Attributing inpatient care to diabetes: The case of Medicare for the elderly in Texas, 1995

Roy R McCandless, The University of Texas School of Public Health


Specific aims. This study estimated the accuracy of alternative numerator methods for attributing health care utilization and associated costs to diabetes by comparing findings from those methods with findings from a benchmark denominator method. Methods. Using Medicare's 1995 inpatient and enrollment databases for the elderly in Texas, the researcher developed alternative estimates of costs attributable to diabetes. Among alternative numerator methods were selection of all records having diabetes as a principal or secondary diagnosis, and a complex ICD-9-CM sorting routine as previously developed for study of diabetes costs in Texas. Findings from numerator methods were compared with those from a benchmark denominator method based on attributable risk and adapted from a study of national diabetes costs by the American Diabetes Association. This study applied age, gender and ethnicity specific estimates of diabetes prevalence taken from the 1987–94 National Health Interview Surveys to person-months of Medicare Part A, non-HMO enrollment for Texas in 1995. Outcome measures were number of persons identified as having diabetes using alternative definitions of the disease; and number of hospital stays, patient days, and costs using alternative methods for attributing care and costs to diabetes. Cost estimates were based on Medicare payments plus deductibles, co-pays and third party payments. Findings. Numerator methods for attributing costs to diabetes produced findings quite different than those from the benchmark denominator method. When attribution was based on diabetes as principal or secondary diagnosis, the resulting estimates were significantly higher than those obtained from the denominator method. The more complex sorting routine produced estimates near the lower boundary for the confidence interval associated with estimates from the benchmark method. Conclusions. Numerator methods employed by previous researchers poorly estimate the costs of diabetes. While crude mathematical adjustment can be made to the respective numerator approaches, a more useful strategy would be to refine the complex sorting routine to include more hospitalizations. This report recommends approaches to improving methods previously employed in study of diabetes costs.

Subject Area

Public health|Gerontology|Welfare|Health care

Recommended Citation

McCandless, Roy R, "Attributing inpatient care to diabetes: The case of Medicare for the elderly in Texas, 1995" (2000). Texas Medical Center Dissertations (via ProQuest). AAI9985996.