Skip to content

10 September 2022

Can we predict the need for long-term care? A case study of the Austrian cash-for-care system using administrative data

Ulrike Famira-Mühlberger , Austrian Institute of Economic Research , Austria

Klaus Nowotny, University of Salzburg
Christine Mayrhuber, Austrian Institute of Economic Research

Abstract

The need for long-term care imposes financial challenges on many elderly and disabled people. To support those in need of long-term care, Austria introduced the "Pflegegeld" (long-term care allowance) in 1993, a needs (but not means) tested cash-for-care transfer. The benefit level depends on the hours of professional care required per month and is inter alia con-tingent on the applicant's health status and their ability to perform (instrumental) activities of daily living. Given the demographic challenges of the coming decades, which are expected to increase the number of people in need of long-term care, financial pressures on public long-term care systems and health budgets will doubtlessly increase. Understanding the empirical relationship between health care services and the long-term care system is thus crucial. This paper studies the relationship between the individual consumption of health services and the receipt of long-term care allowance in Austria. We use detailed administrative data that covers all 550,960 receivers of long-term care allowance in Austria who were 60 years or older between 2016 and 2018. In addition, we also have access to a case-control sample of 435,332 non-receivers matched by age, gender and NUTS-3 region of residence.

The data gives information on age, gender, region of residence and the level of long-term care allowance received (if any), but also on all their doctor visits, hospital stays (including the diagnoses and the types of medical services received) and all the drugs prescribed to them in Austria. We use this detailed information on health services to analyse the first-time receipt of the long-term care allowance, using regularization and supervised machine learning methods to train a classification algorithm to predict a person's prospective care requirements in the short run, based on previous health services, doctor visits and drugs prescribed on an individual level.

Furthermore, in a second step we analyse the correlation between the transfer to a higher care allowance level and the health benefits provided prior to the transfer. In a third step, we analyze how recipients' probability of entering into inpatient care can be explained by the consumption of health services.

Our results show that the first-time receipt of long-term care benefits can be predicted relatively well from the data in the short run. Age, inpatient stays in hospitals, the frequency of contacts with general practitioners and the use of drugs affecting the nervous system correlate most strongly with the subsequent first-time receipt of long-term care benefits. The most important factors preceding the move to a higher long-term care allowance level are the previous long-term care allowance level, age and - again - the use of drugs for the nervous system. The level of long-term care allowance also correlates strongly with entry into inpatient care, as does the frequency of contacts with general practitioners and age. The analysis reveals several avenues for health policy that can be drawn from the conclusions of the paper, but also shows that further research is necessary in order to improve the quality of the predictions.

Slides


Skip to toolbar