Comparison of face-to-face interaction and the electronic medical record for venous thromboembolism risk stratification using the 2005 Caprini score.

Objective: Perioperative venous thromboembolism (VTE) risk can be quantified with the 2005 Caprini score. The Caprini score has previously been validated by review of the electronic medical record (EMR) in >3000 plastic surgery patients. However, the accuracy of Caprini-based risk stratification using the EMR, as opposed to face-to-face contact with the patient, remains unknown.

Methods: Plastic and reconstructive surgery patients who had surgery under general anesthesia, required postoperative admission, and were started on enoxaparin prophylaxis were identified. The 2005 Caprini scores were calculated retrospectively using EMR review only (no direct contact with the patients) to establish cohort 1. The 2005 Caprini scores were calculated prospectively using face-to-face interaction with the patients, followed by EMR review, to establish cohort 2. For all included patients, EMR review or face-to-face screening was personally performed by the authors. We compared the proportions of patients with identified Caprini risk factors and the aggregate risk scores of patients between cohorts.

Results: Complete data were available for 536 unique patients in the EMR review cohort and 207 unique patients in the face-to-face cohort. Patients whose risk scores were calculated face to face had higher Caprini scores than those calculated by EMR review alone. The face-to-face cohort had a higher proportion of patients risk stratified as Caprini 7-8 (29.5% vs 24.8%) and Caprini >8 (26.6% vs 10.5%) compared with the EMR review cohort. Patients risk stratified by face-to-face discussion were significantly more likely to be stratified into a higher risk Caprini stratum. Face-to-face discussion identified a 2-fold increase in patients with personal history of deep venous thrombosis (12.6% vs 6.3%; P = .005), a 3-fold increase in patients with family history of VTE (16.9% vs 5.2%; P < .001), and a 20-fold increase in patients with personal history of multiple lost pregnancies (13.6% vs 0.6%; P < .001) compared with EMR review. Observed differences for family history of VTE and history of pregnancy loss persisted after propensity score analysis, created using component variables in the 2005 Caprini score plus gender; this supports the conclusion that observed differences were not due to site variation or case mix.

Conclusions: When it is used in isolation, the EMR may provide inaccurate estimation of patient-level VTE risk using the 2005 Caprini score. This study demonstrates that EMR review may miss key VTE risk factors, such as personal or family history of VTE, history of pregnancy loss, and others; this omission results in lower estimates of perioperative VTE risk. The importance of provider-patient interaction for accurate VTE risk stratification cannot be overstated.