Back to 2018 Program and Abstracts

---

DEVELOPMENT OF AN OBJECTIVE PREDICTION MODEL FOR PROLONGED POSTOPERATIVE LENGTH OF STAY IN CROHN'S DISEASE PATIENTS UNDERGOING MAJOR ABDOMINAL SURGERY
Fares Ayoub^*1, Amir Kamel², Naueen A. Chaudhry¹, Atif Iqbal³, Ellen M. Zimmermann⁴, Sarah C. Glover⁴, Sanda Tan^3
1Department of Medicine, University of Florida, Gainesville, FL; ²Department of Pharmacy, University of Florida, Gainesville, FL; ³Department of Surgery, University of Florida, Gainesville, FL; ⁴Department of Gastroenterology, Hepatology & Nutrition, University of Florida, Gainesville, FL

Background
Postoperative length of stay (LOS) for patients with Crohn's disease requiring major abdominal surgery is determined by multiple factors. A prolonged LOS leads to higher costs, increased risk of iatrogenic postoperative complications, poorer quality of care, and has serious adverse effects on patient satisfaction and well being. We sought to develop a predictive model for prolonged postoperative LOS based on modern, readily available criteria.
Methods
A retrospective review of patients who underwent major abdominal surgery for a diagnosis of Crohn's disease at a tertiary care university medical center serving a large underserved population between May 2011 and May 2017 was performed. Prolonged post-operative LOS was defined as a LOS more than the median length of stay for the study group (>6 days). Baseline characteristics analyzed included: age, sex, race, smoking history, history of venous thrombosis, American Society of Anesthesiologists physical status class, body mass index, albumin, pre-albumin, C-reactive protein level, steroid use, biologic use, abdominal imaging findings, history of prior intestinal resection, indication for surgery, type of surgery (open versus laparoscopic), urgency of surgery and LOS. Multivariate logistic regression was used to determine risk factors associated with prolonged LOS. A prediction model was developed with each of the significant predictor variables assigned one point, proportional to its regression coefficient.
Results
A total of 149 patients were included with a median age of 39 years and median BMI of 24. Females comprised 55% of the study population. On univariate analysis, African American race (OR 5.92), smoking (OR 2.74), history of venous thrombosis (OR 4.49), absence of preoperative biologic therapy (OR 2.52), presence of intra-abdominal complications [fistula, abscess, obstruction, perforation] on pre-operative imaging (OR 2.79) and open abdominal surgery (OR 2.56) were found to be significant predictors of prolonged postoperative stay. On multivariate analysis, all of the above variables remained significant except for race [Table 1]. A prediction model was developed which defined 3 risk groups: low risk (0-1 points), intermediate risk (2-3 points) and high risk (4 or more points). The risk for prolonged postoperative hospitalization in these three groups was 13.6%, 47.6% and 100%, respectively. The area under the receiver operating characteristic curve was 0.75 [95% CI 0.69-0.82].
Conclusions
Prediction of prolonged postoperative length of stay in Crohn's patients undergoing major abdominal surgery is possible with a practical and objective prediction model. This allows for preoperative risk factor modification and postoperative facilitation of hospital and personnel resource allocation. Validation of our prediction model to confirm its generalizability and performance is needed.

Table 1: Multivariate logistic regression demonstrating odds ratios for risk factors of prolonged post-operative length of stay

	Odds ratio	p	95% confidence interval
African-American race	4.89	0.078	0.83-28.61
History of smoking	2.58	0.016	1.19-5.60
History of venous thrombosis	3.48	0.030	1.13-10.69
Absence of preoperative biologic therapy	3.40	0.011	1.32-8.77
Intra-abdominal complication on pre-operative imaging [fistula, abscess, obstruction, perforation]	2.98	0.030	1.11-8.03
Open surgery	2.56	0.018	1.17-5.60

---

Back to 2018 Program and Abstracts