Joeky Senders

27 Thirty-day outcomes after craniotomy µL). Covariates present in less than one percent of cases or with more than ten percent missing data were excluded from the analysis. Cases with missing data in one of the variables of the multivariable analysis were excluded from the analysis. Outcomes Major complications, extended length of stay, reoperations and readmission with related reasons, and mortality within 30-days were extracted as primary endpoints. Based on a previously published definition, major complication was defined as cardiovascular accident, cardiac arrest, myocardial infarction, deep venous thrombosis, pulmonary embolism, unplanned intubation, failure to wean from ventilator, acute renal failure, sepsis, septic shock, deep incisional or organ space surgical site infection, return to the operating room, or death. 27 Extended length of stay was defined as a total hospital stay of more than ten days. Unplanned reoperation data has been collected by NSQIP since its inception in 2005, but unplanned readmission data has been collected since 2011 regardless of the hospital to which the patient was readmitted. Associated procedures for reoperation, based on CPT codes, and reasons for readmission, based on ICD-9 codes, have been collected since 2012. Non-routine discharge was defined as discharge to another acute care facility, skilled nursing facility, or rehabilitation center. Statistical analysis Statistical analyses were conducted using R 3.3.3 (R Core Team, Auckland, New Zealand). Descriptive statistics were performed on baseline demographics after categorization, together with univariable analysis for each of the primary endpoints by means of logistic regression. Multivariable logistic regression models were constructed using variables screened by univariable analysis for occurrence of major complication, extended length of stay, reoperation, readmission, and death. All potential predictors (p<0.05 in the univariable analysis) were included in the multivariable analysis, but age and gender were included automatically. Potential predictors were excluded from the final model if they demonstrated multicollinearity or had a relative low contribution to model fit. Due to the relatively large sample size of the study population, a p<0.0015 was considered as significant, to decrease the chance of Type I (false-positive) error. This critical value was based on a Bonferroni correction for multiple testing with 33 degrees of freedom (0.05/33=0.0015). A confirmatory analysis was performed for every multivariable model, in which missingness was coded as an additional group to verify if the missing group significantly affected the results. The β-coefficients of the continuous variables in the final model were scaled to represent the odds ratios and confidence intervals of meaningful and interpretable units for age (per ten-year increase), BMI (per five kg/m 2 increase), and operative time (per 60-minute increase).