Predictors of inappropriate emergency department utilization following elective thoracolumbar spine surgery

Background: The incidence of spine surgery continues to rise, representing a significant patient population and a material percentage of healthcare expenditure. While spine surgeries yield a high success rate, recent research indicates that a large number of patients visit the emergency department (ED) postoperatively. This poses a burden to the patient and the healthcare system alike. This study aimed to identify and predict inappropriate ED visits within 6 months after elective thoracolumbar spine surgery. Methods: We identified 987 consecutive patients receiving surgery from 2008 to 2013. Patients with previous spine surgery, malignancy, worker's compensation or spine-related litigation (n = 388) were excluded. Through regional electronic medical record review, we identified 98 ED visits. Comprehensive chart reviews were then conducted for these visits, and 2 fellowshiptrained orthopedic spine surgeons split patients into groups: those who engaged in an inappropriate surgery-related ED visit (n = 80) and those who did not (n = 519). Eighteen visits were determined to be appropriate and/or unrelated. These patients were included in the 519 "no inappropriate visit" group. Age, body mass index (BMI), SF-36 physican and mental summary scores, Oswestry Disability Index (ODI) scores, Numeric Rating Scale back pain and leg pain scores, Charlson Comorbidity Score, levels of intervention, sex, surgeon, marital status, living arrangement, education, work status, smoking, drug and alcohol use, exercise, primary pathology, primary symptom, surgery type (fusion/nonfusion), approach (open/MIS), perioperative adverse events, and family doctor status (yes/no) were measured. We performed t tests on continuous variables and x2 tests on categorical variables. A logistic regression model was then built based on these results and previous research. Results: During the exploratory phase, BMI (t589 = 2.487, p = 0.01) and marital status (x22 = 8.24, p = 0.02, n = 555) were the only 2 significant predictors of an inappropriate visit. The regression model was not statistically significant (p = 0.54). Upon further investigation, marital status was sex-mediated, with divorced/separated women significantly more likely to engage in an inappropriate visit than single or married/engaged/common-law women (x22 = 9.48, p < 0.01, n = 555); there was no effect for men (p = 0.76). Conclusion: While there were 2 significant factors in this model, the R2 values were very low. This, in combina