OBJECTIVES: To explore mortality rates in commercially insured HE patients treated with rifaximin (with/without lactulose) versus lactulose alone using a claims data based proxy for mortality.

METHODS: Commercially-insured adult patients with HE treated with lactulose and/or rifaximin were identified in a large commercial claims database. This exploratory analysis compared time to death between rifaximin episodes (with/without lactulose) and lactulose alone treatment episodes using multivariate Cox proportional hazards models. Month-Year of death was available through the Social Security Administration Death Master File for a portion of the patients (30-50% of expected deaths). This data was used to assess and refine a previously published algorithm (Joyce 2004) for predicting death from claims data by identifying events likely to be fatal within the last month of claim activity. Sensitivity of the published algorithm (Algorithm 1) and a refinement (Algorithm 2) was assessed; the number of predicted deaths was also compared to the expected number of deaths in the sample.

RESULTS: A total of 3,530 patients met selection criteria. There were 416 (11.8%) patients with a valid death month; sensitivity was 94.5% for Algorithm 1 and 84.6% for Algorithm 2. In the full sample between 832-1,387 deaths were expected; Algorithm 1 predicted 1,098 deaths in the sample, and Algorithm 2 predicted 796. Based on the claims-based proxy of mortality, rifaximin episodes were associated with significantly reduced mortality rate compared to lactulose alone (Algorithm 1: hazard ratio [95% CI]=0.75 [0.64-0.88]; Algorithm 2: hazard ratio [95% CI]=0.77 [0.64-0.92]).

CONCLUSIONS: Using all available claims data to build a proxy for mortality in claims data, this analysis showed that rifaximin (with/without lactulose) was associated with improved survival compared to lactulose alone. In the absence of death date, claims-based algorithms can be judiciously used to explore determinants of mortality.