This retrospective population-based cohort study was approved by the Institutional Review Board (No. X-2304-821-901). Data sharing for this initiative was authorized by the Big Data Center of the National Health Insurance Service (NHIS; NHIS-2023-1-525). The requirement for informed consent was deemed unnecessary for data analyses, given that analyses were conducted retrospectively and anonymized data were obtained from the South Korean NHIS database.

The data for this study were obtained from the NHIS, an exclusive public insurance system in South Korea. The NHIS database must contain all disease diagnoses and prescription information for any medication, procedure, or both, as mandated by law. Upon enrollment, individuals are eligible to receive health insurance benefits sponsored by the government. All diagnoses were made using the classification outlined in the 10th Revision of the International Classification of Diseases (ICD-10).

This study initially screened all cranial neurosurgery procedures conducted under general anesthesia from January 1, 2016, to December 31, 2021. Cranial neurosurgeries included 1) brain tumor surgery, 2) decompressive craniotomy or craniectomy, 3) hematoma evacuation, 4) cranioplasty or synostosis, 5) skull base surgery, 6) surgery for arteriovenous malformation (AVM), 7) surgery for brain abscess, 8) burr hole procedure, and 9) other types of brain surgery. If a patient underwent cranial neurosurgery more than once (≥ 2) throughout the study period (six years: 2016–2021), only the first one was included in the analysis. Using these inclusion criteria, we aimed to guarantee that recruited patients possessed comparable characteristics, thus fostering homogeneity. These selection criteria were established for the study population to satisfy the hypothesis that all individuals are treated independently in survival analysis using logistic regression. Pediatric patients younger than 18 years were excluded from study participation. The patients were categorized into two groups based on the anesthetic approach used for cranial neurosurgery: TIVA and inhalation anesthesia. Patients who were administered an inhalational anesthetic (e.g., sevoflurane, desflurane, or isoflurane) were assigned to the inhalation anesthesia group; those continuously infused with propofol for anesthesia were assigned to the TIVA group. The prescription information for each anesthetic agent was used to categorize patients into the TIVA and inhalation anesthesia groups. The use of propofol without a prescription of an inhalation anesthetic was considered TIVA. Conversely, a prescription of an inhaled anesthetic was considered inhalation anesthesia. Prescription of only one or two ampoules of an additional 1% propofol in addition to an inhalation anesthetic was considered the use of propofol for induction, and in this case, the patient was assigned to the inhalation anesthesia group.

Herein, the two study endpoints were 90-day mortality and postoperative complications. Ninety-day mortality was defined as death within 90 days that occurred during hospitalization after cranial neurosurgery. Categorization criteria for postoperative complications were based on a previous report [10]: acute coronary events (I21, I22, and I252), heart failure (HF; I50), pulmonary embolism (I26), acute and subacute hepatic failure (K720), acute kidney injury (AKI; N17), sepsis (A40 and A41), wound infection (T793 and T814), pneumonia (J12 to J18, and J69), and urinary tract infection (UTI; N17). Postoperative complications were extracted using ICD-10 codes during hospitalization after cranial neurosurgeries. The hospitalization period also encompassed the patient’s transfer to another medical facility, specifically for rehabilitation and postoperative care.

Demographic information, including age and sex, was collected. Residence, employment status, and household income level were used as covariates to denote the socioeconomic status of patients. Five distinct categories were established to represent household income levels, one of which incorporated medical assistance programs and a four-quartile ratio. The government classifies impoverished individuals who are incapable of remitting insurance premiums as participants in medical aid programs. Urban regions were allocated to include the capital and other relevant communities; the remaining areas were categorized as rural.

To account for patients’ concomitant conditions, the Charlson Comorbidity Index (CCI) and underlying disability were implemented. The CCI scores at the time of hospital admission were computed using ICD-10 codes entered into the NHIS database (Supplementary Table 1).

In South Korea, the registration of all disabilities in the NHIS database is a requirement for eligibility for various benefits offered by social welfare programs. Each disability must be officially diagnosed by a medical professional who evaluates difficulties experienced during the performance of daily tasks. A detailed classification of disabilities is presented in Supplementary Table 2. The patients were allocated to one of the six severity classifications according to the severity of the condition (1: most severe; 6: least severe): grades one to three were labeled as ‘severe,’ and grades four to six as ‘mild to moderate.’ Additionally, the year and type of cranial neurosurgery were used as covariates. Furthermore, based on the anesthesia prescription codes, we extracted cases registered as emergency surgeries separately that we used as a covariate. This is because the degree of urgency of the surgery can affect the patient’s prognosis and severity. Moreover, we calculated annual case volumes (CVs) of cranial neurosurgery in each hospital during the study period and used it as a covariate because higher surgical volume may impact the improved prognosis in patients who underwent cranial neurosurgery [11]. Patients were categorized into four groups using quartiles based on the annual surgical volume of the hospital where the cranial neurosurgery was performed (Q1 < 114, 114 ≤ Q2 < 218, 218 ≤ Q3 < 330, and Q4 ≥ 330).

To reduce bias in this observational study, we performed propensity score (PS) matching to match the clinicopathological features of the TIVA and inhalation anesthesia groups. Using the nearest neighbor method in a 1:1 ratio, without replacement, and with a caliper width of 0.25, we conducted PS matching, typically employed to decrease confounding in observational investigations [12]. PSs were calculated using logistic regression analysis in a logistic model containing all covariates. The absolute standardized mean difference (ASD) was used to compare the balance in the TIVA and inhalation anesthetic groups before and after PS matching. ASDs between the two groups were set to < 0.1 to establish whether the two groups were well-balanced using PS matching.

The clinicopathological features of the two groups were individually analyzed using the t-test for continuous variables and the Chi-squared test for categorical variables.

To determine whether the TIVA group had a distinct risk of 90-day mortality or postoperative complications compared with the inhalation anesthetic group, the PS-matched cohort was subjected to univariable logistic regression analysis. The findings are presented as odds ratios (ORs) with 95% CIs. For sensitivity analyses, we constructed multivariable logistic regression models to examine whether the results obtained in the PS-matched cohort were generalizable to the entire cohort. Using this sensitivity analysis, it is possible to compensate for the fact that PS matching discards a substantial number of samples. Finally, we performed subgroup analyses for the entire cohort according to emergency surgery or non-emergency surgery and each type of cranial neurosurgery to examine whether the different conditions affected the results. The adjustment models incorporated all covariates, and the Hosmer–Lemeshow statistic was employed to validate the model’s adequacy in terms of goodness of fit. There were no concerns regarding multicollinearity among the variables, as the variance inflation factors were all below 2.0. All statistical analyses were conducted using R software (version 4.0.3; R Foundation). The threshold for significance was set at P < 0.05.

Between January 1, 2016, and December 31, 2021 (six years), 173,385 cases of cranial neurosurgery were recorded in South Korea. After eliminating 24,339 cases of multiple (≥ 2) cranial surgeries in a patient to include only the first episode of cranial neurosurgery, we screened 149,046 patients. After excluding 4,540 pediatric patients aged < 18 years, we included 144,506 patients who underwent cranial neurosurgery. A total of 65,442 patients received TIVA (45.3%), while 79,064 received inhalation anesthesia (54.7%). Fig. 1 presents the inclusion of 97,156 patients (48,578 in each group) after PS matching. Table 1 compares the clinicopathological characteristics of the TIVA and inhalation anesthesia groups before and after PS matching. All ASDs of variables between the two groups were < 0.1, indicating that the two groups were well balanced upon PS matching.

Table 2 presents the results of the PS-matched cohort analyses. The 90-day mortality rates after cranial neurosurgery were 14.0% (6,660/48,578) in the TIVA group and 14.2% (6,779/48,578) in inhalation anesthesia group. In the univariable logistic regression analysis, TIVA was not associated with 90-day mortality compared with inhalation anesthesia (OR: 0.97, 95% CI [0.94, 1.01], P = 0.188). Following cranial neurosurgery, postoperative complication rates were 47.1% (22,411/48,578) and 50.3% (23,912/48,578) in the TIVA and inhalation anesthesia groups, respectively. Univariable logistic regression analysis revealed that the TIVA group had a 12% (OR: 0.88, 95% CI [0.86, 0.90], P < 0.001) lower postoperative complication rate than the inhalation anesthesia group.

Table 3 presents the results of multivariable logistic regression analyses of the entire cohort. TIVA was not associated with 90-day mortality compared with inhalation anesthesia (OR: 0.96, 95% CI [0.94, 1.03], P = 0.278; model 1). However, TIVA was associated with a 19% (OR: 0.81, 95% CI [0.79, 0.83], P < 0.001; model 2) lower postoperative complication rate than inhalation anesthesia. All ORs with 95% CIs of covariates for 90-day mortality in multivariable model 1 are presented in Supplementary Table 3.

Table 4 shows the results of subgroup analyses according to emergency surgery and type of cranial neurosurgery. TIVA was not significantly associated with 90-day mortality in either the emergency (P = 0.298) or non-emergency surgery groups (P = 0.102). Nevertheless, TIVA was associated with a 24% (OR: 0.76, 95% CI [0.67, 0.86], P < 0.001) and 18% (OR: 0.82, 95% CI [0.79, 0.84], P < 0.001) lower postoperative complication rate than inhalation anesthesia in the emergency and non-emergency surgery groups, respectively.