Validity of ascertainment of co-morbid illness using administrative databases: a systematic review

Co-morbid illnesses have a major influence on the epidemiology of infectious diseases. Although International Classification of Diseases (ICD) discharge codes are frequently used to evaluate the presence of co-morbidities in observational research, additional research is required about their validity. We reviewed the evidence supporting the use of routinely coded administrative data for ascertainment of co-morbid diseases with emphasis as it relates to the study of infectious diseases. A systematic Medline, Embase, and bibliographic review were conducted in order to identify and critically appraise published (1990-2008) studies comparing administrative databases with conventional chart review. Twenty-one co-morbidities commonly associated with infectious diseases risk were a priori selected for specific evaluation. Of the 21 co-morbid conditions chosen, only 19 had adequate data available for evaluation. Thirteen studies were included; only one focused on an infectious disease population. Eleven articles validated individual co-morbid conditions data in electronic administrative databases and reported a wide range of pooled sensitivity (13-82%) but overall high pooled specificity (>97%) when compared with medical chart review. Seven articles compared Charlson Co-morbidity Index scores derived from administrative data algorithms as compared with that calculated from medical record review and found that administrative data underscored the index in all articles with kappa agreement ranging from 0.30 to 0.56. The small body of literature published to date suggests that electronic administrative databases have limited validity for co-morbidity evaluation. Studies evaluating administrative database ascertainment of co-morbidities specifically in infectious diseases research are needed.