Skip to content

2014 Conference Presentation

Care models Switzerland

1 September 2014

What is home care policy and how can it be measured? A latent variable model

Judite Gonçalves, Geneva School of Economics and Management, University of Geneva, Switzerland
France Weaver, Geneva School of Economics and Management, University of Geneva, Switzerland


Objective: Home care policy (HCP) is a latent concept. This study investigates what HCP encompasses and proposes a way to measure it as a latent variable. Little is known about the characteristics of HCP and how to measure it comprehensively. Furthermore, in the absence of specific policy changes, HCP is typically measured by observable indicators, such as HC expenditures per capita. The selected observable measures are not motivated and research has shown that results are sensitive to the chosen indicator. The main contributions of our work are to characterize HCP and its dimensions, and to provide an alternative approach to measure HCP, via latent factors.

Data and methods: The analysis is conducted on data from Switzerland, a federal country where HCP is decentralized in 26 cantons (i.e. states). Decentralization results in variation of HCP across cantons and over time. For example in 2012, the canton of Obwalden provided home care (HC) services to 1.3% of its population, and the canton of Jura to 5.5%t. From 1997 to 2012, this measure decreased by 3.5 percentage points in Obwalden and grew by 2.3 percentage points in Ticino. Using HC data for the 26 Swiss cantons over a 16-year period (1997–2012), we conduct two-level exploratory factor analysis to identify the structure of HCP. We consider an initial battery of 32 HC indicators, and keep those that are reflected by any latent factor. We rely on model fit criteria and the interpretability of the factors to determine the appropriate number of latent factors. Then, we test this structure using two-level confirmatory factor analysis. The two-level analyses account for the two sources of variation: between cantons and within cantons over time.

Results: We end up with six indicators and a two-factor structure. The first latent factor reflects indicators such as the number of hours of HC per patient, and can be interpreted as the ‘intensity’ dimension of HCP. The second latent factor reflects indicators such as the percentage of the canton population receiving HC, and can be interpreted as the ‘broadness’ dimension of HCP. The two dimensions are not significantly correlated at the between canton level, but they are strongly and negatively correlated within cantons over time. This suggests that cantons, facing budget constraints, choose either to provide HC to many people (being generous on the broadness dimension) or to provide a high level of care to fewer people (being generous on the intensity dimension).

Policy implications: These results provide a characterization of HCP and document how the two estimated dimensions of HCP – intensity and broadness – relate to each other over time within cantons and across cantons.

Audio   Slides

Skip to toolbar