Using the Nationwide Readmissions Database, we quantitatively and systematically examined the relationship between hospital procedure volume as a linear variable and outcomes in patients undergoing craniotomies for malignant primary or secondary brain tumors. We conducted a continuous volume-outcome multivariate analysis and demonstrated that a significant linear association between increasing institutional volume and patient mortality, major complications, 30-day and 90-day readmissions. After the very high institutional volumes, the effect size of the volume-outcome relationship becomes less pronounced but remains significant. Further, we estimated the risk of mortality, major complications, 30-day and 90-day readmission after resection at respective volume thresholds throughout the reported range of volume. This confirmed that the risk of each outcome improved with volume increase in an additive fashion and can be utilized to compare the risk of each outcome of different institutions at any comparative volume. To our knowledge, prior studies have only confirmed the volume-outcome relationship in this patient cohort by arbitrarily dichotomizing hospital volume into high and low cohorts and our group is the first to use a national database to conduct a multivariate linear analysis to investigate the impact of procedural volume on patient outcomes in a continuous and additive function in this patient cohort [3, 6, 9, 10, 15].

In our multivariate analysis of institutional procedure volume as a continuous variable, we showed that volume was significantly associated with increased likelihood of mortality, major complications, 30-day and 90-day readmissions in an additive fashion. Though prior studies do not evaluate the association between volume and adverse outcome using procedural volume as a linear variable, these studies similarly demonstrated a significant association between high institutional procedural volume and decreased risk of poor outcomes across multiple surgical subspecialties [10, 15,16,17,18,19,20,21]. Among analyses of craniotomies for brain tumors, high-volume hospitals have repeatedly been found to be associated with a decreased likelihood of mortality, major complications, and/or adverse discharge disposition [3, 6, 7, 9, 10, 16]. While the relationship between hospital volume and patient outcome clearly exists, the reason for this association is debated [22]. Prior studies have attributed several reasons for this effect, with one explanation being the “practice-makes-perfect” theory, which presumes that surgeons at high-volume institutions have more opportunity to practice their surgical technique for particular procedures, and are thus more experienced in handling complications and dealing with anomalous anatomy [10, 16, 23, 24]. In the United Kingdom, the National Institute for Health and Care Excellence recommends that neurosurgeons practicing surgical neuro-oncology commit at least half of their practice to treating neuro-oncology patients [7]. On a hospital/systems level, increased accessibility of resources, such as specialized multidisciplinary teams (e.g. neuroradiologists, neuro-oncologists), neurosurgical intensive care units staffed with dedicated neurologically-trained personnel, and advanced diagnostic or therapeutic technologies may also contribute to improved outcomes at higher volume hospitals [7, 10, 25,26,27]. Additionally, lower volume hospitals may be at a disadvantage of having a higher proportion of non-elective cases which can limit preoperative risk optimization and surgical planning, thus contributing to worsened outcomes. Although lower volume hospitals were more likely to have non-elective admission than higher volume hospitals (top 75th percentile) in the present study, the linear association of volume with outcomes persists even into the highest volume hospitals (with low non-elective surgery rate), suggesting that both volume and the non-elective/emergent nature of surgeries may each play a critical role.

Our results add to the prior understanding of the volume-outcome effect. Prior studies in this patient cohort have only shown the relationship between hospital volume and patient outcomes by distinguishing low- and high-volume centers without an evidence-based approach to the specific cutoff [3, 6, 9, 10, 15]. Most of these studies only compared outcomes between the lowest and highest quartile or quintile hospitals [3, 6, 9, 10, 15]. A resulting challenge lies in resulting hypotheses where improved outcomes are significantly associated with institutions performing procedures only over specific volume cutoffs. The present study systematically defined the volume-outcome curve by conducting a multivariate linear analysis and delineated the graded association between patient outcomes and hospital volume. For all outcomes, the continuous volume-outcome curve demonstrated more significant association between linear increase in volume and decreased adverse outcomes during volume increases at lower volumes. Studies using systematic approaches similar to the present study showing that the association between volume and patient outcomes (e.g. mortality, patient safety indicator events) is not dichotomous have been primarily conducted in patients undergoing treatment for neurovascular pathologies (e.g. subarachnoid hemorrhage [SAH], unruptured intracranial aneurysms) [28,29,30,31, 33]. These national database studies observed continuous volume-outcome relationships using adjusted linear models [29,30,31,32,33]. Notably, two of these studies show that the effect size for mortality risk is most prominent at lower volumes and becomes attenuated as volume increases, like the findings of the present study [30, 31]. The limiting volume effect at institutions with higher volumes could potentially be explained by the fact that after a certain threshold, expertise and experience do not significantly differ between these institutions and outcomes are more variable on a case-by-case basis. Higher patient complexity and disease severity at higher volume hospitals could be a contributing factor to limiting improving outcomes at higher volumes as well [6]. Moreover, high-volume hospitals may become overburdened if the caseload exceeds the hospitals’ capabilities and thus attenuate the advantages of high-volume hospitals [34]. Taking together the findings of these studies and the present study, the linear graded relationship may be perhaps a more effective way of understanding the volume-outcome relationship rather than establishing volume cutoffs to dichotomously compare high- and low-volume hospitals.

While increasing volume is associated with a more significant decrease in mortality, major complications, readmission rates at lower volumes, we demonstrate that the additive effect for all outcomes remains significant throughout the entire range of volume. This is notable as prior volume-outcome studies in this patient cohort primarily demonstrated that hospitals with procedure volumes in the highest quartile have better outcomes than those in the lowest quartile [3, 6, 9, 10]. Our results suggest that the likelihood of adverse outcomes examined in this study can be estimated and compared between institutions of different volume thresholds throughout the reported range of volume (as demonstrated in Tables 3, 5, 8, 10), rather than only comparing the highest to the lowest volume cohorts. Successful attempts have been made in other neurosurgical pathologies to find an evidence-based approach to compare institutions of different volumes [28,29,30,31,32]. For instance, a national database study on transsphenoidal pituitary tumor resection outcomes estimated that for every ten additional cases, patients were 11% less likely to have iatrogenic panhypopituitarism and 6% less likely to have diabetes insipidus, suggesting an additive effect [32]. A British study on SAH patients demonstrated that mortality risk decreased in an additive fashion for every additional 100 cases [28]. A national database study on patients undergoing treatment for SAH identified a decrease in inpatient mortality and adverse discharge disposition risk from 20 to 100 cases in an additive fashion as well [30]. This study reported the amplitude of change in risk to be higher at lower volumes similar to the present study [30]. Further, our findings highlight the potential benefits of centralizing care for this patient cohort, which has been shown to positively impact outcomes in this complex patient cohort and more extensively in stroke care [9, 35,36,37]. Finally, our findings suggest continued study of whether patients initially admitted or establishing care at low-volume centers may benefit from referral and transfer to high-volume, comprehensive centers after any necessary acute symptomatic treatment [38, 39].