BACKGROUND: Health care funding levels differ significantly across geographic regions, but there is little correlation between regional funding levels and outcomes of elderly Medicare beneficiaries. Our goal was to determine whether this relationship holds true in a non-Medicare population cared for in a large integrated health care system with a capitated budget allocation system. METHODS: We explored the association between health care funding and risk-adjusted mortality in the 22 Veterans Affairs (VA) geographic Networks over a six-year time period. Allocations to Networks were adjusted for illness burden using Diagnostic Cost Groups. To test the association between funding and risk-adjusted three-year mortality, we ran logistic regressions with single-year patient cohorts, as well as hierarchical regressions on a six year longitudinal data set, clustering on VA Network. RESULTS: A 1000 dollar increase in funding per unit of patient illness burden was associated with a 2-8% reduction in three-year mortality in cross sectional regressions. However, in longitudinal hierarchical regressions clustering on Network, the significant effect of funding level was eliminated. CONCLUSIONS: When longitudinal data are used, the significant cross sectional effect of funding levels on mortality disappear. Thus, the factors driving differences in mortality are Network effects, although part of the Network effect may be due to past levels of funding. Our results provide a caution for cross sectional examinations of the association between regional health care funding levels and health outcomes. Copyright (c) 2006 John Wiley & Sons, Ltd.
BACKGROUND: Health care funding levels differ significantly across geographic regions, but there is little correlation between regional funding levels and outcomes of elderly Medicare beneficiaries. Our goal was to determine whether this relationship holds true in a non-Medicare population cared for in a large integrated health care system with a capitated budget allocation system. METHODS: We explored the association between health care funding and risk-adjusted mortality in the 22 Veterans Affairs (VA) geographic Networks over a six-year time period. Allocations to Networks were adjusted for illness burden using Diagnostic Cost Groups. To test the association between funding and risk-adjusted three-year mortality, we ran logistic regressions with single-year patient cohorts, as well as hierarchical regressions on a six year longitudinal data set, clustering on VA Network. RESULTS: A 1000 dollar increase in funding per unit of patient illness burden was associated with a 2-8% reduction in three-year mortality in cross sectional regressions. However, in longitudinal hierarchical regressions clustering on Network, the significant effect of funding level was eliminated. CONCLUSIONS: When longitudinal data are used, the significant cross sectional effect of funding levels on mortality disappear. Thus, the factors driving differences in mortality are Network effects, although part of the Network effect may be due to past levels of funding. Our results provide a caution for cross sectional examinations of the association between regional health care funding levels and health outcomes. Copyright (c) 2006 John Wiley & Sons, Ltd.