Development and Validation of a Healthcare Utilization-Based Algorithm to Identify Acute Exacerbations of Chronic...

Area of Science:

• Pulmonary Medicine
• Health Informatics
• Data Science

Background:

• Acute exacerbations of chronic obstructive pulmonary disease (AECOPD) are critical events impacting patient outcomes and healthcare management.
• Accurate identification of AECOPD in electronic administrative data is essential for effective population health surveillance and clinical practice improvement.

Purpose of the Study:

• To develop and validate codified algorithms for identifying moderate and severe AECOPD events using administrative data and electronic medical records.
• To assess the performance of these algorithms by calculating positive predictive value (PPV) and negative predictive value (NPV) across two US healthcare systems.

Main Methods:

• Two algorithms were created: one for moderate AECOPD (outpatient/emergency visits with AECOPD codes and antibiotic/steroid prescriptions) and one for severe AECOPD (inpatient visits with relevant codes).
• Algorithm performance was validated through pulmonologist-adjudicated chart reviews, calculating PPV and NPV on a substantial number of events and patient records.

Main Results:

• Both algorithms demonstrated high positive predictive values: 98.3% for moderate AECOPD and 96.0% for severe AECOPD.
• Negative predictive values varied, with 75.0% for moderate AECOPD and 95.0% for severe AECOPD.
• Consistent results were observed across the two participating healthcare systems, supporting external validity.

Conclusions:

• Healthcare utilization-based algorithms can effectively identify moderate and severe AECOPD with high PPV.
• While the algorithm for severe AECOPD showed high NPV, the moderate AECOPD algorithm's NPV was lower.
• The consistency of findings across different healthcare systems validates the generalizability of these AECOPD identification algorithms.