Performance of Risk Assessment Models for VTE in Patients Who Are Critically Ill Receiving Pharmacologic Thromboprophylaxis: A Post Hoc Analysis of the Pneumatic Compression for Preventing VTE Trial.

Abstract

Background: The diagnostic performance of the available risk assessment models for VTE in patients who are critically ill receiving pharmacologic thromboprophylaxis is unclear.

Research question: For patients who are critically ill receiving pharmacologic thromboprophylaxis, do risk assessment models predict who would develop VTE or who could benefit from adjunctive pneumatic compression for thromboprophylaxis?

Study design and methods: In this post hoc analysis of the Pneumatic Compression for Preventing VTE (PREVENT) trial, different risk assessment models for VTE (ICU-VTE, Kucher, Intermountain, Caprini, Padua, and International Medical Prevention Registry on VTE [IMPROVE] models) were evaluated. Receiver-operating characteristic curves were constructed, and the sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios were calculated. In addition, subgroup analyses were performed evaluating the effect of adjunctive pneumatic compression vs none on the study primary outcome.

Results: Among 2,003 patients receiving pharmacologic thromboprophylaxis, 198 (9.9%) developed VTE. With multivariable logistic regression analysis, the independent predictors of VTE were Acute Physiology and Chronic Health Evaluation II score, prior immobilization, femoral central venous catheter, and invasive mechanical ventilation. All risk assessment models had areas under the curve < 0.60 except for the Caprini model (0.64; 95% CI, 0.60-0.68). The Caprini, Padua, and Intermountain models had high sensitivity (> 85%) but low specificity (< 20%) for predicting VTE, whereas the ICU-VTE, Kucher, and IMPROVE models had low sensitivities (< 15%) but high specificities (> 85%). The positive predictive value was low (< 20%) for all studied cutoff scores, whereas the negative predictive value was mostly > 90%. Using the risk assessment models to stratify patients into high- vs low-risk subgroups, the effect of adjunctive pneumatic compression vs pharmacologic prophylaxis alone did not differ across the subgroups (P_interaction > .05).

Interpretation: The risk assessment models for VTE performed poorly in patients who are critically ill receiving pharmacologic thromboprophylaxis. None of the models identified a subgroup of patients who might benefit from adjunctive pneumatic compression.

Clinical trial registration: ClinicalTrials.gov; No.: NCT02040103; URL: www.

Clinicaltrials: gov. ISRCTN44653506.