Does home care policy generosity impact hospitalization?

This study assesses the effect of home care policy (HCP) generosity on inpatient care use. The analysis is conducted on data from Switzerland, a federal country where health care policy is decentralized in 26 cantons (i.e. states). Decentralization results in variation of HCP generosity across cantons and over time. For example in 2007, the canton of Schwyz provided 0.9 hours of home care (HC) per capita, and Jura 3.6 hours. From 1997 to 2007, this measure decreased by 30 percent in Geneva and grew by 282 percent in Ticino. Besides providing care in the preferred setting of the population (at home), HC may partly substitute for more costly forms of care, such as inpatient care. Overall, there is little evidence on the effect of HCP generosity on inpatient care use. Most studies don’t address the potential endogeneity of HCP, use data from social experiments with limited generalizability, or focus on one specific type of HC, such as Medicare home health in the US. The main contribution of our work is to consider all forms of HC services and inpatient care (acute and non-acute), regardless of the payers (public, health insurance, out-of-pocket). We compare the effect of two forms of HC: medically-related, e.g. nurse visits, and non-medically-related HC, mainly help with limitations in the activities of daily living. The former is hypothesized to be a better substitute to inpatient care than the latter.

A two-part model is estimated to determine the impact of canton HCP generosity on the individual likelihood of hospitalization and length of stay (LOS), conditional on hospitalization. The endogeneity of HCP generosity is addressed through an instrumental variable strategy combined with canton and time fixed effects. The variable of interest is logged hours of HC per capita. It is further decomposed into hours of medically- and non-medically-related HC. Control variables include health status, informal care availability, socio-demographics, and indicators of pressure to be at home. The tested instrumental variables are political and fiscal indicators, such as the proportion of left-wing sits in the canton legislative assemblies or the financial resource index used in the equalization between cantons.

The analysis is conducted on three cross-sectional waves of the Swiss Health Survey (1997, 2002 and 2007), which is merged with the indicators of canton HCP. There are 42,500 observations of individuals aged 20+. Some 12 percent of the observations have a hospitalization, with a mean LOS of 10 days. Preliminary results indicate that the instruments pass the specification tests (F-stat=73.8, p-value of overidentifying restrictions=0.38). The null hypothesis of exogeneity of HCP is rejected in the LOS model. HCP generosity does not impact the likelihood of hospitalization but significantly reduces LOS. A 10 percent increase in per capita HC hours reduces LOS by 7 percent. This effect seems to be solely driven by medically-related HC hours. These results suggest that encouraging HC services, in particular medically-related, may be a way to substitute for some forms of inpatient care and help mitigate the expansion in health care costs.