon A. Extended broad-spectrum beta-lactamases conferring transferable resistance to newer beta-lactam agents in Enterobacteriaceae: hospital prevalence and susceptibility patterns. Rev Infect Dis 1988;10:867-78. Crossref

20. Micek ST, Hampton N, Kollef M. Risk factors and outcomes for ineffective empiric treatment of sepsis caused by Gram-negative pathogens: stratification by onset of infection. Antimicrob Agents Chemother 2017;62:e01577-17.

21. Lautenbach E, Patel JB, Bilker WB, Edelstein PH, Fishman NO. Extended-spectrum beta-lactamase-producing Escherichia coli and Klebsiella pneumoniae: risk factors for infection and impact of resistance on outcomes. Clin Infect Dis 2001;32:1162-71. Crossref

22. Eilertson B, Cober E, Richter SS, et al. Carbapenem-resistant Enterobacteriaceae infections in patients on renal replacement therapy. Open Forum Infect Dis 2017;4:ofx216.

23. Tian L, Tan R, Chen Y, et al. Epidemiology of Klebsiella pneumoniae bloodstream infections in a teaching hospital: factors related to the carbapenem resistance and patient mortality. Antimicrob Resist Infect Control 2016;5:48. Crossref

24. Pau CK, Ma FF, Ip M, You JH. Characteristics and outcomes of Klebsiella pneumoniae bacteraemia in Hong Kong. Infect Dis (Lond) 2015;47:283-8. Crossref

25. Li L, Huang H. Risk factors of mortality in bloodstream infections caused by Klebsiella pneumonia: a singlecenter retrospective study in China. Medicine (Baltimore) 2017;96:e7924.

26. Melot B, Colot J, Guerrier G. Bacteremic community-acquired infections due to Klebsiella pneumoniae: clinical and microbiological presentation in New Caledonia, 2008-2013. Int J Infect Dis 2015;41:29-31. Crossref

27. Russo A, Falcone M, Gutiérrez-Gutiérrez B, et al. Predictors of outcome in patients with severe sepsis or septic shock due to extended-spectrum beta-lactamase-producing Enterobacteriaceae. Int J Antimicrob Agents 2018;52:577-85.Crossref

28. Zilberberg MD, Shorr AF, Micek ST, Vazquez-Guillamet C, Kollef MH. Multi-drug resistance, inappropriate initial antibiotic therapy and mortality in Gram-negative severe sepsis and septic shock: a retrospective cohort study. Crit Care 2014;18:596. Crossref

29. Garnacho-Montero J, Gutiérrez-Pizarraya A, Escoresca-Ortega A, Fernández-Delgado E, López-Sánchez JM. Adequate antibiotic therapy prior to ICU admission in patients with severe sepsis and septic shock reduces hospital mortality. Crit Care 2015;19:302. Crossref

30. Antonio M, Gudiol C, Royo-Cebrecos C, Grillo S, Ardanuy C, Carratalà J. Current etiology, clinical features and outcomes of bacteremia in older patients with solid tumors. J Geriatr Oncol 2019;10:246-51. Crossref

31. Tseng CP, Wu HS, Wu TH, Lin YT, Fung CP. Clinical characteristics and outcome of patients with community-onset Klebsiella pneumoniae bacteremia requiring intensive care. J Microbiol Immunol Infect 2013;46:217-23. Crossref

32. Peralta G, Sánchez MB, Roiz MP, Garrido JC, Teira R, Mateos F. Diabetes does not affect outcome in patients with Enterobacteriaceae bacteremia. BMC Infect Dis 2009;9:94. Crossref

33. Filsoufi F, Rahmanian PB, Castillo JG, Mechanick JI, Sharma SK, Adams DH. Diabetes is not a risk factor for hospital mortality following contemporary coronary artery bypass grafting. Interact Cardiovasc Thorac Surg 2007;6:753-8. Crossref

34. Teo BJ, Chong HC, Yeo W, Tan AH. The impact of diabetes on patient outcomes after total knee arthroplasty in an Asian population. J Arthroplasty 2018;33:3186-9. Crossref

35. Stegenga ME, Vincent JL, Vail GM, et al. Diabetes does not alter mortality or hemostatic and inflammatory responses in patients with severe sepsis. Crit Care Med 2010;38:539-45. Crossref

36. Zilberberg MD, Nathanson BH, Sulham K, Fan W, Shorr AF. 30-day readmission, antibiotics costs and costs of delay to adequate treatment of Enterobacteriaceae UTI, pneumonia, and sepsis: a retrospective cohort study. Antimicrob Resist Infect Control 2017;6:124. Crossref

37. Gutiérrez-Gutiérrez B, Salamanca E, de Cueto M, et al. Effect of appropriate combination therapy on mortality of patients with bloodstream infections due to carbapenemase-producing Enterobacteriaceae (INCREMENT): a retrospective cohort study. Lancet Infect Dis 2017;17:726-34. Crossref

38. Cheng WL, Hsueh PR, Lee CC, et al. Bacteremic pneumonia caused by extended-spectrum beta-lactamase-producing Escherichia coli and Klebsiella pneumoniae: appropriateness of empirical treatment matters. J Microbiol Immunol Infect 2016;49:208-15. Crossref

39. Kohler PP, Volling C, Green K, Uleryk EM, Shah PS, McGeer A. Carbapenem resistance, initial antibiotic therapy, and mortality in Klebsiella pneumoniae bacteremia: a systematic review and meta-analysis. Infect Control Hosp Epidemiol 2017;38:1319-28. Crossref

40. Zilberberg MD, Shorr AF, Micek ST, Mody SH, Kollef MH. Antimicrobial therapy escalation and hospital mortality among patients with health-care-associated pneumonia: a single-center experience. Chest 2008;134:963-8. Crossref

41. Raman G, Avendano E, Berger S, Menon V. Appropriate initial antibiotic therapy in hospitalized patients with gram-negative infections: systematic review and meta-analysis. BMC Infect Dis 2015;15:395. Crossref

42. Seymour CW, Gesten F, Prescott HC, et al. Time to treatment and mortality during mandated emergency care for sepsis. N Engl J Med 2017;376:2235-44. Crossref

43. Moreno R, Vincent JL, Matos R, et al. The use of maximum SOFA score to quantify organ dysfunction/failure in intensive care. Results of a prospective, multicentre study. Working Group on Sepsis related Problems of the ESICM. Intensive Care Med 1999;25:686-96. Crossref

44. Minne L, Abu-Hanna A, de Jonge E. Evaluation of SOFA-based models for predicting mortality in the ICU: a systematic review. Crit Care 2008;12:R161.

45. Singer M, Deutschman CS, Seymour CW, et al. The third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA 2016;315:801-10. Crossref

Pill splitting by patients is common globally. A German study observed that 24.1% of all drugs required splitting,1 and a Swiss study found that 10% of all discharged prescriptions contained pill splitting.2 In Sweden, 10% of 600 000 investigated prescriptions required splitting, and over 30% of the Swedish patients stated that they had problems dividing the tablets.3 The observed prevalence of pill splitting has been observed to range from 10% to >35% worldwide1 4 5 6 and was even higher in elderly patients (35-67%).6 7 One of the reasons for pill splitting is cost saving because it may result in institutions not needing to stock too many drug items in their formularies.1 In addition, drug splitting may achieve dose flexibility, particularly for patients requiring frequent dosing adjustment.8 Furthermore, some dosages may not be commercially available, especially those for off-label drug use. In these cases, splitting drugs may be essential.9 10 11 Nevertheless, it can also create other clinical issues including medication non-compliance, difficulties experienced by patients in handling unscored pills, drugs that crumble after splitting, and inappropriate drug splitting of extended release formulations, which may lead to treatment failure or toxicity.12 A published study reported that most hypertensive patients preferred not to split pills and that over 70% of patients were willing to pay more for medications with dosages that they did not need to split.13 Limited published studies have addressed this drug-related problem. In this study, we aimed to identify the effects of pill splitting on drug physiochemical properties and clinical outcomes among elderly cardiac patients.

A parallel design was adopted in this study. Patients were recruited from the Cardiac or Hypertension clinics of the Prince of Wales Hospital, Hong Kong. After medical records review, it was found that lisinopril, amlodipine, simvastatin, metformin, and perindopril were among the most commonly prescribed medications that required splitting in the two clinics. Therefore, patients who needed to split their pills when taking lisinopril, amlodipine, simvastatin, metformin, or perindopril were recruited. Patients were randomised into either group A (pill splitting with self-technique) or group B (pill splitting with instructions and training) after their first clinic visit. All patients were asked to sign an informed consent form before enrolment. Group A patients were asked to split drugs using their own technique and continue until their next clinic visit. Group B patients were given proper instruction by pharmacists or pharmacy students on using pill cutters at their first visit and were asked to cut their pills accordingly until their next clinic visit. Patients watched a 2-minute video that explained the reasons for using pill cutters and described the proper way to open the pill cutter, position the drug in the pill cutter, clean the pill cutter, and store the split pills. Subsequently, the pharmacist answered patients’ enquiries regarding the video or other questions related to pill splitting. The current study did not change any drug or dosage of the patient’s existing treatment regimens. Follow-up clinic visits were scheduled with mean duration between first and follow-up clinic visits of 23.1±7.3 weeks.

Chinese patients aged ≥65 years, both male and female, and currently prescribed one or more of metformin, lisinopril, perindopril, amlodipine, or simvastatin (which require splitting) were included in the current study. Patients with dementia or severe physical limitations such as hemiplegia, blindness, or upper limb contractures were excluded from the study.

In our pilot study, we found that the change in drug assays of metformin, atenolol, and amlodipine varied from 52.7% to 147.2% after splitting.14 In our previously published study, systolic blood pressure decreased significantly (from 152.38±18.80 mm Hg to 147.04±20.72 mm Hg, P=0.021) after pharmacist intervention.15 The sample size for the primary outcome was calculated based on a population standard deviation of 0.75, and that for the secondary outcomes was based on a population standard deviation of 0.85. To achieve a 5% significance level and 80% power, 30 and 45 patients in each group would be needed for the primary and secondary outcomes, respectively. We expected a 10% dropout rate, and therefore, at least 80 patients were recruited for each group. There were 193 total participants in this study.

The participants were randomised into group A or B using a computerised dynamic allocation programme and stratification according to types of medications taken, sex, age, and visit dates to ensure balanced patient allocation. All operations were performed by the same pharmacist who conducted the survey.

The primary outcome was the change of drug content before and after the pill splitting training. At baseline, patients in groups A and B were asked to split three tablets of the drugs that they were currently taking using their usual technique. At follow-up visit, group A patients were asked to split three tablets using their own technique, and group B patients were asked to split three tablets using a pill cutter. Two halved tablets were randomly selected for analysis each time. The halved tablets were weighed, and standard drug assays were performed. The drug content of metformin was assessed by ultraviolet-visible spectrophotometry; that of amlodipine, simvastatin, and perindopril was assessed by ultra-performance liquid chromatography; and that of lisinopril was assessed by high-performance liquid chromatography.

The secondary outcomes were the changes in clinical outcomes between before and after the pill splitting training, including the change in blood pressure measurements, haemoglobin A1c, and cholesterol levels, the changes in patients’ attitudes towards and acceptance of pill splitting, and the changes in patients’ knowledge about pill splitting. Haemoglobin A1c and lipid levels were usually collected in hospital 1 week before patients’ clinic visit. Upon initiating their clinic visits, patients had their blood pressures measured in the hospital’s nurse station before they met their physicians. Blood pressure, haemoglobin A1c, and cholesterol levels were collected at baseline and at follow-up visit, and the patients’ knowledge about and attitudes towards pill splitting were assessed by questionnaires.

Paired-samples t tests were used for intra-group comparisons, and independent-samples t tests were used for inter-group comparisons of mean tablet weight between before and after pill splitting training. Paired-samples t tests were also used to assess the changes in clinical outcomes. McNemar’s test was used for intra-group comparisons, and Fisher’s exact test (two-sided) was used for inter-group comparisons of content uniformity, change in patients’ acceptance towards pill splitting, and change in patients’ knowledge about pill splitting. A P value of <0.05 was considered statistically significant. All analyses were performed using the SPSS statistical programme (Windows version 25.0; IBM Corp, Armonk [NY], United States).

A total of 193 eligible patients were enrolled on or before 17 January 2019, and they had follow-up visits on or before 30 April 2019. The patients were randomised into group A (n=106) and group B (n=87). A total of 101 patients participated, of whom 47 from group A and 54 from group B returned for follow-up visit. Among them, 46 patients from group A and 47 patients from group B provided samples for assay. The primary outcome analysis was conducted on those patients, and the secondary outcomes analysis was conducted on the 101 patients who returned for follow-up visit. The patients’ demographic data are shown in Table 1.

The primary outcome was the change of drug content between before and after the pill splitting training. Patients were asked to split three tablets during each visit. Two halved tablets were randomly selected as samples. The samples were weighed, and assays were performed. The mean weight of the halved tablets at baseline and at follow-up visit is documented in Table 2. Table 3 shows the percentage of halved tablets that were within the assay specifications at baseline and at follow-up visit. The percentage of samples with both halved tablets within range was compared between groups A and B. In group A, the percentage in range increased from 39.13% to 47.82% (P=0.523), and the corresponding increase for group B was from 48.94% to 51.06% (P=1.000). The difference in drug assay results between groups A and B at baseline (P=0.406) and at follow-up visit (P=0.837) also did not reach statistical significance.

The correlation between pill cutting training and clinical outcomes is summarised in Table 4. The mean triglyceride in group B decreased significantly from 1.62±1.05 to 1.36±0.80 mmol/L (P=0.049), whereas the mean heart rate increased significantly from 73.97±11.01 to 77.92±12.72 bpm (P=0.026). In group B, there was also improvement in the mean diastolic blood pressure (from 73.40±14.39 to 73.05±9.33 mm Hg), high-density lipoprotein (from 1.40±0.39 to 1.46±0.44 mmol/L), low-density lipoprotein (from 1.87±0.88 to 1.85±0.73 mmol/L), and total cholesterol (from 3.98±0.93 to 3.91±0.85 mmol/L), but those differences did not reach statistical significance. In the overall cohort, improvements were seen in diastolic blood pressure (from 74.21±12.47 to 74.01±9.98 mm Hg), high-density lipoprotein (from 1.42±0.37 to 1.43±0.40 mmol/L), total cholesterol (from 3.90±0.97 to 3.88±0.96 mmol/L), and triglyceride (from 1.46±0.91 to 1.31±0.72 mmol/L), but the changes did not reach statistical significance.

In total, 57.43% of patients split their pills with their bare hands, followed by pill cutters (24.75%), knives (13.86%), and scissors (10.89%). The major reasons for not using pill cutters included: (1) the current method could split pills evenly (68.18%), (2) using pill cutters was time consuming (34.09%), and (3) the pills could not be split evenly by pill cutters (15.91%). The major reasons for using pill cutters included: (1) pills could be cut evenly (80.95%) and (2) the patient was able to exert force more easily (33.33%). In total, 29.70% and 24.75% of patients found pill splitting troublesome at baseline and at follow-up visit, respectively (the difference was not significant, P=0.063). The three major problems encountered by patients while splitting pills were (1) difficulty splitting the pills evenly (17.00%), (2) the pills easily fragmented (10.00%), and (3) difficulty seeing the pills clearly, as they were too small (9.00%). Overall, 61.00% of the patients claimed that they had no difficulties. Nevertheless, 98.21% preferred to take tablets with exact dosages so that no splitting would be required. Patients’ responses to other questions are listed in Table 5.

Table 6 shows that a significantly higher portion of patients in group B had a correct understanding of the following three questions after training: ‘Using pill cutters allows pills to be divided into more accurate doses’ (from 7.41% to 31.48%; P=0.002); ‘The pills should be put into the triangular tip of the pill cutter’ (from 9.26% to 31.48%; P=0.008); and ‘Pill cutters should be stored in a cool and dry place, away from sun or moisture’ (from 9.26% to 35.19%; P=0.003). In contrast, patients in group A did not show a statistically significant improvement in their understanding of any question. During the interview and evaluation of patients’ knowledge about pill splitting at baseline and at follow-up visit, we did not detect any patients with major physical or cognitive abnormalities.

Tablet splitting is a common practice in in-patient and out-patient settings,16 and it may be desirable in terms of dose adjustment, cost saving, and ease of swallowing.3 11 12 17 18 19 20 Nevertheless, it has been reported that splitting pills may cause drug instability, loss of drug due to powdering, uneven dosage, and reduced drug strength.21 22 It is generally understood that using tablet splitting devices can provide a more consistent dose.10 21 Previous studies have identified some characteristics that might affect the quality of halved tablets. Coated, unscored, and small tablets were found to be more difficult to cut.23 Individual pill cutting skill was another crucial factor that determined tablets’ uniformity.23 In the current study, only 24.75% of patients split pills using pill cutters, and only 14.43% of patients had received pill splitting training. Therefore, it is likely that the drug content in the halved tablets did not reach assay standards.

Previous studies mainly focused on the weight deviations among halved tablets, not on drug content.21 22 One study showed that more than one-third of sampled half-tablets did not meet the United States Pharmacopeia specifications.24 The measured drug content variations among half-tablets were: warfarin sodium (90.01%-109.40%), simvastatin (95.21%-111.35%), metoprolol succinate (82.77%-115.92%), metoprolol tartrate (94.83%-112.37%), citalopram (96.50-111.93%), and lisinopril (81.15%-125.72%). In another study, five of eight drugs failed to meet European Pharmacopoeia recommendations for tablet weight deviation after splitting, with 25% of samples deviating by >15% and 10% of samples deviating by >25%.23 The study drugs used were phenobarbitone (maximum deviation: 80.45%), digoxin (maximum deviation: 56.69%), chloroquine (maximum deviation: 48.97%), atenolol (maximum deviation: 45.37%), and doxycycline (maximum deviation: 43.97%). In the present study, both halves of the tablet were within the assay standard at baseline for 39.13% and 48.94% of the patients in groups A and B, respectively. After training, this percentage increased to 47.82% and 51.06%, respectively, but the improvement was not significant, and the percentage of tablets in range was still relatively low. The results corroborated those of previous studies.

Few studies have examined the effect of patient education on the drug content of split pills.25 26 In the current study, we found no significant improvement in content uniformity after pill splitting training. This may be because our patients were elderly patients who may not have been able to perform the task well after a single training session. Content uniformity after pill splitting may be improved if pills are split by pharmacists or qualified staff. A study of paediatric pharmacists suggested that tablets >8 mm could be split once to achieve an approximate half dose for paediatric use.27 Another study found a significant difference in splitting accuracy between nurses and laypersons.23 Nevertheless, only 39.29% of that study’s patients were eager to partake of pill splitting service, and only 18.18% were willing to pay extra money for it. Therefore, pill cutting service may not be practical without a financial incentive.

Triglyceride levels decreased significantly and heart rate increased significantly in group B patients after the intervention. Nevertheless, we did not evaluate the patients’ diet consumption or exercise levels, which may impact their triglyceride levels. No significant change in clinical outcomes was observed in other groups or other parameters. Because the studied drugs were lisinopril, perindopril, simvastatin, and amlodipine, which are not narrow-therapeutic-index drugs, these results were predictable and coincide with other studies that concluded that drugs with long half-life and wide therapeutic index are less likely to be affected.20 28 In view of the high variability of blood pressure measurements in the clinic, all patients were originally instructed to conduct daily blood pressure measurements at home using a portable blood pressure monitor. However, many patients did not measure their blood pressure daily or did not keep a proper self-record, so the clinical outcomes relied on the readings at clinic visits, which may not be consistent with their usual readings. In addition, management of chronic diseases like hypertension, diabetes mellitus, and dyslipidaemia could be influenced by multiple factors, and 3 months was a relatively short period for observation. The effect of drug content deviation after splitting on clinical outcomes may be more obvious in antibiotics or drugs with narrow therapeutic index (eg, digoxin).23 29 30

In the current study, we focused on the effect of pill splitting on drug content. Nevertheless, pill splitting may have other effects on drugs. The pill may carry a bitter taste, as the coating is broken, and the active ingredients may be more susceptible to moisture after exposure.31 Over 70% of patients prepared a sufficient quantity of pills for more than 1 day each time. In total, 36.63% of patients cut for 2 days to 1 week, 26.73% cut for 1 week to 1 month, and 6.93% cut for more than 1 month each time. Exposing the cut pills for too long may increase the risk of crushing or cracking.19 In total, 83.93% of patients found pill splitting training helpful, and the intervention produced significant improvement in patients’ knowledge about pill splitting. This study has identified the major difficulties encountered by patients and the reasons behind their choices. Those problems should be addressed in future patient education. More than half of patients split pills with their bare hands, and the majority of patients who did not use pill cutters thought their own methods could divide pills evenly and that the use of pill cutters was time consuming. The major obstacles patients faced were the difficulties in splitting pills evenly and that the pills fragmented easily. Overall, 98.21% of patients preferred to take tablets with the exact dosage instead of splitting pills. Previous studies also found that dispensing the exact dosage would be more favourable.23 32 Nevertheless, if pill splitting is unavoidable, pharmacists should encourage patients to split coated, unscored, or irregularly shaped tablets with pill cutters to reduce crushing or fragmenting. Pharmacists should also educate patients about the appropriate way to use and clean pill cutters and remind patients to seek doctors’ or pharmacists’ advice before cutting any pills.21 23 33

This project has several limitations. First, the participants’ dropout rate was high, which might result in attrition bias. Compared with group A, a higher proportion of group B patients returned for follow-up visits. The statistically significant improvement in clinical outcomes among group B patients might be caused by their higher awareness about their own health instead of the effectiveness of the pill splitting method. There were limited human resources to make phone calls to patients between the baseline and face-to-face follow-up visits, which could have served as a reminder for patients to attend follow-up visits and perform home monitoring of their blood pressure and their pill splitting methods.

Second, dietary consumption and exercise levels were not evaluated, even though they may affect the clinical outcomes. Third, participants’ education level, household income, and major caregivers were not collected at baseline. Only approximately 65% of participants who attended follow-up visits provided such information. These confounding factors might affect patients’ ability to understand and memorise the steps of using pill cutters, thus affecting the content uniformity of their split pills. The effects of patients’ characteristics on their knowledge and pill splitting skills were not assessed in the current study.

This study revealed that content uniformity can hardly be achieved after pill splitting by patients. No significant difference in clinical outcomes was observed after pill splitting training. It is preferable for pills with doses that do not require splitting to be provided, considering the assay results and patients’ preference. Currently, there is inadequate patient education about pill splitting. Pharmacists should educate patients to use pill cutters properly if splitting is inevitable.

Concept or design: VWY Lee, FYH Fong, BPY Yan.
Acquisition of data: VWY Lee, FYH Fong, BPY Yan.
Analysis or interpretation of data: JTS Li.
Drafting of the manuscript: JTS Li.
Critical revision of the manuscript for important intellectual content: VWY Lee.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

As an editor of the journal, BPY Yan was not involved in the peer review process. Other authors have disclosed no conflicts of interest.

We thank the technicians at the School of Pharmacy, The Chinese University of Hong Kong for carrying out the drug physiochemical tests and the staff at Prince of Wales Hospital for arranging the logistics for patient counselling.

This study was supported by the Health and Medical Research Fund, Food and Health Bureau, Hong Kong SAR Government (#14152111). The funder had no role in study design, data collection/analysis/interpretation, or manuscript preparation.

Ethical approval was obtained from The Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (Clinical trial registration no.: CREC Ref No. 2017.014). Patient consent was obtained upon enrolment. The trial protocol can be obtained as requested.

1. Quinzler R, Gasse C, Schneider A, Kaufmann-Kolle P, Szecsenyi J, Haefeli WE. The frequency of inappropriate tablet splitting in primary care. Eur J Clin Pharmacol 2006;62:1065-73. Crossref

2. Berg C, Ekedahl A. Dosages involving splitting tablets: common but unnecessary? J Pharm Health Serv Res 2010;1:137-41. Crossref

3. Rodenhuis N, De Smet PA, Barends DM. The rationale of scored tablets as dosage form. Eur J Pharm Sci 2004;21:305- 8. Crossref

4. Rodenhuis N, de Smet PA, Barends DM. Patient experiences with the performance of tablet score lines needed for dosing. Pharm World Sci 2003;25:173-6. Crossref

5. Stanković I, Pešić M, Lalić M. Breaking of tablets as a cost-saving strategy. Timočki Medicinski Glasnik 2007;32:5-10.

6. Fischbach MS, Gold JL, Lee M, Dergal JM, Litner GM, Rochon PA. Pill splitting in a long-term care facility. CMAJ 2001;164:785-6.

7. Denneboom W, Dautzenberg MG, Grol R, De Smet PA. User-related pharmaceutical care problems and factors affecting them: the importance of clinical relevance. J Clin Pharm Ther 2005;30:215-23. Crossref

8. Peek BT, Al-Achi A, Coombs SJ. Accuracy of tablet splitting by elderly patients. JAMA 2002;288:451-2. Crossref

9. Duncan MC, Castle SS, Streetman DS. Effect of tablet splitting on serum cholesterol concentrations. Ann Pharmacother 2002;36:205-9. Crossref

10. Rindone JP. Evaluation of tablet-splitting in patients taking lisinopril for hypertension. JCOM Wayne PA 2000;7:22-30.

11. Gee M, Hasson NK, Hahn T, Ryono R. Effects of a tablet-splitting program in patients taking HMG-CoA reductase inhibitors: analysis of clinical effects, patient satisfaction, compliance and cost avoidance. J Manag Care Pharm 2002;8:453-8. Crossref

12. Fawell NG, Cookson TL, Scranton SS. Relationship between tablet splitting and compliance, drug acquisition costs, and patient acceptance. Am J Health Syst Pharm 1999;56:2542-5. Crossref

13. McDevitt JT, Gurst AH, Chen Y. Accuracy of tablet splitting. Pharmacotherapy 1998;18:193-7.

14. Lee VW, Yu Y, Tang L, Yan B. Impact of pill-splitting training on drug physiochemical properties, compliance and clinical outcomes in elderly population–a cross-over cohort study. Value Health 2013;16:A514. Crossref

15. Lee VW, Pang PT, Kong KW, Chan PK, Kwok FL. Impact of pharmacy outreach services on blood pressure management in the elderly community in Hong Kong. Geriatr Gerontol Int 2013;13:175-81. Crossref

16. Crawford DJ. Is tablet splitting safe? Nursing made incredibly easy 2011;9:18-20. Crossref

17. Cohen CI, Cohen SI. Potential savings from splitting newer antidepressant medications. CNS Drugs 2002;16:353-8. Crossref

18. Stafford RS, Radley DC. The potential of pill splitting to achieve cost savings. Am J Manag Care 2002;8:706-12.

19. Bachynsky J, Wiens C, Melnychuk K. The practice of splitting tablets: cost and therapeutic aspects. Pharmacoeconomics 2002;20:339-46. Crossref

20. Helmy SA. Tablet splitting: is it worthwhile? Analysis of drug content and weight uniformity for half tablets of 16 commonly used medications in the outpatient setting. J Manag Care Spec Pharm 2015;21:76-86. Crossref

21. Verrue C, Mehuys E, Boussery K, Remon JP, Petrovic M. Tablet-splitting: a common yet not so innocent practice. J Adv Nurs 2010;67:26-32. Crossref

22. Rosenberg JM, Nathan JP, Plakogiannis F. Weight variability of pharmacist-dispensed split tablets. J Am Pharm Assoc (Wash) 2002;42:200-5. Crossref

23. Elliott I, Mayxay M, Yeuichaixong S, Lee SJ, Newton PN. The practice and clinical implications of tablet splitting in international health. Trop Med Int Health 2014;19:754-60. Crossref

24. Hill SW, Varker AS, Karlage K, Myrdal PB. Analysis of drug content and weight uniformity for half-tablets of 6 commonly split medications. J Manag Care Pharm 2009;15:253-61. Crossref

25. Gharaibeh SF, Tahaineh L. Effect of different splitting techniques on the characteristics of divided tablets of five commonly split drug products in Jordan. Pharm Pract (Granada) 2020;18:1776. Crossref

26. Ashrafpour R, Ayati N, Sadeghi R, et al. Comparison of treatment response achieved by tablet splitting versus whole tablet administration of levothyroxine in patients with thyroid cancer. Asia Ocean J Nucl Med Biol 2018;6:108-12. Crossref

27. Andersson ÅC, Lindemalm S, Eksborg S. Dividing the tablets for children—good or bad? Pharm Methods 2016;7:23-7. Crossref

28. Chou CL, Hsu CC, Chou CY, Chen TJ, Chou LF, Chou YC. Tablet splitting of narrow therapeutic index drugs: a nationwide survey in Taiwan. Int J Clin Pharm 2015;37:1235-41. Crossref

29. Simpson JA, Watkins ER, Price RN, Aarons L, Kyle DE, White NJ. Mefloquine pharmacokinetic-pharmacodynamic models: implications for dosing and resistance. Antimicrob Agents Chemother 2000;44:3414-24. Crossref

30. White NJ, Pongtavornpinyo W, Maude RJ, et al. Hyperparasitaemia and low dosing are an important source of anti-malarial drug resistance. Malar J 2009;8:253. Crossref

31. Margiocco ML, Warren J, Borgarelli M, Kukanich B. Analysis of weight uniformity, content uniformity and 30-day stability in halves and quarters of routinely prescribed cardiovascular medications. J Vet Cardiol 2009;11:31-9. Crossref

32. Madathilethu J, Roberts M, Peak M, Blair J, Prescott R, Ford JL. Content uniformity of quartered hydrocortisone tablets in comparison with mini-tablets for paediatric dosing. BMJ Paediatr Open 2018;2:e000198. Crossref

33. Allemann SS, Bornand D, Hug B, Hersberger KE, Arnet I. Issues around the prescription of half tablets in Northern Switzerland: the irrational case of quetiapine. Biomed Res Int 2015;2015:602021. Crossref

With the continued increase in life expectancy worldwide, population ageing is an especially marked phenomenon in Asian populations.1 It has been estimated that nearly one in three individuals will be in the geriatric age-group (aged ≥65 years) in Hong Kong within the next 15 years.2 Unfortunately, improved longevity is not necessarily linked with improved health or healthcare. Ageing is an unavoidable process associated with many age-related diseases. For example, “immunosenescence”—the age-related dysregulation of the immune system—increases geriatric patients’ susceptibilities to a myriad of immune-mediated disorders (eg, infection, malignancy, and autoimmunity) and adverse reactions to medications.3 4

Adverse drug reactions (ADRs) are much more common in geriatric patients, such that they cause significant morbidity and mortality, compared with younger patients.5 6 Geriatric patients are much more likely than younger patients to be hospitalised for ADRs.7 In particular, drug allergies (DAs) comprise approximately 6% to 10% of all ADRs and up to 10% of the resulting fatal reactions.6 Despite the severe consequences of genuine DAs, many patients mistakenly report non-immune-mediated ADRs as “allergies”. For example, almost 90% of patients with beta-lactam DA labels were confirmed not to be genuinely allergic in previous studies, although such incorrect DA labels were associated with a multitude of dire clinical consequences.8 9 10 11 12 To the best of our knowledge, although the prevalence of ADRs has been extensively reported in geriatric populations, there have been no adequate studies concerning the epidemiology or burden of DAs in geriatric patients.13 To address this lack of information, we performed a prospective analysis of the prevalence and outcomes of geriatric patients with DA labels in a cohort of hospitalised patients in Hong Kong.

All patients admitted to the acute general medical wards of Queen Mary Hospital from 1 July 2018 to 31 December 2018 were recruited for this study. The Queen Mary Hospital is the only public hospital in the Hong Kong West Cluster, which serves a population of 0.5 million and provides acute medical admissions through its Accident and Emergency Department. After admission, patients may be transferred to other convalescent or rehabilitation units of the Hong Kong West Cluster for further management if deemed unfit for direct discharge from the hospital. Patient age and sex were recorded, as was the presence of DA labels. All patients with DA labels were then followed until discharge; clinical data were anonymously extracted for analyses. Extracted clinical data included patient age and sex, presence and details of DA labels, length of stay (from the day of admission to the day of discharge [including stay at convalescent or rehabilitation hospitals] or death), and discharge outcomes (direct discharge, transfer to another hospital, or death). Details of DA labels were reviewed to ensure that the reported manifestations were consistent with the presence of clinical allergies (ie, immune-mediated hypersensitivity reactions). Manifestations suggestive of other non-immune-mediated ADRs were excluded.

The DA labels were categorised in accordance with the British National Formulary classifications (if available): beta-lactam antibiotics (5.1.1 Penicillins and 5.1.2 Cephalosporins and other beta-lactams), non-beta-lactam antibiotics (5.1 Antibacterial drugs, other than 5.1.1 and 5.1.2), non-steroidal anti-inflammatory drugs (10.1.1 NSAIDs), cardiovascular system (CVS) medications (2 CVS), intravenous contrast media, allopurinol, opioid analgesic (4.7.2 Opioid analgesics), non-opioid analgesics (4.7.1. Non-opioid analgesics and compound prep), antihistamines (3.4.1 Antihistamines), antifungals (5.2 Antifungal drugs), or others. Patients were categorised into either geriatric (aged ≥65 years) or non-geriatric (aged <65 years) groups. Demographic characteristics, clinical outcomes, and prevalences of DA labels were compared between groups.

The Chi squared test and independent samples t tests were respectively used to compare categorical and continuous variables between groups in univariate analysis. A P value of <0.05 was considered statistically significant for the multivariate analysis. IBM SPSS Statistics for Windows (version 20.0; IBM Corp, Armonk [NY], Untied States) was used for all analyses. The study protocol was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster.

There were 4361 admissions involving 3641 patients during the 6-month study period. The male-to-female ratio was 1:1.2. In total, 2522 patients (69.3%) were included in the geriatric group, with mean age 71.56±17.3 years.

In total, 492 patients (13.5%) had DA labels, consisting of 151 non-geriatric patients (30.7%) and 341 geriatric patients (69.3%) [Fig 1].

The overall prevalence of DA labels did not significantly differ between geriatric and non-geriatric patients (13.5% [341/2522] vs 13.5% [151/1119], P=0.976) [Table]. The absolute and proportional prevalences of the top 10 categories of DA labels (in descending order) for geriatric and non-geriatric patients are shown in Figures 2 and 3, respectively.

The majority of patients with DA labels were aged ≥65 years (69.3%, 341/492). The male-to-female ratio did not significantly differ between geriatric and non-geriatric groups. Significantly more patients in the geriatric group had DA labels to CVS drugs (15.5% vs 4.6%, P=0.001), while the proportions of other DA labels were similar between the two groups. Patients in the geriatric group had a significantly lower rate of direct discharge from the hospital (73.0% vs 88.1%, P<0.001). The absolute mortality rate tended to be higher among patients in the geriatric group, but this difference was not statistically significant (10.3% vs 6.0%, P=0.123).

Although the prevalence and consequences of overall ADRs have been extensively investigated, this is the first study to specifically examine the epidemiology and outcomes of geriatric patients with DA labels. In our cohort, 13.5% of all hospitalised geriatric patients had DA labels; the leading causes of DA labels were comparable between geriatric and non-geriatric patients. Notably, there were significantly more labelled DAs to CVS medications and significantly more adverse clinical outcomes in geriatric patients.

Adverse drug reactions are defined as any “appreciably harmful or unpleasant reaction” to medications which can occur through various immunological or non-immunological mechanisms.14 Drug allergies or “hypersensitivity reactions” comprise type B (non-dose-related) ADRs, which result from specific immune-mediated responses to a medication. An important problem is that many non-immune-mediated ADRs are often clinically misinterpreted or incorrectly recorded as “allergies”. Although the initial DA reactions may be immunological, genuine allergies may gradually wane and warrant re-evaluation. For example, the vast majority of patients with beta-lactam allergies lose skin testing sensitivity over an interval of 10 years.15 16 Similarly, mild delayed (presumptively T-cell-mediated) reactions do not consistently recur upon re-exposure.17 18 Often, DA labels present in the medical records of geriatric patients have not undergone appropriate allergy testing to verify whether these labels remain accurate. Geriatric patients also have had more time and events to become sensitised or develop ADRs which may be interpreted as allergies; these labels may not be entirely correct for some patients. Overall, our study confirms the presence of the high burden of DA labels in geriatric patients and corresponding worse clinical outcomes (ie, significantly lower rate of direct discharge from the hospital) compared with non-geriatric patients. This highlights the urgent need to expand the availability of allergy testing for this vulnerable population.19 20

As expected, beta-lactam antibiotics constituted the leading cause of DA labels in both patient populations in our study (61.5% [131/213] in the geriatric group and 53.4% [47/88] in the non-geriatric group). In Hong Kong, the prevalence of reported beta-lactam antibiotic allergy is approximately 2% with a cumulative incidence approaching 10 per 100 000 population.21 Beta-lactam DA labels are known to have clinically significant consequences including the use of broad-spectrum antibiotics, enhanced microbial resistance, greater number of Clostridium difficile infections, and expansions of multidrug-resistant organisms.10 11 12 In Hong Kong, Chen et al22 found that the prevalence of methicillin-resistant Staphylococcus aureus was 30.1% among older adults living in residential care homes. The presence of DA labels greatly restricts the repertoire of first-line antibiotics for such patients. Beta-lactams remain the most effective first-line treatment for many bacterial infections including methicillin-sensitive Staphylococcus aureus; in agreement with our findings, the unnecessary use of alternatives leads to worse patient outcomes, especially in the vulnerable geriatric population.

Furthermore, we observed a significantly greater proportion of reported DAs to CVS drugs among geriatric patients. We postulate that this is related to the substantially greater burden of CVS diseases and exposure to CVS drugs in geriatric patients, compared with other conditions.23 24 As previously mentioned, although greater exposure to CVS drugs theoretically increases the risk of genuine DAs, “allergy” labels could be the result of incorrectly interpreted ADRs. For example, the incidences of angiotensin-converting enzyme inhibitor treatment–related cough and angioedema (non-immune-mediated ADRs) increase with age.25 Regardless of their accuracy, this greater proportion of DA labels to CVS drugs is likely to further restrict therapeutic options and elicit CVS complications in geriatric patients. The accuracies of these labels and their specific effects on CVS complications warrant dedicated studies in the future.

This study had some important limitations. First, a higher rate of other adverse clinical outcomes (such as recurrent admissions and mortality) was evident among patients in the geriatric group, although this was not statistically significant. This trend may have constituted a type II statistical analysis error due to inadequate sampling and observational design. Second, we only analysed geriatric and non-geriatric patients with DA labels, although geriatric patients may have worse clinical outcomes regardless of DA status. We were also unable to analyse individual DAs or manifestations within the CVS subgroup. Nonetheless, our findings highlight the vulnerability of this specific geriatric population and emphasise the need for future prospective studies. Third, although all DAs were recorded only after confirmation by the patients’ attending doctors and reported manifestations were screened by an allergist during data collection, we were unable to ascertain the accuracy of the DA labels. Comprehensive evaluations of suspected DAs often require allergological confirmation with skin and/or drug provocation tests, which is especially difficult in frail older adults. A follow-up study to identify the impacts of genuine allergies and incorrectly interpreted adverse clinical outcomes is currently in progress. Lastly, the results of our study were from a single-centre cohort of hospitalised patients and allergy records may have been influenced by local physician practices. Additional multicentre studies, including patients in the ambulatory setting, are needed to corroborate the external validity of our findings.

To the best of our knowledge, this is the first report concerning the epidemiology and outcomes of geriatric patients with DA labels. More than 13% of hospitalised geriatric patients had DA labels; the leading causes of reported DAs in these patients were similar to those of non-geriatric patients in the same hospital. We also observed significantly more reported DAs to CVS drugs, as well as worse clinical outcomes (ie, more frequent transfer to convalescent or rehabilitation facilities) among patients in the geriatric group. Additional dedicated studies are required to confirm the burden and accuracy of DA labels among the already-vulnerable geriatric population.

Concept or design: PH Li.
Acquisition of data: PH Li.
Analysis or interpretation of data: PH Li, HY Chung.
Drafting of the manuscript: PH Li, CS Lau.
Critical revision of the manuscript for important intellectual content: All authors.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Ethical approval was obtained from HKU/HKW Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (HKU/HA HKW IRB), Ref UW 18-669.

1. Balachandran A, de Beer J, James KS, van Wissen L, Janssen F. Comparison of population aging in Europe and Asia using a time-consistent and comparative aging measure. J Aging Health 2020;32:340-51. Crossref

2. Census and Statistics Department, Hong Kong SAR Government. Hong Kong Population Projections 2015 Available from: https://www.censtatd.gov.hk/hkstat/sub/sp190.jsp?productCode=B1120015. Accessed 26 May 2020.

3. Pawelec G. Age and immunity: what is “immunosenescence”? Exp Gerontol 2018;105:4-9. Crossref

4. De Martinis M, Sirufo MM, Ginaldi L. Allergy and aging: an old/new emerging health issue. Aging Dis 2017;8:162-75. Crossref

5. Davies EA, O’Mahony MS. Adverse drug reactions in special populations—the elderly. Br J Clin Pharmacol 2015;80:796-807. Crossref

6. Lazarou J, Pomeranz BH, Corey PN. Incidence of adverse drug reactions in hospitalized patients: a meta-analysis of prospective studies. JAMA 1998;279:1200-5. Crossref

7. Muehlberger N, Schneeweiss S, Hasford J. Adverse drug reaction monitoring—cost and benefit considerations. Part I: frequency of adverse drug reactions causing hospital admissions. Pharmacoepidemiol Drug Saf 1997;6 Suppl 3:S71-7. Crossref

8. Li PH, Siew LQ, Thomas I, et al. Beta-lactam allergy in Chinese patients and factors predicting genuine allergy. World Allergy Organ J 2019;12:100048. Crossref

9. Mattingly TJ 2nd, Fulton A, Lumish RA, et al. The cost of self-reported penicillin allergy: a systematic review. J Allergy Clin Immunol Pract 2018;6:1649-54.e4. Crossref

10. MacFadden DR, LaDelfa A, Leen J, et al. Impact of reported beta-lactam allergy on inpatient outcomes: a multicenter prospective cohort study. Clin Infect Dis 2016;63:904-10. Crossref

11. Macy E, Contreras R. Health care use and serious infection prevalence associated with penicillin “allergy” in hospitalized patients: a cohort study. J Allergy Clin Immunol 2014;133:790-6. Crossref

12. Blumenthal KG, Lu N, Zhang Y, Li Y, Walensky RP, Choi HK. Risk of meticillin resistant Staphylococcus aureus and Clostridium difficile in patients with a documented penicillin allergy: population based matched cohort study. BMJ 2018;361:k2400. Crossref

13. Ventura MT, Scichilone N, Paganelli R, et al. Allergic diseases in the elderly: biological characteristics and main immunological and non-immunological mechanisms. Clin Mol Allergy 2017;15:2. Crossref

14. Edwards IR, Aronson JK. Adverse drug reactions: definitions, diagnosis, and management. Lancet 2000;356:1255-9. Crossref

15. Trubiano JA, Adkinson NF, Phillips EJ. Penicillin allergy is not necessarily forever. JAMA 2017;318:82-3. Crossref

16. Blanca M, Romano A, Torres MJ, et al. Update on the evaluation of hypersensitivity reactions to betalactams. Allergy 2009;64:183-93. Crossref

17. Mori F, Cianferoni A, Barni S, Pucci N, Rossi ME, Novembre E. Amoxicillin allergy in children: five-day drug provocation test in the diagnosis of nonimmediate reactions. J Allergy Clin Immunol Pract 2015;3:375-80.e1. Crossref

18. Bourke J, Pavlos R, James I, Phillips E. Improving the effectiveness of penicillin allergy de-labeling. J Allergy Clin Immunol Pract 2015;3:365-74.e1. Crossref

19. Chan YT, Ho HK, Lai CK, et al. Allergy in Hong Kong: an unmet need in service provision and training. Hong Kong Med J 2015;21:52-60. Crossref

20. Lee TH, Leung TF, Wong G, et al. The unmet provision of allergy services in Hong Kong impairs capability for allergy prevention-implications for the Asia Pacific region. Asian Pac J Allergy Immunol 2019;37:1-8.

21. Li PH, Yeung HH, Lau CS, Au EY. Prevalence, incidence, and sensitization profile of beta-lactam antibiotic allergy in Hong Kong. JAMA Netw Open 2020;3:e204199. Crossref

22. Chen H, Au KM, Hsu KE, et al. Multidrug-resistant organism carriage among residents from residential care homes for the elderly in Hong Kong: a prevalence survey with stratified cluster sampling. Hong Kong Med J 2018;24:350-60. Crossref

23. Islam MM, Valderas JM, Yen L, Dawda P, Jowsey T, McRae IS. Multimorbidity and comorbidity of chronic diseases among the senior Australians: prevalence and patterns. PLoS One 2014;9:e83783. Crossref

24. Morgan TK, Williamson M, Pirotta M, Stewart K, Myers SP, Barnes J. A national census of medicines use: a 24-hour snapshot of Australians aged 50 years and older. Med J Aust 2012;196:50-3. Crossref

25. Alharbi FF, Kholod AA, Souverein PC, et al. The impact of age and sex on the reporting of cough and angioedema with renin-angiotensin system inhibitors: a case/noncase study in VigiBase. Fundam Clin Pharmacol 2017;31:676-84.Crossref

The rapid and devastating spread of the coronavirus disease 2019 (COVID-19) pandemic caught the global community unprepared and overwhelmed the disease control measures of many nations. Measures deployed in Hong Kong to control the spread of COVID-19 were less stringent than those adopted in other nations; however, they proved to be effective.1 Territory-wide lockdowns, curfews, and the controversial surveillance of smartphone data for contact tracing purposes were all avoided. A recent local study showed that behavioural changes were the key factors associated with limiting the spread of COVID-19 and seasonal influenza.1

Community masking by healthy individuals is controversial and opinions on its effectiveness or necessity differ among health authorities worldwide.2 Since the severe acute respiratory syndrome (SARS) epidemic in 2003, the population of Hong Kong has maintained a strong masking culture. Although masking in crowded areas has always been voluntary in Hong Kong, rates of 61% and 79% during the SARS outbreak were recorded in two previous studies.3 4 In the present study, we aimed to explore the masking behaviour of pedestrians in crowded areas, as well as the attitude of the population towards community masking, during the initial spread of COVID-19 in Hong Kong.

This study protocol was approved by the Human Subjects Ethics Application Review board of The Hong Kong Polytechnic University and complied with the Declaration of Helsinki. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for cross-sectional studies was implemented in the drafting of this article. The study protocol consisted of two parts: an observational study (Part 1) and an online survey (Part 2).

The masking behaviour of pedestrians in Hong Kong was observed between 1 and 29 February 2020. Eleven well-populated locations across Hong Kong were selected for observation. Observation sites on the main street of each location were chosen based on pedestrian throughput, and the ability to observe pedestrians in a clear, unhindered manner without interrupting the flow of traffic. Three observation sessions per day were conducted at 12:00-14:30 (lunch time), 14:30-17:00 (afternoon), and 17:00-19:30 (evening). To reduce selection bias, the location and observation times were randomly preselected on the evening before each study by drawing two sets of shuffled opaque envelopes containing the session time and locations.

On days with sufficient rain to warrant umbrella use, further observations were terminated. The data collected until the occurrence of rain were included for analysis. One observer was allocated to each session and the following data were collected: frequency of masking, type of mask worn, and number and type of erroneous masking practices. Only pedestrians walking in one direction were observed to prevent duplicate counting. The criteria for seven common types of masking errors were based on deviations from the surgical mask use guidelines published by the Hong Kong Centre for Health Protection.5 The definitions for masking errors were standardised before the study by consensus among the observers after a field test.

Members of the Hong Kong population were included in an online survey of behaviour and views on community masking, which was conducted between 23 March and 14 April 2020. Consent was implied in the voluntary participation and completion of the survey. Personal information was not collected; however, demographic details (eg, age, gender, education level, and whether the respondents were healthcare students or professionals [HCSPs]) were recorded. Short-term visitors were excluded by targeting only respondents who had lived in Hong Kong for the preceding 6 months.

A link to the online survey was distributed via various means such as social media and email. The survey required approximately 5 minutes to complete. A 5-point Likert scale was used to grade respondents’ perceptions of mask efficacy for protection and the prevention of community spread, evaluation of mask performance, and confidence in mask selection and correct use of mask. Respondents were also asked about mask reuse habits and storage, and information sources for COVID-19–related health matters, including their perceived reliability of those sources. A sample questionnaire is shown in the online supplementary Appendix.

Statistical analysis was performed by an independent statistician using IBM SPSS Statistics (Windows version 21.0, IBM Corp. Armonk [NY], United States). Univariate logistic regression and univariate ordinal regression were used to explore associations with binary and ordinal outcomes, respectively. Crude odds ratios (ORs) for each demographic variable were calculated from univariate analysis. Multivariate regression analyses were then performed, including all demographic variables, and the adjusted OR was estimated for each demographic variable. Descriptive statistics were used to provide an overview of the observations; the findings of regression analysis were presented as ORs and 95% confidence intervals (CIs). A P value of <0.05 was considered statistically significant.

In total, 10 211 pedestrians were observed over 25 sessions. The masking rate was 94.8%; most pedestrians wore disposable surgical masks (83.7%), and a small number wore N95 respirators (0.7%). The remaining pedestrians wore an assortment of face masks made of fabric or neoprene rubber; a few wore gas masks.

Among pedestrians wearing surgical masks, masking errors were observed in 1113 (13.0%) individuals (Table 1). The most common errors observed included: mask worn too low, exposing the nostrils and mouth (42.5%), or mask worn inside-out/upside-down (35.5%). A less common but serious error was the absence of hand hygiene after touching their masks (16.4%).

A total of 3199 respondents completed the survey. Data from 74 non-residents were excluded, and the remaining 3125 responses were analysed. Female respondents comprised a larger proportion (65.7%), education status was at diploma level or above for 76.9% of the respondents, and 18.6% were HCSPs (Table 2).

Most online respondents were confident of using a face mask correctly (96.9%) and believed in its efficacy for protection (94.6%) and the prevention of community spread (96.6%). Most respondents indicated that they clearly understood the functions of the different types of masks available (83.6%) and were confident of their ability to evaluate those masks (77.1%). Multivariate ordinal regression analyses showed that HCSPs were associated with greater confidence (OR=1.62; 95% CI=1.34-1.95; P<0.001) of using a face mask correctly; while increasing age was associated with lower confidence (OR=0.87; 95% CI=0.81-0.94; P<0.001 for each successive age-group). However, no significant associations were found with education level or gender.

In all, 78.9% of the respondents reused their face masks and stored them using a variety of methods: between tissue papers (74.1%), in paper envelopes (46.5%), in plastic bags (22.9%), left on table (6.5%), and in plastic containers (6.0%) [Table 3]. Multivariate logistic regression analyses concerning reuse of face masks showed greater likelihood among respondents with higher education level (OR=1.14; 95% CI=1.05-1.24; P=0.003 for each successive education level), older age (OR=1.48; 95% CI=1.35-1.62; P<0.001 for each successive age-group), and HCSP status (OR=1.61; 95% CI=1.26-2.07; P<0.001). Gender was not associated with mask reuse.

More respondents reported the acquisition of information on mask usage via social media (65.9%) and online newspapers (47.9%), compared with official government websites (23.2%) [Table 4]. Ratings for reliability were highest for government websites (69.9%), followed by online (64.6%) and printed (64.5%) newspapers (Table 5).

Compared with other respondents, HCSPs had a better understanding of mask types and indications for each type (multivariate ordinal regression: OR=1.93; 95% CI=1.60-2.34; P<0.001). They were more likely to choose surgical masks (multivariate logistic regression: OR=2.43; 95% CI=1.04-5.72; P=0.041) and less likely to choose N95 respirators (OR=0.54; 95% CI=0.44-0.68; P<0.001) for community use. They were also more likely to reuse their face masks (OR=1.61; 95% CI=1.26-2.07; P<0.001) and store them between tissues (OR=1.80; 95% CI=1.47-2.21; P<0.001), but less likely to use paper envelopes (OR=0.80; 95% CI=0.65-0.98; P=0.030).

Our study investigated pedestrian mask use and public perceptions of community masking during the initial spread of COVID-19 in Hong Kong. The masking rate was high, and the public was confident in mask efficacy for protection and the prevention of community spread. Most pedestrians wore a surgical mask, but a small proportion of these masks were worn incorrectly. Surprisingly, the mask reuse rate was high, and varying methods were used for storage. Although government websites were considered reliable, social media was more popular as the source of information regarding masking and health-related matters.

Population-level behavioural insights are essential for coordinating an effective and coherent infection control strategy.6 Previous events and disease outbreaks have considerably influenced the masking culture in Hong Kong. Similar to covering up when coughing or sneezing, wearing a mask in the community or workplace when unwell became a part of recent social etiquette following the SARS outbreak. During the SARS epidemic, a public hospital became a source of community spread, prompting the government to enforce and promote community protective behaviour thereafter, particularly in public hospitals. These efforts included strong recommendations for hospital visitors to wear masks, as well as the widespread availability of hand sanitisers in strategic areas. Whilst evidence supporting these practices remains controversial, these recommendations have positively influenced the attitude and behaviour of the general public towards mass masking and hand hygiene for protection during disease outbreaks. These events may explain the high voluntary masking rate that we have recorded in this study. A high masking rate was also previously noted during the SARS outbreak in Hong Kong.2 3 Notably, face masks are also commonly used to protect against air pollution, particularly during hazy weather and within high traffic areas.

The issue of community masking was controversial particularly during the early stages of the COVID-19 pandemic, such that conflicting recommendations were issued by various health authorities and public figures worldwide. At the time of this study, the World Health Organization recommended against community masking because of insufficient evidence regarding its effectiveness, the potential for a false sense of security, and the stressed supply of surgical masks for hospital use.7 However, this stance on effectiveness should have been considered in the context of clinical outcome studies,8 which were based largely on the spread of influenza and would not necessarily be applicable to the spread of COVID-19. Furthermore, reports of disease spread involving pre-symptomatic carriers of COVID-19 were not considered.

Although the surgical mask was originally designed for the protection of patients during surgery, its role in reducing wound infection is not fully established and has been contentious.9 The rationale for masking later shifted to protection for the wearer, although evidence to support this perspective is equally tenuous.10 Under standardised simulated conditions, laboratory studies have shown that surgical masks are effective in limiting both inbound and outbound transmission of aerosol particles.11 Thus, wearing a face mask will limit the spread of droplets during coughing or sneezing from both symptomatic and asymptomatic carriers; it will also protect the mucosa of the nostrils and mouth of the wearer from droplets and aerosols.12 Contrary to the claims by the World Health Organization7 that wearing a mask may create a false sense of security leading to the abandonment of other protective behaviours, voluntary mask use in crowded areas was shown to encourage protective behaviour and performance of hand hygiene.13 The high masking rate in Hong Kong may be an intangible factor that enabled indirect control of community spread by preventing viral shedding from asymptomatic carriers. Multiple clusters of infections have occurred in locations with poor masking or social distancing,14 15 suggesting that these measures are important. Since April 2020, the World Health 16 has updated its guidelines to recommend the use of non-medical masks among the general public when there is a limited capacity to implement other containment measures.

A substantial proportion of pedestrians (13.0%) wore their surgical face masks incorrectly, which may have limited the protective efficacy of these masks. Most commonly, they were worn too low, ‘upside-down’ or ‘inside-out’. A surgical mask consists of an inner water-absorbing layer and an outer water-repelling filter, which are horizontally pleated to create rows of gutters for expansion and to catch moisture. A mask worn inside-out accumulates moisture on the facial side, which is uncomfortable. This increases the likelihood that users will touch and rub their faces, leading to self-contamination or temporary mask removal. Additionally, a mask worn inside-out may trap droplets from surrounding people within the outward-facing water-absorbing layer. A mask worn too low on the face exposes the nostrils or mouth, which are mucosal surfaces vulnerable to droplets and airborne contamination. Although unlikely, this error may arise from semantics—the Chinese term for ‘face mask’ literally means ‘mouth cover’, which may have misled users into believing that this type of coverage was its sole purpose. We examined the packaging of various brands of surgical masks sold locally and found that very few provided instructions for correct use. Instructions were previously considered unnecessary because surgical masks were intended for use by HCSPs; however, many users now are members of the general public. Manufacturers are encouraged to provide clear instructions on their packaging and print a symbol on each mask to prevent users from wearing masks inside-out.

We were alarmed by the high reuse rate of disposable masks. Some reasons were obvious, such as a supply shortage, compounded by panic buying that leads to price inflation. However, the masking rate might have been lower if the masks were not reused. Many users were probably aware that the masks should not be reused, but our findings should serve as a ‘reality check’.17 Surprisingly, mask reuse was more common among HCSPs; this may have been related to greater confidence in their ability to handle a potentially contaminated mask, as well as the belief that the causative virus (severe acute respiratory syndrome coronavirus 2) will degrade on non-living surfaces over time.18 Although there is no evidence of increased disease spread, the potential for contamination from poor handling is obvious.

Various mask storage methods, such as within tissue papers, in paper envelopes, in plastic bags, and in containers, were described. Recent evidence suggests that the severe acute respiratory syndrome coronavirus 2, which causes COVID-19, is more stable on smooth non-porous surfaces; thus, it may be safer to store masks in paper material (eg, tissues or envelopes) where it will dry effectively.18 19 It has also been reported that the virus can be inactivated at 70°C in approximately 5 to 30 minutes.18 20 This information will be useful should the reuse of surgical masks be necessary during an exceptional shortage; moreover, input from infectious disease experts on the appropriate handling techniques is likely to provide considerable value. Recently, Hong Kong residents were issued reusable face masks with antimicrobial properties for community use.19 21 Despite the high cost of such masks, this may be the solution to face mask shortage issues; it may also preserve medical face masks for hospital use, as recommended by the World Health Organization.16

Social media was the most common source of health information but was regarded as the least reliable source. Although official government websites were regarded as the most reliable sources, many respondents chose convenience over perceived reliability when sourcing health information. However, the potential for misinformation is an important concern and conflicting advice may create distrust, thereby interfering with the establishment of a coherent response to the pandemic. Because of the popularity of social media, authorities should utilise these platforms as a supplement to their standard websites for better public exposure and communication concerning health-related information. There is a clear need to address the issues that we have identified. Correct masking technique will reduce wastage and prevent self-contamination through mishandling.

There were several limitations to this study. First, because of its observational nature, we were unable to determine why some pedestrians did not wear masks (eg, whether this was related to availability or choice). Second, our findings may not be sufficiently representative of other less crowded areas in Hong Kong. Third, the respondents to our online survey were limited to those with internet access, which might have prevented inclusion of individuals who were older, less educated, or more vulnerable. Telephone and face-to-face interviews may provide sufficient data concerning older people and individuals with low socio-economic status. Lastly, we did not identify the respondents of our survey; thus, multiple responses could have been submitted by some users. Nevertheless, this is the largest behavioural study thus far to explore some of the issues on the use of face masks during the initial spread of COVID-19 in Hong Kong.

This study provided behavioural insights and attitudes on community masking in a region that has successfully managed the initial spread of COVID-19 through a combination of public health and behavioural interventions. Members of the Hong Kong population are highly motivated to engage in masking practices and believe in its effectiveness for protection and the prevention of disease spread. However, a high face mask reuse rate and incorrect masking techniques were observed. Information on government websites should be enhanced and linked to social media to improve accessibility and provide suitable guidance for the general public.

Concept or design: VCW Tam, CPL Chan, HKW Law, SWY Lee.
Acquisition of data: VCW Tam, HKW Law, SWY Lee.
Analysis or interpretation of data: VCW Tam, SWY Lee.
Drafting of the manuscript: VCW Tam, SY Tam, ML Khaw, SWY Lee.
Critical revision of the manuscript for important intellectual content: All authors.

All authors had full access to the data, contributed to the study, approved the final version for publication, and take responsibility for its accuracy and integrity.

All authors have disclosed no conflicts of interest.

We thank Miss Abbie Chan, Department of Health Technology and Informatics, The Hong Kong Polytechnic University, for her assistance in data collection throughout the study; Mr Alex Nicol for his professional advice with manuscript preparation and statistical analysis; and Dr Wai-kwong Poon for his expert advice with research design and manuscript revision.

A letter reporting preliminary findings of part of the present study was published in Tam VC, Tam SY, Poon WK, Law HK, Lee SW. A reality check on the use of face masks during the COVID-19 outbreak in Hong Kong. EClinicalMedicine. 2020 Apr 24;22:100356. https://doi.org/10.1016/j.eclinm.2020.100356

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

This study was approved by the Human Subjects Ethics Application Review Board of Hong Kong Polytechnic University (Ref HSEARS20200213002-01). Participation in the survey was voluntary, and consent was implied from completion of the survey.

1. Cowling BJ, Ali ST, Ng TW, et al. Impact assessment of non-pharmaceutical interventions against coronavirus disease 2019 and influenza in Hong Kong: an observational study. Lancet Public Health 2020;5:e279-88. Crossref

2. Cheng KK, Lam TH, Leung CC. Wearing face masks in the community during the COVID-19 pandemic: altruism and solidarity. Lancet 2020 Apr 16. Epub ahead of print. Crossref

3. Leung GM, Quah S, Ho LM, et al. A tale of two cities: community psychobehavioral surveillance and related impact on outbreak control in Hong Kong and Singapore during the severe acute respiratory syndrome epidemic. Infect Control Hosp Epidemiol 2004;25:1033-41. Crossref

4. Tang CS, Wong CY. Factors influencing the wearing of facemasks to prevent the severe acute respiratory syndrome among adult Chinese in Hong Kong. Prev Med 2004;39:1187-93. Crossref

5. Centre for Health Protection, Department of Health, Hong Kong SAR Government. Use mask properly. 2015. Available from: https://www.chp.gov.hk/files/pdf/use_mask_properly.pdf. Accessed 12 Aug 2020.

6. Sadique MZ, Edmunds WJ, Smith RD, et al. Precautionary behavior in response to perceived threat of pandemic influenza. Emerg Infect Dis 2007;13:1307-13. Crossref

7. World Health Organization. Advice on the use of masks in the context of COVID-19: interim guidance. 5 June 2020. Available from: https://apps.who.int/iris/bitstream/handle/10665/332293/WHO-2019-nCov-IPC_Masks-2020.4-eng.pdf. Accessed 5 Jun 2020.

8. Orr NW. Is a mask necessary in the operating theatre? Ann R Coll Surg Engl 1981;63:390-2.

9. Romney MG. Surgical face masks in the operating theatre: re-examining the evidence. J Hosp Infect 2001;47:251-6. Crossref

10. Lipp A. The effectiveness of surgical face masks: what the literature shows. Nurs Times 2003;99:22-4.

11. van der Sande M, Teunis P, Sabel R. Professional and home-made face masks reduce exposure to respiratory infections among the general population. PloS One 2008;3:e2618. Crossref

12. Leung NH, Chu DK, Shiu EY, et al. Respiratory virus shedding in exhaled breath and efficacy of face masks. Nat Med 2020;26:676-80. Crossref

13. Wada K, Oka-Ezoe K, Smith DR. Wearing face masks in public during the influenza season may reflect other positive hygiene practices in Japan. BMC Public Health 2012;12:1065. Crossref

14. Centre for Health Protection, Department of Health, Hong Kong SAR Government. CHP investigates 10 additional cases of novel coronavirus infection. 2020. Updated 9 Feb 2020. Available from: https://www.info.gov.hk/gia/general/202002/09/P2020020900704p.htm. Accessed 20 Mar 2020.

15. Centre for Health Protection, Department of Health, Hong Kong SAR Government. CHP investigates 43 additional cases of COVID-19. 2020. Updated 26 Mar 2020. Available from: https://www.info.gov.hk/gia/general/202003/26/P2020032600765p.htm. Accessed 10 Apr 2020.

16. World Health Organization. Advice on the use of masks in the community, during home care and in health care settings in the context of the novel coronavirus (2019-nCoV) outbreak: interim guidance, 29 January 2020. Available from: https://apps.who.int/iris/handle/10665/330987. Accessed 17 Aug 2020.

17. Tam VC, Tam SY, Poon WK, Law HK, Lee SW. A reality check on the use of face masks during the COVID-19 outbreak in Hong Kong. EClinicalMedicine 2020;22:100356. Crossref

18. Chin AW, Chu JT, Perera MR, et al. Stability of SARS-CoV-2 in different environmental conditions. Lancet Microbe 2020;1:e10. Crossref

19. van Doremalen N, Bushmaker T, Morris DH, et al. Aerosol and surface stability of SARS-CoV-2 as compared with SARS-CoV-1. N Engl J Med 2020;382:1564-7. Crossref

20. Nathan N. Waste not, want not: re-usability of N95 masks. Anesth Analg 2020;131:3. Crossref

21. Hong Kong SAR Government. Government to distribute free reusable masks to all citizens. 2020. Updated 5 May 2020. Available from: https://www.info.gov.hk/gia/general/202005/05/P2020050500692.htm. Accessed 3 Aug 2020.