Creating a predictive model and online calculator for high-value care outcomes following glioblastoma resection: incorporating neighborhood socioeconomic status index.

PURPOSE

Social determinants of health including neighborhood socioeconomic status, have been established to play a profound role in overall access to care and outcomes in numerous specialized disease entities. To provide glioblastoma multiforme (GBM) patients with high-quality care, it is crucial to identify predictors of hospital length of stay (LOS), discharge disposition, and access to postoperative adjuvant chemoradiation. In this study, we incorporate a novel neighborhood socioeconomic status index (NSES) and develop three predictive algorithms for assessing post-operative outcomes in GBM patients, offering a tool for preoperative risk stratification of GBM patients.

METHODS

Adult GBM patients who underwent surgical resection from a single center were identified; NSES was identified via patient street address of residence, with lower scores representing disadvantaged neighborhoods. Multivariate logistic regression analysis was used to predict high value care outcomes. The Hosmer-Lemeshow test was used to assess model calibration.

RESULTS

A total of 467 patients were included, with a mean age of 59.85 ± 13.21 years and 58.7% being male. The mean NSES for our cohort was 63.77 ± 14.91, indicating that the majority resided in neighborhoods with a higher socioeconomic status compared to the national average NSES of 50. One hundred nine (23.3%) patients had extended LOS, 28.9% had non-routine discharge, and 19.1% did not follow the Stupp protocol following surgery. On multivariate regression, worse NSES was significantly and independently associated with extended LOS (OR = 0.981, p = 0.026), non-routine discharge disposition (OR = 0.984, p = 0.033), and non-compliance with the Stupp protocol (OR = 0.977, p = 0.014). Our three models predicting high-value care outcomes had acceptable C-statistics > 0.70, and all models demonstrated adequate calibration (p > 0.05). Final models are accessible via online calculator. https://neurooncsurgery4.shinyapps.io/GBM_NSES_Caclulator/ CONCLUSION: NSES scores are readily available and may be utilized via our open-access calculators. After external validation, our predictive models have the potential to assist in providing patients with individualized risk estimates for post-operative outcomes following GBM resection.