Linking Disaster Predictions to Health Care Strain and Costs: A Novel Military-Civilian Case Study

Broad predictions about disaster health care needs are insufficiently granular to estimate system impacts. Historical utilization data could refine predictions, but disaster patients differ systematically from usual health care. This study matches civilian health care utilization data to predicted disaster patient characteristics and validates the method, using the theoretical example of mass military patient transfer to civilian hospitals.

An ICD-10 code sorting algorithm was developed, categorizing each ICD-10 code into one of 13 broad stakeholder-predicted categories. Blinded clinicians validated each categorization. Healthcare Cost and Utilization Project (HCUP) and civilian hospital billing data were used to match category/ICD-10 code pairs to Diagnosis-Related Groups (DRG) to understand utilization for each disaster injury category.

Agreement was excellent (Cohen's ĸ = 0.86; 99.2% agreement among ≥2/3 clinicians). The resulting ICD-10 codes-disaster injury category crosswalk was applied to 1,945,272 HCUP inpatient encounters. Most disaster injury categories corresponded exactly to one DRG; some DRGs, e.g., multi-system trauma, corresponded to multiple disaster injury categories. Length of stay and payer varied by disaster injury category and HCUP vs hospital billing data.

This method refines broad predictions about disaster epidemiology using linkage to granular civilian health care data; it can improve readiness by accurately modeling disaster care and reimbursement.