Background: Studies of interventions to decrease rates of surgical site infections (SSIs) must include thousands of patients to be statistically powered to demonstrate a significant reduction. Therefore, it is important to develop methodology to extract data available in the electronic medical record (EMR) to accurately measure SSI rates. Prior studies have created tools that optimize sensitivity to prioritize chart review for infection control purposes. However, for research studies, positive predictive value (PPV) with reasonable sensitivity is preferred to limit the impact of false-positive results on the assessment of intervention effectiveness. Using information from the prior tools, we aimed to determine whether an algorithm using data available in the Veterans Affairs (VA) EMR could accurately and efficiently identify deep incisional or organ-space SSIs found in the VA Surgical Quality Improvement Program (VASQIP) data set for cardiac and orthopedic surgery patients. Methods: We conducted a retrospective cohort study of patients who underwent cardiac surgery or total joint arthroplasty (TJA) at 11 VA hospitals between January 1, 2007, and April 30, 2017. We used EMR data that were recorded in the 30 days after surgery on inflammatory markers; microbiology; antibiotics prescribed after surgery; International Classification of Diseases (ICD) and current procedural terminology (CPT) codes for reoperation for an infection related purpose; and ICD codes for mediastinitis, prosthetic joint infection, and other SSIs. These metrics were used in an algorithm to determine whether a patient had a deep or organ-space SSI. Sensitivity, specificity, PPV and negative predictive values (NPV) were calculated for accuracy of the algorithm through comparison with 30-day SSI outcomes collected by nurse chart review in the VASQIP data set. Results: Among the 11 VA hospitals, there were 18,224 cardiac surgeries and 16,592 TJA during the study period. Of these, 20,043 were evaluated by VASQIP nurses and were included in our final cohort. Of the 8,803 cardiac surgeries included, manual review identified 44 (0.50%) mediastinitis cases. Of the 11,240 TJAs, manual review identified 71 (0.63%) deep or organ-space SSIs. Our algorithm identified 32 of the mediastinitis cases (73%) and 58 of the deep or organ-space SSI cases (82%). Sensitivity, specificity, PPV, and NPV are shown in Table 1. Of the patients that our algorithm identified as having a deep or organ-space SSI, only 21% (PPV) actually had an SSI after cardiac surgery or TJA. Conclusions: Use of the algorithm can identify most complex SSIs (73%–82%), but other data are necessary to separate false-positive from true-positive cases and to improve the efficiency of case detection to support research questions.

Funding: None

Disclosures: None