Development and Testing of a Risk Assessment Model for Venous Thrombosis in Medical Inpatients: the Medical Inpatients and Thrombosis (MITH) Study Score

Multiple government organizations (i.e. the Joint Commission in the United States and the National Institute for Health and Clinical Excellence in the United Kingdom) mandate venous thrombosis (VT) risk assessment for hospitalized patients and provision of VT prophylaxis, however there are no validated VT risk assessment models (RAM) available for use in medical inpatients.

Between January 2002 and June 2009 all cases of VT complicating medical admissions were identified using ICD-9 codes and confirmed by medical record review at a 500 bed teaching hospital. Two controls without VT were frequency matched to each case by admission service (medicine, cardiology, and oncology) and admission year. VT required positive imaging or autopsy. Medical history, presenting conditions, and use of VT prophylaxis in cases and controls were assessed by chart review. Weighted logistic regression was used to calculate odds ratios (OR) and the Taylor series method for 95% confidence intervals (CI) accounting for VT prophylaxis use (both mechanical and pharmacologic). A RAM was developed using clinical judgment and sequentially adding risk factors into a multivariable model. A point value was assigned for each risk factor by dividing the b coefficients' by the lowest b coefficient value and rounding to the nearest integer. To validate the model, the 95% CI for the C-statistic was calculated using bootstrapping with 1000 replicate samples.

299 cases of VT and 601 matched controls were reviewed. The rate of VT per 1000 admissions (95% CI) was 4.6 (3.9, 5.4). Table 1 presents the RAM with the point value for each risk factor. The c-statistic for the model was 0.73 (95% CI 0.70, 0.76). Using a cut-off of ≥2 points as high risk, 79% of cases and 39% of controls were classified as high risk. The probability of VT in the absence of VT prophylaxis for a score <2 was 1.5 (95% CI 1.0, 2.3) per 1000 admissions and for a score ≥2 was 8.8 (95% CI 4.1, 18.8) per 1000 admissions. To evaluate a score assessed by clinical characteristics only, we assessed a score with the same risk factors but removing platelet count and white cell count from the model. The C-statistic was 0.71 (95% CI 0.68, 0.74) and 74% of cases and 39% of controls were high risk. Stratification by admission service or admission to an intensive care unit did not affect interpretation of the results.

We present an internally validated RAM that assesses the risk of VT complicating medical admission. The score is simple, relies only on information easily known at the time of admission, and could be incorporated into an electronic medical record. It will allow clinicians to assess VT risk at admission for medical inpatients and weigh the risks and benefits of pharmacologic VT prophylaxis. The RAM will enable further studies to determine optimal VT prevention strategies in medical inpatients.

No relevant conflicts of interest to declare.