RESUMEN
BACKGROUND: Surgical intensive care unit (SICU) patients are known to be at high risk for venous thromboembolism (VTE). The 2005 Caprini Risk Assessment Model (RAM) predicts VTE risk in surgical patients. However, a physician's ability to accurately complete this RAM and the effect that inaccurate RAM completion might have on VTE risk remain unknown. STUDY DESIGN: Between 2009 and 2012, physicians completed a 2005 Caprini score for all SICU admissions at our institution. For comparison, we used a previously validated, computer-generated score. Regression-based techniques examined the effect of inadequate risk stratification on inpatient VTE risk, when controlling for other confounders. RESULTS: Among 3,338 consecutive SICU admissions, 55.2% had computer-generated scores that were higher than the physician-reported score, and 20.6% of scores were equal. Computer-generated scores were higher than physician-reported scores for both median (6 vs 5) and interquartile range (5 to 8 vs 3 to 7). Inter-rater reliability between the 2 scores was poor (kappa = 0.238). Risk score underestimation by ≥2 points was significantly associated with inpatient VTE (7.67% vs 4.59%, p = 0.002). Regression analysis demonstrated that each additional day's delay in chemoprophylaxis (odds ratio [OR] 1.05, 95% CI 1.01 to 1.08, p = 0.011) and under-risk stratification by ≥2 points (OR 2.46, 95% CI 1.53 to 3.96, p < 0.001) were independent predictors of inpatient VTE, as were higher admission APACHE score, personal history of VTE, recent pneumonia, and younger age. CONCLUSIONS: Physicians under-risk stratify SICU patients when using the 2005 Caprini RAM. As hospitals incorporate electronic medical records into daily practice, computer-calculated Caprini scores may result in more accurate VTE risk stratification. Inadequate VTE risk assessment and delay to chemoprophylaxis carry independent and significant increased risk for VTE.