More than half of the patients still suffer from moderate to severe postoperative acute pain, which causes not only suffering and interferes with recovery but also is associated with several adverse outcomes, including delirium, complications of the respiratory and circulatory system, and even the development of persistent chronic pain after surgery. Health services for the aged are nowadays an important public health concern in China since more than twenty million old patients underwent surgeries each year. The fragility of this particular population places the old patients at a disadvantage, susceptible to severe adverse effects due to the inappropriate perioperative pain management, which results in high individual and societal costs. The toll of poorly managed postoperative pain in Chinese aged patients will be very high, and resolving pain should be a moral obligation of all healthcare providers. Hence, more effective and safer management of postoperative acute pain is of high priority in aged patients. Though multiple previous studies have reported correlates of postoperative pain, they focused on the predictors for identifying the patients at risk for moderate to severe postoperative pain but neglected to devise practice patterns that might mitigate this risk. Furthermore, the intensity of postoperative acute pain is diverse among different countries and territories in spite of the comparable common risk factors for postoperative pain, that hints the inherent characters exist in a specific population that would account for the differences in the intensity of postoperative acute pain. Therefore, it is critically needed that to infer a predictive model of postoperative acute pain from a selected data set specific to Chinese aged patients which adopts a hierarchical approach guaranteeing a high priority of the modifiable variables. That could change the concepts of perioperative pain management and provide strategies that improve clinical care of Chinese aged patients with postoperative acute pain.

To identify perioperative modifiable practice patterns that may provide Chinese aged patients with better care by ameliorating acute postoperative pain.

Data for this analysis were collected by PAIN OUT quality improvement network, a multinational perioperative pain registry, providing a standardized methodology to assess perioperative pain management practices and multi-dimensional pain-related patient-reported outcomes (PROs) . The international pain outcomes questionnaire (IPO-Q) that was validated in English and has been translated into Chinese by standardized methodology was used to evaluate the PROs. The PAIN OUT methodology for collecting and evaluating perioperative pain on postoperative day 1 (POD1) is registered at ClinicalTrials.gov (NCT02083835) . Approvals were obtained approval for collecting non-identified patient data by all collaborators from the local ethics committees. A data set specific to Chinese aged patients who underwent orthopedic surgery was selected to reduce the influence of potential confounders in this study. Patients could be eligible if they fulfilled the inclusion criteria as follows: (1) underwent any kind of inpatient orthopedic surgery; (2) 65 years or older; (3) were on POD1 and returned to the ward from the post-anesthesia care unit (PACU) for at least six hours; (4) agreed to take part in the survey. Patients whose data about the primary outcome were missing were excluded from this study. The multiple imputation (MI) technique was used to handle random missing data, by which 5 MI data sets were established for the following analysis. Hence the consequential statistics adopted in this study were pooled ones. The outcome was the presence of moderate to severe pain on POD1, which was according to the "worst pain intensity" derived from PROs. The "worst pain intensity" is scored using an 11-point numerical rating scale (NRS, 0=null, 10=worst possible) . In this study, moderate to severe pain was defined as NRS of "worst pain intensity" was more than 3, and an NRS of 3 or less was considered as identifying patients with mild or no pain. Patient characteristics and perioperative clinical data were used as candidate predictors, which comprised gender, age, body mass index (BMI) , psychiatric comorbidities (consisting of schizophrenia, substance abuse of drugs, alcohol use disorder, current smoker, and affective disorders such as depression, anxiety, phobia, post-traumatic stress disorder, and bipolar disorder) , chronic pain (persistent painful condition for 3 months or more or receipt any opioid before the current admission) , type of surgery, perioperative use of regional anesthetic techniques (composed of peripheral neural blockades and neuraxial nerve blockades alike) , intraoperative administration of nonopioid [such as nonsteroidal anti-inflammatory drugs (NSAIDs) , acetaminophen, ketamine, and et al], wound infiltration with local anesthetics, postoperative administration of cyclooxygenase inhibitors, participation in decisions about pain treatment, and receipt of information about pain treatment options. Multivariate logistic regression in combination with bootstrapping techniques was used to infer a model predictive of postoperative acute pain. Briefly, the association between every candidate predictor and the outcome was estimated first by bivariate analysis using the univariate logistic regression. Then besides the pre-selection based on P values, all candidate predictors that were considered as causal variables based on the clinical logic were included in the multivariable logistic regression using an enter modeling. A hierarchical approach in the modeling was adopted in which modifiable variables were of high priority. Then, candidate parameters were manually deleted one by one according to the P-value of the log-likelihood ratio test. To reduce the influence of potential confounders and hold a better competence in prediction, a more liberal P-value was used in this study to keep parameters in the final model. Ultimately, the multivariable model was reduced manually till only predictors with P <0.10 were included. A Hosmer-Lemeshow test was used to estimate the calibration of the final model, and the area under the receiver operating characteristic curve (AUC) was used to evaluate the discrimination. The internal validation was performed using the bootstrapping technique. The BCa method was conducted to draw random bootstrap samples with replacement (1 000 replications) from the MI data sets by which resampling was stratified. The AUC of the adjusted model derived from the bootstrapped data was expected as the estimate of the performance that the prediction model could gain in similar populations.

From May 2014 to November 2019, 2 244 patients from 13 hospitals were approached, of which 455 patients qualified for analysis. 1 751 patients were younger than 65 years old. Twenty-five patients did not give consent to participating in the survey. Thirteen patients were excluded from this study due to the data missing about the primary outcome. The median age of the patients included in the analysis was 70 (67, 76) years old. 292 (64.2%) patients were female, and 163 (35.8%) patients were male. 259 (56.9%) patients suffered from moderate to severe pain after surgery, and 196 (43.1%) patients reported mild or no pain according to the NRS of the "worst pain intensity" derived from PROs. The final model presented accredited reliability and goodness of fit (Hosmer and Lemeshow test P = 0.215) , which included seven parameters that comprised age (P = 0.091, OR=1.03, 95% CI: 1.00-1.06) , body mass index (P = 0.085, OR=0.95, 95% CI: 0.89-1.01) , chronic pain or receipt of opioid pharmacotherapy before admission (P = 0.011, OR=1.68, 95% CI: 1.12-2.51) , perioperative regional anesthesia (P = 0.097, OR=1.43, 95% CI: 0.94-2.18) , wound infiltration (P = 0.002, OR=0.453, 95% CI: 0.274-0.749) , receipt of NSAIDs pharmacotherapy after surgery (P = 0.008, OR=0.579, 95% CI: 0.387-0.867) , receipt of information about pain treatment (P = 0.003, OR=0.537, 95% CI: 0.359-0.804) . Other predictors consisting of gender (P = 0.358) , psychiatric comorbidities (P = 1.000) , type of surgery (P = 0.174) , intraoperative administration of nonopioid (P = 0.010) , and receipt of information about pain treatment options (P = 0.258) , of which some even seemed relevant in the univariable analysis such as intraoperative administration of nonopioid, were no independent predictors in the final multivariable analysis and excluded from the model. The AUC of the final model was 0.661 (95% CI: 0.611-0.711) . After bootstrapping and adjustment for over-optimism or overfitting, the AUC of the adjusted model was quite slightly decreased to 0.660 (95% CI: 0.638-0.682) . Three modifiable practice variables were demonstrated to be associated with decreasing the presence of moderate to severe pain on POD1 significantly, which consisted of wound infiltration, receipt of NSAIDs pharmacotherapy after surgery, and receipt of information about pain treatment.

n More than half of the Chinese age patients suffer from moderate to severe acute pain after orthopedic surgery, in spite of the progressive understanding of the physiology of postoperative pain and the development of lots of new analgesia procedures. The presented analysis revealed modifications that can be implemented in clinical practice and may improve postoperative acute pain management of the Chinese aged patients who underwent orthopedic surgery. Since the detailed type of surgery did not impact significantly on the outcome of the worst pain intensity on POD1, it might be rational to generalize to the Chinese aged patients undergoing other major procedures, though further studies will be necessary to confirm that.