Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

The purpose of this paper is to estimate changes in the trajectory of health care use that result from changes in drinking behaviors caused by the COMBINE therapies. Introduction: Reductions in health care use over time following alcohol dependence treatment have long been used to justify funding of alcohol treatment (Holder and Blose, 1992; Booth et al., 1997; Goodman et al., 2000; Kane et al., 2004). Despite the reliance on trajectories of health care use inherent in these analyses, most studies in this literature have relied on simple fixed-effect time trends to describe changes over time in health care use. In this paper, we model these changes using parallel process latent growth models (LGM) that explicitly tie changes in health care use to changes in alcohol behaviors. Hypotheses: We will test 3 hypotheses: H1: Greater reductions in the severity of alcohol dependence will result in greater reductions in health care use. H2: Pharmacotherapies, by themselves, will cause a one time downward shift in health care use trajectories with no effect on the slope of the trajectories. H3: Behavioral interventions, either alone or in combination with pharmacotherapies, will reduce the slope of health care use trajectories. H4: Behavioral interventions and/or pharmacotherapies will alter the relationship between changes in alcohol use over time and changes in health care use over time. Approach: We postulate that health care use is determined, in part, by the underlying health of individuals that evolves over time following a latent growth model that includes an individual-level latent intercept and an individual-level latent slope. The latent growth in health is influenced by changes in drinking behaviors and alcohol dependence caused by the COMBINE interventions. Data analysis: We will estimate the LGM using structural equation modeling. We will use the health care use items form the Economic Form 90 to measure changes in health care use. Based on our previous work linking economic outcomes to measures of dependence (Bray et al., forthcoming), we will use the ADS to measure dependence. We will use the investigator release data to measure changes in health care use that occur between the end of treatment (week 16) and week 68.