Healthcare spending in Hong Kong continues to grow, as demonstrated by an 8.6% increase (to HK$69.7 billion [US$8.99 billion]) from fiscal year 2018 to fiscal year 2019.1 According to internal statistics drawn from the Clinical Data Analysis and Reporting System of the Hospital Authority (HA), the annual number of emergency department (ED) visits in all public EDs in the New Territories West Cluster increased by 11.5% (to 410 707) between 2018-19 and 2019-20. Similar trends have been observed in many industrialised countries.2 Moreover, the proportion of older adults (age ≥65 years) is projected to increase from 17.2% in 2019 to 25.1% in 2029.3 The increasing demand for ED services may result in diminished quality of care and negative outcomes for vulnerable patients.

Frequent ED use is often regarded as a major component of ED utilisation burden. Substantial gaps remain in research concerning individuals who frequently use EDs in Hong Kong. Efforts to understand this population and the impacts of frequent use on ED utilisation burden are needed to identify interventions that can improve healthcare delivery, both in and out of the EDs.

This region-based study analysed individuals who frequently used public EDs and explored specific interventions that may reduce ED visits by these patients.

This multi-centre retrospective cohort study analysed ED visits from all acute care hospitals located within the New Territories West Cluster (serving a population of 1.14 million in 20183). During the study period, only three EDs (in Tuen Mun Hospital, Pok Oi Hospital, and Tin Shui Wai Hospital4) provided acute emergency care in this region. All three EDs are public and managed by the HA, the statutory body that manages all public hospitals in Hong Kong. In the study region, individuals who called ambulances were typically taken to one of these three EDs.

The study population included all patients, regardless of age, who visited any of these three EDs during the fiscal year from 1 April 2018 to 31 March 2019. Patients without acceptable identification documents were excluded.

Usage classes were defined according to the total number of ED visits by an individual patient to any of the 18 public EDs in Hong Kong5 during the study period, using cut-off values commonly reported in the literature to facilitate comparison6: normal users (NUs) had ≤3 visits,6 frequent users (FUs) had 4 to 9 visits,6 and high-intensity users (HIUs) had ≥10 visits.7

Two additional groups were defined: non-ED users had no ED visits over a 12-month period, whereas superusers had very high numbers of ED visits (≥35 over a 12-month period, almost 1 visit every 10 days).

Data were collected from the Clinical Data Analysis and Reporting System of the HA, a comprehensive electronic patient database that includes patient demographics, diagnoses, surgical records, ED visits, hospitalisation episodes, and radiological investigations from all public hospitals and most public clinics in Hong Kong. It allows episode-based and patient-based analyses according to user-defined criteria. For this study, multiple visits were linked using a unique record linkage number. Population demographic data were obtained from the 2016 Population By-census8 and other official government statistics.3 9

Two dimensions of analysis were performed on the three classes NUs, FUs, and HIUs: patient-based and episode-based, in which patients and individual visits were the respective units of analysis.

Demographic variables were assigned based on values reported at the index visit; they included age, ethnicity (Chinese/non-Chinese), sex (male/female), institutionalisation (residency in an old-age home or not), and ED payment status (exempt or not). In Hong Kong, identity card holders and residents aged <11 years are charged HK$180 (US$23.2) per ED visit. Civil servants and their family members, individuals receiving social security assistance from the government, and individuals in vulnerable groups (low income, chronically ill, and older adults with minimal income/assets) are exempt from the requirement to pay for ED visits. To evaluate ED utilisation trends, data were retrieved regarding utilisation in the 12-month period before the study.

In Hong Kong, all hospital admissions and ED visits (excluding patients who leave before consultation) are coded using an appropriate International Classification of Diseases, Ninth Revision (ICD-9) diagnosis recommended by the World Health Organization.9 A list of diagnoses that represent chronic conditions with significant morbidity was compiled and classified into nine categories with reference to ICD-9 (online supplementary Appendix 1).9 All in-patient procedures, except minor bedside and clinical procedures, are also appropriately coded and classified as minor, intermediate, major, or ultra-major10; they are then divided into elective and emergency categories. These coding processes, and the processes described below, are routine procedures that undergo strict internal auditing for completeness and accuracy. This study analysed ICD-9 diagnosis codes, as well as major and ultra-major procedure records, from hospital and clinic encounters in the 5 years prior to the study period. The full list of major and ultra-major operations included in the current study is mostly based on the HA’s List of Private Services.10

Deaths occurring in Hong Kong were retrieved from the Hong Kong Death Register. Two-year mortality was defined as death from any cause between 1 April 2018 and 31 March 2020.

Variables in the episode-based analysis included age, ambulance utilisation, triage category (in descending order of urgency: critical, emergency, urgent, semi-urgent, and non-urgent), imaging performed in ED (plain radiography and computed tomography), and visit details (to be discussed below). Admission rates and median length of stay for admitted patients were calculated.

For each episode, the attending specialty, trauma status, and ICD-9 diagnosis or chief complaint were recorded by the attending emergency physician. The attending specialty was regarded as the specialty to which a patient’s chief complaint belongs. Four main specialties (medicine and geriatrics, general surgery, orthopaedics, and paediatrics) were included in the data analysis. The ICD-9 diagnosis codes were grouped into musculoskeletal pain (eg, joint pain, cervicalgia, and lumbago) and minor infections (eg, acute upper respiratory tract infection). A full list of diagnoses within each of these two categories is included in online supplementary Appendix 2.9

Descriptive statistics were compiled for the demographic characteristics of non-ED users, NUs, FUs, and HIUs, with reference to data from the 2016 Population By-census.8 Age-specific distributions of ED visits and numbers of ED visits per capita were analysed. In the patient-based analysis, demographic and clinical characteristics of NUs, FUs, and HIUs were compared by univariate analysis with appropriate statistical tests. The episode-based analysis compared age distribution, ambulance use, triage category, investigations performed, principal diagnosis during each ED visit, and admission statistics among NUs, FUs, and HIUs. The clinical and demographic characteristics of superusers were also analysed.

Two models were used to identify independent predictors of FU and HIU statuses, compared with NU status and specific numbers of ED visits. Multinomial logistic regression was conducted, using the patient as the unit of analysis, to compare the FU and HIU groups with the reference group. Zero-truncated Poisson regression was performed to predict the independent association of each predictor with the number of ED visits. All bivariate predictors associated with the outcome (P<0.1) were entered into the model and used as categorical variables. Age-stratified subgroup analysis (≤17 years, 18-64 years, and ≥65 years) was performed to identify independent predictors of FU and HIU statuses in each age-group. P<0.05 was regarded as the threshold for statistical significance in all analyses. Zero-truncated Poisson regression was performed with R11 with vector generalised linear and additive model.12 All other statistical analyses were conducted using SPSS (Windows version 25.0; IBM Corp, Armonk [NY], United States).

As shown in the Figure, the number of per capita ED visits was the highest in the youngest and oldest age-groups (0.66 for age 0-4 years and 1.02 for age ≥85 years); patients aged 10 to 54 years had a stable and low number of per capita ED visits (<0.3).

During the study period, 215 862 patients with valid identity documents accessed EDs in the study region; there were 371 915 visits in total (Table 1). Frequent users and HIUs constituted small percentages of the total number of patients (8.2% and 0.8%, respectively), but they represented larger percentages of visits (21.3% and 5.9%, respectively). The total number of ED visits by a single HIU ranged from 10 to 263.

Compared with NUs (median age, 47 years), FUs and HIUs were older (55 and 52 years, respectively). Frequent users and HIUs more often had underlying chronic illnesses, such as cardiac, respiratory, neurological, and gastrointestinal/hepatobiliary diseases. Mental health disorders and substance abuse were also much more prevalent among FUs and HIUs. Moreover, 23.5% and 59.0% of FUs and HIUs were FUs or HIUs in the previous year, implying habitual attendance behaviour. Frequent users and HIUs also had higher all-cause mortality rates, compared with NUs.

Reasons for ED visits differed among NUs, FUs, and HIUs (Table 2). Normal users were much more likely to visit the ED for an injury, whereas primary diagnoses related to musculoskeletal pain were more common in HIUs. Levels of urgency on presentation also differed among NUs, FUs, and HIUs. More visits by FUs were triaged as critical, emergency, or urgent (38.1%), compared with visits by NUs (31.8%). However, a lower percentage of visits by HIUs were triaged as urgent or above (28.6%), compared with visits by NUs; more HIUs attended EDs by calling an ambulance (HIUs: 21.2% vs NUs: 17.9%).

Multinomial logistic regression and zero-truncated Poisson regression showed similar independent predictors of FU and HIU statuses (Table 3). Payment exemption and number of visits in the prior year were the strongest predictors. Residency in an old-age home was a risk factor for FU status (adjusted odds ratio [OR]=1.131) but a protective factor for HIU status (adjusted OR=0.48). Non-Chinese ethnicity, pre-existing systemic diseases including cardiac, respiratory, gastrointestinal/hepatobiliary and renal, mental health disorders, and substance abuse were risk factors for frequent ED use. Age-stratified analysis (Table 4) showed similar predictors among the three groups. Respiratory diseases and gastrointestinal/hepatobiliary diseases were prevalent in paediatric FUs; neurological diseases and mental health disorders were prevalent in both paediatric FUs and paediatric HIUs. Non-Chinese ethnicity was a risk factor for FU and HIU statuses among patients aged <18 years (adjusted ORs=1.949 and 2.107, respectively), but it was not associated with ED use in older patients (age >64 years).

Superusers had particularly high prevalences of mental health disorders (26.9%) and gastrointestinal/hepatobiliary diseases (36.5%) [Table 5]. Many of their visits were less serious or complex, compared with visits by NUs, FUs, and HIUs. Only 9.4% of visits by superusers required the use of ambulance services. Most visits by superusers were triaged as semi-urgent or non-urgent (90.9%). Musculoskeletal pain and minor infections represented 54.3% of the principal diagnoses for superusers.

The characteristics of individuals who frequently use EDs in Hong Kong have not been extensively investigated. Frequent users and HIUs constituted 9% of all ED patients and represented 27.2% of all visits. Region-based studies in developed countries have revealed similar findings.6 7 13 14 15 16 Although frequent ED use is often assumed to indicate inappropriate use, our data do not support this assumption. The health statuses and presentation conditions of FUs strongly suggest that these individuals have greater health needs than the rest of the population. Frequent users were more likely to have underlying illnesses, require ambulance services, have visits triaged as urgent or above, require hospital admission, and have a longer length of hospitalisation. In contrast, HIUs had visits triaged as less urgent, required less investigations, often attended for conditions (eg, musculoskeletal pain and minor infections) that could potentially be managed in primary care clinics, and had lower admission rates compared with both NUs and FUs. These trends were much more pronounced among superusers.

High percentages of visits involved musculoskeletal pain and minor infections, conditions that could potentially be managed in primary care clinics. Such visits could be related to a misunderstanding of ED function, a misperception of condition severity, or use of the ED as a substitute for unavailable forms of care. Policies to decrease ED utilisation burden should focus on improving healthcare delivery services and primary care; such policies are likely to benefit NUs with similar needs. Recent advances in information technology and the widespread use of smartphones offer many opportunities to improve information dissemination and increase accessibility to alternative sources of care.

Our findings suggest disproportionate use of ED services by ethnic minorities. The number of Hong Kong residents of non-Chinese ethnicity rose by 70% over a 10-year period; they constituted 8% of the total population in 2016.8 The Chinese literacy rate was <20% in some age-groups.8 More frequent ED use may result from language barriers to access of care, limited knowledge of local healthcare alternatives, or cultural differences in health-seeking behaviour. Effective policy measures targeting these populations should promote an understanding of the local healthcare system through informative advertisements about alternative services offered; measures should also reduce language barriers (eg, through translator services or the provision of online booking instructions in other languages).

Frequent users constitute a sicker population requiring medical care that cannot be easily provided in primary care clinics. They need support to facilitate integration and maintenance in the community, which would reduce the need for emergency medical care at EDs. The burden placed on EDs may be reduced by developing new community-based support systems, such as enhanced coverage of community nursing services and extended service hours (eg, weekends and public holidays); dedicated multidisciplinary teams to provide rehabilitation services and caregiver training in the community; and centres for the provision of coordinated community services. Eligible geriatric patients could benefit from a holistic community geriatric care strategy triggered by an ED visit, with protocol-driven assessments in the ED to identify potentially reversible risk factors for subsequent deterioration and repeated ED visits. This strategy should include comprehensive geriatric assessment prior to ED discharge, focused on factors such as fall risk, delirium screening, and frailty assessment. In addition to physical assessments, psychosocial assessments and support may improve patient outcomes and minimise repeated ED visits. Pre-discharge care planning that empowers patients and family members to seek help from non-ED sources may also prevent repeated ED visits. Thus far, there remains uncertainty about the effectiveness of community-based interventions for people with multimorbidity because of the relatively small number of randomised controlled trials focused on this area of healthcare.17

Reformation of the chronic disease service model may improve quality of care and reduce repeated ED visits. Patient-based care, rather than disease-based care, may be beneficial. Patients with multimorbidity (especially older adults) receive medical treatment through multiple specialty or subspecialty clinics, which can result in fragmented and duplicative care. It is not uncommon for patients with chronic diseases to attend the ED for minor problems and questions about their chronic diseases because they cannot find an alternative source of medical advice. Case management helps improve outcomes in some chronic diseases.18 Efforts to strengthen the abilities of primary care clinics to function as ‘case managers’ for patients with chronic diseases may improve quality of care and reduce the number of ED visits.

In this study, mental health disorders and substance abuse were significant predictors of frequent ED use. A previous work indicated that one in seven Hong Kong residents aged 16 to 75 years has anxiety, depression, or another common mood disorder.19 The HA is the main specialist service provider for patients with mental health disorders. It provides in-patient facilities, day hospitals, specialist out-patient clinics, and community outreach services. The HA is experiencing increased demand for specialist mental health services,20 which may be causing patients to use EDs instead. Emergency department visits and readmissions for psychiatric problems may be reduced by reforming the current service model to expand community psychiatric services, with a focus on personalised care for psychiatric patients and their caregivers through a case management approach that facilitates community re-integration and strengthens recovery. Enhanced screening to identify early features of mental health disorders may allow earlier detection and treatment, thereby reducing ED utilisation. Patients and caregivers should receive education about health-seeking behaviours during instances of acute deterioration (eg, using a 24-hour psychiatric advisory hotline or undergoing urgent assessment at a psychiatric specialist out-patient clinic), rather than simply using the ED as a safety net.

This study was limited to the three EDs in the New Territories West Cluster. Its findings may not be generalisable to other regions in Hong Kong with different demographics, health-seeking behaviours, and socio-economic statuses. Also, this study only investigated ED visits within a specific time period and did not consider past or future periods. Thus, it may have underestimated ED visits for patients who were born or died during the study period.

By reviewing diagnosis codes, we were able to include many visits and patients in the analysis; however, we could not analyse individual charts. Although coding is a routine component of hospital procedures, codes are only required for the current condition or presenting problem. Generally, coding is not mandatory for appointments at out-patient clinics. This difference in coding information may have led to underestimation of patient comorbidities. To mitigate this possibility, we examined all diagnosis codes from the past 5 years to acquire a more complete representation of underlying comorbidities.

Because this study excluded patients without valid identification documents, homeless persons may have been underrepresented. These individuals potentially have a heavier disease burden and a disproportionate share of frequent visits. The study may also have excluded visitors to Hong Kong and individuals who do not have residency status.

Frequent users and HIUs are a small but diverse population that represents a substantial proportion of annual ED visits. Demographic factors, economic considerations, and medical conditions all contribute to increased numbers of ED visits. Our data suggest that there are many opportunities for improvement via streamlining and enhancement of healthcare delivery to reduce ED utilisation.

Concept or design: PYT Ng, CT Lui.
Acquisition of data: PYT Ng, CT Lui.
Analysis or interpretation of data: PYT Ng, CT Lui.
Drafting of the manuscript: PYT Ng, CT Lui.
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 Mr Chun-ho Lam, Statistical Officer at Research Assist Team of the New Territories West Cluster of Hospital Authority, Hong Kong, for his advice on the zero-truncated Poisson regression model.

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

This research was approved by the New Territories West Cluster Research Ethics Committee of Hospital Authority, Hong Kong (Ref No.: NTWC/REC/19081). Informed patient consent was waived by the Committee due to the retrospective nature of the study.

1. Hospital Authority, Hong Kong SAR Government. Hospital Authority annual report 2018-2019. Available from: https://www.ha.org.hk/ho/corpcomm/AR201819/PDF/HA_Annual_Report_2018-2019.pdf. Accessed 1 May 2020.

2. Pines JM, Hilton JA, Weber EJ, et al. International perspectives on emergency department crowding. Acad Emerg Med 2011;18:1358-70. Crossref

3. Data.gov.hk, Hong Kong SAR Government. Projections of population distribution 2021-2029. Available from: https://data.gov.hk/en-data/dataset/hk-pland-pland1-projections-of-population-distribution-2021-to-2029. Accessed 17 Jul 2023.

4. Hospital Authority, Hong Kong SAR Government. List of all accident & emergency departments. 2023. Available from: https://www.ha.org.hk/visitor/ha_visitor_index.asp?Content_ID=200246&lang=ENG. Accessed 19 Jul 2023.

5. Hospital Authority, Hong Kong SAR Government. Introduction of clusters: New Territories West Cluster. 2023. Available from: https://www.ha.org.hk/visitor/ha_visitor_index.asp?Content_ID=10181&Lang=ENG&Dimension=100&Parent_ID=10084. Accessed 19 Jul 2023.

6. Krieg C, Hudon C, Chouinard MC, Dufour I. Individual predictors of frequent emergency department use: a scoping review. BMC Health Serv Res 2016;16:594. Crossref

7. Dr Foster UK. High intensity users: reducing the burden on accident & emergency departments. Available from: https://www.telstrahealth.com/content/dam/telstrahealth/pdf-downloads/Dr-Foster_High-Intensity-Users-Report.pdf. Accessed 10 Aug 2023.

8. 2016 Population By-census, Hong Kong SAR Government. District profiles. Available from: https://www.bycensus2016.gov.hk/en/bc-dp.html. Accessed 1 May 2020.

9. World Health Organization & International Conference for the Ninth Revision of the International Classification of Diseases. Manual of the international statistical classification of diseases, injuries, and causes of death: based on the recommendations of the ninth revision conference, 1975, and adopted by the Twenty-ninth World Health Assembly, 1975 revision. 2021. Available from: https://apps.who.int/iris/handle/10665/40492. Accessed 19 Jul 2023.

10. Hospital Authority, Hong Kong SAR Government. Operations. List of private services. Available from: https://www3.ha.org.hk/fnc/Operations.aspx?lang=ENG. Accessed 19 Jul 2023.

11. R Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing. Available from: https://www.R-project.org/. Accessed 17 Jul 2023.

12. Yee TW. Vector generalized linear and additive models: with an implementation in R. Springer. Available from: https://link.springer.com/book/10.1007/978-1-4939-2818-7. Accessed 17 Jul 2023. Crossref

13. Hunt KA, Weber EJ, Showstack JA, Colby DC, Callaham ML. Characteristics of frequent users of emergency departments. Ann Emerg Med 2006;48:1-8. Crossref

14. Lee WL, Chen WT, Hsiao FH, Huang CH, Huang LY. Characteristics and resource utilization associated with frequent users of emergency departments. Emerg Med Int. 2022;2022:8064011. Crossref

15. Leporatti L, Ameri M, Trinchero C, Orcamo P, Montefiori M. Targeting frequent users of emergency departments: prominent risk factors and policy implications. Health Policy 2016;120:462-70. Crossref

16. Fuda KK, Immekus R. Frequent users of Massachusetts emergency departments: a statewide analysis. Ann Emerg Med 2006;48:9-16. Crossref

17. Smith SM, Wallace E, O’Dowd T, Fortin M. Interventions for improving outcomes in patients with multimorbidity in primary care and community settings. Cochrane Database Syst Rev 2021;1:CD006560. Crossref

18. Reilly S, Miranda-Castillo C, Malouf R, et al. Case management approaches to home support for people with dementia. Cochrane Database Syst Rev 2015;1:CD008345. Crossref

19. Lam LC, Wong CS, Wang MJ, et al. Prevalence, psychosocial correlates and service utilization of depressive and anxiety disorders in Hong Kong: the Hong Kong Mental Morbidity Survey (HKMMS). Soc Psychiatry Psychiatr Epidemiol 2015;50:1379-88. Crossref

20. Food and Health Bureau, Hong Kong SAR Government. Mental health review report. Available from: https://www.fhb.gov.hk/download/press_and_publications/otherinfo/180500_mhr/e_mhr_full_report.pdf. Accessed 1 May 2020.