Validation of the Diagnosis Procedure Combination Database in Japan for ICU Research: A Multicenter Comparison with...

The Diagnosis Procedure Combination (DPC) database is Japan's most widely used administrative inpatient dataset, supporting epidemiological and health services research. While its validity is established for various diagnoses and procedures, accuracy for intensive care unit (ICU) variables has not been directly evaluated using a clinical registry as the gold standard.

We conducted a multicenter retrospective validation study using four Japanese ICUs. Patient records from the national ICU registry, the Japanese Intensive Care Patients Database (JIPAD), were matched with corresponding DPC data, retaining only successfully matched patients to evaluate coding accuracy rather than case ascertainment. We assessed binary variables (demographics, comorbidities, diagnoses, interventions, ICU admission, mortality) and continuous variables (demographics, Sequential Organ Failure Assessment [SOFA] scores). We calculated sensitivity and specificity for binary variables and intraclass correlation coefficients (ICCs) for continuous variables.

We included 14,070 ICU admissions. Most binary variables, including demographics, major diagnostic categories, ICU interventions, and mortality, had high sensitivity and specificity (≥80%). Comorbidity specificity exceeded 95%, but sensitivity was <30% for several diseases. Sensitivity was 83.3% for invasive mechanical ventilation but low for noninvasive positive pressure ventilation (5.5%) and high-flow nasal cannula (36.1%), although specificity was high across respiratory supports (97.3%-99.9%). SOFA scores showed moderate agreement (ICC, 0.61). Sensitivity/specificity were 99.1%/100.0% for in-hospital mortality and 96.2%/99.9% for ICU mortality.

The DPC administrative inpatient database accurately captures most key ICU variables, but comorbidities, noninvasive respiratory support, and SOFA scores require caution. These findings support its use for ICU clinical and epidemiological research in Japan.