Monitoring capacities have always been a very limited commodity. Efficient risk-based distribution to patients is not always easy in day-to-day clinical practice. In most neurosurgical departments patients are monitored on the ICU or IMC after elective craniotomies. However, there are no recommendations for the duration and kind of postoperative monitoring for patients concerned. Consequently, patients at risk are not monitored as a priority.

In this study, we were able to develop a prediction score that provides a practical approach for individual risk assessment of patients undergoing elective craniotomy. This score can improve the efficiency of the allocation of the valuable resource of monitoring capacities by prioritization of patients at high risk for postoperative complications after elective craniotomies.

In our cohort, in 13.8% of cases, a complication requiring monitoring occurred. The majority (64.7%) of the complications had a pulmonary cause. We found that a 12-h monitoring period is sufficient to capture 90% of all incidents. We identified age, the CCI score, duration of surgery, vestibular schwannomas as the treated pathology, blood loss, and the ASA score as independent predictors of postoperative complications. However, the combination these predictors, excluding blood loss, formed the best prediction model for postoperative complications with regard to the AIC. Based on these parameters, we developed the presented three-tiered risk classification and validated it internally. This score can then be employed for a preoperative risk stratification, thereby optimizing the use of ICU beds and prioritizing patients with a calculated high risk.

The postoperative monitoring of patients who have undergone elective neurosurgical intracranial procedures is a routine standardized procedure that is conducted in most neurosurgical departments. It is standard practice for hospitals to have established standard operating procedures, which provide guidance on bed allocation and the duration of monitoring. However, these workflows are not standardized and vary in international and national comparison. The decision-making process is based on internal agreements and the experiences of senior neurosurgeons and anesthetists.

Nevertheless, some attempts have been made to evaluate the necessity of ICU monitoring for neurosurgical patients, while clear recommendations have not been established so far [4, 5, 9, 15]. Sioshansi et al. analyzed their cohort of 200 patients after lateral skull base approaches and found that 17 (8.5%) of them experienced an adverse event. This was in the context of mandatory intensive care monitoring in patients. It is notable that the majority of complications were described as hypertensive urgencies. There were no mortalities [5].

Ziai et al. analyzed a group of 158 patients after cranial tumor resection and found that of 158 patients, 23 required monitoring in the ICU for more than 24 h. As potential predictors for the necessity to exceed 24-h monitoring due to complications or instability, radiologic findings, large intraoperative blood loss, fluid requirements, and the decision to keep the patient intubated at the end of surgery were identified [4].

A group of patients undergoing elective craniotomies was evaluated by Bui et al. They found that 12.5% of 343 patients were planned for postoperative ICU monitoring because of the length of surgery or anesthetic risks. In only 2% of cases, an unexpected admission or transfer to the ICU was necessary because of slow neurological recovery and extensive intraoperative blood loss. In 3% of cases, a medical emergency team was called for the management of complications on the ward [15]. In the study by Lonjaret et al. 16% of patients presented with a neurological complication, which may justify monitoring for early detection [16]. In a review of literature from 2001 to 2021, testing alternative postoperative pathways excluding ICU monitoring, Azad et al. found that only 2% of 2469 patients needed a readmission to the ICU [17].

The present analysis focused on complications that occurred within the first 24 h following craniotomy. The rationale for selecting this temporal range is that the early postoperative phase is commonly perceived as a period of heightened risk for postoperative hematoma, edema, and the eventual need for hemodynamic support [18]. As shown in the Results, the mean monitoring time at our department was 26.6 h. Although there is no recommendation for the duration of stay on the ICU/IMC, our monitoring times appear to align with the commonly held understanding of the early postoperative phase as a period of increased risk [3].

In our study the complication rate was 13.8%. A majority (64.7%) of complications were pulmonary in nature, such as pneumonia (34.5%) and hydropic decompensation (peripheral and pulmonary edema, pleural effusion) (10.1%). A minor proportion (2.8%) required monitoring because of neurological deterioration resulting from edema, minor hemorrhage, or hydrocephalus. Twenty patients underwent revision surgery owing to hemorrhage, edema, hypoperfusion, and hydrocephalus whereas ten were readmitted because of generalized seizures causing impaired consciousness and respiratory failure.

It is already known that certain factors increase the risk of postoperative complications. Sioshansi et al. conducted an analysis of a cohort of patients with purely oncological diagnoses undergoing neurosurgical procedures. Their findings indicated that tumor mass and preexisting hypertension were independent predictors of adverse outcomes [5]. Lohmann et al. identified the number of secondary diagnoses, BMI, and incision closure time as risk factors for adverse events in their evaluation of elective craniotomies [9]. Lonjaret et al. observed that patients undergoing posterior fossa surgery were more likely to be readmitted to the ICU and proposed that ICU monitoring should be prolonged following this type of surgery [16]. Our data do not support this statement, as the location of supratentorial or infratentorial showed only a trend of correlation with the development of complications.

In our study, six factors demonstrated predictive value in the multivariate analysis: CCI, ASA score, the patients’ age, schwannoma as the treated pathology, duration of surgery, and blood loss. The first four parameters were associated with the patients’ conditions and were known preoperatively. These factors could permit the preliminary allocation of monitoring capacities to be made based on preoperative data. The latter two parameters can be evaluated during the operative procedure or subsequently. A prolonged surgical procedure and significant blood loss would justify postoperative monitoring, even in the absence of high-risk stratification, by using solely preoperative factors.

It is noteworthy that other parameters, such as the type of surgery, the location of the intracranial pathology, BMI, obstructive sleep apnea, sex, and the level of postoperative monitoring (IMC or ICU) did not correlate with the risk for postoperative complications. Interestingly, the risk for a complication increased with the time of monitoring. This could depict the correct choice based on experience and clinical aspects to monitor certain patients postoperatively and not to transfer them too early to the ward.

Badenes et al. aimed to define criteria to allocate ICU beds to neurosurgical patients at risk of complications [19]. Moreover, an attempt was made to evaluate the parameters with the greatest impact. Nevertheless, the evidence available was insufficient to create a standardized protocol. Lohmann et al. created a Surgical Outcome Risk Score (SOS) score to predict postoperative complications in neurooncological patients with intracranial tumors. The score included the size of the tumor, BMI, and the time required from incision to closure. The authors subsequently evaluated an existing score systems on a group of patients with spinal or cranial tumors with a view to predict infections, the risk of reoperation, and adverse events. The Nursing care level, length of stay on the ICUscore (NonInfECT), Leukocytosis, ECOG on admission, Urgency of surgery and Cutting-suture time of index surgery (LEUCut), and Leukocytosis, length of stay in the ICU, Nursing care level, and CRP on admission (LINC) score systems were found to be effective in predicting the occurrence of complications of various types [9]. An external validation of the score for assessment of predictive value is pending [9].

In the present study, it was possible to calculate a threshold for three of the identified parameters that leads to a significantly higher risk for complications. The calculated threshold of approximately 200 min from incision to closure can be used to stratify patients into a high-risk group preoperatively if the planned duration of the operation exceeds this value. In addition, it can be employed for intraoperative decision-making in the event of unexpected prolongation of the operative duration. In the event of blood loss exceeding approximately 500 mL, admission to the ICU should be considered. Furthermore, a CCI of more than 6 points was identified as the level of comorbidities that is associated with a higher perioperative risk and could be used preoperatively as a cutoff for decision-making.

Considering these new insights, it was imperative to conduct a validation and weighting of influencing factors to create a reliable prediction score. In the course of an extended statistical analysis, the prediction score was the most reliable when based on age, CCI, duration of surgery, ASA score, and treatment of a vascular pathology. The model demonstrated a satisfactory predictive capacity, with an AUROC of 0.78. The scoring system allows the determination of the grade of risk for postoperative complications. The score encompasses parameters delineating the patients’ health status, the nature of the pathology, and the duration of surgery, which represents an element of intraoperative influence. It was our objective to provide a reliable assessment of the risk of developing postoperative complications. The score is easy to calculate, and the requisite parameters are used in clinical practice.

Regarding the results, the period with the highest risk for adverse events was the first postoperative 5.7 h. An SD of 4.4 h indicates that a maximum of 14.5 h of monitoring was necessary to detect 97.5% of complications. The mean monitoring time in our cohort was approximately 26 h, which is more than 100% longer than the expected duration.

In their analysis, Mirza et al. examined 421 patients, of whom 25 had an indication for ICU admission and were already planned for this. Only four patients who were transferred postoperatively to the general ward had to be readmitted to the ICU because of complications. All these complications occurred during the first postoperative hours. The study concluded that after 4 h of monitoring in the postanesthetic care unit, the risk of a severe complication was very low [11].

A similar study was conducted by Pendharkar et al., which involved the transfer of 63 young patients (aged < 65 years and ASA class of 1, 2, or 3) after elective craniotomy to the ward for hourly neurological examination for 8 h. Only 11% of patients were admitted to the ICU. No complications were observed after 8 h [13].

In light of the findings of other published studies and the present investigation, it can be concluded that a monitoring period exceeding 10 h is not obligatory. However, to detect 90% of all incidents, a monitoring time of 12 h is warranted. To capture 90% of the complications, the mean time of 5.5 h needs to be added to twice the SD (5.5 h + 2 × 3.5 h). The implementation of a guideline regulating the time of monitoring and a subsequent rapid transfer to the ward would have a considerable impact on the valuable resources of ICU or IMC capacity and specialized medical personnel.