A Novel Risk Stratification Tool to Predict Hospital Length of Stay after Surgery for Meningioma

Background Meningiomas are common intracranial tumors that often necessitate surgical intervention. There is a need for risk stratification tools that accurately predict individual patient outcomes after surgery for meningioma.

Methods We developed prognostic models to estimate time to favorable discharge after surgery for meningioma in adults (≥ 18 years) by using data from the National Inpatient Sample, a prospective database that captures discharge data from all hospitals in the United States, from January 2009 to December 2013. Favorable discharge was defined as any routine discharge to home after surgery. A variety of preoperative clinical and socioeconomic factors, present at the time of hospital admission, were evaluated in two models: the full model contains 37 predictors, while the reduced model contains 13 modifiable risk factors, age, race, and gender. Cox proportional hazard regression was used for model development. Predictive accuracy was measured using a concordance index. Both models underwent internal bootstrap validation to assess their future performance on other populations. After model validation, an electronic tool was developed using the reduced model to predict hospital length of stay and individual patient discharge probabilities.

Results Of the 10,757 included patients, 60% were favorably discharged after surgery with a median length of stay of 3 days (interquartile range: 2–5 days). The strongest predictors of increased time to favorable discharge in both models were increasing age, recent weight loss, electrolyte disorders, coagulopathy, and pulmonary circulation disorders. Each of the models accurately estimated time to favorable discharge and predicted individual patient discharge probabilities (concordance index across models: 0.68–0.72; Fig. 1). The reduced model performed as well as the full model. The electronic tool accurately predicted time to favorable discharge based on patients' preoperative characteristics (Fig. 2).

Conclusion We developed two prognostic models that predict time to favorable discharge after surgery for meningioma. These models accurately estimate length of stay and discharge probabilities for individual patients prior to surgery. The electronic risk stratification tool may ultimately be used to guide preoperative clinical decision making.

Fig. 1 Probability of favorable discharge on or before day 3. Calculated from the reduced model. Each panel shows the probabilities calculated for a white male patient with and without the listed comorbidity when all other comorbidities are absent. Dark center line indicates estimated probabilities; lighter colored bands denote confidence bands. Y-axis represents the probability of a favorable discharge on or before day 3; x-axis represents age.

Fig. 2 Example of the risk stratification tool. For a 54-year-old black male with a medical history of diabetes and hypertension, the median time to favorable discharge is hospital day 6. The graph represents cumulative probability of discharge by hospital days.