Using ICD Codes Alone May Misclassify Overdoses Among Perinatal People

Area of Science:

• Perinatal Health
• Public Health Surveillance
• Medical Informatics

Background:

• Rising perinatal drug overdose rates necessitate improved monitoring strategies.
• Accurate identification of overdose events during pregnancy and postpartum is critical.
• Medication for opioid use disorder (MOUD) is a key intervention for pregnant individuals.

Purpose of the Study:

• To evaluate the accuracy of ICD-9/10-CM codes for detecting drug overdose events in pregnant individuals receiving MOUD.
• To determine the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of these codes.
• To assess code performance during pregnancy and postpartum periods.

Main Methods:

• Analysis of electronic health records (EHR) for 3,911 pregnancies with MOUD and known outcomes (2014-2021).
• Comparison of ICD-9/10-CM codes against abstracted EHR data for drug overdose events.
• Utilized CDC drug overdose case definitions for categorization.
• Analyses conducted between May 2023 and May 2024.

Main Results:

• During pregnancy, ICD-9/10-CM code sensitivity for overdose was 32.7% (NPV 99.0%, specificity 98.5%, PPV 23.4%).
• Postpartum, code sensitivity was 30.9% (NPV 98.8%, specificity 98.4%, PPV 25.9%).
• Low sensitivity and PPV indicate undercounting of overdose events using these codes.

Conclusions:

• ICD-9/10-CM codes demonstrate low sensitivity and PPV for identifying perinatal drug overdoses in individuals with MOUD.
• High specificity and NPV suggest codes are reliable for ruling out overdose when absent.
• Integrating additional EHR data and external sources is recommended for comprehensive overdose surveillance.