Because of the continuum between different stages of the inflammatory response from SIRS to sepsis and septic shock (20), early diagnosis of postoperative SIRS is critical to initiate timely interventions to prevent septic shock and improve clinical outcomes in elderly patients. Heretofore, the early prediction of postoperative SIRS has been challenging, and there have been no reliable and accurate methods to predict SIRS in the elderly. In this study, we identified 6 feature variables that have strong independent discriminatory power for SIRS with maximal AUC values. Furthermore, we constructed a nomogram composed of the 6 features that had high sensitivity and specificity to distinguish elderly patients at high risk for postoperative SIRS and to alert clinicians to provide early interventions.
Our novel nomogram has important implications for public health policy, clinical practice, and the informed consent process. Firstly, the early identification of elderly patients at risk of progression to postoperative SIRS and the generation of an individualized probability for each patient can be achieved by using our validated prediction tools. Implementation will potentially lead to better management and optimal use of medical resources, and will be critical to improve the survival rates of high-risk patients, especially those in surgical ICUs. Secondly, all of the variables integrated in the predictive model are measured routinely during the perioperative management for elderly surgical patients in China where both the elective patients and emergency patients admitted to the hospital have value of preoperative Albumin. To further facilitate its external validation and application, we have established an online risk calculator (http://wb.aidcloud.cn/zssy/SIRS.html)(19) and it has been accessible for all the peers in daily clinical practice. Moreover, the model can also be easily incorporated into EHR and HIS systems, thus having straightforward applicability, and thus enabling the integration of a risk prediction tool as a clinical decision support aid in perioperative elderly patient care(21). Additionally, our nomogram can be used as a decision support tool in the informed consent process by providing adequate information regarding risks and benefits defined by using a personalized medicine approach to estimate the individualized probability of postoperative SIRS.(22).
To our knowledge, this was the first study to develop a model and construct a quantitative nomogram to predict the probability of postoperative SIRS in aged patients. Because of the critical significance of the early prediction of postoperative SIRS, the identification of predisposing factors is of crucial importance. Several risk factors have been associated with postoperative SIRS including mannose-binding lectin deficiency(23), high levels of circulating GM-CSF + CD4 + T cells(24), bacteriuria and renal stone size (25), diabetes mellitus, and the intraoperative use of an intra-aortic balloon pump(26). However, their predictive values are limited because measurements of most of these parameters are generally not available or easily obtainable in routine testing, or only pertain to particular surgical operations. These limitations preclude their general application to the geriatric population. Fernando et al. developed a model using a Bayesian network approach to predict SIRS in patients admitted to ICU with acute infections (27). In comparison, a nomogram could present a quantitative and practical prediction tool for risk stratification for patients at risk for postoperative SIRS, with the ability to generate a patient-specific numerical probability of a clinical event by integrating diverse prognostic and determinant variables(28, 29).
Wang et al. developed a nomogram to predict SIRS after transrectal ultrasound-guided prostate biopsy (11). However, it was evaluated in a single-center study with a small sample size and without external validation. In the current study, we enrolled 16141 patients aged ≥ 65 years from two medical centers and divided the patients into training and validation cohorts. We trained the logistic regression model on 19 statistically significant features and incorporated 6 comprehensive and easily accessible variables to form our nomogram, which performed well as evidenced by the AUCs of 0.800 and 0.823 in the training and validation cohorts, respectively. Furthermore, the calibration curves of our nomogram demonstrated an agreement between predictions and actual observations.
The 6 easily-accessible features incorporated into our novel predictive model included preoperative fever, preoperative albumin level, ASA classification, intraoperative total infusion volume, surgical duration, and postoperative ICU admission; most of which have been associated with SIRS (30–35). All of the variables were chosen based on the results of resampling methods, previous literature, and clinical experience. Although some of the factors may be very practice dependent, for instance, the total volume of infusion and postoperative ICU admission are strong predictors that heavily weighted in the nomogram, we have tried to validate the utility of this nomogram in another 1105 patients in 2020 and found it performed well in predicting postoperative SIRS and provide clinicians with practical guidelines to facilitate early diagnosis and proactive interventions in elderly patients to avoid worsen complications, especially for the clinicians in surgical ICU and for the aged patient with preoperative fever, a preoperative albumin level < 30 g/L, an ASA classification of III/IV/V, an intraoperative total infusion volume > 2000 ml, a surgical duration of > 200 min, and who is admitted to ICU after surgery (Fig. 3).
Several limitations in this study should be addressed. Firstly, the retrospective study design may be prone to collection and entry bias, as well as residual confounding. Secondly, we did not report the incidence of sepsis in the study since that the key data needed for sepsis-3 criteria to diagnose sepsis is missing due to the retrospective design. Thirdly, as the elderly patients receiving regional anesthesia in our hospital are generally in relatively good conditions and often require a short and minor operation that might have lower risk of postoperative SIRS, we only enrolled the patients with general anesthesia and endotracheal intubation in the study. Fourthly, despite having high sensitivity and specificity to identify elderly patients at high risk for postoperative SIRS, the model may potentially miss important factors that could not be accounted for in the retrospective analysis such as genetic or clinical factors that might be even more important to predict postoperative SIRS(26). Future prospective studies are needed to collect more clinical and genomic information to predict an individual patient's predisposition to SIRS more precisely.