Factors associated with persistent high healthcare service utilisers in Singapore: A population health analysis

ABSTRACT

Introduction: Albeit comprising a small portion of the hospital population, persistent high utilisers (PHUs) contribute disproportionately to healthcare expenditures. Amid rising healthcare costs and an ageing population, this study examines factors associated with PHUs among residents in eastern Singapore.

Method: This is a retrospective study of eligible patients at Changi General Hospital in Singapore between 1 January 2020 and 31 December 2022. The study included Singapore citizens who utilised any services offered by CGH. Patients were classified as PHUs if their annual healthcare expenditure exceeded SGD3700 for 3 consecutive years. Demographics, healthcare utilisation patterns and clinical profiles were compared, and multivariable analyses were conducted to identify factors associated with PHUs.  

Results: There were 267,838 eligible patients identified, with 5316 (2%) classified as PHUs. PHUs accounted for 18.4% of the total healthcare expenditure, with the highest costs attributed to inpatient services, followed by outpatient services. PHUs were more likely to be older, male, non-Chinese and of lower socioeconomic status. Conditions with the strongest association with PHUs were mental health disorders, chronic obstructive pulmonary disease, osteoporosis, asthma and renal diseases. Inpatient discharges from renal medicine, psychological medicine and general/geriatric medicine wards had the strongest association with PHUs. Utilisation of allied health services had the highest odds of being a PHU in outpatient care.

Conclusion: This study identified key factors associated with PHUs, providing invaluable insights into the planning of population health services within the hospital’s geographical region. Targeted service development and process improvements of medical care can help mitigate persistent high utilisation.

CLINICAL IMPACT

What is New

• This is the first study in Singapore to identify persistent high healthcare service utilisers from the lens of national-level healthcare utilisation.
• The findings identified key factors associated with persistent high healthcare utilisation.

Clinical Implications

• This study further underscored the importance of advancing integrated care between hospitals, community and primary care in the management of patients with complex care to manage persistent high utilisation of healthcare services for conditions beyond diabetes, hypertension and dyslipidaemia.

Singapore’s healthcare system comprises private and public sector providers,¹ financed through a combination of government subvention and subsidies, personal medical savings and funds, health insurance and a government endowment fund.² Public sector primary care, intermediate and long-term care (ILTC) facilities, and hospitals are grouped and managed by 3 healthcare clusters that receive annual subvention and subsidies from the government for the provision of medical services. In 2023, Singapore’s Ministry of Health (MOH) launched the multi-year Healthier SG (a national initiative to focus on preventive healthcare) strategy,³ with 1 key feature to support the 3 healthcare clusters in managing primary care, ILTC and hospitals to care for the residents within their catchment areas. Supported by capitation funding based on the residential population, the strategy focuses on right-siting, service efficiency and preventive care.³ Patients may seek care from private or public providers island-wide, where transfer payments will be made between clusters.³ To optimise the use of capitation funds,⁴ the healthcare clusters would need to identify and implement targeted interventions for high utilisers (HU), reduce healthcare costs and allow resource reallocation. Addressing HUs minimises cost variability and stabilises resource allocation under the capitation funding model.

Patients with complex needs are often HUs of healthcare services,^5,6 with various definitions used to identify them.⁷udies showed that most HUs reduce their utilisation over time,^8,9 but a subgroup of them termed persistent HUs (PHUs),^10-12 continue high utilisation for more than 2 years.^10,13 Although PHUs represent a small proportion of the population, they incur disproportionate healthcare expenditures.^{10,11,14,15,24,25} Early identification of PHUs through various definitions aims to reduce future utilisation.^16,17 However, heterogeneity in definitions, study populations and healthcare financing models makes direct comparisons and application to the Singapore context difficult. 

Although some studies in Singapore have characterised PHUs, they focused on either inpatient settings¹⁸ or specific conditions,¹⁹ defining high utilisation based on expenditures, a minimum length of stay or a minimum number of emergency department (ED) visits. While effective for capturing PHUs with high inpatient costs, these approaches may miss other healthcare utilisation patterns such as high outpatient clinic visits or medication costs without inpatient admissions. Moreover, with residents given the flexibility to seek care across multiple institutions nationwide, such a method may not adequately identify PHUs with dispersed utilisation patterns.

With these in mind, this study aimed to examine factors associated with PHUs among residents in eastern Singapore by adopting a different approach to identify PHUs from the lens of national-level healthcare utilisation. Findings from this study could offer potential insights on possible early identification of PHUs, targeted interventions for PHUs and strategic healthcare resource reallocation to reduce their healthcare utilisation. Overall, this study comprised 2 main components. First, it identified PHUs residing within the catchment area of Changi General Hospital (CGH)—a public tertiary hospital in eastern Singapore—from 2017 to 2019, from the national registry. Thereafter, the annual gross bill threshold incurred by these PHUs in CGH was derived and applied to identify PHUs who had visited CGH in 2020–2022.

METHOD

Study design, settings and participants

This retrospective cohort study included all CGH patients who utilised the hospital’s services (i.e. ED, inpatient, day surgery, specialist outpatient clinics and community health services—referred to as outpatient services) between 1 January 2020 and 31 December 2022, excluding non-Singapore citizens and those deceased before 1 January 2023.

Definitions of PHUs

Although utilisation and cost are both common approaches, used either in combination or separately in defining PHUs,²⁰ an expenditure-based approach (surrogate for cost) was employed instead of a utilisation-based approach, basing the expenditure on a patient’s gross healthcare bills, for a few reasons. First, a utilisation-based approach may not accurately capture instances of costly healthcare use, such as short inpatient stays that result in disproportionately high bills. Second, in the absence of cost data reflecting the actual expenses incurred by the healthcare system in delivering services, expenditure serves as the most appropriate proxy for assessing patients’ healthcare utilisation.²¹

This study classified Singapore citizens as HUs in a year if their annual gross healthcare bills were within the 10th decile of the Singapore population in the same year. These gross healthcare bills referred to gross bills (inclusive of any government subsidies or third-party payments) captured in MOH’s Population Health Staple Dataset (PHSD), which recorded all healthcare services utilised by residents at all public healthcare institutions and care settings in Singapore, subsidised primary care and step-down care services (offered by ILTC facilities), inpatient, day surgery and selected outpatient services in private hospitals.

First, PHSD was used to identify PHUs who resided within the catchment area of CGH from 2017 to 2019. Then, the threshold of annual gross bills—which were incurred by these PHUs in CGH—was estimated at SGD3700; this was used to identify PHUs among patients who had visited CGH in 2020–2022. Details of how the threshold was derived from PHSD are described in Supplementary Appendix S2 and Fig. S1. Therefore, a patient was classified as an HU that year if he/she had an annual expenditure of at least SGD3700 in CGH. PHUs were then defined as patients who were HUs for 3 consecutive years from 2020 to 2022 and non-PHUs otherwise. A flowchart of this methodology is summarised in Fig. 1.

Fig. 1. Flowchart for PHU determination.

In this study, CGH data (instead of the national registry) were used for a few reasons. First, institutional data can be timely accessed with minimal delays and administrative barriers. Second, internal data autonomy enables further qualitative exploration, such as case note reviews, to identify care gaps and understand PHU profiles. Last, institutional data offered higher granularity, capturing details like inpatient discharge specialities, outpatient visit characters and a detailed breakdown of bills by service type. This allowed deeper insights often missed in the national registry. 

Data sources and variables

All datasets were obtained from CGH’s administrative databases. Visit-level data were extracted for inpatient admissions, ED visits, outpatient specialist visits and day surgery admissions. Bills data were extracted by service types (e.g. gross bill for specific ward types during inpatient admissions). Patient-level variables were aggregated annually (2020–2022) for variables such as (1) demographics, (2) healthcare service utilisation (i.e. total gross bills, number of visits/admissions and visit/admission details by service type) and (3) disease conditions.

Patients’ age, sex, ethnicity, resident status and housing type (based on the housing type with the largest proportion in the block via patients’ postal code data²²) were studied. To estimate the density of PHU in each geographic location, a population-related census by geographical location was obtained from the website by a government agency in Singapore and referenced to the 2020 data.²² All healthcare expenditures were adjusted to 2022 Singapore dollars.²³

Dates of expenditures for emergency and outpatient services were based on attendance dates, while those for inpatient services were based on admission dates. Annual healthcare expenditure (gross bills and number of visits/admissions) was estimated by summing gross bills, including subsidies and third-party payments, by year. Patients’ disease conditions were identified based on inpatient discharge diagnosis codes (International Classification of Diseases, Tenth Revision), broadly categorised as detailed in Supplementary Table S3. Inpatient discharge specialities and outpatient specialities were collated to assess associations with PHU status.

Statistical methods

Categorical variables were summarised as frequencies and percentages, while continuous variables as mean and standard deviation (SD). Comparisons between PHUs and non-PHUs were tested with chi-square tests and two-sample t-test for categorical and continuous variables, respectively. For healthcare utilisation-related data, the mean with SD and median with interquartile range of the total healthcare expenditures (2020–2022) were presented, and differences were tested with the Mann-Whitney U-test. Multivariable logistic regressions were conducted to study the association between PHU status with different disease conditions, healthcare utilisation patterns and demographic characteristics. Statistical significance was set at 5% and analyses were performed using R version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Patient population

From 1 January 2020 to 31 December 2022, 376,877 patients visited CGH. After excluding non-Singapore citizens (25.6%) and those who passed away before 1 January 2023 (3.3%), 267,838 patients remained in the study cohort. Of these, 5316 were identified as PHUs (1.98%) and 262,522 as non-PHUs (98.0%) (Fig. 2). Fig. 3 illustrates expenditure differences over 3 years for PHUs and non-PHUs, and their distribution across service types. Across the 3 years, despite comprising only 1.98% of the study cohort, PHUs incurred disproportionately higher total healthcare expenditure at SGD298.8 million (18.4% of CGH’s total expenditure), compared to the 99.1% of non-PHUs at SGD1329.5 million. The mean total expenditure for PHUs was SGD56,200 per patient, 11 times that of non-PHUs, at SGD5064.

Fig. 2. Flowchart of patients included in the study cohort.

Fig. 3. Breakdown of total bill cumulated from 2020 to 2022 for services utilised by PHUs versus non-PHUs.

Patient population characteristics

Table 1 compares the demographic profiles of PHUs and non-PHUs. The mean age of PHUs was statistically higher than non-PHUs at 66.3 (SD=14.6) and 48.9 (SD=19.7), respectively (P<0.001). The proportion of males was higher in PHUs (58%) compared to non-PHUs (56%). The ethnic distribution of patients was different between PHUs and non-PHUs (P<0.001) with more non-Chinese (Indian, Malay and other ethnic groups) in the PHU group (39%) as compared to the non-PHU group (33%). The distribution of housing types was statistically different (P<0.01); PHUs (34%) lived in 1- to 3-room flats (a proxy for lower socioeconomic status [SES]), which was statistically higher compared to non-PHUs (23%).

Table 1. Demographic comparison of PHUs and non-PHUs.

Table 2 summarises differences in healthcare service utilisation between PHUs and non-PHUs. Among the cohort, 27% had inpatient admissions (89% of PHUs and 26% non-PHUs). PHUs had 3.5 times more admissions and a mean inpatient expenditure of 4.6 times higher than non-PHUs. The proportion of urgent admissions was similar between groups despite statistical significance. Still, PHUs had an average length of stay 4.0 times longer than non-PHUs, with higher usage across all ward types. Regarding ED utilisation, 60% of the study cohort visited the ED (90% of PHUs and 59% of non-PHUs). PHUs had 4.0 times higher mean ED expenditures and 3.76 times more visits than non-PHUs. Self-referred ED visits were higher among PHUs (73.9%) than non-PHUs (64.9%), but the difference was not statistically significant. For outpatient services, 80% of the cohort utilised them, including all PHUs and 80% of non-PHUs. PHUs had 7.5 times higher mean outpatient expenditures and 6.2 times more visits than non-PHUs. The mean number of no-show visits was 3.1 times higher in PHUs than in non-PHUs. Last, 21% of the study cohort underwent day surgery (58% of PHUs and 21% of non-PHUs). PHUs had 1.8 times higher mean day surgery expenditure and 1.7 times more day surgery than non-PHUs.

Table 2. Service utilisation and bill size comparison of PHU and non-PHUs.

Among those with inpatient admission, all diseases were significantly associated with being a PHU, except dyslipidaemia and myocardial infarction (MI) (Table 3 [item A]). Specifically, mental health disorders (presence of either schizophrenia, major depressive disorder, anxiety disorder or bipolar disorder) showed the strongest association (odds ratio [OR]=3.48; 95% confidence interval [CI]=[2.97, 4.08]), followed by chronic obstructive pulmonary disease (COPD) (OR=2.69; 95% CI=[2.29, 3.17]) and osteoporosis (OR=2.64; 95% CI=[2.36, 2.94]). Table 3 (item B) shows that renal medicine or nephrology department (OR=14.1; 95% CI=[12.2, 16.2]) had the strongest association with being a PHU, followed by the psychological medicine (OR=11; 95% CI=[7.86, 15.3]), general medicine or geriatric medicine (OR=4.9; 95% CI=[4.55, 5.28]). Table 3 (item C) highlights that visits to allied health services had the highest odds of being a PHU (OR=23.2; 95% CI=[20, 27]). Across the 3 logistic regression models, older age, male sex, non-Chinese and lower socioeconomic status had significantly higher odds of being a PHU.

Table 4 details the mean bills for each service type and services within the service types by PHUs and non-PHUs, while Supplementary Fig. S2 details the proportion of services spent for each service type by groups. For inpatient admission, ward services and investigations constitute the largest portion of bills, with PHUs spending 4.13 and 3.69 times more, respectively, compared to non-PHUs. In the ED, attendance fees and investigation dominated costs, with PHUs spending 4.77 and 3.75 times, respectively than non-PHUs. In outpatient care, PHUs primarily spent on medications, investigations and consultations, whereas non-PHUs spent most on investigations. PHU medication bills were 14.8 times higher than non-PHUs. For day surgery, facility fees accounted for the largest share, with PHUs spending 1.7 times more than non-PHUs. The heat map in Supplementary Fig. S3 shows the density of PHUs per 1000 residents in subzones with over 3000 residents. Bedok North (7.83), Tampines East (7.16) and Kembangan (7.12) had the highest PHU densities.

Table 3. Distribution of (A) disease profile, (B) inpatient discharge specialties, (C) outpatient specialties by groups and adjusted OR from respective logistic regression models with 95% CI for diseases associated with being a PHU.

Table 4. Mean bills breakdown by service types and different services (in SGD).

DISCUSSION

Key results

This study identified factors significantly associated with being a PHU, including older age, male sex, non-Chinese and lower SES. Despite only comprising 2% of the study cohort, they accounted for nearly 20% of the total healthcare expenditure. PHUs spent 4.6, 4.0, 7.5 and 1.8 times more on inpatient, ED, outpatient and day surgery services respectively, compared to non-PHUs. Mental health disorders had the strongest association with being a PHU, followed by COPD and osteoporosis. Renal medicine was the inpatient specialty most strongly associated with PHUs, while utilisation of allied health services in outpatient settings had the strongest association.

Interpretation

This study’s analysis reflected the high prevalence of cardiovascular and other chronic medical conditions in PHUs, which was commonly observed across studies.^5,10,11,18 While diseases such as diabetes, hypertension and dyslipidaemia are more prevalent in this study’s population settings, the analysis saw specifically that mental health issues, COPD and osteoporosis had the highest odds of being a PHU, highlighting that prioritisation of the needs of these conditions to address high utilisation would be useful. While the prevalence of mental health disorders (5.0%) among PHUs was lower than other chronic medical conditions, its strong association (OR 3.48) was observed, in line with many studies in US Medicaid and total population studies.^7,24

Renal conditions were the most prevalent among the diseases with the highest odds of association with being a PHU. This was aligned with the findings of other studies given that patients often had worsening symptoms with progression of chronic kidney disease, with other possible complications such as acute kidney injury, heart failure, MI, stroke and anaemia requiring outpatient, ED visits and inpatient admissions.^25-27 Utilisation of allied health services in outpatient settings was strongly associated with PHU status, which was unsurprising given that PHUs usually have functional impairments or psychosocial challenges, requiring intervention by allied health professionals. This study further underscored the importance of advancing integrated care between hospitals, community and primary care, e.g. via patient-centred care models^12,30-33 in the management of patients with complex care. Integrated care models through community sited services can reduce reliance on hospital-based allied health intervention and address persistent high utilisation more effectively. Future studies could benefit from cross-jurisdictional comparisons to understand how differences in healthcare systems, social practices and cultural structures influence healthcare utilisation.

Similar to other studies, PHUs in this study were older, lived in smaller flat sizes and have a higher prevalence of chronic diseases as well as mental health disorders. While the OR for being a PHU was higher in those above the age of 41, the OR was lower for those above 80 years compared to those aged between 41 and 80 years. While a previous study in the US²⁸ attributed similar findings to mortality effects, this explanation was not applicable to this study as patients who were deceased prior to 2023 were excluded. The mechanism underlying this observation warrants further investigation by future studies. The study’s findings also align with a previous study in Singapore²⁹, which identified total length of stay in the last 12 months and days following the last non-elective admission as key predictors of frequent hospital use. This reinforces the value of historical inpatient utilisation in identifying high-needs patients. Last, useful insights on the distribution of PHUs across the catchment area of CGH were provided to aid regional health management. The identification of the subzones with a higher density of PHUs would facilitate the assessment and development of service needs in these areas with stakeholders in medical care as well as social services.

Strength

Defining PHU in this study presented methodological challenge due to substantial heterogeneity in existing literature on study population selection criteria, PHU definitions and healthcare expenditure metrics.³ To the authors’ knowledge, only 1 other study from Singapore conducted analysis on PHUs in a 1000-bed academic medical centre,¹⁸ using the annual threshold of the top 10% of inpatient expenditure (SGD8150) to define a PHU (if they were HUs for 3 consecutive years). A novel threshold-setting approach, which considered the residents’ total healthcare expenditure and those who had significant hospital service utilisations, was undertaken. In this study, the threshold for an HU in the hospital was established at SGD3700, taking reference from the healthcare expenditure of the resident population across all healthcare services in hospitals and the community. While the derived threshold was lower, this reflects a larger denominator with further lower-cost users. Additionally, policy shifts between the 2 periods (2005–2013 and 2020–2022) such as the expansion of Community Health Assist Scheme³⁰ in 2019, a Singapore government programme that provides subsidies for Singapore citizens at participating clinics, may have expanded subsidies and shifted care towards the community, reducing hospital-centric expenditures and lowering top-decile thresholds in this analysis. Further, this study considered all healthcare services within CGH apart from just inpatient admission, although inpatient admissions were the main contributor to the healthcare expenditures. This approach revealed that high outpatient service utilisation contributed substantially to total expenditure, with some patients classified as PHU based on outpatient service utilisation alone. This approach, which expanded the study population beyond a focus on specific chronic conditions,¹⁹ captured patient groups who might otherwise slip through the gaps in care, particularly those with high healthcare needs that do not fit neatly into specific disease categories. Last, the analysis of visit-level and billing-item-level data provided granular insights into utilisation patterns that may not be discernible in national-level datasets.

This study identified key factors associated with PHUs. These factors can inform clinical interventions and system-level strategies to address PHUs. Once these interventions and strategies are identified, relevant resources can be scoped and set aside so that the interventions and strategies can be implemented efficiently to address PHUs. These resources may include (1) manpower and information technology (IT) infrastructure needed for the development of predictive tools for early identification of at-risk PHU patients using predictive tools; (2) manpower, workflow and IT integration to enhance coordination between hospital, community and primary care, particularly in managing mental health, renal care and rehabilitation, so that the fragmentation of care can be reduced and unnecessary utilisation can be prevented. Additionally, the geospatial insights on PHU concentration across subzones offer a basis for prioritising service development and outreach efforts within the catchment area in Singapore. Admittedly, the findings may not be generalisable to other countries due to underlying differences in healthcare practices and policies. Nevertheless, the reported methodology offers a systematic framework for studying PHUs, and this framework can be readily employed by other researchers in other countries who also aim to develop clinical interventions and system-level strategies that address PHUs.

Limitations

This study is not without its limitations. First, the analysis was restricted to PHUs who utilised CGH’s services, excluding those residing in CGH’s catchment area but sought care elsewhere. Second, while many CGH patients were discharged to community hospitals for continued care, expenditure data from these facilities were unavailable for a comprehensive analysis. Third, healthcare utilisation patterns would usually fluctuate with patient, policy and environmental factors, which should be considered when the results from the analysis is generalised. Also, the cross-sectional design of this study using calendar year cut-offs may have resulted in patients who were ill and consumed a series of sequential services spanning the year-end to fall below the HU threshold, compared to a patient with similar care journey beginning earlier in the year.

Additionally, the derivation of the annual expenditure threshold did not account for expenditures incurred in private services due to data limitations. This could partly explain the association of low SES with PHU as higher SES residents may tend to seek care in private institutions. The macro-costing approach used, although suitable for estimating overall healthcare utilisation, unfortunately does not allow attribution of expenditure to specific conditions. As such, observed associations with individual chronic diseases, may reflect the cumulative cost burden of multimorbidity rather than isolated disease effects. The study period was also affected by the COVID-19 pandemic, which may have influenced healthcare utilisation patterns. However, the recency of the data for this study context (requiring 3 years of data) ensures relevance despite this limitation. Additionally, findings from these patients identified during the pandemic are likely to offer more valuable and relevant insights, as they are more likely to be clinically complex (given that many avoided healthcare services during the pandemic). 

CONCLUSION

This study highlighted PHUs as a small but distinct patient group with disproportionately high healthcare utilisation across all service types. Their profile (i.e. older age, male sex, lower socioeconomic status, chronic disease burden and strong associations with mental health disorders, renal conditions and allied health services used) highlighted the complexity of their needs. These findings can inform clinical interventions and system-level strategies to address PHUs. Early identification of at-risk patients using predictive tools can support timely care planning. Enhanced coordination between hospital, community and primary care, particularly in managing mental health, renal care and rehabilitation, may reduce the fragmentation of care and prevent unnecessary utilisation.

Supplementary materials
Appendix S1. Population Health Staple Dataset (PHSD): The Singapore National Healthcare Expenditure Registry dataset.
Appendix S2. Threshold estimation.
Fig. S1. Breakdown of eligible residents from the study cohort in DS1.
Fig. S2. Bills breakdown by PHUs and non-PHUs for the different services for (A) inpatient admission, (B) emergency department, (C) outpatient and (D) day surgery.
Fig. S3. Heat map for the no. of resident PHUs per 1000 residential population in the respective subzones.
Table S1. Comparison of catchment area demographics with study population demographics.
Table S2. Determination of AET in DS1.
Table S3. ICD-10 codes for broad disease conditions.
Table S4. Fully adjusted logistic regression models with 95% CI (A) for diseases, (B) inpatient discharge specialty and (C) outpatient specialty.
Table S5. Subgroup analysis: PHUs versus non-PHUs by sex (female and male), age group (above 60 and 60 and below) and ethnicity (Chinese, Malay, Indian and Others).

REFERENCES

1. Ministry of Health, Singapore. Upcoming Healthcare Facilities from 2025 onwards. https://www.moh.gov.sg/seeking-healthcare/find-a-facility-or-service/types-of-medical-facilities-and-services/find-a-medical-facility/facilities-from-2025-onwards/. Accessed 18 August 2025.
2. Ministry of Health, Singapore. Government Health Expenditure and Healthcare Financing. https://www.moh.gov.sg/others/resources-and-statistics/government-health-expenditure-and-healthcare-financing. Accessed 18 August 2025.
3. Ministry of Health, Singapore. The White Paper on Healthier SG, 21 September 2022. https://www.healthiersg.gov.sg/resources/white-paper/. Accessed 1 March 2024.
4. Ministry of Health, Singapore. Capitation, 10 May 2023. https://www.moh.gov.sg/newsroom/capitation. Accessed 30 December 2024.
5. Zook CJ, Moore FD. High-cost users of medical care. N Engl J Med 1980;302:996-1002.
6. Reid R, Evans R, Barer M, et al. Conspicuous consumption: characterizing high users of physician services in one Canadian province. J Health Serv Res Policy 2003;8:215-24.
7. Wammes JJG, van der Wees PJ, Tanke MAC, et al. Systematic review of high-cost patients’ characteristics and healthcare utilisation. BMJ Open 2018;8:e023113.
8. Chakravarty S, Cantor JC. Informing the Design and Evaluation of Superuser Care Management Initiatives: Accounting for Regression-to-the-Mean. Med Care 2016;54:860-7.
9. Johnson TL, Rinehart DJ, Durfee J, et al. For many patients who use large amounts of health care services, the need is intense yet temporary. Health Aff (Millwood) 2015;34:1312-9.
10. Wodchis WP, Austin PC, Henry DA. A 3-year study of high-cost users of health care. CMAJ 2016;188:182-8.
11. Figueroa JF, Zhou X, Jha AK. Characteristics And Spending Patterns of Persistently High-Cost Medicare patients. Health Aff (Millwood) 2019;38:107-14.
12. Longden T, Wong CY, Haywood P, et al. The prevalence of persistence and related health status: An analysis of persistently high healthcare costs in the short term and medium term. Soc Sci Med 2018;211:147-56.
13. Chang HY, Boyd CM, Leff B, et al. Identifying Consistent High-cost Users in a Health Plan: Comparison of Alternative Prediction Models. Med Care 2016;54:852-9.
14. Cohen SB. The Concentration and Persistence in the Level of Health Expenditures over Time: Estimates for the U.S. Population, 2012–2013. September 2015. In: Statistical Brief (Medical Expenditure Panel Survey (US)) [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2001.
15. Jiang L, Qiu Q, Zhu L, et al. Identifying Characteristics Associated with the Concentration and Persistence of Medical Expenses among Middle-Aged and Elderly Adults: Findings from the China Health and Retirement Longitudinal Survey. Int J Environ Res and Public Health 2022;19:12843.
16. Lee JY, Muratov S, Tarride JE, et al. Managing High‐Cost Healthcare Users: The International Search for Effective Evidence‐Supported Strategies. J Am Geriatr Soc 2018;66:1002-8.
17. Howson SN, McShea MJ, Ramachandran R, et al. Improving the Prediction of Persistent High Health Care Utilizers: Retrospective Analysis Using Ensemble Methodology. JMIR Med Inform 2022;10:e33212.
18. Ng SHX, Rahman N, Ang IYH, et al. Characterising and predicting persistent high-cost utilisers in healthcare: a retrospective cohort study in Singapore. BMJ Open 2020;10:e031622.
19. Tan JK, Quan L, Salim NNM, et al. Machine Learning–Based Prediction for High Health Care Utilizers by Using a Multi-Institutional Diabetes Registry: Model Training and Evaluation. JMIR AI 2024;3:e58463.
20. Berkman ND, Chang E, Seibert J, et al. Management of High-Need, High-Cost Patients: A “Best Fit” Framework Synthesis, Realist Review, and Systematic Review [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2021.
21. Schousboe JT, Paudel ML, Taylor BC, et al. Estimating True Resource Costs of Outpatient Care for Medicare Beneficiaries: Standardized Costs versus Medicare Payments and Charges. Health Services Research 2016;51:205-19.
22. Gov.sg. Sold and Rented HDB Properties and Facilities. Updated 15 July 2015. https://data.gov.sg/datasets?topics=housing&page=3&resultId=187. Accessed 10 June 2022.
23. Ministry of Health, Singapore. Consumer Price Indices (CPI). https://www.healthhub.sg/a-z/health-statistics/consumer-price-indices-cpi. Accessed 10 June 2022.
24. Banerjee S, Alabaster A, Kipnis P, et al. Factors associated with persistent high health care utilization in managed Medicaid. Am J Manag Care 2021;27:340-4.
25. Sanchez JJG, James G, Carrero JJ, et al. Health Care Resource Utilization and Related Costs of Patients With CKD From the United States: A Report From the DISCOVER CKD Retrospective Cohort. Kidney Int Rep 2023;8:785-95.
26. Jha V, Al-Ghamdi SMG, Li G, et al. Global Economic Burden Associated with Chronic Kidney Disease: A Pragmatic Review of Medical Costs for the Inside CKD Research Programme. Adv Ther 2023;40:4405-20.
27. Chen W, Howard K, Gorham G, et al. Costs and healthcare use of patients with chronic kidney disease in the Northern Territory, Australia. BMC Health Serv Res 2024;24:791.
28. Yoon J, Chee CP, Su P, et al. Persistence of High Health Care Costs among VA Patients. Health Serv Res 2018;53:3898-916.
29. Ta A, Goh HL, Tan CS, et al. Identifying prospective frequent readmitters for hospital to home intervention using machine learning. Int J Integr Care 2019;19:228.
30. Ministry of Health, Singapore. Speech by Mr Gan Kim Yong, Minister for Health, at the Community Health Assist Scheme (CHAS) Carnival, 27 July 2019. https://www.moh.gov.sg/newsroom/speech-by-mr-gan-kim-yong-minister-for-health-at-the-community-health-assist-scheme-(chas)-carnival-27-july-2019. Accessed 30 June 2025.
31. gov.sg. Master Plan 2019 Planning Area Boundary (No Sea). https://data.gov.sg/collections/2104/view. Accessed 1 May 2022.
32. Singapore Department of Statistics. Population Trends, 2024. https://www.singstat.gov.sg/publications/population/population-trends. Accessed 18 August 2025.
33. OneMap Singapore. Population Query. https://www.onemap.gov.sg/apidocs/populationquery. Accessed 15 May 2025.