Variation in Hospital Thromboprophylaxis Practices for Abdominal Cancer Surgery

Abstract

Introduction

Venous thromboembolism remains a prominent cause of morbidity and mortality following cancer surgery. Although evidence-based guidelines recommend major cancer surgery thromboprophylaxis starts before incision and continues at least 7–10 days postoperatively, the extent to which the guidelines are followed is unknown. We assessed variation in thromboprophylaxis practices for abdominal cancer surgery in a regional surgical collaborative.

Methods

We studied abdominal resections for primary gastrointestinal, hepatopancreatobiliary (HPB), and neuroendocrine malignancies in the Michigan Surgical Quality Collaborative from July 2012 to September 2013 (N = 2967 patients in 52 hospitals). We obtained detailed perioperative and postoperative pharmacologic and mechanical thromboprophylaxis information for patients without documented exemptions (e.g., active bleeding, allergy), and compared differences in procedure mix and operative complexity across hospitals based on their perioperative thromboprophylaxis rates. Additionally, we surveyed hospitals to identify variations in perioperative practice and barriers to prophylaxis administration.

Results

Overall, 40.4 % of eligible patients had perioperative pharmacologic thromboprophylaxis for abdominal cancer surgery, and 25.3 % of the highest-risk patients had evidence of inadequate postoperative prophylaxis (under-prophylaxis, either by dose or duration). Hospital perioperative thromboprophylaxis rates ranged from 0 to 96.1 %, and postoperative thromboprophylaxis rates ranged from 73.9 to 100 %. Epidural use was not independently associated with hospital pharmacologic thromboprophylaxis rates.

Conclusions

Fewer than half of patients undergoing abdominal cancer surgery receive perioperative thromboprophylaxis, and there is wide variation in hospital thromboprophylaxis utilization despite strong evidence-based guidelines supporting its use.

Appendices

Appendix 1

See Table 3.

Table 3 Procedure groups with relevant CPT codes; patient diagnoses with relevant ICD-9 codes

Appendix 2

Venous Thromboembolism Risk Models

First, we estimated patient VTE risk using multivariable logistic regression models that included patient age, sex, race, and their interactions, as well as body mass index, cancer diagnosis, history of VTE, functional status, hypertension, smoking status, ascites, chronic immunosuppression, preoperative dyspnea, pneumonia, ventilator dependence, or dialysis requirement, procedure type, and urgency of operation as covariates. The model c-statistic was 0.802 and demonstrated good calibration across deciles of risk according to the Hosmer–Lemeshow test.

In a second model, we used a previously published VTE risk calculator (REF) that assigns patients points based on the presence of the following: age ≥ 60 years; body mass index ≥ 40, male sex, preoperative sepsis, personal or family history of VTE, and active cancer. Patients are then classified as low, intermediate, or high risk based on their total risk score.