B Mozes1, E Shabtai, D Zucker. 1. Gertner Institute for Health Services Research, Chaim Sheba Medical Center, Tel Hashomer, Israel.
Abstract
OBJECTIVES: To identify patient attributes that were associated with increased mortality; variables that were associated with process of care that were correlated with mortality; and outlier centers after adjustment for patient attributes. DESIGN: Standard interviews were conducted by trained nurses with all patients. Detailed information regarding primary renal diagnosis, comorbidity, and results of laboratory tests were obtained from the medical charts. The vital status of the patients was obtained from the records of each of the centers. We used the Cox hazard method to identify variables that correlated with a 1-year mortality. Centers with observed mortality exceeding the 95% confidence interval (CI95) of the expected probability of death were marked as outliers. SETTING: Seven dialysis centers located in large teaching hospitals in Israel. PATIENTS: The current study included patients > 16 years of age who had undergone hemodialysis > 4 weeks prior to the day of data collection. RESULTS: The study included 564 patients. Significant differences were found in patient demographics and process variables among the centers. The following variables correlated with mortality; diabetes (odds ratio [OR], 2.03; CI95, 1.28-3.21); ischemic heart disease (OR, 2.2; CI95, 1.39-3.49); each year of age (OR, 1.04; CI95, 1.02-1.06); each 1 g% of albumin (OR, 0.51; CI95, 0.30-0.86). The average observed mortality in all centers was 17.4%. After adjustment for casemix, one center showed excess mortality (24% observed compared to 15% expected after adjustment for patient attributes; CI95, 6.2-23.7). CONCLUSIONS: The ability to compare mortality rates among dialysis centers to detect possible quality outliers depends on thorough consideration of patient attributes and random variation.
OBJECTIVES: To identify patient attributes that were associated with increased mortality; variables that were associated with process of care that were correlated with mortality; and outlier centers after adjustment for patient attributes. DESIGN: Standard interviews were conducted by trained nurses with all patients. Detailed information regarding primary renal diagnosis, comorbidity, and results of laboratory tests were obtained from the medical charts. The vital status of the patients was obtained from the records of each of the centers. We used the Cox hazard method to identify variables that correlated with a 1-year mortality. Centers with observed mortality exceeding the 95% confidence interval (CI95) of the expected probability of death were marked as outliers. SETTING: Seven dialysis centers located in large teaching hospitals in Israel. PATIENTS: The current study included patients > 16 years of age who had undergone hemodialysis > 4 weeks prior to the day of data collection. RESULTS: The study included 564 patients. Significant differences were found in patient demographics and process variables among the centers. The following variables correlated with mortality; diabetes (odds ratio [OR], 2.03; CI95, 1.28-3.21); ischemic heart disease (OR, 2.2; CI95, 1.39-3.49); each year of age (OR, 1.04; CI95, 1.02-1.06); each 1 g% of albumin (OR, 0.51; CI95, 0.30-0.86). The average observed mortality in all centers was 17.4%. After adjustment for casemix, one center showed excess mortality (24% observed compared to 15% expected after adjustment for patient attributes; CI95, 6.2-23.7). CONCLUSIONS: The ability to compare mortality rates among dialysis centers to detect possible quality outliers depends on thorough consideration of patient attributes and random variation.