33. Utilizing predictive analytics to understand the influence of modifiable risk factors on readmission in patients receiving lumbar spinal fusion

Abstract

BACKGROUND CONTEXT Within the last several decades, the field of preventive medicine has grown immensely and is now considered the first line of defense against disease and disability. Although many factors lie beyond the scope of patient control, a small number of modifiable risk factors remain within reach. Yet, research outlining the influence of modifiable risk factors on the perioperative well-being of lumbar spinal fusion patients continues to be limited. PURPOSE Here, we develop predictive models to analyze the influence of modifiable risk factors on postfusion outcomes and complications using data from a national database. STUDY DESIGN/SETTING Retrospective cohort analysis using the 2016 National Readmission Database. PATIENT SAMPLE Using the 2016 National Readmission Database, we conducted a retrospective cohort analysis of 110,396 patients who received a lumbar spinal fusion procedure. Modifiable risk factors (MRF) were defined as smoking, alcohol use, malnutrition, and elevated lipid panel. From this cohort, 47,374 (42.9%) patients with at least one MRF were identified using ICD-10 codes. OUTCOME MEASURES Associated costs, perioperative complications, and nonelective readmission at 30, 60, and 90 days. METHODS We used propensity score matching to obtain a group of 47,374 one-to-one age- and sex-matched control patients from the initial 110,396 patient lumbar spinal fusion cohort with no MRFs for comparison. Data were collected for those readmitted within 30, 60, and 90 days, and all patients with routine or elective follow-ups were excluded. Statistics were conducted using RStudio, and predictive algorithms were developed using generalized Gaussian logistic regression models. Significance was determined by Welch Two Sample t-testing, Tukey multiple comparison of means, and Wald testing. RESULTS The percent of female patients in the MRF-present and control groups were 51.6% and 52.2% respectively. The average age of the MRF cohort was found to be 62.36 years, and the average age of the propensity matched control group was 62.15 years at the time of lumbar fusion. Significant differences in readmission between the two groups were found at 30 days (MRF: 5.21%, Control: 4.47%, p CONCLUSIONS The presence of MRFs significantly increased the rate of nonelective readmissions in patients undergoing lumbar spinal fusion. In fact, the number of present MRFs at the time of primary admission significantly increase the hospital LOS, total surgical cost, and future readmission rate at 30, 60, and 90 days. Predictive algorithms are promising new developments that may allow for surgical optimization in patients admitted with multiple MRFs and may find significant patterns that traditional statistical methods may overlook. Preventive approaches aimed at minimizing the presence of MRFs should be considered in patients undergoing lumbar spinal fusion procedures. FDA DEVICE/DRUG STATUS This abstract does not discuss or include any applicable devices or drugs.