BMJ Open. For peer review only - Journal: BMJ Open. Manuscript ID: bmjopen

Transcription

1 Relationship between use of general practice and health care costs at the end of life: data linkage study in New South Wales, Australia Journal: BMJ Open Manuscript ID: bmjopen Article Type: Research Date Submitted by the Author: -Jul-0 Complete List of Authors: Tran, Bich Falster, Michael; University of New South Wales, Centre for Big Data Research in Health Girosi, Federico; Centre for Health Research, University of Western Sydney Jorm, Louisa; University of New South Wales, Centre for Big Data Research in Health <b>primary Subject Heading</b>: Health services research Secondary Subject Heading: Health services research Keywords: Health economics < HEALTH SERVICES ADMINISTRATION & MANAGEMENT, Organisation of health services < HEALTH SERVICES ADMINISTRATION & MANAGEMENT, PUBLIC HEALTH

2 Page of BMJ Open Relationship between use of general practice and health care costs at the end of life: data linkage study in New South Wales, Australia Bich Tran, Michael O Falster, Federico Girosi and Louisa Jorm Centre for Big Data Research in Health, UNSW Australia, Kensington, Australia; Centre for Health Research, School of Medicine, University of Western Sydney, Campbelltown, Australia Keywords: general practitioners (GP); health expenditures; end of life; and Up Study; linkage study. Manuscript word count: 0 Corresponding author: Bich Tran Ph.D., Centre for for Big Data Research in Health, Faculty of Medicine, UNSW Australia, Australia. Phone: + 00, bich.tran@unsw.edu.au.

3 Page of ABSTRACT Objective: This analysis investigated the relationships between healthcare expenditures in the last months of life and use of general practitioner (GP) services in the preceding -month period among older residents of New South Wales, Australia. Methods: Questionnaire data (00-00) for more than 0,000 people aged years and over were linked to individual hospital and death records (to December 0) and cost data. For, participants who died during follow-up, generalised linear mixed models were used to explore the relationships between costs of hospital, emergency department (ED) and Medicare-funded outpatient and pharmaceutical services in the last months of life and quintile of GP use in the prior months. Analyses were adjusted for age at death, sex, educational level, language, private health insurance, household income, self-reported health status, functional limitation, psychological distress, number of comorbidities and geographic clustering. Results: Almost % of decedents had at least one hospitalisation in the end of life and the mean (median) of total cost for each person in this period was AUD 0, (,). There was no significant difference in the hospital cost, including cost for preventable hospitalisations, across quintiles of GP use. Participants in the lowest quintile of GP use incurred more ED costs, but ED costs were similar across the other quintiles of GP use. End-of-life costs for Medicarefunded outpatient services and pharmaceuticals increased steeply according to quintile of GP use. Conclusion: In the Australian setting, there was no association between use of GP services in the lead up to end-of-life and hospital inpatient costs in the -month end-of-life period, but there was significant association with higher costs for outpatient services and pharmaceuticals. However, there was some indication that limited GP access might increase ED use at end-oflife.

4 Page of BMJ Open INTRODUCTION The costs of health care rise dramatically at the end of life, especially in developed countries., Most of these costs are spent caring for older people, such that in Australia almost % of total hospital expenditure is attributable to care for people aged years and over in their last year of life. Although the cost may vary amongst countries, the overall health care expenditures for these older people in the last year of life are estimated at about % of the total health budget. The high costs of health care at the end of life have focussed attention on how these costs might be contained, and better end-of-life care delivered, through provision of primary, community and palliative care services. However, little information exists about the potential to reduce end-of-life costs through better management in primary care in the lead up to end of life. A study of almost 0,000 deceased Medicare beneficiaries aged years and over in the United States found that greater numbers of visits to primary care physicians in the year prior to the -month end-of-life period were associated with lower total health care costs at the end of life, and with fewer preventable hospitalisations for congestive heart failure and chronic obstructive pulmonary disease. While the notion that increased use of primary care may lead to lower hospital cost is appealing, it is not known whether these findings hold in settings with greater access to publically-funded health care, such as Australia. Therefore, this study investigated the relationship between health care expenditures in the last six months of life and use of general practitioner (GP) services in the prior months, using data from a large cohort of older residents of New South Wales, Australia. METHODS Participants This analysis was nested within the Assessing Preventable Hospitalisation InDicators (APHID) study. APHID uses linked survey and administrative data for participants in the Sax Institute s and Up Study, a prospective cohort of,0 men and women aged years and over and residents in New South Wales (NSW), Australia. Participants entered the Study by completing a mailed self-administered questionnaire at study entry (between February 00 and April 00) and providing written consent for long-term follow-up and linkage of their health information to a range of routine health databases. People residing in non-urban areas and those aged 0 years and over were oversampled. The overall response rate for the and Up Study was estimated to be % and the study included about 0% of the NSW population aged and over. Ethics approval for the APHID study was obtained from the NSW Population and Health Services Research Ethics Committee, Aboriginal Health and Medical Research Council of NSW Ethics Committee, and the University of Western Sydney Ethics Committee. The conduct of the and Up Study was approved by the University of New South Wales Human Research Ethics Committee. Data sources

5 Page of NSW Registry of Births, Deaths and Marriages (RBDM) The NSW RBDM captures details of all deaths registered in NSW. In this analysis, we used death registrations for participants in the and Up Study who died up to December 0. Cause of death was not available at the time of this analysis. The NSW Admitted Patient Data Collection (APDC) The APDC is a routinely collected census of hospital separations (discharges, transfers and deaths) from all NSW public and private sector hospitals and day procedure centres. The APDC data used in this analysis related to all separations for and Up Study participants in the last months of life (hereafter referred to as end of life ). The NSW Emergency Department Data Collection (EDDC) The EDDC provides information about patient presentations to the emergency departments (ED) of urban and large regional public hospitals across New South Wales which cover approximately three quarters of all ED visits in the state. Data used in this analysis were those presentations for and Up Study participants in the end of life. Medicare Australia Medicare Australia is the country s universal health insurance scheme and administers claims for subsidised medical care including GP consultations under the Medical Benefits Schedule (MBS) and for pharmaceutical products under the Pharmaceutical Benefits Schedule (PBS). Only services attracting subsidy are included. Unreferred attendances by GPs or practice nurses (on behalf of a medical practitioner) in the months before end of life, including those occurring in residential aged care facilities (hereafter referred to as GP visits ), were identified according to the definition used by the National Health Performance Authority. Quintiles of GP visits were generated based on the distribution of GP visits for all participants in the months leading up to the end of life. All data from the MBS and PBS, except for claims of inhospital services in the MBS, were used to calculate expenditures for outpatient health care and pharmaceutical services in the end of life. MBS services were further categorised into Specialist services based on broad type of service item codes. 0 Data linkage Linkage of RBDM and APDC data was performed by the Centre for Health Record Linkage (CHeReL) ( using probabilistic record linkage methods and commercially available software (ChoiceMaker; ChoiceMaker Technologies Inc.). CHeReL quality assurance data show false positive and negative rates for data linkage of 0.% and less than 0.%, respectively. Linkage of Medicare Australia data was performed by the Sax Institute, using a unique identifying number that was provided to the Commonwealth Department of Human Services. Eligibility criteria and Up Study participants who died less than months after study entry were excluded from the analysis to ensure that baseline data reflected health prior to, rather than during, the end-of-life period. Moreover, participants who held a Department of Veterans Affairs (DVA)

6 Page of BMJ Open healthcare card were excluded, because Medicare data do not capture all services provided to these cardholders. Those having no claims to the Medicare system during the entire linkage (from June 00 to December 0) were also excluded. Variables Socio-demographic and health characteristics of participants were derived from the selfreported baseline questionnaire of the and Up Study (available at These included age, sex, educational level, language spoken at home, health insurance status and annual household income. Health characteristics collected included self-reported health status, level of functional limitation, level of psychological distress and number of comorbidities (heart disease, high blood pressure, stroke, diabetes, blood clot, asthma, Parkinson s disease and any cancer except skin cancer). Geographic area of residence was classified according to Statistical Local Areas (SLAs) defined using boundaries from the 00 Australian Census. Outcomes The outcomes examined, all for the last months of life, were (i) hospital costs for all inpatient services including hospital costs for preventable hospitalisations (classified using the ICD-0- AM diagnosis codes specified in the 0 Australian National Healthcare Agreement indicator Selected potentially preventable hospitalisation ) ; (ii) total ED costs; (iii) total MBS costs, including costs for GP and specialist consultations; (iv) total PBS costs; and (v) total costs, the sum of (i)-(iv). Hospital costs were assigned to each inpatient episode using the Australian Refined Diagnosis-Related Group (AR-DRG)-specific average cost reported in the National Hospital Cost Data Collection (NHCDC) for the specific year of admission. Cost for each ED presentation was estimated using visit type, triage category and separation mode. Costs for MBS and PBS were the cumulative amount of subsidy paid by Medicare. All expenditures were converted to 0 Australian dollars (AUD) using the consumer price index (CPI) for all groups. Statistical analysis Socio-demographic and health characteristics of decedents were summarised using simple descriptive statistics and compared across quintiles of GP use using Pearson s Chi-square statistics. Differences in the average of health service use including number of admissions, ED presentations, number of claims and bed days, number of MBS services and all PBS items in the end of life by quintile of GP visits was tested using the non-parametric Wilcoxon-Mann- Whitney test. Mean total health care costs, and separate for all hospitalisations including preventable hospitalisations, ED presentations, MBS and PBS, were estimated using generalised linear mixed models with a log link and gamma distribution. A sensitivity analysis was used to test if there was an association between quintile of GP visits and number of claims or MBS cost provided for each GP visit at the end of life. The covariates included in the models for estimating health care costs were age at death, sex, educational level, language other than English spoken at home, private health insurance, annual household income, self-reported health status, functional limitation, psychological distress and number of comorbidities. Participants were clustered within geographic areas using a random intercept across the Statistical Local Area of residence (N=).

7 Page of To investigate the effects of baseline health status on health care expenditures at the end of life, stratified analyses according to number of comorbidities (none,, and more than ) and self-reported health status (excellent/very good/good vs. fair/poor) were performed. Stata statistical software (version., StataCorp LP) with gllamm package was used to perform multivariate analyses of cost and quintiles of GP services utilisation; and SAS statistical software (version., SAS Institute Inc) for all other analyses. All analyses tests were -sided and P value <0.0 was considered statistically significant. RESULTS Characteristics of decedent participants The analysis included, participants in the and Up Study who died during follow-up, after excluding those who were holders of a DVA health care card (N=,), died less than months after study entry (N=,0) or had possibly inconsistent linkage (N=0). Average follow-up time from study entry to death for decedents in this cohort was approximately. years (range: to. years). The average age at study entry and death was years (range: -0) and years (range: -0 years), respectively. Women comprised 0% of total decedents. Decedents had on average. GP visits (median, range 0-) in the months prior to the last months of life. In this period, % of decedents had no GP visit, % had - visits, % had - visits, % had -0 visits and 0% had more than 0 visits. Decedents in the upper quintile of GP visits were significantly older, were more likely to report poor health status and had higher levels of functional limitation at baseline (Table ) compared to decedents in the bottom quintile. There were no associations between other socio-demographic or health characteristics and different groups of GP use (Table ). In this decedent cohort, there were a total of,00 hospital admissions for any cause in the last months of life, of which, (%) were classified as preventable hospitalisations. On average, each decedent had. hospital admissions (median ) during the end of life period and 0. preventable hospitalisations (median 0) (Table ). Almost % of decedents (N=,) had at least one hospitalisation at the end of life and % (N=) had at least one preventable hospitalisation, while approximately % (N=,) had at least one presentation to ED. There was a significant increase in the average number of hospital admissions and bed days in the upper quintiles of GP use compared to the lowest for preventable hospitalisations (Table ). In contrast, the average number of ED presentations was higher in the lowest quintile of GP use compared to the other quintiles. There was a strong and positive association between numbers of MBS and PBS claims at end of life and quintile of GP visits in the previous months before end of life (Table ). The positive association was also observed even if we took into account number of claims by number of GP visits in order to capture the intensity of services provided for each visit (data not shown).

8 Page of BMJ Open Health care costs Expenditures were right skewed with the mean (median) for the total health care cost in the end of life being AUD 0, (,). Mean (median) costs for all hospitalisations, preventable hospitalisations, ED presentation, MBS and PBS at the end of life for each decedent were AUD, (,0), AUD, (0), AUD (), AUD () and AUD (0), respectively. The average cost for GP services per person was twice as much as for specialist services (Table ). On average, expenditure for hospital services accounted for % of total end-of-life health care costs. There was no significant trend in the unadjusted or adjusted hospital cost (p trends=0. or 0., respectively) and cost for preventable hospitalisations (p trends=0. or 0.0, respectively) at the end of life across quintiles of GP visits (Table and Figure ). However, there was a significant inverse trend between increasing GP visits and adjusted ED cost (p trend = 0.0), driven by greater use of ED services in the lowest quintile of GP use compared with the other quintiles (Table and Figure ). There was a steep increase in MBS and PBS costs according to number of GP visits (P trends <0.00). Separate analyses for MBS costs for GP consultations, specialist consultations and all other claims showed similar patterns (Table ). The association was also observed for MBS cost per each GP visit. Overall, there was a significant positive association between quintile of GP visits and total health care cost, with this result driven by the costs for outpatient services and pharmaceuticals rather than hospital costs. Stratified analyses according to number of comorbidities and self-reported health status showed no differences between strata or compared to the un-stratified results (Supplementary Table ). DISCUSSION This analysis showed that greater use of GP services in the year leading up to the final months of life (end-of-life) was associated with greater total healthcare costs at the end of life, with this result driven by the costs for outpatient services and pharmaceuticals. There was no association between GP use leading up to end of life and hospital inpatient costs at the end of life. These findings contrast starkly with results from the only previous study, to our knowledge, that has directly addressed this issue, in a large sample of US Medicare beneficiaries, which found an inverse relationship between GP use and total costs. A striking difference between the two studies was probably due to the different patterns of GP use: in this study, % of decedents had at least GP visits or more in a year prior to end of life, compared with only % of the US decedents. This is however consistent with the overall higher number of annual GP visits per capita in Australia (.) versus in the US (.). Thus it is plausible that differences in access to primary care between Australia and the US may explain the discrepant results.

9 Page of The potential for preventive care to avert end-of-life hospital costs would be expected to be greatest for those admissions that are considered to be preventable through primary care. About % of the end-of-life admissions in our analysis met the current Australian definition for a preventable hospitalisation, as used nationally to monitor primary care performance. However, the absence of a relationship between quintile of GP visits and end-of-life costs, either for all hospital costs or preventable hospitalisations, suggests that if GPs have a role in preventing these hospitalisations, it is likely to lie much earlier in life and in the causal pathways for these conditions, consistent with findings regarding the key roles of patient socio-demographic factors and adverse health-related behaviours. 0, Preventable hospitalisation at the end-of-life may in fact represent appropriate inpatient care for those with deteriorating health and at high risk of mortality. On the other hand, hospitalisation at the end-of-life for these conditions could also reflect the shortage of home- and communitybased palliative services. Consideration should therefore be given to applying an upper age limit, or excluding admissions occurring in the end-of-life period, when using preventable hospitalisations as a performance measure for primary care. There was no existing study of the relationship between GP visits and ED cost at the end of life with which to compare with our results. We could find only one study for cancer patients at the end of life suggesting that a greater level of continuity of care with their primary care provider was associated with fewer presentations to ED and thus lower cost. We found that decedents who used no GP services in the lead up to end of life had more ED visits at the end of life. These decedents, who were older on average than others, may have included a greater proportion of residents of residential aged care, for whom there is growing concern that current delivery of GP services is not optimal. Greater utilisation of GP services in the lead up to end-of-life was associated with higher use of pharmaceutical and outpatient services, including GP and specialist visits, in the last months of life. This association is consistent with patterns of healthcare expenditure in the general population, and may reflect the continued engagement of patients with out-of-hospital care. Only % of decedents in this cohort visited their GP in the end-of-life period, compared to % who had at least one ED presentation and % who had at least one hospital admission. The greater use of hospital services than primary health care by decedents is likely to reflect the care needs of people who are terminally ill. This study had several strengths. It used detailed questionnaire data from a large prospective cohort study, with comprehensive ascertainment of health care utilisation and costs from administrative databases, eliminating the potential for recall bias. While previous Australian studies of end-of-life health care costs have reported only hospital costs,,, this study captured inpatient, ED, outpatient services and pharmaceuticals costs. Nonetheless, only Medicare-subsidised outpatient expenditures were included, and costs for non-admitted community based services, such as home-based palliative care, or dispensing of non-subsided pharmaceuticals, were not captured. This might reduce the generalisability of the findings to other populations with different health care systems. In addition, the incompleteness of data

10 Page of BMJ Open collection for presentations to all EDs in NSW may have resulted in underestimation of ED costs. Moreover, measures of comorbidities and health status used baseline questionnaire data, so did not reflect incident conditions or deteriorating health during follow-up. This study found that in the Australian setting, greater use of GP services in the lead up to endof-life had no impact on hospital costs in the -month end-of-life period, but was associated with higher costs for outpatient services, pharmaceuticals and overall total health care cost. There was some indication that limited GP access might increase ED use at end of life. The findings do not preclude a key role for GP care in containing end-of-life costs, for example through discussion of end-of life treatment preferences and advance care planning, delivering care in residential aged care settings and participating in home-based palliative care. Acknowledgements The APHID investigator team comprises Louisa Jorm, Alastair Leyland, Fiona Blyth, Robert Elliot, Kirsty Douglas, Sally Redman, Marjon van der Pol, Michael Falster, Bich Tran, Neville Board, Danielle Butler, Douglas Lincoln, Sanja Lujic, Damilola Olajide, Deborah Randall, Kim Sutherland and Diane Watson. This research was completed using data collected through the and Up Study ( The and Up Study is managed by the Sax Institute in collaboration with major partner Cancer Council NSW; and partners: the National Heart Foundation of Australia (NSW Division); NSW Ministry of Health; beyondblue; NSW Government Family & Community Services Carers, Ageing and Disability Inclusion; and the Australian Red Cross Blood Service. We thank the many thousands of people participating in the and Up Study Competing interests The authors have declared that no competing interests exist. Funding Source The study was funded by a National Health and Medical Research Council Partnership Project Grant (#0) and by partner agencies the Australian Commission on Safety and Quality in Health Care, the Agency for Clinical Innovation and the NSW Bureau of Health Information. The funders play no role in approving the publications. Data sharing statement The APHID study dataset has been constructed with the permission of each of the custodians of the respective source datasets and with specific ethical approval. The dataset could potentially be made available to other researchers if they obtain the necessary approvals. More information about these approvals is available from the authors on request. Contributors BT conducted the data analyses and drafted the manuscript. MF extracted data for analysis, assisted with statistical methods and data presentation, reviewed and edited the manuscript.

11 Page 0 of FG advised on the statistical methods, assisted with data interpretation and edited the manuscript. LJ conceived, designed and managed the study, obtained funding, reviewed and edited the manuscript and provided overall supervision. BT, MF and LJ have full access to the data. All authors reviewed and approved the final manuscript. Key messages In the Australian health care system, greater visits to general practitioner (GP) in the lead up to end-of-life had no impact on hospital inpatient costs in the -month end-of-life period, but there was significant association with higher costs for outpatients services and pharmaceuticals. There was some evidence that limited GP access might be associated with higher visit to emergency department. Strengths and limitations of the study This is the large-scale study to explore the longitudinal relationship between use of primary services and health care expenditures at the end of life, using detailed individual-level information about potential confounders and health service use Limitation include the use of administrative claims data containing only limited information about quality of primary care services and cause of death. The use of self-reported data for covariates may also introduce some bias. REFERENCES. Lubitz JD, Riley GF. Trends in Medicare payments in the last year of life. The New England journal of medicine ;():0-. Stooker T, van Acht JW, van Barneveld EM, et al. Costs in the last year of life in The Netherlands. Inquiry : a journal of medical care organization, provision and financing 00;():-0. Kardamanidis K, Lim K, Da Cunha C, et al. Hospital costs of older people in New South Wales in the last year of life. Med J Aust 00;():-. Swerissen H, Duckett S. Dying well. Melbourne: The Grattan Institute. Available at 0. Rosenwax LK, McNamara BA, Murray K, et al. Hospital and emergency department use in the last year of life: a baseline for future modifications to end-of-life care. Med J Aust 0;():0-. Kronman AC, Ash AS, Freund KM, et al. Can primary care visits reduce hospital utilization among Medicare beneficiaries at the end of life? Journal of general internal medicine 00;():0-. Jorm LR, Leyland AH, Blyth FM, et al. Assessing Preventable Hospitalisation InDicators (APHID): protocol for a data-linkage study using cohort study and administrative data. BMJ open 0;(). Banks E, Redman S, Jorm L, et al. Cohort profile: the and up study. Int J Epidemiol 00;():-. METeOR. Sevice event-general Practitioner (GP) service, total number NN[N]. METeOR identifier. National Health Performance Authority, Standard //0. Available at 0. Medicare Australia Statistics. Appendix : Mapping of Medicare items to Broad Type of Service. Australian Goverment. Department of Human Services. Last updated March 0. Available at 0

12 Page of BMJ Open os_map&start_dt=0&end_dt=0.. Stewart A, Kamberg CJ. Physical functioning measures. In: Steward A, Ware J, eds. Measuring Functioning and Well-Being: the Mecial Outcomes Study Approach: Durham (NC): Duke University Press.,.. Kessler R, Mroczek D. Final version of our Non-Specific Psychological Distress Scale [memo dated /0/]. Ann Arbor (MI): Survey Research Center of the Institute for Social Research: University of Michigan.. Trewin D. Statistical Geography Volume - Australian Standard Geographical Classification (ASGC). Canberra: Australian Bureau of Statistics, ABS Catalogue No..0, 00.. National Healthcare Agreement: PI -Selected potentially preventable hospitalisations, 0. Australian Institute of Health and Welfare. Available at: National Hospital Cost Data Collection. The Department of Health. Australia Government. URL Costs of Care Standards 00/0. NSW Department of Health. Document Number GL0_00. File Number H/. Available at Australian Bureau of Statistics, June 0. ABS 0.0 Tables &, series AC. Available at: Rabe-Hesketh S, Skrondal A, Pickles A. Maximum likelihood estimation of limited and discrete dependent variable models with nested random effects. Journal of Econometrics 00;():0-. Thomson S, Osborn R, Squires D, et al. International Profiles of Health Care Systems. The Commonwealth Fund Tran B, Falster MO, Douglas K, et al. Health behaviours and potentially preventable hospitalisation: a prospective study of older Australian adults. PloS one 0;():e. Falster MO, Jorm LR, Douglas KA, et al. Sociodemographic and Health Characteristics, Rather Than Primary Care Supply, are Major Drivers of Geographic Variation in Preventable Hospitalizations in Australia. Med Care 0;():-. Katteri R, Anikeeva O, Butler C, et al. Potentially avoidable hospitalisations in Australia: Causes for hospitalisations and primary health care interventions. PHC RIS Policy Issue Review Adelaide: Primary Health Care Research & Information Service 0. Rhee JJ, Zwar N, Vagholkar S, et al. Attitudes and barriers to involvement in palliative care by Australian urban general practitioners. Journal of palliative medicine 00;():0-. Burge F, Lawson B, Johnston G. Family physician continuity of care and emergency department use in end-of-life cancer care. Med Care 00;():-00. Hillen JB, Reed RL, Woodman RJ, et al. Hospital admissions from residential aged care facilities to a major public hospital in South Australia (-00). Australasian journal on ageing 0;0():0-. National Health Performance Authority 0, Healthy Communities: Frequent GP attenders and their use of health services in 0.. Moorin RE, Holman CD. The cost of in-patient care in Western Australia in the last years of life: a population-based data linkage study. Health policy 00;():0-0. Calver J, Bulsara M, Boldy D. In-patient hospital use in the last years of life: a Western Australian population-based study. Aust N Z J Public Health 00;0():-

13 Page of Table : Characteristics of decedents by quintile of GP visits in the -month period preceding the last months of life. Total decedent cohort Quintile of GP visits N (.) (.) 0 (.) (.) 0 (0.) Age at death (years) - (.) (.) (.) 0 (.) (.) (.) - (.) (.0) (.) (.) 0 (.) (.) - (.) (.) (.) 0 (.) (.) (.) - (.) 0 (.) (.) (.) (.) 0 (.) + (.) (0.) (.) 00 (.) 0 (.) 00 (.) P value ref <0.00 <0.00 <0.00 <0.00 Sex Male (.) (.) (0.) (0.0) (0.) (0.) Female (0.) 0 (.) 0 (.) (0.0) (.) 0 (.) P value ref Language English 0 (.) (.) (.) 0 (.) 0 (.) (.) Other (0.) (0.) (0.) (0.) (0.) 0 (0.) Missing (0.0) (0.0) (0.0) 0 (0.0) 0 (0.0) 0 (0.0) P value ref Education Did not complete high school (.) (.) (.) 0 (.) 0 (.) 0 (.) High school/apprenticeship (.) (.) (.) (.) (.0) (.) University or higher (.) (.) (.0) (.) (.) 0 (.) Missing (.0) (.) (.) (.) 0 (.0) (.) P value ref Remoteness Major cities (.) 0 (0.) 0 (.) (.) (.) (.) Inner regional 0 (.) 0 (.) 0 (.) 000 (.) 0 (.) (.) Outer regional (.) (.) (.0) (.) 0 (.) (.) Remote/Very remote (.) (.) (.) (.) (.) (.) Missing (0.0) (0.0) 0 (0.0) (0.) 0 (0.0) (0.0) P value ref Private health insurance Private extras (.) 0 (.) (.) 0 (.) (.) 0 (.0) Private no extras 0 (.) (.) (.) (.) (.) (.) Health care card 0 (.) 0 (.) (.) 0 (.) (.) 0 (.) None (.) (.) (.0) (.) 0 (.) (.) P value ref Household annual income <$0,000 (0.) (.) (0.) (0.) (0.) (.) $0,000-$, (.) (.) (.) (.) (.) (.) $0,000-$, 0 (0.) 00 (.) (0.) (.0) (0.) 0 (0.) $0,000-$, (.) (.) (.) (.) (.) 0 (.) $0,000 or more (.) 0 (.) (.) (.) (.) 0 (.) Prefer not to answer 0 (.) 0 (.) (.) (.) (.) 0 (.) Missing (.0) (0.) (0.) (0.) (.) 0 (.) P value ref Self-reported health status Excellent 0 (.) (.0) (.) (.0) (.0) (.) Very good (.) (0.) (0.) 0 (.) (0.0) (.) Good (.) (.) (.) 0 (.0) 00 (.) 0 (.)

14 Page of BMJ Open Fair (.) (.0) (.0) (.) (.) 0 (.) Poor (.0) (.) (.) (.) (.) (.) Missing (.) (.) 0 (.0) (.) 0 (.) 0 (.) P value ref 0.0 <0.00 <0.00 <0.00 Number of comorbidities None (.) 0 (0.) 00 (.) (.) (.) (.0) (.) 0 (.) 00 (.) 0 (.) 0 (.) 0 (.) (.) (.) (.) (.) (.) (.) or more (0.) (.) (0.) (0.) (.) 0 (0.) P value ref Functional limitation No limitation 0 (.) (.) 0 (.) (.) (.) (.) Minor limitation (.) (.) (.) (.) (.) (.) Mild limitation (.) 0 (.) (.) (.) (.) (0.) Moderate limitation 0 (.) (0.) (.) (.) (.) 0 (0.) Severe limitation (.) 0 (.) (.) (.) (.) (.) Missing 0 (.) 0 (.0) (.) (.) (.) (.) P value ref <0.00 Psychological distress Low distress 0 (.) (.) (.) 0 (0.0) (0.) (0.) Moderate distress 0 (.) (.) (.) (.) (.0) (.) High distress (.) (.) (.) (.) 0 (.) (.) Very high distress (.) (.) (.) (.) (.) (.) Missing (.) (.) (.) 0 (.) (.) (.) P value ref Numbers in parenthesis represent the proportion of decedent in the specified quintile. P values were estimated excluding missing records.

15 Page of Table. Average utilisation of health care services in the last months of life, by quintile of GP visits in the preceding months Average utilisation of health care service for the decedent cohort Average utilisation of health care service by quintile of GP visit, mean (median) Hospital bed days All hospital admissions.0 (). (). (). ().0 (). () P value ref Preventable hospitalisations. (0). (0). (0).0 (0). (0). (0) P value ref Number of hospital admissions All hospital admissions. (). (). (). (). (). () P value ref <0.00 Preventable hospitalisations 0. (0) 0. (0) 0. (0) 0. (0) 0. (0) 0. (0) P value ref Number of ED presentations. (). (). (). (). (). () P value ref Number of MBS claims # All claims. () 0. (0). (0). ().00 (). () P value ref <0.00 <0.00 <0.00 <0.00 # GP. () 0. (0) 0. (0). ().0 (). () P value ref <0.00 <0.00 <0.00 <0.00 # Specialist 0. (0) 0.0 (0) 0. (0) 0. (0) 0.0 (0). () P value ref <0.00 <0.00 <0.00 <0.00 # All other.0 (0) 0.0 (0). (0). (). (). () P value ref <0.00 <0.00 <0.00 <0.00 Number of PBS claims. (0) 0. (0). (0). (0).0 (). () P value ref <0.00 <0.00 <0.00 <0.00

16 Page of BMJ Open Table. Mean health care expenditure ($AUD) and its % confidence interval (CI) in the last months of life, by quintile of GP visits in the preceding months Mean health care expenditure (% CI) for the decedent cohort Mean health care expenditure (% CI) by quintile of GP visits P trend Total health care costs Unadjusted 0 (0-0) (-0) (0-0) 0 (0-0) 00 (-) (0-) <0.00 Adjusted 0 (0-0) (-) (-) 0 (0-0) 0 (0-00) 0 (-) <0.00 All hospital admissions Unadjusted (-) (-0) (-) (0-) (0-) (-) 0. Adjusted (-) 0 (-) (0-0) (-) (-00) (0-) 0. Preventable hospitalisation Unadjusted (0-) (-) (-) 0 (-0) (-) 0 (-) 0. Adjusted 0 (-) 0 (-) (-) 0 (0-) 0 (0-) 0 (0-) 0.0 All ED presentations Unadjusted (-) 0 (-0) (-) (-) (-) (-00) 0.0 Adjusted (0-) 0 (0-0) (-) (-) (-) (-) 0.0 All MBS claims Unadjusted (0-). (.-.) (-) (-) (-) (-0) <0.00 Adjusted (-). (.-.) (-) (-0) (-) 00 (0-0) <0.00 GP MBS claims Unadjusted (-). (.-.). (.-.). (.0-.) (-) (-) <0.00 Adjusted (-). (.-.). (.-.). (.-.) (-) (-) <0.00 Specialist MBS claims Unadjusted. (.-.0). (.-.). (0.-0.). (.-.). (.-0.) (-) <0.00 Adjusted. (.-.0). (.-.). (.-.). (.-.). (.-.) (-) <0.00 All other MBS claims Unadjusted (-). (.-.). (.-.) (0-) (-) (0-0) <0.00 Adjusted (-). (.-.). (.-00.) (-) 00 (-0) (0-) <0.00 All PBS claims Unadjusted (0-). (.-0.0). (.-0.) (-) (-) (-) <0.00 Adjusted (-). (.-.). (.0-.) 0 (-0) (-) (-) <0.00 Values were adjusted mean of costs either unadjusted or adjusted for age at death (0-year groups), sex, language, education, private health insurance, number of comorbidities, self-reported health status, functional limitation, psychological distress and random effect of statistical local area. P trends were estimated for each comparison using the lowest quintile as a reference group.

17 Page of Figure legends Figure. Adjusted* mean health care expenditures in the last months of life, by quintile of GP visits in the preceding month. *Adjusted for age at death (0-year groups), sex, language, education, private health insurance, number of comorbidities, self-reported health status, functional limitation, psychological distress and random effect of statistical local area.

18 Page of BMJ Open x0mm ( x DPI)

19 Page of Supplementary table : Mean health care expenditure ($AUD) and its % confidence interval (CI) in the last months of life, by quintile of GP visits in the preceding months, stratified by number of comorbidities and self-reported health status at study entry Mean health care Mean health care expenditure (% CI) by quintile of GP visits P trend expenditure (% CI) for the decedent cohort Comorbidity=0 Total cost (-0) (0-) (-) 00 (-0) 0 (0-0) 0 (0-) 0.00 Hospital cost (-) (-0) (00-) (-) (0-) (-0) 0.0 Preventable hospitalisation (-) 0 (-0) 00 (0-) 0 (-) 0 (0-) (-) 0. ED cost (-) 0 (-) (-) 0 (-0) 0 (-) (-) 0. MBS cost (-). (.-.) (-) (-) (-0) (-) <0.00 PBS cost Comorbidity= Total cost 00 (-00) (0-) 0 (-) (-) (0-) 0 (-) 0.00 Hospital cost (-) (0-0) 0 (-0) (000-) (-00) (-) 0. Preventable hospitalisation (0-) (-) 0 (-) (-) (-) (-0) 0. ED cost (-) 0 (-) (-) 0 (0-) (-) (-) 0.0 MBS cost (-). (.-.) (-) (-) (-) 0 (0-0) <0.00 PBS cost Comorbidity= Total cost 0 (0-0) 0 (0-0) 0 (00-0) (-00) (0-) (-) 0. Hospital cost (0-) (-) 0 (0-) 0 (0-0) 0 (-) (-) 0. Preventable hospitalisation (-) (0-) (-) (-) 0 (-) (-) 0. ED cost (-) 0 (00-0) (0-) (-) (-) (-) 0.0 MBS cost (-) 0. (0.-.) (-) (-) 0 (-) 00 (0-0) <0.00 PBS cost Comorbidity=+ Total cost (-) 0 (0-) (-) (-0) (-00) (0-) 0.0 Hospital cost 0 (00-0) 00 (00-0) 0 (0-0) 0 (00-0) (-) 0 (00-0) 0. Preventable hospitalisation (-) (-) (0-) 0 (0-) (-0) (-) 0.0 ED cost 0 (0-0) 00 (0-0) 0 (0-00) 0 (0-0) (-) 0 (0-00) 0. MBS cost (0-). (.-.) (-) (-) (-) 00 (0-0) <0.00 PBS cost (-). (.-.) (-) (-) (-) (0-) - Self-reported health = excellent/very good/good Total cost 0 (0-0) (0-) (-) 00 (-00) 0 (000-0) (0-) 0.00 Hospital cost 0 (-) (0-) (0-) (-) (0-) 0 (-) 0.

20 Page of BMJ Open Preventable hospitalisation (-) (0-0) (0-0) 0 (-0) (0-) (-) 0. ED cost (-) 0 (00-0) (0-) (-) (-) (-) 0.0 MBS cost (-). (.-.) (-) (-) (-) 0 (0-0) <0.00 PBS cost (0-). (.0-.) (-0) (-) (-) (0-) <0.00 Self-reported health=fair/poor Total cost 00 (0-0) 0 (-) 00 (00-0) 0 (00-0) (0-0) (0-0) 0.0 Hospital cost 0 (-0) (-) (-) (-0) (-) (-) 0. Preventable hospitalisation 0 (-) 0 (-) (0-) (-0) (0-) (-) 0. ED cost 00 (00-0) 0 (0-0) 0 (0-0) 0 (-) (-) 0 (00-0) 0.0 MBS cost (0-). (.-.) (-0) (-) (-) (-) <0.00 PBS cost indicates the convergence did not achieved for estimates of cost due to possible missing in any random geographic area(s).

21 Page 0 of STROBE Statement checklist of items that should be included in reports of observational studies Item No Recommendation Title and abstract (a) Indicate the study s design with a commonly used term in the title or the abstract Introduction (b) Provide in the abstract an informative and balanced summary of what was done and what was found Background/rationale Explain the scientific background and rationale for the investigation being reported Objectives State specific objectives, including any prespecified hypotheses Methods Study design Present key elements of study design early in the paper Setting Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection Participants (a) Cohort study Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up Case-control study Give the eligibility criteria, and the sources and methods of case ascertainment and control selection. Give the rationale for the choice of cases and controls Cross-sectional study Give the eligibility criteria, and the sources and methods of selection of participants (b) Cohort study For matched studies, give matching criteria and number of exposed and unexposed Case-control study For matched studies, give matching criteria and the number of controls per case Variables Clearly define all outcomes, exposures, predictors, potential confounders, and effect Data sources/ measurement modifiers. Give diagnostic criteria, if applicable * For each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group Bias Describe any efforts to address potential sources of bias Study size 0 Explain how the study size was arrived at Quantitative variables Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen and why Statistical methods (a) Describe all statistical methods, including those used to control for confounding Continued on next page (b) Describe any methods used to examine subgroups and interactions (c) Explain how missing data were addressed (d) Cohort study If applicable, explain how loss to follow-up was addressed NA Case-control study If applicable, explain how matching of cases and controls was addressed Cross-sectional study If applicable, describe analytical methods taking account of sampling strategy (e) Describe any sensitivity analyses

22 Page of BMJ Open Results Participants * (a) Report numbers of individuals at each stage of study eg numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed (b) Give reasons for non-participation at each stage NA (c) Consider use of a flow diagram NA Descriptive data * (a) Give characteristics of study participants (eg demographic, clinical, social) and information on exposures and potential confounders (b) Indicate number of participants with missing data for each variable of interest (c) Cohort study Summarise follow-up time (eg, average and total amount) Outcome data * Cohort study Report numbers of outcome events or summary measures over time Case-control study Report numbers in each exposure category, or summary measures of exposure Cross-sectional study Report numbers of outcome events or summary measures Main results (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (eg, % confidence interval). Make clear which confounders were adjusted for and why they were included (b) Report category boundaries when continuous variables were categorized (c) If relevant, consider translating estimates of relative risk into absolute risk for a meaningful time period NR Other analyses Report other analyses done eg analyses of subgroups and interactions, and sensitivity analyses Discussion Key results Summarise key results with reference to study objectives Limitations Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias Interpretation 0 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence Generalisability Discuss the generalisability (external validity) of the study results Other information Funding Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based *Give information separately for cases and controls in case-control studies and, if applicable, for exposed and unexposed groups in cohort and cross-sectional studies. Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely available on the Web sites of PLoS Medicine at Annals of Internal Medicine at and Epidemiology at Information on the STROBE Initiative is available at

23 Relationship between use of general practice and health care costs at the end of life: data linkage study in New South Wales, Australia Journal: BMJ Open Manuscript ID bmjopen r Article Type: Research Date Submitted by the Author: 0-Oct-0 Complete List of Authors: Tran, Bich Falster, Michael; University of New South Wales, Centre for Big Data Research in Health Girosi, Federico; Centre for Health Research, University of Western Sydney Jorm, Louisa; University of New South Wales, Centre for Big Data Research in Health <b>primary Subject Heading</b>: Secondary Subject Heading: Keywords: Health services research Health services research, General practice / Family practice, Health economics Health economics < HEALTH SERVICES ADMINISTRATION & MANAGEMENT, Adult palliative care < PALLIATIVE CARE, PRIMARY CARE

24 Page of BMJ Open Relationship between use of general practice and health care costs at the end of life: data linkage study in New South Wales, Australia Bich Tran, Michael O Falster, Federico Girosi and Louisa Jorm Centre for Big Data Research in Health, UNSW Australia, Kensington, Australia; Centre for Health Research, School of Medicine, University of Western Sydney, Campbelltown, Australia Keywords: general practitioners (GP); health expenditures; end of life; and Up Study; linkage study. Abstract word count: 00 Manuscript word count: 0 Corresponding author: Bich Tran Ph.D., Centre for for Big Data Research in Health, Faculty of Medicine, UNSW Australia, Australia. Phone: + 00, bich.tran@unsw.edu.au.

25 Page of ABSTRACT Objective: This analysis investigated the relationships between healthcare expenditures in the last months of life and use of general practitioner (GP) services in the preceding -month period among older residents of New South Wales, Australia. Methods: Questionnaire data (00-00) for more than 0,000 people aged years and over were linked to individual hospital and death records and cost data. For, participants who died during follow-up, generalised linear mixed models were used to explore the relationships between costs of hospital, emergency department (ED) and Medicare-funded outpatient and pharmaceutical services in the last months of life and quintile of GP use in the - months before death. Analyses were adjusted for age at death, sex, educational level, language, private health insurance, household income, self-reported health status, functional limitation, psychological distress, number of comorbidities and geographic clustering. Results: Almost % of decedents had at least one hospitalisation in the last months and the mean (median) of total cost for each person in this period was AUD 0, (,). There was no significant difference in the hospital cost, including cost for preventable hospitalisations in the last months of life, across quintiles of GP use in the - months before death. Participants in the lowest quintile of GP use incurred more ED costs, but ED costs were similar across the other quintiles of GP use. Costs for Medicare-funded outpatient services and pharmaceuticals increased steeply according to quintile of GP use. Conclusion: In the Australian setting, there was no association between use of GP services in the - months before death and hospital costs in the last months, but there was significant association with higher costs for outpatient services and pharmaceuticals. However, there was some indication that limited GP access might be associated with increased ED use at end-oflife.

26 Page of BMJ Open INTRODUCTION The costs of health care rise dramatically at the end of life, especially in developed countries., Most of these costs are spent caring for older people, such that in Australia almost % of total hospital expenditure is attributable to care for people aged and over in their last year of life, which is estimated to be about % of the total health budget. These high costs could at least in part reflect unnecessary and expensive treatments for those at the end of life. The high costs of health care at the end of life have focussed attention on how these costs might be contained, and better end-of-life care delivered, through provision of primary, community and palliative care services. However, little information exists about the potential to reduce end-of-life costs through better management in primary care in the lead up to end of life. A study of almost 0,000 deceased Medicare beneficiaries aged years and over in the United States found that greater numbers of visits to primary care physicians in the year prior to the -month end-of-life period were associated with lower total health care costs at the end of life, and with fewer preventable hospitalisations for congestive heart failure and chronic obstructive pulmonary disease. The applicability of these findings to a country such as Australia, where the government provides universal health coverage for its citizens and there is no charge for treatment at public hospitals, is unknown. In a system with greater access to publically-funded health care, such as Australia, it is possible that those who use more primary care services prior to the end-of-life might also use more care at the end of life. Therefore, this study investigated the relationship between health care expenditures in the last months of life and use of general practitioner (GP) services in the - months before death, using data from a large cohort of older residents of New South Wales, Australia. METHODS Participants This analysis was nested within the Assessing Preventable Hospitalisation InDicators (APHID) study. APHID uses linked survey and administrative data for participants in the Sax Institute s and Up Study, a prospective cohort of,0 men and women aged years and over and residents in New South Wales (NSW), Australia. Participants entered the Study by completing a mailed self-administered questionnaire at study entry (between February 00 and April 00) and providing written consent for long-term follow-up and linkage of their health information to a range of routine health databases. People residing in non-urban areas and those aged 0 years and over were oversampled. The overall response rate for the and Up Study was estimated to be % and the study included about 0% of the NSW population aged and over. Ethics approval for the APHID study was obtained from the NSW Population and Health Services Research Ethics Committee, Aboriginal Health and Medical Research Council of NSW Ethics Committee, and the University of Western Sydney Ethics Committee. The conduct of the