Objective: To validate the estimates of the prevalence of multimorbidity based on administrative hospital discharge data, with medical records and chart reviews as benchmarks.
Design: Retrospective cohort study.
Setting: Medical division of a tertiary care teaching hospital.
Participants: A total of 170 medical inpatients admitted from the emergency unit in January 2009.
Main measures: The prevalence of multimorbidity for three different definitions (≥2 diagnoses, ≥2 diagnoses from different ICD-10 chapters, and ≥2 medical conditions as defined by Charlson/Deyo) and three different data sources (administrative data, chart reviews, and medical records).
Results: The prevalence of multimorbidity in medical inpatients derived from administrative data, chart reviews and medical records was very high and concurred for the different definitions of multimorbidity (≥2 diagnoses: 96.5%, 95.3%, and 92.9% [p = 0.32], ≥2 diagnoses from different ICD-10 chapters: 86.5%, 90.0%, and 85.9% [p = 0.46], and ≥2 medical conditions as defined by Charlson/Deyo: 48.2%, 50.0%, and 46.5% [p = 0.81]). The agreement of rating of multimorbidity for administrative data and chart reviews and administrative data and medical records was 94.1% and 93.0% (kappa statistics 0.47) for ≥2 diagnoses; 86.0% and 86.5% (kappa statistics 0.52) for ≥2 diagnoses from different ICD-10 chapters; and 82.9% and 85.3% (kappa statistics 0.69) for ≥2 medical conditions as defined by Charlson/Deyo.
Conclusion: Estimates of the prevalence of multimorbidity in medical inpatients based on administrative data, chart reviews and medical records were very high and congruent for the different definitions of multimorbidity. Agreement for rating multimorbidity based on the different data sources was moderate to good. Administrative hospital discharge data are a valid source for exploring the burden of multimorbidity in hospital settings.