Potential Errors in Health Disparities Research Resulting from Lack of Unique Patient Identifiers: Analysis of Diabetes-related Preventable Hospitalizations.

Abstract

All-payer, population-level hospital discharge data have been used to identify health disparities across racial/ethnic and other demographic groups. However, researchers are often unable to identify unique patients in the data sets if a unique patient identifier is not provided. The lack of the unique patient identifier can result in biased estimates of research outcomes using discharge data. This could then mislead the researchers, public, or policy-makers who utilize such biased results. This study examined estimation bias of health disparities due to rehospitalizations considering diabetes-related preventable hospitalizations using 6 years of state-level data from Hawai'i Health Information Corporation. Different analyses methods showed different probabilities of having multiple visits by age, race/ethnicity and payer subgroups. Charge analysis results also showed that ignoring the multiple visits could result in significance error. For a patient with multiple hospitalizations, rehospitalizations are often dependent upon the discharge status of previous visits, and the independence assumption of the multiple visits may not be appropriate. Ignoring the multiple visits in population-level analyses could result in severe health disparities significance errors. In this hospitalization charge analysis, the Chinese group was not significantly different than the White group (relative risk ratio - RR: [95% CI]: 0.93 [0.80, 1.08]), while the difference was signficant (RR [95% CI]: 0.86 [0.77,0.96]) when the multiple visits were ignored.