Background: Quality monitoring is increasingly important to support and assure sustainability of the orthopedic practice. Surgeons in nonacademic settings often lack resources to accurately monitor quality of care. Widespread use of electronic medical records (EMR) provides easier access to medical information, facilitating its analysis. However, manual review of EMRs is highly inefficient. Artificial intelligence (AI) software allows for the development of algorithms for extracting relevant complications from EMRs. We hypothesized that an AI-supported algorithm for complication data extraction would have an accuracy level equal to or higher than manual review after total hip arthroplasty (THA).
Methods: A total of 532 consecutive patients underwent 613 THA between January 1 and December 31, 2017. A random derivation cohort (100 patients, 115 hips) was used to determine accuracy. After generation of a gold standard, the algorithm was compared to manual extraction to validate performance in raw data extraction. The full cohort (532 patients, 613 hips) was used to determine recall, precision, and F-value.
Results: AI accuracy was 95.0%, compared to 94.5% for manual review (P = .69). Recall of 96.0% (84.0%-100%), precision of 88.0% (33%-100%) and F-measure of 0.85 (0.5-1) was achieved for all adverse events. No adverse events were recorded in 80.6%, 1.3% required reintervention and 18.1% had "transient" events.
Conclusion: The use of an automated, AI-supported search algorithm for EMRs provided continuous feedback on the quality of care with a performance level comparable to manual data extraction, but with greater speed. New clinical information surfaced, as 18.1% of patients can be expected to have "transient" problems.
Keywords: artificial intelligence; clinical research informatics; complication reporting; search algorithm; total hip arthroplasty.
Copyright © 2019 Elsevier Inc. All rights reserved.