Introduction: With benchmarking of outcomes becoming increasingly critical in a setting of accountable healthcare, we propose a validated clinical apparatus (interactive risk-calculator) to predict outcomes following craniotomy for acoustic neuroma (AN) excision.
Methods: An observational, cohort study involving patients in the HCUP NIS database undergoing craniotomy for AN excision.
Outcome endpoints: Discharge disposition, LOS, charges, and post-operative complications (cardiac; neurological; venous-thromboembolism, facial nerve palsy; hydrocephalus).
Exposures: Demographics (age, sex, race, insurance, socioeconomic status); Hospital characteristics (bedsize, region); Comorbidities (Stroke, seizure disorder, NF-2, sensorineural hearing loss, CAD, hypertension, VHD, coagulopathy, anemia, hypercholesterolemia, COPD, DM, hyponatremia, & obesity)
Statistical methods: Multivariable models were constructed to identify association of exposures to predict individual outcomes. Regression diagnostics, c-statistics, and calibration were performed for each models. C-statistics corrected for over-fitting bias using stratified bootstrap and leave-out cross validation using 1000 sub-samples. As sensitivity analyses, all models fitted with generalizing estimating equations using sandwich-covariance matrix estimator to restrict clustering of similar outcomes within hospitals. A clinical tool was developed to estimate the predicted-risk of each outcome at 5% type-I error for retaining variables. Model-based multiple-imputation approach was preferred over traditional deletion methods, single, cold and deck imputation methods.
Results: 2,983 patients underwent craniotomy for AN excision at over 200 centers. Mean age: 49.71± 13.30 years, and 54% were female, 80% Caucasian. Private insurance covered 75%, Medicare 13.5% and Medicaid 5.5%. 30% patients were operated in the South, 29% in the West, 24% in Midwest and 17% in the Northeast. Forest plots depicting risk of exposure variables on outcomes are plotted (Figure 1-3). An interactive-calculator is developed with an ability to predict outcomes (Figure 4)
Conclusions: Our study provides individualized estimates of the risks of postoperative complications. The developed and validated clinical apparatus could potentially aid in risk-stratification, shared-decision making, strengthening referral patterns, and pre-surgical counselling for complex cases.
Patient Care: Shared decision making, outcome prediction, risk stratification, strengthening referrals for complex cases, and exploring alternative treatment methods where surgery may lead to poor outcomes postoperatively
Learning Objectives: By going through the presented app, the participants should be able to:
1.) Experience the ability to predict outcomes in patients undergoing surgery for with acoustic neuroma on a hand-held device.
2.) Imbibe clinical decision making using predictive models and tools in complex cases and even using it for referring patients that preoperatively could be assessed to incur poor outcome post-surgery and use alternative treatment methods (eg. gamma knife where indicated).