Examining the high users of hospital resources: implications of a profile developed from Australian health insurance claims dataJoanna Khoo A B C , Helen Hasan A and Kathy Eagar A
A Australian Health Services Research Institute, Building 234 (iC Enterprise 1), Innovation Campus, University of Wollongong, Wollongong, NSW 2522, Australia. Email: email@example.com; firstname.lastname@example.org
B Capital Markets Cooperative Research Centre Health Market Quality Research Program, Capital Markets Cooperative Research Centre Ltd., Level 3, 55 Harrington Street, Sydney, NSW 2000, Australia.
C Corresponding author. Email: email@example.com
Australian Health Review - https://doi.org/10.1071/AH17046
Submitted: 24 February 2017 Accepted: 28 September 2017 Published online: 13 November 2017
Journal compilation © AHHA 2017 Open Access CC BY-NC-ND
Objective To develop and examine a profile of the demographic, hospital admission and clinical characteristics of high users of hospital resources within a cohort of privately insured Australians.
Methods Hospital admissions claims data from a group of private health insurance funds were analysed. The top 1% of hospital users were selected based on three measures of resource utilisation: number of admissions, total bed days and total insurance benefits paid. The demographic, hospital admission and clinical characteristics data were compared for these three measures of resource utilisation.
Results Compared with the general insured population, the three high-use cohorts are older, have more public hospital admissions and have more same-day admissions. The three high-use cohorts have the same top five principal diagnosis categories. These five categories account for more than two-thirds of admissions. The top 1% of users is responsible for a large proportion of total resource utilisation, accounting for 13% of total costs and 21% of total bed days.
Conclusions The highest users of hospital resources have a distinct profile, accounting for a large proportion of total resource utilisation for a narrow range of health conditions. The age and hospital admission profile of this group suggest both policy and service considerations for the targeting of interventions to support this high-needs group.
What is known about this topic? Statistics are regularly published on the uptake and use of private health insurance in Australia but there is little detailed information on resource utilisation in specific subgroups, particularly those with the highest levels of hospitalisation.
What does this paper add? This paper provides a profile of high resource utilisation among a privately insured cohort, describing demographic, hospital admission and clinical characteristics across three measures of resource utilisation. Patterns of use are detailed in this profile, for example the top 1% of users have a higher proportion of public hospital admissions as a private patient. The clinical profile of admissions was similar for the three measures of resource utilisation and there was considerable overlap in the individuals categorised in each high-use group.
What are the implications for practitioners? The narrow demographic and clinical profile of the high resource utilisation groups shows a chronic disease burden that is different to the focus of current chronic disease policy measures. The high-use conditions identified in this study are less amenable to preventive measures and new strategies may be required to target this high-needs group.
Additional keywords: chronic disease, diagnosis, need, private.
Health insurance arrangements in Australia are distinctive. Medicare is a universal, tax-financed insurance system that covers the whole population for certain health services. Private health insurance is voluntary but government policies subsidise the industry and encourage uptake, particularly among high-income earners. Private health insurance can be used in both private and public hospitals and both supplements and duplicates Medicare for hospital care.1,2 Initiatives of the Federal Government to reverse declining private health insurance membership in the late 1990s and early 2000s mean that people take out insurance for reasons unrelated to their health3 and, as a result, may have perverse motivations to have higher levels of health service use.4,5
There is a wide body of research comparing health services utilisation of those with and without private health insurance in Australia. Those with private health insurance have been found to have higher levels of health service utilisation than those without private health insurance, particularly in relation to surgical procedures.6,7 Studies have also looked at the usage of hospital services as a public or private patient for specific types of conditions, such as nervous, respiratory and circulatory conditions, although these studies have not focused on the use of private health insurance specifically.8,9
Although critiques of government policies that support the private health insurance industry are plentiful,10–13 there has been surprisingly little research on the patterns of hospital utilisation associated with private health insurance, and none that looks at the characteristics associated with the highest use of hospital resources among the insured population. This is an important area of investigation because following the introduction of the Private Health Insurance Act (2007), private health insurers have been able to offer a wider range of disease management and hospital substitution services with expected benefits of improved member health status and a reduction in hospitalisations and associated hospital claims.14
International research on disease management programs suggests that targeting those most at risk is an important implementation strategy to ensure that programs are most effective, from both a health outcomes and cost perspective.15,16 Although private health insurers in Australia generally do not have detailed health and medical information on their members, they do have information on hospital admissions, used for the payment of insurance claims. Given this, investigation into the patterns of hospital resource utilisation among the insured population could inform disease management and other support strategies by identifying who has the highest level of resource utilisation and for what conditions people are seeking care.
To create a profile of the highest users of hospital resources within the insured population, this study selected three top-1% samples using three different measures of resource utilisation: number of admissions; total bed days; and total benefits paid. These three measures have been used in a range of studies in both Australia and internationally to measure resource utilisation but are generally not used together.8,17–19
This study used de-identified insurance claims data from 13 Australian private health insurance funds. The dataset covered a period of more than 5 years, with the earliest hospital discharge date being 1 September 2009 and the latest date being 2 June 2015. The dataset contained 1 387 173 admissions relating to 405 428 individuals.
The dataset included items at the level of member, hospital admission and claims. All original identifiers were removed from the variables before researcher access and replaced with encrypted identifiers to protect the privacy of individuals and organisations in the dataset. The variables used in the analysis were:
Member information: member identifier, fund identifier, member year of birth, member sex
Admission administrative information: admission identifier, date of admission, date of separation, hospital type, and same-day status
Admission clinical information: diagnosis, code index (code index of 1 used to indicate principal diagnosis)
Claims information: benefits paid (used to derive total member benefits paid)
The level of insurance cover of the members was not included so the exact effect, although not expected to be material, of different types of insurance policies on the services claimed is not known.
From the full study population, three cohorts were identified as the top 1% of users based on resource utilisation defined either by total number of admissions, total bed days or total benefits paid. Table 1 provides further information on the method of calculation and selection criteria for each of the ‘high-use’ groups.
As 405 428 people had a hospital admission, a 1% sample is 4054 people. Due to the number of people that shared cut-off values for admissions and length of stay, the three cohort sizes vary slightly. The cohort sizes are included in Table 2.
A limitation of the dataset used for this analysis is the completeness of clinical information, particularly relating to public hospital admissions. Although the Hospital Casemix Protocol is the national standard for information provided from hospitals to both private health insurers and the Australian Government,20 public hospitals do not have to provide the complete set of Hospital Casemix Protocol data to insurers. Xie and colleagues also noted this issue in a recent study using data obtained directly from another Australian private health insurance fund.21 Although admission administrative information for public hospitals was complete, admission clinical information was only 48% complete. As a result, public hospital information was excluded from the analysis of clinical information.
After cohort selection, descriptive statistics were generated to profile the three high-use cohorts and the full study population. For demographic statistics, sex and age distribution of the cohorts were analysed. Hospital admissions were analysed by hospital type (public hospital, private hospital or private day hospital) and by same-day or overnight admission status.
Where the data were available, the principal diagnosis of each private hospital admission was investigated (principal diagnosis data were available for 95.9% of private admissions). The percentage of hospital admissions included in the clinical profile is detailed in Table 2.
The principal diagnosis for each admission was grouped according to the 21 International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM) chapters.22 For all cohorts, chapter 21 (‘Factors affecting health status and contact with health services’) had the highest proportion of admissions. Due to the heterogeneous nature of this ICD-10-AM chapter, specific ICD-10-AM codes for principal diagnosis for this chapter were investigated. It was found that three ICD-10-AM codes (Z509: care involving use of rehabilitation procedure, unspecified, Z511: pharmacotherapy session for neoplasm, and Z491: extracorporeal dialysis) made up the majority of chapter 21 items (more than 90% for the three high-use cohorts). Although these codes do not represent specific health conditions, there are circumstances in which the Australian coding standards require specific therapies or interventions to be coded as the principal diagnosis, rather than a health condition.23 The coding of these three items represents three such circumstances. These three codes were treated as their own diagnosis grouping in the clinical profile analysis due to the large proportion of admissions in which they were listed as the principal diagnosis.
The resource utilisation for each high-use cohort was then compared with that of the full study population. Finally, overlap in individual membership of the cohorts was investigated by identifying the common member identifiers that appeared in each high-use cohort. All statistical analysis was performed using R version 3.2.2 (The R Foundation, Vienna, Austria).
This study received ethics approval from the University of Wollongong and Illawarra Shoalhaven Local Health District Health and Medical Human Research Ethics Committee on 12 July 2016 (No: HE16/211).
Demographic and hospital admission profile of high-use cohorts
Table 2 summarises key demographic and hospital admission characteristics of each of the three high-use cohorts and the full study population. As may be expected, the high-use cohorts are older than the full study population with approximately double the proportion of individuals aged 65 years and over (high-use cohorts ranged from 60% – 72% of members aged 65 years and over compared with 32% for the full study population). The high-cost cohort has the highest proportion of individuals aged 65 years and over at 72%. There are a greater proportion of females in all high-use groups, apart from the high-cost group.
Each of the high-use cohorts has a higher proportion of admissions to public hospitals compared with the full study population (16% of all admissions). For the bed days and high-cost cohorts, public hospital admissions as a proportion of total hospital admissions represent about one-quarter of all admissions (27% and 24% respectively). The proportion of same-day admissions for the high-use cohorts is higher than the total study population. The highest proportion of same-day admissions is for the high-admission cohort, with 80.3% of their admissions being same-day.
Clinical profile of high-use cohorts
Figure 1 shows the top five principal diagnosis categories for hospital admissions to a private facility for each of the high-use groups and the total study population where data were available.
The same five categories account for the top principal diagnoses for admissions in all high use cohorts – mental health, dialysis, rehabilitation, pharmacotherapy and neoplasms. With the exception of mental health and dialysis, the same principal diagnosis categories are also represented in the top five categories for the total population dataset. Mental-health-related conditions account for the highest proportion of admissions in all high-use cohorts, but if the categories of pharmacotherapy and neoplasms are combined then cancer-related diagnoses are responsible for the highest proportion of admissions for the high-admission and high-cost cohorts.
Although there are 21 ICD-10-AM chapters, the top five principal diagnosis categories, from only three chapters, make up a remarkably high proportion of all admissions in the high-use cohorts – 67.8% for high-cost, 74.4% for high-admission and 78.1% for high-bed-days.
The contribution of high-use cohorts to overall resource utilisation
The high-use cohorts represent the top 1% of individuals using the most resources however measured, but they account for much more than 1% of total resource utilisation. The highest proportion overall is for bed days, with the high-bed-days cohort representing 21.2% of total bed days. The high-admission cohort represents 19.0% of total admissions and the high-cost cohort represents 13.3% of total costs.
The relationship between the three measures of high resource utilisation
Figure 2 depicts the relationship between each of the three high-use cohorts and the degree of overlapping membership of the three cohorts. The figure demonstrates a high degree of variation in resource utilisation even within the top 1% cohorts. The 13% of individuals that are represented in all three high-use cohorts have an average number of hospital admissions 10 times the number of those that are only represented in the high-cost or high-bed-days cohorts. Similarly, the average bed days of this subgroup are four times the number of those only represented in the high-admission and high-cost cohorts. There is greater overlap in membership of the high-cost and high-bed-days cohorts than the high-admission cohort.
This study examines the demographic, hospital admission and clinical characteristics associated with the highest levels of hospital resource utilisation among a privately insured cohort. Although many studies have examined differences in demographic and hospitalisation characteristics between those with and without private health insurance in Australia, the differences within the privately insured population has not been explored in detail. As might be expected, those with the highest levels of hospital resource utilisation are, on average, older than the general insured population. They also have a higher rate of public hospital admissions and a higher rate of same-day admissions.
The higher rate of public hospital admissions may be influenced by several factors including: a higher rate of admission via emergency department presentations (which are concentrated in public hospitals); previous selection of private health insurance in a public hospital making it more likely that the payment category is selected on a subsequent admission; or the location in which services are provided, such as rural and regional areas that have fewer private hospitals.
The profile reported in the present study complements national hospitalisation statistics that analyse the reason for admissions to public and private hospitals24 by focusing on the top 1% of hospital users with private health insurance and the conditions that users are seeking treatment for in the private sector. The findings of this study are relevant to policy reform of the private health insurance industry that is being actively considered by the Private Health Ministerial Advisory Committee, established in 2016 to provide the Federal Government with advice on private health insurance reforms.25
A notable finding is that the proportion of same-day admissions for each of the high-use cohorts is higher than the rate for the full study population. Although this in part reflects the frequency with which some services need to be delivered (for example, dialysis several times per week), it also reflects the high proportion of people requiring frequent, but usually time-limited, services, such as mental health programs, rehabilitation and chemotherapy. Although the clinical profile was only based on admissions to private hospitals, which have higher rates of surgical procedures,6,24 these top categories represent conditions requiring primarily non-surgical interventions.
Despite some differences in the demographic and admission characteristics of the three high-use cohorts, the same top five principal diagnosis categories were found for each group and in each case they account for more than two-thirds of hospital admissions. Few studies have compared the three measures of hospital resource utilisation used in the present study. Similar to the present study but for a different outcome of interest, a recent study by Cheng and colleagues also found that using different hospital resource utilisation measures – admissions numbers, bed days and costs – produced similar results in relation to public and private hospital utilisation for heart disease patients.8 The degree of cohort membership overlap between the different high-resource-utilisation cohorts may explain the similar results between resource utilisation measures.
The study results suggest that the highest users of hospital resources have a distinct profile. The most resources are being used for a narrow range of health conditions and use is highly concentrated within a small group of fund members accounting for a large proportion of total resource utilisation. These results are similar to an examination of a Victorian public health organisation in which it was found that 20% of costs are spent by 3% of the population.26
There is a strong focus within the Australian health sector at present on reducing potentially preventable hospital admissions and moving care for chronic conditions outside of hospitals to the community.27 Although the results of both insurer- and government-funded chronic disease management programs in Australia have been mixed,28–30 the findings of the present study indicate that the heaviest users of hospital services are seeking services for a narrow range of conditions that are not the target of traditional disease management programs. The major reasons for hospitalisation – including chemotherapy, dialysis and most rehabilitation – in the top 1% of privately insured hospital users are not preventable.
Furthermore, the majority of hospital admissions are on a same-day basis, which raises questions as to how, and if, admissions could be further reduced or made more efficient. Although only a narrow set of health conditions account for the majority of hospital admissions among high-resource-utilisation groups, these conditions relate to a diverse set of care needs, so different strategies will likely be required. This is similar to the conclusion drawn by Wodchis and colleagues in a recent study of high health costs in Canada.17
The present study used insurance claims data from a group of private health insurance funds so was not able to investigate hospital admissions for the cohort as a public patient or paid by another party, such as third-party compensation schemes. This is an area for future research. The private health insurance funds that contributed data to this study only represent ~10% of the private health insurance market, so it needs to be considered whether results are representative of the wider Australian private health insurance industry. However, the age profile of the study population is similar to publicly available statistics published by the regulator of the private health insurance industry.31
Another limitation of this study relates to the completeness of clinical information as outlined earlier. It is worth noting that the insurance claims data used for this study is the main source of data available to private health insurance funds for the purposes of understanding and targeting health interventions towards their members. The fact that a substantial proportion of this data has missing diagnosis and clinical information is a concern, particularly as many insurers seek to increase disease management services.
Through examination of the highest users of healthcare resources, this study found that private health insurance claims for hospital admissions are concentrated to a small number of users who are seeking care for a narrow range of conditions. In contrast to most hospital-admission-reduction strategies that focus on chronic diseases such as heart disease and diabetes, the high-use conditions identified in this study are not easily preventable through disease management interventions that focus on self-management. In many cases, treatments for these conditions already have a level of efficiency being provided as same-day admissions. Further research could build on the findings of this study by investigating the incidence of comorbidity of conditions among the top hospital users and examining specific conditions, such as the detailed patterns of service use for mental health patients, where there are differences in the classification of hospital episodes between public and private sectors.
The authors declare there are no competing interests.
The authors wish to acknowledge the financial support and the data analytics platform provided by the Capital Markets Cooperative Research Centre under the Health Market Quality Program and our industry partner HAMB Systems Ltd. The authors acknowledge Mr Nicolas Fenwick for assistance in the preparation of figures for this manuscript.
References Colombo F, Tapay N. Private health insurance in Australia: a case study. OECD Health Working Papers. No. 8. Paris: OECD Publishing; 2003. Available at: http://www.oecd-ilibrary.org/social-issues-migration-health/private-health-insurance-in-australia_478608584171 [verified 18 January 2017]
 Segal L. Why it is time to review the role of private health insurance in Australia? Aust Health Rev 2004; 27 3–15.
| Why it is time to review the role of private health insurance in Australia?CrossRef |
 Fiebig DG, Savage E, Viney R. Does the reason for buying health insurance influence behaviour? Sydney: Centre for Health Economics Research and Evaluation, University of Technology Sydney; 2006. Available at: https://www.uts.edu.au/sites/default/files/wp2006_1.pdf [verified 18 January 2017]
 O’Loughlin MA. Conflicting interests in private hospital care. Aust Health Rev 2002; 25 106–17.
| Conflicting interests in private hospital care.CrossRef |
 Savage E, Wright DJ. Moral hazard and adverse selection in Australian private hospitals: 1989–1990. J Health Econ 2003; 22 331–59.
| Moral hazard and adverse selection in Australian private hospitals: 1989–1990.CrossRef |
 Brameld K, Holman DA, Moorin R. Possession of health insurance in Australia – how does it affect hospital use and outcomes? J Health Serv Res Policy 2006; 11 94–100.
| Possession of health insurance in Australia – how does it affect hospital use and outcomes?CrossRef |
 Hindle D, McAuley I. The effects of increased private health insurance: a review of the evidence. Aust Health Rev 2004; 28 119–38.
| The effects of increased private health insurance: a review of the evidence.CrossRef |
 Cheng TC, Palangkaraya A, Yong J. Hospital utilization in mixed public–private system: evidence from Australian hospital data. Appl Econ 2014; 46 859–70.
| Hospital utilization in mixed public–private system: evidence from Australian hospital data.CrossRef |
 Gu M, Johar M. Profiling hospital utilization in a mixed public–private system. Appl Econ 2017; 49 361–75.
| Profiling hospital utilization in a mixed public–private system.CrossRef |
 Duckett SJ, Jackson TJ. The new health insurance rebate: an inefficient way of assisting public hospitals. Med J Aust 2000; 172 439–42.
| 1:STN:280:DC%2BD3czivVKhtg%3D%3D&md5=8d3b5c1d4c3d70f6d5be1d96fd9f801bCAS |
 Butler JR. Policy change and private health insurance: did the cheapest policy do the trick? Aust Health Rev 2002; 25 33–41.
| Policy change and private health insurance: did the cheapest policy do the trick?CrossRef |
 Thomas PE. Reflections on the role of less-than-comprehensive (exclusionary) private health insurance hospital products in the Australian healthcare system. Aust Health Rev 2012; 36 273–6.
| Reflections on the role of less-than-comprehensive (exclusionary) private health insurance hospital products in the Australian healthcare system.CrossRef |
 Eckermann S, Sheridan L, Ivers R. Which direction should Australian health system reform be heading? Aust N Z J Public Health 2016; 40 7–9.
| Which direction should Australian health system reform be heading?CrossRef |
 Biggs A. Chronic disease management: the role of private health insurance. Canberra: Department of Parliamentary Services; 2013. Available at: http://www.aph.gov.au/About_Parliament/Parliamentary_Departments/Parliamentary_Library/pubs/rp/rp1314/ChronDisease [verified 18 January 2017]
 Aljutaili M, Becker C, Witt S, Holle R, Leidl R, Block M, Brachmann J, Silber S, Bestehorn K, Stollenwer B. Should health insurers target prevention of cardiovascular disease? A cost-effectiveness analysis of an individualised programme in Germany based on routine data. BMC Health Serv Res 2014; 14 263
| Should health insurers target prevention of cardiovascular disease? A cost-effectiveness analysis of an individualised programme in Germany based on routine data.CrossRef |
 Russell LB. Preventing chronic disease: an important investment, but don’t count on cost savings. Health Aff (Millwood) 2009; 28 42–5.
| Preventing chronic disease: an important investment, but don’t count on cost savings.CrossRef |
 Wodchis WP, Austin PC, Henry DA. A 3-year study of high-cost users of health care. CMAJ 2016; 188 182–8.
| A 3-year study of high-cost users of health care.CrossRef |
 Roos N, Burchill C, Carriere K. Who are the high hospital users? A Canadian case study. J Health Serv Res Policy 2003; 8 5–10.
| Who are the high hospital users? A Canadian case study.CrossRef |
 Harris ML, Dolja-Gore X, Kendig H, Byles JE. End of life hospitalisations differ for older Australian women according to death trajectory: a longitudinal data linkage study. BMC Health Serv Res 2016; 16 484
| End of life hospitalisations differ for older Australian women according to death trajectory: a longitudinal data linkage study.CrossRef |
 Department of Health. Hospital Casemix Protocol. Canberra: Department of Health; 2016 Available at: http://www.health.gov.au/internet/main/publishing.nsf/Content/health-casemix-data-collections-about-HCP [verified 18 January 2017]
 Xie Y, Schreier G, Chang DCW, Neubauer S, Liu Y, Redmond SJ, Lovell NH. Predicting days in hospital using health insurance claims. IEEE J Biomed Health Inform 2015; 19 1224–33.
| Predicting days in hospital using health insurance claims.CrossRef |
 National Casemix and Classification Centre (NCCC). International Statistical Classification of Diseases and Related Health Problems, Australian Modification and Australian Refined Diagnosis Groups (Eighth Edition). Wollongong: NCCC, Australian Health Services Research Institute, University of Wollongong; 2013.
 National Casemix and Classification Centre (NCCC). Australian coding standards (Eighth Edition). Wollongong: NCCC, Australian Health Services Research Institute, University of Wollongong; 2013.
 Australian Institute of Health and Welfare (AIHW). Admitted patient care 2014–15: Australian hospital statistics. Health services series no. 68. Cat. no. HSE 172. Canberra: AIHW; 2016.
 Department of Health. Summary of the second meeting of the Private Health Ministerial Advisory Committee. Canberra: Department of Health; 2016. Available at: http://www.health.gov.au/internet/main/publishing.nsf/Content/phmac [verified 18 January 2017]
 Heslop L, Athan D, Gardner B, Diers D, Poh BC. An analysis of high-cost users at an Australian public health service organization. Health Serv Manage Res 2005; 18 232–43.
| An analysis of high-cost users at an Australian public health service organization.CrossRef |
 Australian Institute of Health and Welfare (AIHW). National Healthcare Agreement. Canberra: AIHW; 2017). Available at: http://meteor.aihw.gov.au/content/index.phtml/itemId/630028 [verified 18 January 2017]
 Morello RT, Barker AL, Watts JJ, Bohensky MA, Forbes AB, Stoelwinder J. A Telephone support program to reduce costs and hospital admissions for patients at risk of readmissions: lessons from an evaluation of a complex health intervention. Popul Health Manag 2016; 19 187–95.
| A Telephone support program to reduce costs and hospital admissions for patients at risk of readmissions: lessons from an evaluation of a complex health intervention.CrossRef |
 Billot L, Corcoran K, McDonald A, Powell-Davies G, Feyer A-M. impact evaluation of a system-wide chronic disease management program on health service utilisation: a propensity-matched cohort study. PLoS Med 2016; 13 e1002035
| impact evaluation of a system-wide chronic disease management program on health service utilisation: a propensity-matched cohort study.CrossRef |
 Hamar GB, Rula EY, Coberley C, Pope JE, Larkin S. Long-term impact of a chronic disease management program on hospital utilization and cost in an Australian population with heart disease or diabetes. BMC Health Serv Res 2015; 15 174
| Long-term impact of a chronic disease management program on hospital utilization and cost in an Australian population with heart disease or diabetes.CrossRef |
 Australian Prudential Regulation Authority (APRA) Private Health Insurance Membership and Benefits: September 2016. Sydney: APRA; 2016. Available at: http://www.apra.gov.au/PHI/Publications/Pages/Private-Health-Insurance-Membership-and-Benefits.aspx [verified 18 January 2017].