論文ID: 27011
Background
Accurate identification of fetal outcomes among hospitalized pregnant women is essential for perinatal research. Japan’s Diagnosis Procedure Combination (DPC) database may enable outcome ascertainment, but the validity of DPC-derived algorithms remains unclear. This study aimed to develop and validate DPC-based algorithms to identify abortions and stillbirths.
Methods
This retrospective cohort study included hospitalized pregnant women at Kyoto Min-iren Chuo Hospital (January1, 2017–November 30, 2021) and Kobe City Medical Center General Hospital (October 1, 2018–March 31, 2024). We created expert-defined algorithms to predict fetal outcomes, including abortions/stillbirths. Predictions were compared using pregnancy and neonatal summary tables and electronic medical records. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated with 95% confidence intervals.
Results
At Kyoto Min-iren Chuo Hospital, 865 pregnancies included 131 reference abortions/stillbirths (92 induced abortions). Sensitivity was 0.98 (95% CI, 0.94–1.00), specificity 1.00 (95% CI, 1.00–1.00), PPV 1.00 (95% CI, 0.97–1.00), and NPV 1.00 (95% CI, 0.99-1.00). At Kobe City Medical Center General Hospital, 2,694 pregnancies included 43 reference abortions/stillbirths (12 induced abortions). Sensitivity was 0.63 (95% CI, 0.47–0.77), specificity 1.00 (95% CI, 1.00–1.00), PPV 1.00 (95% CI, 0.87–1.00), and NPV 0.99 (95% CI, 0.99–1.00).
Conclusions
The DPC-derived algorithms demonstrated excellent specificity, PPV, and NPV, although the sensitivity varied considerably across hospitals. These findings support the use of inpatient DPC data for pharmacoepidemiological research, provided that institution-specific coding practices are considered. Accurate estimation of outcome prevalence may require site-specific algorithm recalibration.
