Introduction
Postoperative respiratory function disorders remain one of the most significant problems in anesthesiology and intensive care [1]. Postoperative pulmonary complications (POPC) have a significant impact on perioperative morbidity and mortality, increase the likelihood of re-hospitalization and contribute to a longer hospital stay [2], thereby making a significant contribution to health care costs [3–6]. The frequency of POPC in the general surgical population ranges from 2.0 % to 5.6 %, and during operations on the upper abdominal tract and chest it varies from 20 to 70 % [7–10]. It is important to note that almost 25 % of postoperative deaths occurring in the first week after surgery are associated with POPC [10].
One of the leading places in the structure of POPC is occupied by postoperative pneumonia (PP). Currently, PP accounts for about 50 % of all nosocomial pneumonias, and the frequency of its development ranges from 1.5 to 15.8 % [9, 11–13]. PP adversely affects the outcome of treatment of surgical patients and even threatens their lives. It has been reported that the mortality associated with PP among surgical patients ranges from 9 to 50 %, and its level depends on the type of operation [14]. There is also evidence that it adversely affects the early postoperative recovery of patients and the long-term quality of life. In addition, PP can significantly lengthen the stay of surgical patients in the hospital, increase the frequency of repeated transfers to intensive care units, the frequency of repeated operations and mortality [14, 15], which leads to an increase in medical expenses on average by 2–10 times [12, 15]. Given its prevalence and the increase in the number of surgical procedures, predicting its development is an urgent task that allows taking measures to reduce the frequency of its occurrence by influencing the risk factors of PP that can be corrected, or increasing alertness and conducting more thorough monitoring in patients with conditions that cannot be changed [16, 17].
Preoperative risk assessment requires a structured approach and the use of scales to identify risk factors for the development of PP, which is a significant part of the POPC. These scales include A.M. Arozullah PP risk index [18], P.K. Gupta calculators for PP [19], V. Russotto PP prediction model [14], K. Kawasaki PP prediction nomogram [20], Y. Takesue PP risk calculator [21], H. Baba PP risk factors [22] (appendix).
A.M. Arozullah PP risk index [18], includes the type of surgery, age, functional condition, weight loss, chronic obstructive pulmonary disease (COPD), general anesthesia, sensitivity disorder, cerebral circulation disorder, urea nitrogen level in the blood, blood transfusion, emergency surgery, long-term steroid intake, smoking and alcohol abuse. Patients were divided into five classes using risk index indicators. This scale showed good prognostic significance (AUROC = 0.817), but it was too time-consuming and inconvenient for routine clinical practice.
The P.K. Gupta calculator [19] presents seven preoperative predictors for PP such as age, functional class according to the classification of the American Society of Anesthesiologists (ASA), COPD, functional dependence, preoperative sepsis, smoking before surgery and type of surgery. This scale had good prognostic significance (AUROC = 0.855).
The V. Russotto prediction model [14] identifies five factors independently related to PP: functional dependence, preoperative oxygen saturation of blood (SpO2), intraoperative administration of colloids, intraoperative transfusion of blood preparations and the area of operation. The regression model with five variables was of good predictive value (C-statistics 0.89) and calibration (Hosmer—Lemeshow χ2 = 6.69, p = 0.572).
K. Kawasaki [20] demonstrated that age, male gender, a history of cerebrovascular diseases, Brinkman index ≥ 900 and upper median laparotomy were independent prognostic factors of PP. The nomogram showed good predictive value with a compliance index of 0.77.
Eighteen characteristics, including gender, COPD, sepsis, and functional dependence, six laboratory parameters and two intraoperative factors were used by Y. Takesue [21] to calculate the risk of PP. This scale had good prognostic significance (AUROC = 0.826). However, this model was evaluated only in patients after gastroenterological operations and in the Japanese population and proved to be too time-consuming and inconvenient for routine clinical practice.
In the paper of H. Baba et al. [22] it was shown that significant predictors of postoperative PP were low forced vital capacity and low forced expiratory volume in 1 second, malnutrition (low serum albumin and low nutritional status control indicators and values of the prognostic nutrition index), esophagectomy, surgery of the upper gastrointestinal tract and non-laparoscopic surgery. This scale had satisfactory prognostic significance (AUROC = 0.709). In addition, it was also too time-consuming and inconvenient for routine clinical practice.
In contrast to the developed models for predicting postoperative PP, which have limitations [14, 18–22], POPC prediction models have been developed and confirmed on a large group of patients [23, 24]. However, they predicted all DISEASES, such as respiratory failure, PP, pleural effusion, atelectasis, bronchospasm, pneumothorax, and aspiration pneumonitis, without singling out the risks of PP development separately.
Despite their value, these assessment systems do not provide local specialists with a reliable and permanent method by which to stratify the risk of developing PP in our population.
Purpose of the study
Thus, the main purpose of this work is to develop a model for predicting PP based on the identification of risk factors for its development.
Materials and methods
The data of the STOPRISK study on the perioperative parameters of 6844 patients operated on abdominal and pelvic organs from 32 centers in 21 cities representing 8 federal districts for the period from July 1, 2019 to June 30, 2022 were analyzed [25]. All patients signed a voluntary informed consent to participate in the study.
Criteria for inclusion in the study: patients over 18 years of age undergoing planned surgical interventions on abdominal organs, whose physical status corresponds to classes I–III according to ASA.
Exclusion criteria: acute massive blood loss, aspiration, bronchospasm, anaphylactic reactions, malignant hyperthermia, transurethral and transvaginal operations, operations on peripheral vessels and heart, thoracic operations, neck, head operations, traumatological operations.
Criteria for non-inclusion in the study: lack of informed consent of the patient, inability to assess the factors included in the study (lack of data).
Estimated outcomes
The 30-day mortality and PP were evaluated, according to the definition of the working group of the European Society of Anesthesiology and the European Society for Intensive Care [26]. According to the definitions of 2015, pneumonia is defined by chest X-ray, with at least one of the changes such as infiltration, consolidation, cavity; plus one of the conditions such as presence of fever more than 38 °C for no other reason, the number of leukocytes less than 4 or more than 12 × 109/l; plus at least 2 of the following signs: purulent/altered sputum, increased secretion/aspiration of contents, cough/shortness of breath/tachypnea, wheezing/bronchial breathing or deterioration of gas exchange [26].
All patients included in the study, depending on the presence of PP, were divided into 2 groups: patients with PP (n = 53); patients without PP (n = 6791).
Statistical analysis
Statistical data analysis was performed using MedCalc (MedCalc Software Ltd, Belgium) version 19.1.3.
Data with a normal distribution is presented as an average value ± standard deviation, data with a distribution other than normal is presented as a median (25–75 percentiles).
At the first stage of the study, a comparison was made between the group with PP and without PP of the initial data of patients, as well as factors related to surgery and anesthesia. The Fisher exact test was used to compare qualitative variables, and the Mann—Whitney test was used for quantitative variables. In all cases, the p level of less than 0.05 was considered statistically significant [27].
At the second stage of the study, a logistic regression analysis was carried out to assess the contribution of factors to the development of the outcome (by the method of simultaneous inclusion of independent variables): the odds ratio (OR) and 95 % confidence interval (CI) were estimated. Independent variables were introduced into the model if their statistical significance was revealed during two-dimensional analysis (p < 0.05).
At the third stage of the study, the construction of a model for forecasting PP was conducted based on the data of multidimensional logistic regression analysis. The predictive value of the obtained model was evaluated using ROC analysis and determination of the area under the ROC curve (AUROC). The AUROC result of 0.70–0.79 was considered to have a satisfactory prognostic value, the result of 0.80–0.89 was considered to have a good prognostic value, and the result of 0.9 or more was evaluated as having excellent prognostic value.
At the final stage, the obtained model was compared with the prediction models of PP by other authors found in the world literature. This comparison was conducted by comparing the ROC curves constructed for each model.
Results
When analyzing the frequency of occurrence of PP and 30-day mortality, the following data were obtained. Postoperative pneumonia was found in 53 patients (0.77 %). The fatal outcome occurred in 39 patients. In patients with PP, it was noted in 15 cases (28.3 %), and in patients without PP it was seen in 24 cases (0.4 %) (p < 0.05 according to the exact Fisher criterion). In all cases, acute cardiovascular insufficiency was the cause of death, developing against the background of decompensation of chronic pathology or complications of the postoperative period. Pneumonia, as an independent pathology, was not the direct cause of death in any of the cases.
When comparing the group of patients with PP and without PP, the following data were obtained (Table 1).
| Parameters | Pneumonia | Without pneumonia | p |
|---|---|---|---|
| Gender | |||
| male | 39.6 % | 64.8 % | 0.00024* |
| female | 60.4 % | 35.2 % | |
| BMI | 26.1 (23–31.6) | 26.9 (23.5–30.9) | 0.7652 |
| Age | 63 (53–69.3) | 56 (42–65) | 0.0010# |
| Duration of the operation | 225 (133.8–361.3) | 80 (55–130) | < 0.0001# |
| Operational risk | |||
| low | 3.8 % | 39.1 % | < 0.0001# |
| medium | 54.7 % | 51.8 % | |
| high | 41.5 % | 9.1 % | |
| HT | 75.5 % | 50.1 % | 0.00026* |
| CHD | 43.4 % | 19.1 % | 0.00008* |
| CHF | 45.3 % | 20.1 % | 0.00004* |
| CF | 13.2 % | 6.5 % | 0.083 |
| COPD | 20.8 % | 5.2 % | 0.00008* |
| Smoking | 34 % | 12.1 % | 0.00003* |
| CKD | 9.4 % | 3.6 % | 0.045* |
| ACVA | 5.7 % | 2.2 % | 0.114 |
| Partial and full functional dependence | 11.3 % | 3.3 % | 0.0088* |
| Diabetes mellitus | 20.8 % | 8.8 % | 0.006* |
| Cancer disease | 49.1 % | 22 % | 0.00002* |
| Intake of beta blockers | 34 % | 20.6 % | 0.025* |
| Intake of ACE inhibitors | 54.7 % | 34.2 % | 0.003* |
| Intake of statins | 11.3 % | 9 % | 0.473 |
| Intake of anticoagulants | 24.5 % | 16.9 % | 0.142 |
| Intake of diuretics | 24.5 % | 9.5 % | 0.0016* |
| Intake of bronchodilators | 13.2 % | 1.6 % | 0.000028* |
| Insulin injections | 3.8 % | 1.6 % | 0.204 |
| Intake of oral hypoglycemic drugs | 17 % | 5.8 % | 0.0034* |
| Intake of iron drugs | 13.2 % | 2.6 % | 0.00048* |
| Antibiotic prophylaxis | 88.7 % | 85.6 % | 0.694 |
| ASA Class | |||
| I | 1.9 % | 18.9 % | < 0.0001# |
| II | 17 % | 53.4 % | |
| III | 81.1 % | 27.7 % | |
| Revised cardiac risk index | 1 (1–2) | 0 (0–1) | < 0.0001# |
| Rod test | 33 (24.8–39) | 39 (31–45) | 0.0007# |
| Hemoglobin, g/l | 124 (112.8–134.3) | 133 (123–143) | 0.0001# |
| Type of anesthesia | |||
| combined | 49.1 % | 75.6 % | < 0.0001* |
| neuroaxial | 0 % | 7.6 % | |
| combined | 49.11 % | 11.7 % | |
| total intravenous | 1.9 % | 5.1 % | |
| Intraoperative blood loss, ml | 200 (150–300) | 50 (30–100) | < 0.0001# |
| Infusion rate, ml/min | 11.1 (8–15.2) | 12.5 (8.3–18.8) | 0.1067 |
| Infusion volume | 2500 (2000–3000) | 1000 (750–1500) | < 0.0001# |
| The need for vasopressors | 34 % | 4 % | < 0.000000001* |
| The need for blood transfusions | 20.8 % | 2.2 % | 0.000000025* |
| Decurarization | 5.7 % | 15.5 % | 0.054 |
| Monitoring of neuromuscular conduction | 3.8 % | 12.7 % | 0.059 |
A multidimensional analysis of factors independently related to PP is presented below (Tables 2 and 3). This analysis includes the risk factors responsible for the development of PP related to both the patient's condition and surgical intervention.
Table 2. Multivariate analysis of factors independently associated with postoperative pneumonia
| Factor | Ratio | St. error | Wald Test | p |
|---|---|---|---|---|
| Gender | 0.47513 | 0.29595 | 2.5774 | 0.1084 |
| BMI | − 0.037705 | 0.025856 | 2.1265 | 0.1448 |
| Age | − 0.0096934 | 0.013512 | 0.5146 | 0.4731 |
| Duration of the operation | 0.0062108 | 0.00098144 | 40.0459 | < 0.0001 |
| Operational risk | 0.85190 | 0.25749 | 10.9457 | 0.0009 |
| HT | − 0.47176 | 0.40071 | 1.3860 | 0.2391 |
| CHD | − 0.17143 | 0.37196 | 0.2124 | 0.6449 |
| CHF | 0.53332 | 0.36229 | 2.1670 | 0.1410 |
| Cardiac arrythmia | − 0.055035 | 0.43850 | 0.01575 | 0.9001 |
| COPD | − 0.96695 | 0.36693 | 6.9444 | 0.0084 |
| Smoking | 0.77606 | 0.33465 | 5.3778 | 0.0204 |
| CKD | 0.36335 | 0.48839 | 0.5535 | 0.4569 |
| ACVA | 0.16474 | 0.62281 | 0.06997 | 0.7914 |
| Partial and full functional dependence | 1.30942 | 0.45622 | 8.2377 | 0.0041 |
| Diabetes mellitus | 0.42001 | 0.35807 | 1.3759 | 0.2408 |
| Cancer disease | 0.69374 | 0.29638 | 5.4789 | 0.0192 |
| Intake of beta blockers | − 0.12871 | 0.32112 | 0.1607 | 0.6886 |
| Intake of ACE inhibitors | 0.23609 | 0.30151 | 0.6131 | 0.4336 |
| Intake of statins | − 0.47503 | 0.45517 | 1.0892 | 0.2967 |
| Intake of anticoagulants | − 0.059603 | 0.33649 | 0.03138 | 0.8594 |
| Intake of diuretics | 0.67223 | 0.34786 | 3.7346 | 0.077 |
| Intake of bronchodilators | 1.67761 | 0.59517 | 7.9450 | 0.0048 |
| Insulin injections | 0.63539 | 0.74319 | 0.7309 | 0.3926 |
| Intake of oral hypoglycemic drugs | 0.97194 | 0.37986 | 6.5469 | 0.0105 |
| Iron supplementation in connection with perioperative anemia | 1.69945 | 0.42362 | 16.0938 | 0.0001 |
| Antibiotic prophylaxis | − 0.60409 | 0.46024 | 1.7228 | 0.1893 |
| ASA Class | 1.50175 | 0.23593 | 40.5175 | < 0.0001 |
| Revised cardiac risk index | 0.27811 | 0.15728 | 3.1266 | 0.0490 |
| Rod test | − 0.024284 | 0.013482 | 3.2369 | 0.0497 |
| Hemoglobin | − 0.0068877 | 0.0077703 | 0.7857 | 0.3754 |
| Type of anesthesia | 0.32595 | 0.14137 | 5.3158 | 0.0211 |
| Intraoperative blood loss | 0.00085206 | 0.00020566 | 17.1645 | < 0.0001 |
| Infusion rate | − 0.060128 | 0.022771 | 6.9726 | 0.0083 |
| Infusion volume | 0.00043990 | 0.000073537 | 35.7853 | < 0.0001 |
| The need for vasopressors | 2.16390 | 0.32480 | 44.3861 | < 0.0001 |
| The need for blood transfusions | − 0.91392 | 0.46619 | 3.8432 | 0.0499 |
| Decurarization | − 0.96631 | 0.59866 | 2.6054 | 0.1065 |
| Monitoring of neuromuscular conduction | − 1.02227 | 0.72810 | 1.9713 | 0.1603 |
| Variable | OR | 95% CI | p |
|---|---|---|---|
| Gender | 1.4851 | 0.8264–2.6687 | 0.1084 |
| BMI | 0.9642 | 0.9165–1.0145 | 0.1448 |
| Age | 0.9897 | 0.9637–1.0165 | 0.4731 |
| Duration of the operation | 1.0064 | 1.0045–1.0083 | < 0.0001 |
| Operational risk | 2.2684 | 1.3703–3.7552 | 0.0009 |
| HT | 0.6444 | 0.2931–1.4169 | 0.2391 |
| CHD | 1.0836 | 0.5140–2.2847 | 0.6449 |
| CHF | 1.7046 | 0.8380–3.4674 | 0.1410 |
| Cardiac Arrythmia | 1.1124 | 0.4713–2.6257 | 0.9001 |
| COPD | 2.6299 | 1.2812–5.3986 | 0.0084 |
| Smoking | 2.1729 | 1.1277–4.1870 | 0.0204 |
| CKD | 1.4381 | 0.5522–3.7456 | 0.4569 |
| ACVA | 1.1791 | 0.3479–3.9966 | 0.7914 |
| Partial and full functional dependence | 3.7040 | 1.5147–9.0576 | 0.0041 |
| Diabetes mellitus | 1.5220 | 0.7544–3.0704 | 0.2408 |
| Cancer disease | 2.0012 | 1.1194–3.5774 | 0.0192 |
| Intake of beta blockers | 0.8792 | 0.4686–1.6498 | 0.6886 |
| Intake of ACE inhibitors | 1.2663 | 0.7013–2.2866 | 0.4336 |
| Intake of statins | 0.6219 | 0.2548–1.5176 | 0.2967 |
| Intake of anticoagulants | 0.9421 | 0.4872–1.8220 | 0.8594 |
| Intake of diuretics | 1.9586 | 0.9905–3.8730 | 0.077 |
| Intake of bronchodilators | 5.3528 | 1.6671–17.1870 | 0.0048 |
| Insulin injections | 1.8878 | 0.4399–8.1015 | 0.3926 |
| Intake of oral hypoglycemic drugs | 2.6431 | 1.2554–5.5648 | 0.0105 |
| Iron supplementation in connection with perioperative anemia | 5.4709 | 2.3849–12.5504 | 0.0001 |
| Antibiotic prophylaxis | 0.5466 | 0.2218–1.3471 | 0.1893 |
| ASA Classs | 4.4895 | 2.8273–7.1289 | < 0.0001 |
| Revised сardiac risk index | 1.3206 | 0.9703–1.7975 | 0.0490 |
| Rod test | 0.9760 | 0.9506–1.0021 | 0.0497 |
| Hemoglobin | 0.9931 | 0.9781–1.0084 | 0.3754 |
| Type of anesthesia | 1.3853 | 1.0501–1.8277 | 0.0211 |
| Intraoperative blood loss | 1.0009 | 1.0004–1.0013 | < 0.0001 |
| Infusion rate | 0.9416 | 0.9005–0.9846 | 0.0083 |
| Infusion volume | 1.0004 | 1.0003–1.0006 | < 0.0001 |
| The need for vasopressors | 8.7050 | 4.6057–16.4530 | < 0.0001 |
| The need for blood transfusions | 0.4009 | 0.1608–0.9998 | 0.0499 |
| Decurarization | 0.3805 | 0.1177–1.2301 | 0.1065 |
| Monitoring of neuromuscular conduction | 0.3598 | 0.0864–1.4990 | 0.1603 |
According to the multidimensional logistic analysis, there were selected the variables that were significantly associated with the development of PP (p < 0.05) (Table 4).
| Factor | OR | 95% CI | p |
|---|---|---|---|
| Duration of the operation | 1.0060 | 1.0038–1.0083 | < 0.0001 |
| Smoking | 2.6699 | 1.4220–5.0130 | 0.0022 |
| Full functional dependence | 9.5848 | 1.1072–82.9745 | 0.0401 |
| Intake of bronchodilators | 7.0942 | 2.8467–17.6793 | < 0.0001 |
| Iron supplementation in connection with perioperative anemia | 3.2400 | 1.2971–8.0927 | 0.0118 |
| ASA Class III | 4.1745 | 1.9983–8.7206 | 0.0001 |
| The need for vasopressors | 4.1256 | 2.1220–8.0210 | < 0.0001 |
| High operative risk | 6.5411 | 1.3146–32.5475 | 0.0215 |
| Type of anesthesia | 1.2222 | 0.8833–1.6912 | 0.2258 |
| Cancer disease | 0.4527 | 0.2327–0.8806 | 0.0596 |
| ASA Class I | 0.8443 | 0.1008–7.0750 | 0.8760 |
| COPD | 0.8725 | 0.2906–2.6193 | 0.8079 |
| Intake of oral hypoglycemic drugs | 2.0167 | 0.8890–4.5747 | 0.0933 |
| Revised cardiac risk index | 1.0249 | 0.5067–2.0734 | 0.9454 |
| Infusion rate | 1.0069 | 0.9554–1.0611 | 0.7972 |
| Infusion volume | 1.0002 | 0.9998–1.0005 | 0.3188 |
| Intraoperative blood loss | 0.9993 | 0.9986–1.0001 | 0.0992 |
| The need for blood transfusions | 2.1758 | 0.8238–5.7469 | 0.1167 |
| Rod test | 0.9887 | 0.9625–1.0157 | 0.4089 |
| Partial functional dependence | 1.4545 | 0.6006–3.5221 | 0.4064 |
At the next stage, risk factors with a significant influence on the frequency of its development (p < 0.05) were selected to build a model for predicting the development of PP according to regression analysis (Table 5).
| Factor | OR | 95% CI | p |
|---|---|---|---|
| Duration of the operation | 1.0060 | 1.0038–1.0083 | < 0.0001 |
| Smoking | 2.6699 | 1.4220–5.0130 | 0.0022 |
| Full functional dependence | 9.5848 | 1.1072–82.9745 | 0.0401 |
| Intake of bronchodilators | 7.0942 | 2.8467–17.6793 | < 0.0001 |
| Iron supplementation in connection with perioperative anemia | 3.2400 | 1.2971–8.0927 | 0.0118 |
| ASA Class III | 4.1745 | 1.9983–8.7206 | 0.0001 |
| The need for vasopressors | 4.1256 | 2.1220–8.0210 | < 0.0001 |
| High operative risk | 6.5411 | 1.3146–32.5475 | 0.0215 |
To assess the prognostic significance of the developed forecasting model, a ROC analysis was performed with the determination of the area under the ROC curve (AUROC). The following data were obtained: AUROC = 0.904; st. error — 0.0197; 95% CI 0.897–0.911 (Figure 1).
Fig. 1. Analysis of the ROC curve of the PP forecasting model
According to the analysis of the ROC curve, a cut-off point was determined and groups of high (probability of developing PP more than 1.2 %) and low risk of developing PP (1.2 % or less) were identified (sensitivity 79.2 %, specificity 86.9 %). Retrospectively, considering the obtained model, 933 patients were assigned to the high-risk group for the development of PP, the incidence of PP was 4.5 %. In the low-risk group of PP there were 5911 patients, the incidence of PP was 0.19 % (Table 6).
| Risk of PP developing | Incidence of PP |
|---|---|
| Low risk of PP developing (n = 5911) | 0.19 % |
| High risk of PP developing (n = 933) | 4.5 % |
When conducting ROC analysis with the factors used in the study by P.K. Gupta et al. [19], the following data were obtained in our patient population: AUROC = 0.852; st. error — 0.023; 95% CI 0.843–0.861 (Figure 2).
Fig. 2. ROC curve analysis of the Gupta forecasting model
When comparing the data obtained during the ROC analysis of the developed PP forecasting model and the P.K. Gupta model (Figure 3), the following results were obtained (Tables 7 and 8). As follows from these figures, the developed forecasting model has greater significance compared to the P.K. Gupta model (p < 0.05) (Table 8).
Fig. 3. Comparison of ROC curves of the developed forecasting model and the Gupta model
| Model | AUC | St. error | 95% CI |
|---|---|---|---|
| Gupta | 0.852 | 0.0234 | 0.843–0.860 |
| STOPRISK | 0.904 | 0.0197 | 0.897–0.911 |
| Parameters | Meaning |
|---|---|
| Difference | 0.0527 |
| Standard error | 0.0155 |
| 95% CI | 0.0223–0.0830 |
| z statistics | 3.401 |
| Significance level | p = 0.0007 |
Discussion
In the studied patient population, the incidence of PP was 0.77 %, eight independent variables associated with its development were determined.
The available models for predicting PP have their limitations. A.M. Arozullah et al. [18] retrospectively analyzed data from 100 medical centers performing extensive operations. The main limitation of this study was the low generalizability since the study cohort consisted entirely of male patients. In addition, the retrospective analysis had limitations regarding missing data and potential misclassification of events. Finally, the use of a large database led to the discovery of a number of significant predictors, followed by the development of a scale that is difficult to work with in everyday practice.
P.K. Gupta et al. [19] also used a retrospective database to identify PP risk factors and build a model. Although the authors included a wider range of patient characteristics, this study also has limitations inherent in retrospective analysis.
When conducting ROC analysis with the factors used in the study by P.K. Gupta et al., the following data were obtained in our patient population: AUROC = 0.852; st. error — 0.023; 95% CI 0.843–0.861. The authors of the model in the initial study obtained comparable results: AUROC = 0.855 [19]. The prediction model we developed had statistically significantly greater predictive value compared to the P.K. Gupta model.
Functional dependence, defined as a reduced ability to perform daily activities, is a recognized factor that increases morbidity and mortality [28]. In previously developed models, it was also mentioned as a risk factor for the development of PP [18, 19]. Moreover, patients with poor functional status have an increased risk of developing aspiration pneumonia [29]. They are more likely to live in nursing homes and long-term treatment facilities and are susceptible to colonization and infection with multidrug-resistant pathogens [30].
High operational risk, which characterizes extensive abdominal operations on the upper abdominal tract, was determined as a predictor of PP. Comparable results were also obtained in previously developed models assessing the risk of PP [18, 19]. Laparoscopic abdominal surgeries expose patients to a relatively lower surgical risk, and, when possible, laparoscopic interventions may be considered to reduce the risk of PP. In addition, prolonged surgical intervention (more than 3 hours) is an independent risk factor for POPC and PP [31].
Patients with COPD are more likely to develop PP due to impaired mucociliary clearance. In addition, smoking is also a risk factor for the development of PP [19, 32]. The risk is reduced to a minimum if a patient gives up smoking 6 months prior to surgery, but the increased risk of PP persists for 1 year [15].
Preoperative anemia increases the risk of developing PP, which is consistent with recent studies defining anemia as a predictor of an unfavorable outcome in critical and postoperative patients. Even a minimal degree of anemia is associated with a significant increase in the risk of 30-day postoperative mortality and adverse cardiac events [33], although there is still no clear evidence that preoperative blood transfusion can reduce the risk.
Of course, the use of the PP prediction model in clinical practice is important for identifying patients with a high risk of developing this pathology. This will allow the anesthesiologist-resuscitator to optimize the perioperative management of this category of patients, showing clinical caution [15, 17]. It is necessary to continue clinical studies that contribute to the development of measures aimed at correcting the risk factors included in this model.
The main advantages of our model are the prospective design of the study, a wide range of factors included in the analysis and a large sample of patients who underwent abdominal surgical interventions of varying severity and risk.
Study limitations
Despite the strengths of the study, it has a number of limitations. Thus, one of the limitations of the study is the possibility of a systematic selection error, since the centers were recruited on a voluntary basis, and different centers could provide different levels of medical care. Four years have passed since the initial data were collected, and during this period the tactics of perioperative management of patients could change. In addition, even though the data set was quite complete and included a large number of perioperative variables, some comorbidities, such as obstructive sleep apnea, were not included. This model must undergo internal and external validation to further evaluate its effectiveness and predictive value.
Conclusion
Eight independent variables associated with postoperative pneumonia were identified: duration of surgery, smoking, complete functional dependence, perioperative anemia requiring the use of iron preparations, intraoperative use of vasopressors, ASA functional class 3, intake of bronchodilating drugs for COPD, high operational risk. The model for predicting postoperative pneumonia has excellent prognostic significance (AUROC = 0.904).
Disclosure. A.I Gritsan is the Vice-President of the all-Russian public organization “Federation of anesthesiologists and reanimatologists”; K.M. Lebedinskii is the President of the all-Russian public organization “Federation of anesthesiologists and reanimatologists”; I.B. Zabolotskikh is the First Vice-President of the all-Russian public organization “Federation of anesthesiologists and reanimatologists”. Other authors declare that they have no competing interests.
Author contribution. All authors according to the ICMJE criteria participated in the development of the concept of the article, obtaining and analyzing factual data, writing and editing the text of the article, checking and approving the text of the article.
Registration of the study. The study was registered in the international database https://clinicaltrials.gov under the auspices of the All-Russian Public Organization “Federation of Anesthesiologists and Reanimatologists” (principal investigator I.B. Zabolotskikh), study number NCT03945968.
Ethics approval. Not required.
Funding source. This study was not supported by any external sources of funding.
Data Availability Statement. The data that support the findings of this study are available from the corresponding author upon reasonable request.
