Multisite validation of a simple electronic health record algorithm for identifying diagnosed obstructive sleep apnea

Area of Science:

• Medical Informatics
• Sleep Medicine
• Health Services Research

Background:

• Obstructive sleep apnea (OSA) is a common condition impacting public health.
• Accurate identification of OSA cases in electronic health records (EHR) is crucial for research and clinical care.
• Existing methods for case identification may lack efficiency or broad applicability.

Purpose of the Study:

• To evaluate a straightforward algorithm for identifying diagnosed obstructive sleep apnea (OSA) cases using EHR data.
• To assess the algorithm's performance across multiple US health systems.

Main Methods:

• A retrospective analysis of EHR data from six US health systems was conducted.
• The algorithm defined OSA cases based on specific ICD-9/ICD-10 codes for sleep apnea on separate dates.
• Chart reviews were used to calculate positive predictive value (PPV) and negative predictive value (NPV).

Main Results:

• The algorithm demonstrated excellent performance, with an overall PPV of 97.1% and NPV of 95.5%.
• High PPV and NPV (≥90%) were observed across most sites, with one site showing a slightly lower PPV.
• Modifying the algorithm to require more diagnostic codes improved PPV at the lower-performing site but reduced the number of identified cases.

Conclusions:

• A simple EHR-based algorithm effectively identifies diagnosed OSA cases with high accuracy.
• This algorithm has significant potential for large-scale EHR-based OSA research.
• Further research is needed to address undiagnosed disease in EHR-defined noncases.