The study protocol was approved by the Biomedical Research Ethics Committee of Peking University Third Hospital, Beijing, China (Chairperson Prof Chunli Song) on April, 2022 (2022 [158-02]). Due to the retrospective design and that no patient follow-up was performed, the Ethics Committee agreed to waive the written informed consent from the patients. The investigators who performed the data collection were blinded to the objective of the study and received strict training sessions.
Patient selection
This retrospective cohort study reviewed the electronic medical records of elderly patients (≥65 years of age) who underwent abdominal surgery (including urologic and general surgical procedures) from January 2018 to December 2019, in Peking University Third Hospital. Patients with incomplete or missing peri-operative data were excluded. All personal information on patients was kept confidential.
Measurement of RAI-rev score
RAI-rev score was calculated by evaluating 11 variables derived from the Veterans Affairs or American College of Surgeons National Surgical Quality Improvement Projects (VASQIP/ACS-NSQIP) datasets, i.e., age, sex, cancer, poor appetite, unintentional weight loss, renal failure, congestive heart failure, shortness of breath, residence other than independent living, cognitive decline, and functional status [7-9]. Total score ranges from 0 to 81, with higher scores indicating more severe frailty. Details on the weight of each item are listed in Supplemental Digital Content (SDC) 1. Ifa patient experienced more than one surgical procedure during the hospital stay and had multiple preoperative RAI-rev scores, only the first round of the surgery and the corresponding preoperative RAI-rev score were analyzed.
Other baseline characteristics not covered by the RAI-rev were gathered, including body mass index (BMI), smoking and drinking status, major comorbidities, ASA physical status classification, and main laboratory test results. Intra-operative factors were also extracted and included type of surgery categorized by operative stress [15], urgency status of surgery (emergency or elective), anesthetic methods, duration of surgery, estimated blood loss, and intra-operative blood transfusion. The operative stress levels of surgical procedures were stratified using the Operative Stress Score (OSS), i.e., OSS1, very low stress; OSS 2, low stress; OSS 3, moderate stress; OSS 4, high stress; and OSS 5, very high stress [15].
Postoperative outcomes
The primary outcome was the occurrence of MMM during hospitalization, i.e., grade III or greater complications according to the Clavien-Dindo (CD) scoring system (SDC 2) [16]. For patients with multiple complications, we included the most severe complication for analysis. The diagnostic criteria for major complications are listed in SDC 3. The secondary outcome was the development of life-threatening complications and mortality, i.e., CD IV or greater complications.
Statistical analysis
The baseline and perioperative variables were compared between patients with MMM and those without. Continuous variables were analyzed with the independent samples t-test or Mann-Whitney U test; the Kolmogorov-Smirnov test was performed to check for normality. Categorical variables were analyzed using c2 tests, continuity-corrected c2 tests, or Fisher's exact tests. Time-to-event variables were analyzed with Kaplan-Meier survival analysis ( Log-Rank test).
We used multivariable logistic regression models to investigate the association between RAI-rev scores and outcomes. Peri-operative variables that might be associated with the development of MMM were screened using univariable logistic regression analyses and tested for multicollinearity. Independent variables with P values <0.10 in univariable analyses and those considered clinically significant were entered into a multivariable logistic regression model to identify the association of RAI-rev scores with the MMM risk. Similarly, another multivariable logistic regression model was constructed to investigate the relationship between RAI-rev scores and life-threatening complications and mortality. The 11 variables included in the RAI-rev were not separately enrolled in either univariable or multivariable analyses. The Hosmer-Lemeshow test was used to confirm the goodness of fit of the multivariable logistic regression models.
The predictive performances of RAI-rev scores alone and the combination of RAI-rev scores, ASA classification, operative stress, and urgency status of surgery were assessed using the receiver-operating characteristic (ROC) curve analysis. The area under the curve (AUC) and 95 % confidence interval (CI) were used to assess the discriminative power (ability to classify correctly) of these risk factors for outcomes. The relevant predictive parameters, including sensitivity, specificity, positive and negative predictive values (PPV and NPV), were calculated for different thresholds of RAI-rev scores. For all analyses, two-tailed P values<0.05 were considered significantly statistical. All statistical analyses were performed with the SPSS version 26.0 (IBM Corp., Armonk, NY, USA).
Although the sample size was not estimated in advance, 258 cases of MMM and 16 independent variables included in the corresponding multivariable logistic regression model, as well as 178 cases of life-threatening complications and mortality and 15 independent variables included in the corresponding multivariable model, meet the requirement of the "ten events per variable" rule [17]. Therefore, the sample size (2225) of our study was sufficient and could guarantee the reliability and validity of the regression estimates.