(1205) Accuracy of diagnosis codes in administrative data for measuring shoulder dystocia

Objective:
Research on shoulder dystocia often utilizes International Classification of Diseases (ICD) diagnosis codes to identify this obstetric emergency. However, the accuracy of ICD codes for measuring shoulder dystocia has not been validated. Potential misclassification can hinder understanding of its prevalence and risk factors and thus limit ability to develop interventions. This study aimed to examine the accuracy of ICD diagnosis codes in administrative data for measuring shoulder dystocia.

Study Design:
This is a retrospective cohort study of all patients with vaginal delivery of a singleton, vertex gestation from 1/1/2016 to 12/31/2024 at a large, academic health system. Shoulder dystocia was identified in two ways: 1) based on ICD diagnosis codes (ICD-9: 660.4*/ICD-10: O66.0) in administrative data, and 2) medical chart abstraction by trained staff in a departmental obstetrics database. The latter was used as the reference standard for our evaluation. Shoulder dystocia identified based on ICD diagnosis codes was compared against the reference standard to determine its sensitivity, specificity, and positive and negative predictive values.

Results:
The overall sample included 30,774 patients (characteristics summarized in Table 1). Among them, shoulder dystocia was identified in 717 (2.3%) deliveries based on ICD codes and 708 (2.3%) based on chart abstraction. ICD codes had a sensitivity of 93.1% and specificity of 99.8% in identifying shoulder dystocia, along with a positive and negative predictive value of 91.9% and 99.8%, respectively (Table 2).

Conclusion:
In this first study to evaluate the accuracy of administrative codes for shoulder dystocia, ICD-9 and -10 codes performed with high specificity and relatively high sensitivity at the study institution. Further evaluation in other data sources is warranted in order to support the broader use of these codes for research purposes.