(C-136) Analysis of risk factors associated with 12-month incidence of pseudarthrosis following lumbar spinal fusion

Background: Healthcare databases, such as hospital databases, don’t include radiographic images that are used to make a definitive diagnosis of pseudarthrosis. Instead, healthcare databases may include diagnosis codes for patients affected with the condition. Pseudarthrosis occurs when the spine fails to fuse after surgery, which can cause burdensome pain to the patient. This study used a diagnosis code as a marker of pseudoarthrosis occurrence. The research aims to shed light on key risk factors for pseudarthrosis utilizing the diagnostic code.

Objectives: To evaluate risk factors associated with 12-month pseudarthrosis after lumbar fusion procedure.

Methods: A retrospective, observational cohort study using logistic regression for risk factor evaluation was performed. Patients with elective lumbar spinal fusion procedures between years 2015-2023 and discharged from hospitals with continuous data availability post-fusion were identified within the PINC AI™ Premier Healthcare Database. Index date was defined as the admission date for the fusion. Patients were followed for pseudarthrosis from the day after index date to 365 days post-index. Demographic, clinical, procedure, and provider information were the explanatory variables and twelve-month pseudarthrosis was the outcome. A random sample of 20% patients without the outcome was selected to balance the outcome distribution. Logistic regression model performance was evaluated using area under the receiver operating characteristic curve (ROC AUC).

Results: We present preliminary findings of this study. The cohort identified for this study included 397,229 patients. After under-sampling, the cohort size was 81,279 which included 69,088 patients without and 12,191 patients with pseudarthrosis. The top five most significant predictors of pseudarthrosis (p≤0.05), ordered by odds ratio (OR), were patients with presence of index pseudarthrosis diagnosis (OR=5.07, 95% confidence interval (CI): 4.62, 5.56), surgeries within the lumbosacral spine at index (OR=1.31, 95% CI: 1.22,1.40), patients with Medicare versus non-Medicare insurance (OR=1.29, 95% CI: 1.18, 1.41), patients assigned as Male gender (OR=1.29, 95% CI: 1.20,1.38) and patients with two or more versus one lumbar spinal level operated on at index (OR=1.27, 95% CI: 1.18,1.37). The ROC AUC of this model was 69.7%.

Conclusions: This preliminary analysis of pseudarthrosis risk factors identified demographic, clinical, and procedural factors as potential predictors of pseudarthrosis occurrence. Future work will include using additional models and techniques to understand the most robust model for prediction of pseudarthrosis occurrence when imaging data are not available.