Systematic screening of secondary diagnoses in medicare administrative data to identify candidate risk factors for the principal diagnosis

Abstract PURPOSE: Secondary diagnoses in Medicare hospital discharge claims may include risk factors for the principal diagnosis. However, risk ratios for the principal diagnosis as a function of secondary diagnoses cannot be calculated because no comparable data exist for beneficiaries who are not hospitalized. METHODS: Hospital discharge rates, as proxies for incidence rates, can be calculated by race and sex from Medicare claims and denominator files. If the prevalence of a risk factor is higher in one population group than another, that risk factor will be overrepresented among patients from the group at higher risk. RESULTS: This imbalance is reflected in what is termed the odds difference, OD  = [( r  +  r ′)/ r ][ f 2 /(1− f 2 )− f 1 /(1− f 1 )], in which r is the background incidence rate, and r ′ is the additional risk conferred by a factor that is present in fractions f 1 and f 2 in the two groups. Unlike the risk ratio, the odds difference can be calculated from claims data. Given f 1 and f 2 , the odds difference is directly proportional to the risk ratio, RR  = ( r  +  r ′)/ r . CONCLUSIONS: Ranking common secondary diagnoses by the magnitude of their odds difference between groups with disparate discharge rates for a given principal diagnosis may disclose secondary diagnoses that merit evaluation as candidate direct or indirect risk factors.