Hospital readmissions confer a significant contribution to healthcare costs, especially in neurosurgical procedures. Studies have estimated the cost of neurosurgical readmissions to be greater than $45,000 USD per stay [10], with some cases approaching $90,000 USD [9]. For perspective, examination of a state-wide California hospital database determined that a 40% reduction in the number of 30-day readmissions for brain tumor patients undergoing craniotomy would eliminate 606 hospitalizations and alleviate $12 million in cost [10].

Factors affecting hospital readmissions after craniotomies for tumor resection have been studied in the adult population, citing insurance type and patient comorbidities as major influences of 30- and 90-day readmissions [3]. The literature for the pediatric population, however, is limited. To our knowledge, this is the first publication addressing readmissions after elective craniotomy for tumor resection. Our overall 90-day readmission rate was 14%. We identified several variables on bivariate testing, but multivariate analysis resulted in only four that were statistically significant.

In our cohort, categorical age 0 to < 5 years was statistically significant for readmission compared to the other age groups. Additionally, continuous age was significant on bivariate analysis but could not be analyzed in multivariate analysis. One could posit that surgical morbidities are more likely in children of younger age. A study by Chotai et al. assessing pediatric hospital readmissions after neurosurgical procedures also found that children of younger age were more likely to be readmitted than older ones [16]. Similarly, Liu et al. found that children aged 0–3 were significantly more likely than older children to be readmitted within 30 days after craniotomy for tumor [17]. Just as plausible is that the younger the child, the more difficult it is to determine the underlying cause of any postoperative issues, leading to a lower threshold for caregivers to bring that child back to the hospital and be readmitted, whether or not a postoperative problem truly exists.

We found that for each additional minute of surgical time, the chances of readmission within 90 days increased by 0.2%. The relationship between duration of craniotomy surgery and readmission risk is not well established in the literature. In a study of 9799 pediatric neurosurgical procedures, Sherrod et al. reported that each additional hour of surgery incurred a 5% increase in 30-day readmission risk on multivariate analysis [18]. Conversely, Chotai et al. found no association between surgical time and 90-day readmission among all pediatric neurosurgical procedures [16]. The more complex a craniotomy, the more time it will take which in turn may make the postoperative course longer, with higher risk of postoperative issues. Further investigation into the association between craniotomy surgery time and readmission risk is needed to better elucidate this relationship.

Like readmission, LOS is a frequently evaluated core metric, and prolonged LOS has been associated with increased likelihood of readmission in both adult [2, 3, 6] and pediatric patients [16, 17, 19, 20]. In general pediatric surgery, readmitted patients had an average LOS twice as long as patients who were not readmitted [19]. The same trend has been previously reported in neurosurgical patients, with longer LOS associated with readmission [16].

In our previous publication [21], we defined extended LOS as greater than 7 days, which is why we decided to analyze categorical LOS in the multivariate analysis and not continuous LOS. While continuous LOS and LOS > 7 days were both significant on bivariate analysis, LOS > 7 days lost its significance on multivariate analysis. Length of ICU stay also lost its significance in multivariate analysis. In the pediatric general surgery population, the relationship between the length of ICU stay and readmission events has not been clear, with their association varying between studies [16, 22]. It’s conceivable that a longer hospital stay may be reflective of a more difficult postoperative course, whether it be from symptom control (e.g., pain, nausea), complexity of the surgery, social factors or other reasons. These same reasons may persist upon discharge and be the reasons for readmission. The relationship between LOS and readmissions requires further examination.

While the occurrence of one or more POE (medical or surgical, planned or unplanned) did not maintain its significance on multivariate analysis, it is worth noting that readmitted patients had higher rates of POEs than non-readmitted patients. On the other hand, having to return to the OR with neurosurgery due to a POE during index admission was found to be predictive of 90-day readmission. There is limited data in the pediatric neurosurgical literature examining the relationship between multiple index operations and risk of readmission. In a study of adults undergoing any neurosurgical procedure, an increased number of operations during the index admission was predictive of neurosurgery-related readmission within 30 days [2]. It follows that patients with POEs serious enough to require additional surgical intervention prior to discharge are more likely to have a complicated post-operative course and thus are more prone to readmission.

It is possible that combining all POEs into a single group and analyzing the significance of “any POE” diminished the impact of certain POEs that are more consequential than others. In our recent publication evaluating extended LOS, POE was the strongest predictor, increasing the odds of LOS > 7 days by almost 30-fold [21]. Other studies have reported POEs as leading causesof hospital readmission after pediatric brain tumor surgery. For instance, Janjua et al., using the Nationwide Readmissions Database (NRD), found that in children undergoing intracranial tumor surgery, CNS-related complications, surgical site infections, and hydrocephalus accounted for the majority of reasons leading to readmission events [23]. Sletvold et al., assessing 30-day post-operative outcomes after pediatric intracranial tumor surgery at a single institution, found that CSF leakage and headache/nausea were the most common reasons for 30-day readmission [24].

We found that a high tumor grade and specific diagnoses of embryonal tumor, high-grade glioma, and medulloblastoma were predictive of future readmission events on bivariate analysis, but only high tumor grade remained significant on multivariate analysis. With respect to specific tumor types, only craniopharyngioma—compared to low grade gliomas—was a predictor of 90-day readmission on multivariate analysis, despite not being significant on bivariate analysis. While similar studies in the pediatric population are lacking, studies of adult patients have also found a diagnosis of malignant tumors to be predictive of a 30-day readmission event [8], while another adult study did not find any significant correlation between 30-day readmission and specific tumor type [11]. It is well established that craniopharyngioma resection, even if limited/subtotal, can be challenging surgically and associated with a number of potential postoperative issues. So, while it stands to reason that the average craniopharyngioma surgery is inherently more complicated than the average low grade glioma resection, we are likely only scratching the surface of what is undoubtedly a complex interconnection between tumor pathology and readmission risk.

The major strength of this study is the quality of the data, prospectively obtained by a single individual over more than a decade. In conducting retrospective analysis of this data, it is notable that variables, such as preoperative neurologic status, precise tumor location, tumor volume, molecular classification of tumors, or goal(s) of surgery, would have added greater depth to our analysis as they could potentially impact readmission status. Regarding our definition of complication as equivalent to that of “unexpected POE”, we acknowledge that determining what is expected versus unexpected introduces an element of subjectivity and therefore potential bias. However, all data was collected by a sole research coordinator with extensive neurosurgical experience as a clinical nurse and not by the surgeon. We feel this resulted in uniform data collection while limiting surgeon bias. Additionally, as this is a single institution study, generalizability will be limited. Lastly, population characteristics are likely different to a certain degree at our institution due to the historically high volume of brain tumor patients and partnership with a quaternary-level children’s cancer hospital.