PPCs are the most common complications in patients undergoing general anesthesia, and according to a review of the literature, the current incidence of PPCs is approximately 40%,17 and the incidence of PPCs in the cardiothoracic and abdominal surgery population is higher (20%-70%) due to geographical differences and disease types.18–20 Critically ill patients often have complex conditions, complex treatment measures, and a higher risk of postoperative complications. The results of this study showed that the incidence of PPCs in critically ill patients was 20.4%. PPCs adversely affect the clinical course in each patient, and if patients at high risk of PPCs can be identified quickly and effectively, clinical care providers can be guided to take targeted preventive measures as early as possible, thereby reducing the occurrence of PPCs, which is conducive to reducing overall mortality and social and family economic burden.
In the present study, LAS VEGAS, ARISCAT, and CHI-BPRI were used to simultaneously predict the risk of PPCs in critically ill patients. Some scales mentioned above included patients' history of respiratory infections as one of the predictive factors. In the LAS VEGAS scale, history of respiratory infection was once considered as one of the predictive factors, but the results of the study showed that this factor was not an independent risk factor for PPCs, which is highly consistent with the present study. The reason why patients with preoperative lung infections or chronic lung diseases were excluded from the present study is it would affect the predictive accuracy of the scale, as these patients may have been diagnosed with pulmonary complications preoperatively.The CHI-BPRI scale mentions postoperative mechanical ventilation as one of the predictive factors; therefore, to minimize bias and to ensure that the study covered all high-risk groups, most of the patients in this study were admitted to the ICU before postoperative extubation, and some were transferred to the ICU after extubation.
As shown in Table 2, LAS VEGAS and ARISCAT performed well in the prediction, but each scale’s predictive value is determined by both sensitivity and specificity, and the best diagnostic cutoff value should be taken when considering the predictive value of the scale. The Youden index was calculated by SPSS software and an Excel sheet; the cutoff point was determined according to the highest value of the Youden index; and the AUC and its difference were applied to compare the results of the ROC curves. The optimal cutoff values provide very reasonable help for clinical staff to predict the risk of PPCs in patients. In this study, the best sensitivity (73.7) and specificity (47) were obtained with a cutoff value of 25 for the LAS VEGAS scale, and the best sensitivity (68.3) and specificity (51.5) were obtained with a cutoff value of 27 for the ARISCAT. For the CHI-BPRI scale, the best sensitivity (51.5) and specificity (61.9) were obtained with a cutoff value of 43. However, the CHI-BPRI score was not statistically significant in predicting the risk of postoperative complications in critically ill patients (P > 0.05), and the efficacy has yet to be studied and will not be discussed at this time. If the scale scores of LAS VEGAS and ARISCAT are greater than 25/27, patients may have a higher risk of PPCs. The sensitivity value of the LAS VEGAS scale at the best cutoff value was higher than that of ARISCAT, indicating a lower rate of missed diagnosis, but the corresponding specificity was lower than that of the other two scales, suggesting that the scale has a larger scope for inclusion in screening high-risk patients and is prone to an increased rate of misdiagnosis.In previous studies by our subject group,21 we have applied the LAS VEGAS Risk Scale to critically ill patients, and found that LAS VEGAS has predictive efficacy for the PPCs. It is suggested that the LAS VEGAS may be selected as a predictive scale for PPCs in the clinic to identify the risk of PPCs at an early stage.
As shown in Table 3, the OR value for predicting postoperative complications in patients based on the LAS VEGAS risk score was 2.258 (1.416, 3.601), and patients with scores greater than 25 were 2.258 times more likely to have postoperative complications than patients with scores less than 25. The OR value for predicting postoperative complications in patients based on the ARISCAT score was 1.387 (0.630,3.056) with a 95% CI containing 1. It cannot yet be stated that patients with scores greater than 27 are more likely to have postoperative complications than those with scores less than 27. The analysis may be because the ARISCAT scores are composed of the patient's preoperative characteristics, such as age, presence of comorbidities, preoperative surgical characteristics, and type and duration of surgery, but they fail to use intraoperative indicators, such as events related to intraoperative ventilation and the use of intraoperative vasoactive drugs. Regardless of the prediction method used, the predictor with higher scores should be considered. In this study, the LAS VEGAS and ARISCAT cutoff values were 25 and 27 points, respectively. In LAS VEGAS, the intraoperative use of supraglottic devices was a protective factor that subtracted 6 points from the total score.In the LAS VEGAS risk score, intraoperative desaturation (SpO2 less than 92% for more than 2 minutes) accounted for 12 points, operative duration ≥ 135 minutes accounted for 11 points, and type of surgery as emergency accounted for 9 points, whereas in the ARISCAT scores, preoperative SpO2 ≤ 90% accounted for 24 points, operative duration greater than or equal to 180 minutes accounted for 23 points, and age ≥ 80 years accounted for 16 points. These scores easily reach the cutoff values of the scale, implying that the level of risk for PPCs will skyrocket if the patient has any of these high-risk factors. Therefore, the patients who have a long operating time, are older, and undergo emergency surgery, may be the patients with a high prevalence of PPCs that we need to be concerned about. In this case, we can promote the use of intraoperative supraglottic devices. In addition, we should pay close attention to the preoperative and intraoperative SpO2 changes in patients to recognize high-risk patients in time.
In scale prediction, the AUC reflects the overall accuracy of the scale prediction, and Hou et al. stated that the larger the value of the area under the AUC is, the higher the diagnostic accuracy.22 The diagnostic accuracy is low when 0.5 < AUC < 0.7, moderate when 0.7 < AUC < 0.9, high when AUC > 0.9, and of no diagnostic significance when AUC is close to 0.5.23 Regarding the AUCs in this study, all the values were between 0.5 and 0.7, but the AUCs for predicting PPCs in critically ill patients were not significantly different (all P > 0.05). This result is in line with the European Society of Anesthesiology clinical trial study and can be considered to indicate some screening value but low diagnostic accuracy regarding the development of PPCs in critically ill patients.24 The AUC of the LAS VEGAS scale for predicting PPCs was 0.78 with a validation subsample of 0.72. In terms of predicting PPCs, Neto et al concluded that the LAS VEGAS score is a simple risk score with moderate discriminatory performance.11 The factors affecting the development of PPCs in critically ill patients are diverse, and the results of this study may be related to other factors, such as individual treatment and disease confounding. The ARISCAT is mainly used in patients undergoing noncardiothoracic surgery and may have low predictive efficacy for the development of PPCs in critically ill patients.25, 26 Factors such as patient affiliation and duration of surgery directly affect the ARISCAT risk scale score, which reduces its sensitivity and specificity. The poor predictive performance of the CHI-BPRI may be related to the fact that the CHI-BPRI originated from a multicenter study of adult patients after noncardiac surgery in China; although there are similarities between the studies in terms of geographic region, ethnicity, and psychosocial background of patients, our study involved critically ill patients from a single center.
The risk factors for the occurrence of PPCs may play different roles in influencing outcomes as follow-up time increases. Previous literature has examined the incidence or mortality of PPCs at 5, 30, or 90 days postoperatively in different populations,11,13,14 and in the present study, due to the uncertainty of the duration of ICU admission of critically ill patients, the follow-up time was not monitored as one of the influencing factors, which may have also influenced the results of the study. Any risk scoring scale is only valid for a specific group of people or a specific period of time, and cannot be a universal tool. It is suggested that healthcare professionals should consider various factors when making clinical decisions in order to take scientific and effective measures to treat and care for patients. Therefore, a new risk assessment system that includes more risk factors and is applicable to Chinese critically ill patients must be developed.