Regional Variation of Cost of Care in the Last 12 Months of Life in Switzerland: Small-area Analysis Using Insurance Claims Data

Abstract

Background: Health care spending increases sharply at the end of life. Little is known about variation of cost of end of life care between regions and the drivers of such variation. We studied small-area patterns of cost of care in the last year of life in Switzerland.

Methods: We used mandatory health insurance claims data of individuals who died between 2008 and 2010 to derive cost of care. We used multilevel regression models to estimate differences in costs across 564 regions of place of residence, nested within 71 hospital service areas. We examined to what extent variation was explained by characteristics of individuals and regions, including measures of health care supply.

Results: The study population consisted of 113,277 individuals. The mean cost of care during last year of life was 32.5k (thousand) Swiss Francs per person (SD=33.2k). Cost differed substantially between regions after adjustment for patient age, sex, and cause of death. Variance was reduced by 52%-95% when we added individual and regional characteristics, with a strong effect of language region. Measures of supply of care did not show associations with costs. Remaining between and within hospital service area variations were most pronounced for older females and least for younger individuals.

Conclusions: In Switzerland, small-area analysis revealed variation of cost of care during the last year of life according to linguistic regions and unexplained regional differences for older women. Cultural factors contribute to the delivery and utilization of health care during the last months of life and should be considered by policy makers.

FIGURE 1

Mean cost [in 1000s Swiss Francs (CHF)] of care in the last 12 months of life among 113,277 individuals across 564 regions of place of residence, Switzerland 2008–2010. HSA indicates hospital service area.

FIGURE 2

Spatial distribution of cost ratios adjusted for age and cause of death (model 2, left column) and fully adjusted (right column) for males. Regional values derived from exponentiated random effects across 564 regions of place of residence, nested within 71 hospital service areas (HSAs). Cost ratio represents the ratio by which regional costs are higher or lower when compared with the national mean; for instance, coefficient of 1.2 indicates 1.2 times (or 20%) higher cost.

FIGURE 3

Spatial distribution of cost ratios adjusted for age and cause of death (model 2, left column) and fully adjusted (right column) for females. Regional values derived from exponentiated random effects across 564 regions of place of residence, nested within 71 hospital service area (HSAs). Cost ratio represents the ratio by which regional costs are higher or lower when compared with the national mean; for instance, coefficient of 1.2 indicates 1.2 times (or 20%) higher cost.

FIGURE 4

Association of cost of care in the last 12 months of life across selected covariates of model 4 (fully adjusted) across sex and age strata. Estimates additionally adjusted for age, civil status, nationality, religion, Swiss-SEP index, and level of urbanization. Exponentiated coefficients of regressing inverse hyperbolic sine transformed overall cost of care during the last 12 months of life representing cost ratios across analyzed variables. Cost ratio represents the ratio by which costs in specific category are higher or lower when compared with the national mean; for instance, coefficient of 1.2 indicates 1.2 times (or 20%) higher cost. All density of supply measures divided into quintiles of spatial units. CHD indicates coronary heart disease; CVD, cardiovascular disease; COPD, chronic obstructive pulmonary disease; DNS, diseases of the nervous system; MBD, mental and behavioral disorders; ref., reference category. See main text for International Statistical Classification of Diseases-10 codes of the causes of death. See Tables 7–10, Supplemental Digital Content 1, http://links.lww.com/MLR/B255 for exact estimates of coefficients across all covariates.