We conducted the analysis using the data from Non-cardiac Operation registry (NoCop) database for our institution. The NoCop registry is a single-center registry with de-identified data for 203,787 consecutive adult patients who underwent non-cardiac surgery, between January 2011 and June 2019. Data were extracted automatically using Clinical Data Warehouse that contains electronic hospital records for more than four million patients, including more than 900 million laboratory findings and 200 million prescriptions. The system can also pull data on mortalities from the National Population Registry of the Korea National Statistical Office using unique personal identification numbers.

Investigators independent of our study and blinded to the results used preoperative evaluation sheets and operation records to organize information about relevant variables. We also used International Classification of Diseases-10 codes to organize preoperative diagnoses. Non-diabetic patients and patients with preoperative diagnoses of delirium or dementia were excluded from this study. We divided our study population into groups and conducted two separate analyses after stratifying patients based on the presence of acute and chronic hyperglycemia. The acute hyperglycemia analyses considered patients with a blood glucose level measured within 24 h before surgery, and the chronic hyperglycemia analyses considered patients with a hemoglobin A_1c (HbA_1c) level measured within three months before surgery.

Acute hyperglycemia was defined as at least one fasting blood glucose level above 140 mg/dl (7.77 mmol/L) or a random blood glucose level above 180 mg/dl (9.99 mmol/L) within 24 h before surgical incision. These levels were set according to the guidelines recommended by the American Diabetes Association and the American Association of Clinical Endocrinology [14]. Chronic hyperglycemia was defined as HbA_1c level above 6.5% within three months before surgery [15]. Following institutional protocol, glucose concentrations were selectively measured during preoperative evaluation for patients with extensive medical histories or undergoing high-risk surgeries. Surgical risk was stratified according to the European Society of Cardiology/European Society of Anesthesiology guidelines on non-cardiac surgery [16]. According to these guidelines, surgeries were categorized into low risk, intermediate risk, and high risk based on the probability of cardiovascular risk occurrence: less than 1% as low risk, 1%–5% as intermediate risk, and over 5% as high risk. Patient physical status was classified by each attending anesthesiologist according to the American Society of Anesthesiologists Physical Status Classification System [17].

The primary endpoint was postoperative delirium, diagnosed by a psychiatrist using the Confusion Assessment Method (CAM), within 30 days after surgery. During postoperative care, patients with acute confusion or behavioral change were examined by the department of neuropsychology at the discretion of the attending clinician. For secondary endpoints, we analyzed mortality during one and three years of follow up.

For baseline characteristics, continuous variables are presented as means ± standard deviation (SD) or medians with interquartile ranges and were compared using a t-test or Mann-Whitney test, as appropriate. Categorical variables are reported as numbers and percentages and were compared using the Chi-square or Fisher’s exact test, as appropriate. Statistical significance was set at P < 0.05. We used inverse probability weighting (IPW) to balance the distribution of all available covariates between the groups while preserving the total number of patients. Specifically, we considered all available potential confounding factors in our analysis to evaluate the impact of hyperglycemia on postoperative delirium such as baseline characteristics (gender, age, and body mass index), and previous medical history such as hypertension, chronic kidney disease, cardiovascular conditions, and chronic obstructive pulmonary disease. We also retained blood test results, electrolyte levels, intraoperative event, and surgical risk. Stabilized weights of these variables inversely proportional to the marginal probability of hyperglycemia were used [18]. Through this method, the weights assigned to patients with higher values were the inverse of the propensity score, while patients with lower values received corresponding weights. The standardized mean difference was calculated to assess the balance of the covariates, and an absolute standardized difference value less than 10% was considered negligible. This diminished the confounding effect of relevant variables while estimating the effect of hyperglycemia on the outcomes. We then compared the risk of outcomes using a Cox proportional hazard regression model to compare the risk of outcomes between the two groups. The results were presented using hazard ratio (HRs) with 95% CIs. Kaplan-Meier curves were generated, and the log-rank test was used to compare the cumulative incidence of delirium. The log-rank test was employed to compare the curves between the hyperglycemia and non-hyperglycemia groups. All statistical analyses were performed using R 4.2.0 (http://www.R-project.org/), and the ‘ipw’ package was employed for IPW adjustment [19].