This retrospective cohort study was initially conducted in November 2020 and was approved by The Institutional Ethics Committee of The Faculty of Medicine, Prince of Songkla University, Songkhla, Thailand, on April 23, 2021 (EC 64–080-8–4). The requirement for informed consent was waived by the Human Research Ethics Committee, Faculty of Medicine, Prince of Songkla University, owing to the retrospective nature of the study. The data were accessed on April 24, 2021, after obtaining approval from the Ethics Committee. All data were fully anonymized before being accessed by the investigators.

Participants

The anesthesia database from the Hospital Information System (HIS) was used to identify all parturients who underwent cesarean delivery under spinal anesthesia at Songklanagarind Hospital, Prince of Songkla University, between November 1, 2019, and October 31, 2020. High sensory level cases were searched in the HIS using a quality assurance process by nurse anesthetists. After a high anesthesia sensory level or analgesia level was identified, the researchers (BB and PB) assessed the accuracy of the data by reviewing the electronic anesthetic records (a PDF file scanned into the hospital information system). We excluded twin pregnancies to examine risk factors for singleton pregnancies.

Anesthesia practice and standard operating procedures

The routine anesthesia practices as well as the standard operating procedures for spinal anesthesia at the Songklanagarind Hospital is described. The basic parameters monitored during spinal anesthesia included noninvasive blood pressure, electrocardiography, oxygen saturation by pulse oximetry, urine output, and sensory block height. All spinal anesthesia procedures were performed by a resident anesthetist or by certified anesthesiologist staff. Given that Songklanagarind Hospital is a university hospital, first-year anesthesia residents with less than 1 year of experience were supervised by anesthesiologist staff. Prior to spinal anesthesia induction, 1000 mL isotonic crystalloid was administered to every parturient. Spinal anesthesia procedures were performed with the patient in the lateral decubitus position, and the patient was placed in the supine position immediately after the procedure. Routinely, a 27G Quincke spinal needle was used in every parturient. However, in the case of more than two unsuccessful attempts, a larger-gauge needle was considered. Baseline blood pressure was measured before subarachnoid block administration and subsequently at 1 min intervals for 15 min. Thereafter, it was measured every 3–5 min.

Block height was tested by performers based on the loss of pinprick sensation after the spinal block, and recorded on the anesthetic record form. Data including the number of blocks, spinal block performer, approach to the block, dose of 0.5% hyperbaric bupivacaine, and dose of intrathecal morphine were extracted from anesthetic records. Routine practice entails the use of 0.5% hyperbaric bupivacaine 10–11 mg with intrathecal morphine 0.1–0.2 mg in all parturients, unless the decision is changed by certified anesthesiologist staff. If the parturient received more than one spinal block, the total dose of 0.5% hyperbaric bupivacaine and intrathecal morphine was documented. We derived the total ephedrine dose and total intraoperative crystalloid solution, and estimated the intraoperative blood loss. Fetal birth weight and Apgar scores at 1 and 5 min were also documented. We recorded the intraoperative time (from successful spinal block until the end of the operation), postanesthetic care unit (PACU) time, and block level before discharge from the PACU.

Outcomes of the study

The primary outcome was the incidence of high spinal block. A high spinal block was defined as a loss of pinprick sensation (anesthesia) at T4 or higher or a decrease in pinprick sensation (analgesia) higher than T4, 15 min after spinal anesthesia induction [12]. The secondary outcomes were the possible complications of high spinal block. The possible sequence after high spinal block—hypotension (defined as a decrease in mean arterial pressure by > 30% of baseline within the first 15 min after spinal block) [13], bradycardia (HR < 60 bpm) [14], respiratory distress (SpO2 < 95% on room air), and loss of consciousness—was reviewed. A failed spinal block was defined as one that required general anesthesia with endotracheal intubation [15]. Hypotension was managed by intravenous bolus of isotonic crystalloid and ephedrine 6 mg at the discretion of the anesthesiologist in charge.

Explanatory and potential confounding variables

The explanatory variables collected and used as potential predictors of high spinal sensory block included patient-related risk factors, surgery-related risk factors, and anesthesia-related risk factors. Patient-related risk factors included gestational age (GA), height, pre-pregnancy weight, post-pregnancy weight (weight on the day of cesarean delivery), comorbidities (preeclampsia, eclampsia, and gestational diabetes mellitus), and fetal birth weight. Surgery-related risk factors included emergency surgery and indications for emergency cesarean delivery. Anesthesia-related risk factors included the American Society of Anesthesiologists (ASA) physical status, median/paramedian approach, puncture site, bupivacaine dose, and spinal block performer. The practitioners who induced spinal anesthesia were first-, second-, and third-year anesthesia residents or certified anesthesiologist staff.

Statistical analysis

Descriptive statistics are presented as frequencies with percentages and medians with interquartile ranges. The chi-square test or Fisher’s exact test was used to compare categorical variables, as appropriate. The data were analyzed for normality of distribution using Shapiro–Wilk test. The Continuous variables were compared using Student's t-test for normally distributed data and the Wilcoxon rank-sum test for non-normally distributed data. Collinearity diagnostics and a bivariate correlation matrix were evaluated for each variable. We used the optimal cutoff point on the receiver operating characteristic (ROC) curve to transform the continuous variables to categorical variables. Then, those cutoff points were determined if they were significantly associated with the outcome. All variables with a p-value < 0.2 in the univariate analysis were included in the initial multivariate logistic regression model. Using a backward selection procedure, the final regression model was determined by selecting the model with the lowest Akaike information criterion value at each step, even though some nonsignificant variables remained. Using the Youden index, the optimal cut-off point was derived from the final model. The association of each factor with the outcome was considered statistically significant if the likelihood ratio test p-value was < 0.05. The strengths of the associations are presented using adjusted odds ratios and 95% confidence intervals. Data were analyzed using R version 4.3.1 (R Core Team [2022], Vienna, Austria).

Risk prediction score

The risk prediction score for high spinal block was developed using the coefficients of the significant covariates in the final logistic regression model [16,17,18]. Scores were obtained by multiplying each coefficient by 10 and then rounding the result to the nearest integer. Model discrimination performance was examined using the area under the ROC curve, yielding a sensitivity and specificity based on the optimal cutoff point of the risk score.

Sample size determination

We estimated the prevalence of the primary outcome (high sensory block) in the exposure group (potential predictors) to be 50%, while the prevalence of the outcomes in the non-exposure group was estimated to be 14%, with a ratio of non-exposure to exposure of 5:1 using a significance level of 0.05 within a 95% confidence interval, and a power of 80%. The calculated sample sizes in the exposure and non-exposure groups were 17 and 85, respectively. The prevalence of high spinal block at our hospital was found to be approximately 20% in a retrospective review conducted in 2019. Accordingly, at least 510 participants were required to determine the risk factors. Hence, 1 year of data collection was deemed adequate to recruit 638 participants and compensate 20% for missing data.