A novel online calculator to predict nonroutine discharge, length of stay, readmission, and reoperation in patients undergoing surgery for intramedullary spinal cord tumors

BACKGROUND CONTEXT: Intramedullary spinal cord tumors (IMSCTs) are rare tumors associated with significant morbidity and mortality. Surgical resection is often indicated for symptomatic lesions but may result in new neurological deficits and decrease quality of life. Identifying predictors of these adverse outcomes may help target interventions designed to reduce their occurrence. Nonetheless, most prior studies have employed population-level datasets with limited granularity. PURPOSE: To determine independent predictors of nonroutine discharge, prolonged length of stay (LOS), and 30 day readmission and reoperation, and to deploy these results as a web-based calculator. STUDY DESIGN: Retrospective cohort study PATIENT SAMPLE: A total of 235 patients who underwent resection of IMSCTs at a single comprehensive cancer center. OUTCOME MEASURES: Nonroutine discharge, prolonged LOS, 30 day readmission, and 30 day reoperation METHODS: Patients who underwent surgery from June 2002 to May 2020 at a single tertiary center were included. Data was collected on patient demographics, clinical presentation, tumor histology, surgical procedures, and 30 day readmission and reoperation. Functional status was assessed using the Modified McCormick Scale (MMS) and queried preoperative neurological symptoms included weakness, urinary and bowel dysfunction, numbness, and back and radicular pain. Variables significant on univariable analysis at the α≤0.15 level were entered into a stepwise multivariable logistic regression model. RESULTS: Of 235 included cases, 131 (56%) experienced a nonhome discharge and 68 (29%) experienced a prolonged LOS. Of 178 patients with ≥ 30 days of follow-up, 17 (9.6%) were readmitted within 30 days and 13 (7.4%) underwent reoperation. Wound dehiscence (29%) was the most common reason for readmission. Nonhome discharge was independently predicted by older age (OR=1.03/year; p<.01), thoracic location of the tumor (OR=2.36; p=.01), presenting with bowel dysfunction (OR=4.09; p=.03), and longer incision length (OR=1.44 per level; p=.03). Independent predictors of prolonged LOS included presenting with urinary incontinence (OR=2.65; p=.05) or a higher preoperative white blood cell count (OR=1.08 per 10³/μL); p=.01), while GTR predicted shorter LOS (OR=0.40; p=.02). Independent predictive factors for 30 day unplanned readmission included experiencing ≥1 complications during the first hospitalization (OR=6.13; p<.01) and having a poor (A-C) versus good (D-E) baseline neurological status on the ASIA impairment scale (OR=0.23; p=.03). The only independent predictor of unplanned 30 day reoperation was experiencing ≥1 inpatient complications during the index hospitalization (OR=6.92; p<.01). Receiver operating curves for the constructed models produced C-statistics of 0.67-0.77 and the models were deployed as freely available web-based calculators (https://jhuspine5.shinyapps.io/Intramedullary30day). CONCLUSIONS: We found that neurological presentation, patient demographics, and incision length were important predictors of adverse perioperative outcomes in patients with IMSCTs. The calculators can be used by clinicians for risk stratification, preoperative counseling, and targeted interventions.