Development of an algorithm to identify preoperative medical consultations using administrative data

Population-level studies of preoperative care often rely on administrative data. These datasets lack specific codes to distinguish preoperative consultations from nonoperative visits. Prior research has shown that internal medicine specialists may improve outcomes through preoperative evaluations. However, no prior work had resolved how to accurately identify these consultations in claims data. This gap motivated the need for a reliable method to detect preoperative consultations. Existing studies use chart abstraction, which is time-consuming and not scalable. Researchers needed a claims-based solution that could be applied broadly. The absence of a validated algorithm limited the ability to assess preoperative care quality. This study aimed to address that limitation by developing a new approach.

The goal was to create a claims-based algorithm for identifying preoperative medical consultations. These consultations are critical for optimizing surgical outcomes but are difficult to track in administrative data. The study focused on elective noncardiac surgeries with intermediate-to-high risk. Researchers wanted to ensure the algorithm could be applied across different jurisdictions. The specific problem was the lack of specific codes for preoperative visits. The motivation was to enable population-level assessments of preoperative care. The study aimed to improve the accuracy of administrative data analysis. This would support future research on surgical outcomes and care quality.

The study used a cross-sectional design with data from Ontario, Canada. Researchers selected 606 patients aged over 40 who underwent elective surgery. Medical records were abstracted to identify preoperative consultations. These were compared with data from physician claims and hospital discharge records. The team tested various combinations of physician types and time windows. The optimal algorithm included consultations by specific specialists within four months of surgery. Sensitivity, specificity, and predictive values were calculated for each model. The final model was validated using random sampling and statistical confidence intervals.

The optimal algorithm identified 90% of true preoperative consultations. It had a 92% specificity, meaning few false positives were included. The positive predictive value was 93%, indicating high accuracy. The negative predictive value was 90%, suggesting reliable exclusion of non-consultations. The algorithm included consultations by cardiologists, internists, and others within four months. These specialists were most frequently involved in preoperative evaluations. The model outperformed other combinations of physician types and timeframes. The results suggest the algorithm can be used in population-level studies.

The authors proposed that their algorithm can accurately identify preoperative consultations in administrative data. They emphasized the need for linked healthcare datasets to apply the method. The study showed that specific physician types and time windows are effective. The researchers suggested this approach could enhance evaluations of preoperative care. They noted that the algorithm's performance was consistent across multiple metrics. The findings may support future studies on surgical outcomes and care quality. The authors proposed that the algorithm could be adapted for other jurisdictions. They suggested further validation in different healthcare systems.