Introduction
Postoperative pulmonary complications (PPCs) are common in surgical populations and closely related to increased perioperative morbidity and mortality, prolonged hospitalization, elevated healthcare costs, and poor long-term survival [1–3]. Major abdominal surgery, in particular, is associated with a high rate of PPCs ranging from 11% to 45%, depending on the study population and diagnostic criteria [4–6]. PPCs are associated with surgical factors (e.g., procedure-related stress responses, inflammation, and diaphragmatic dysfunction) and patient-level factors (e.g., age-related respiratory system degeneration and accumulated comorbidity burdens) [7–9]. These factors render older patients undergoing major abdominal surgery particularly vulnerable to PPCs. Accurate risk prediction of PPCs is crucial for identifying high-risk individuals who may benefit from interventions, highlighting the need to identify novel predictive markers for PPCs.
Frailty, a geriatric syndrome characterized by impaired stress tolerance due to decreased physiological reserves and accumulated deficits, is increasingly recognized as an important risk factor for poor postsurgical outcomes [10–14], including PPCs [15–17]. Previous studies examining the effects of frailty on PPCs have used physical frailty phenotype-based or questionnaire-based tools to assess frailty [15,16]. However, these assessment tools have limitations and may not apply to older patients in some cases. For example, disability or communication barriers are often observed in older patients; however, physical phenotype-based tools may demonstrate “ceiling effects” under disability contexts [18], and questionnaire-based tools may face additional implementation challenges from communication barriers. We therefore reviewed existing frailty instruments for a practical alternative and identified the five-item modified frailty index (mFI-5), a tool based on the “accumulated deficits” model [19]. The mFI-5 evaluates frailty by measuring five medical and functional deficits and has been validated as a perioperative risk stratification tool in several surgical cohorts [14,19–21]. However, its association with PPCs in elderly individuals remains unclear. Furthermore, whether including frailty can effectively improve the predictive ability of conventional risk factors for PPCs has not been clarified.
Given the high prevalence of PPCs among older patients undergoing major abdominal surgery and the fact that these patients may benefit greatly from preoperative risk stratification for PPCs, our study focused on this surgical population to determine the relationship between preoperative frailty, as measured using the mFI-5, and PPCs. We further explored whether frailty improved the predictive performance of traditional risk factors for PPCs.
Materials and Methods
Study design and ethical approval
The study protocol for this retrospective cohort study was approved by the Biomedical Research Ethics Committee of Peking University First Hospital (approval number: 2023-504-002) and registered on the Chinese Clinical Trial Registry (ChiCTR2400084118). The requirement for written informed consent was waived because of the retrospective study design. This study conformed to the Declaration of Helsinki, 2013.
Study population
This study was conducted at a general tertiary hospital in Beijing, China. Older patients (aged ≥ 65 years) who underwent major abdominal surgery under general anesthesia between January 1, 2018, and December 31, 2021, were screened for eligibility. Major abdominal surgery was defined as a surgical procedure that involved the manipulation of an intra-abdominal organ (including gastrointestinal, urological, or gynecological organs) and was accompanied by a moderate, high, or very high degree of physiological stress as measured by the Operative Stress Score (OSS) rating system of surgery [22]. The OSS is a new surgical risk scoring system that stratifies common surgeries into five risk levels (very low, low, moderate, high, and very high operative stress levels) based on the degree of surgery-related physiological stress. Patients were excluded if they met any of the following criteria: (1) combined thoracic and abdominal surgery such as esophagectomy, (2) organ transplantation, (3) unplanned reoperation for a previous postoperative complication during the same hospitalization, (4) discharge within 24 hours after surgery, and (5) a lack of important data.
Data collection
We reviewed and collected baseline and perioperative data using our institution’s electronic medical record system. To minimize the risk of diagnostic bias, we strictly trained and assigned different researchers to extract the baseline/intraoperative and postoperative data separately. Baseline data included demographic characteristics (age, sex, and body mass index [BMI]), the American Society of Anesthesiologists physical status (ASA-PS), comorbidities (chronic obstructive pulmonary disease [COPD], current pneumonia, asthma, interstitial lung disease, obstructive sleep apnea, hypertension requiring medication, congestive heart failure, coronary artery disease, previous stroke, diabetes mellitus, moderate or severe chronic kidney disease [CKD], chronic hepatic insufficiency, connective tissue disease, and malignant tumor), functional status (independent or dependent), recent unintentional weight loss, respiratory infection in the last month, preoperative oxygen saturation on room air, smoking and drinking status, chronic corticosteroid therapy, and the most recent laboratory test results (hemoglobin and serum albumin, sodium, and potassium levels). Additionally, we calculated the Assess Respiratory Risk in Surgical Patients in Catalonia (ARISCAT) score for each patient. The ARISCAT score is a risk prediction tool extensively used for PPCs derived from the following seven variables: age, preoperative oxygen saturation, respiratory tract infection in the last month, preoperative anemia, location of surgery, operative duration, and emergency surgery [23]. Intraoperative data included the urgency status of the surgery (emergency or non-emergency), location of the surgery (upper or lower abdomen), operative duration, stratification of operative risk (moderate, high, or very high stress) [22], methods of anesthesia, duration of mechanical ventilation, tidal volume, respiratory rate, positive end-expiratory pressure, fraction of inspired oxygen, crystalloid and artificial colloid fluid infusion, urine output, estimated blood loss, and intraoperative blood transfusion.
Frailty assessment
Frailty was assessed using the mFI-5, which is derived from the following five items: COPD or current pneumonia, congestive heart failure, hypertension requiring medication, diabetes mellitus, and dependent functional status. The five parameters were determined according to National Surgical Quality Improvement Program definitions [11]. Each item was assigned the same weight of 1 point, and the mFI-5 score was calculated by adding all the points. The score ranged from 0 to 5, with higher scores indicating increasing frailty [19].
Outcome measurements
The primary outcome was in-hospital PPCs. PPCs were defined according to previously established diagnostic criteria and included at least one of the following conditions during the hospital stay: respiratory infection, respiratory failure, pleural effusion, atelectasis, pneumothorax, bronchospasm, and aspiration pneumonitis (Supplementary Table 1) [23]. PPC events were identified based on clinical symptoms, imaging results, laboratory and physical examinations, and corresponding medical interventions documented in the electronic medical record system. The severity of PPC events was graded based on the Clavien-Dindo classification, and only those classified as grade II or higher were included in the analysis (Supplementary Table 2). For patients who developed multiple eligible PPC events, we only analyzed the most severe event (i.e., the event with the highest Clavien-Dindo grade). If one patient exhibited two or more eligible PPC events with identical highest Clavien-Dindo grades, we incorporated these events into the cumulative event count while ensuring that the patient was not double-counted in the statistical analyses.
The secondary outcomes included postoperative intensive care unit (ICU) admission, all-cause in-hospital mortality, and length of hospital stay.
Statistical analysis
The balance of baseline and perioperative data between patients with and those without PPCs was assessed using the standardized mean difference (SMD), which was calculated as the difference in means, medians, or proportions divided by the pooled standard deviation. Variables with an absolute value of SMD (|SMD|) ≥ 0.10 were considered imbalanced between the two groups.
The association between the mFI-5 score and PPCs was examined using logistic regression models and restricted cubic spline analyses. First, we performed univariate logistic regression analyses to calculate the unadjusted odds ratio (OR) of the mFI-5 score in predicting PPCs and to screen potential risk factors for PPCs. We then conducted multivariate logistic analyses using the Enter method to estimate the adjusted OR of the mFI-5 score for predicting PPCs. To verify the robustness of the association between the mFI-5 score and PPCs, we conducted multiple multivariate analyses across the entire cohort and performed subgroup analyses. Specifically, we established four multivariate logistic regression models: (1) multivariate model 1 was adjusted for the conventional PPC risk index (i.e., the ARISCAT score); (2) multivariate model 2 was adjusted for three baseline characteristics (age, sex, and the ASA-PS); (3) multivariate model 3 was adjusted for baseline (age, sex, and the ASA-PS) and surgical (emergency surgery, location of surgery, and operative duration) characteristics; and (4) multivariate model 4 included all the potential risk factors with P values < 0.10 screened by the univariate analyses. We further conducted subgroup analyses based on sex (male or female) and operative risk level (either moderate stress level or high and very high stress level) separately by employing multivariate logistic analyses to examine the robustness of the results observed in the full model (i.e., multivariate model 4) of the entire cohort. In the above analyses, the mFI-5 score was analyzed as a categorical variable and the five factors covered by the mFI-5 were not analyzed separately. We additionally carried out a multivariate analysis of the individual components of mFI-5 to explore their independent relationship with PPCs by adjusting for confounders included in multivariate model 4. To avoid multicollinearity, we conducted a multicollinearity test on independent variables using variance inflation factor (VIF) filtering and removed the variable(s) with a VIF > 10 from the multivariate analyses. Finally, we employed a multivariate-adjusted restricted cubic spline curve based on the logistic regression model (adjusting for confounders included in multivariate model 4) to visualize the dose-response relationship between the mFI-5 score and PPCs.
We further explored whether adding the mFI-5 score to the basic PPC risk models showed an incremental predictive value for PPCs. The basic risk models included the following: basic model 1 (ARISCAT score), basic model 2 (age, sex, and the ASA-PS), basic model 3 (age, sex, the ASA-PS, emergency surgery, location of surgery, and operative duration), and basic model 4 (all confounders included in multivariate model 4). After integrating the mFI-5 score into the basic models, we estimated improvements in the following: (1) discrimination using the change in areas under receiver operating characteristic curve (AUROCs) with DeLong’s test, (2) calibration using the Hosmer-Lemeshow test, (3) goodness of fit based on the likelihood ratio χ2 test, (4) explained variance (i.e., the explanatory capacity) of the model using the change in the Nagelkerke R2 value, and (5) reclassification using categorical and continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI) [24]. A brief explanation of the reclassification metrics is provided in Supplementary Material 1.
Two-tailed P values < 0.05 were considered statistically significant. Statistical analyses were performed using SPSS (version 25.0; IBM SPSS, Inc.), MedCalc (version 19.05; MedCalc Software Ltd.), and Python 3.11.5 (https://www.python.org/).
Results
Of the 4185 elderly patients screened for eligibility, 887 were excluded for the following reasons: undergoing combined thoracic and abdominal surgeries (n = 163) or unplanned reoperations (n = 56), discharge within 24 h after surgery (n = 34), or a lack of important data (n = 678). Among the excluded patients, 14 met both the exclusion criteria of combined thoracic and abdominal surgery and unplanned reoperation and 30 met both the exclusion criteria of discharge within 24 hours after surgery and lack of important data. The remaining 3298 patients were included in the analysis (Fig. 1).
Clinical characteristics
A total of 351 patients (10.6%) developed PPCs during their hospital stay, including 164 with respiratory infections, 95 with respiratory failure, 68 with pleural effusions, 55 with atelectasis, 4 with pneumothorax, 13 with bronchospasm, and 2 with aspiration pneumonitis (Table 1). The Clavien-Dindo classification of these pulmonary complications is detailed in Supplementary Table 3.
The median (Q1, Q3) age of the study population was 72 (68, 78) years and 52.2% of the patients were male. A total of 960 (29.1%), 1,358 (41.2%), 784 (23.8%), 166 (5.0%), and 30 (0.9%) patients had mFI-5 scores of 0, 1, 2, 3, and 4, respectively. Patients with PPCs had higher mFI-5 scores than those without PPCs, and each component of the mFI-5 was more frequently observed in patients with PPCs (all |SMD| ≥ 0.10; Table 2). Imbalances in age, BMI, ASA-PS, ARISCAT scores, prevalence of obstructive sleep apnea, previous stroke, moderate or severe CKD, chronic hepatic insufficiency, recent unintentional weight loss, low baseline blood oxygen saturation, current smoker status, anemia, hypoalbuminemia, and hyponatremia (all |SMD| ≥ 0.10; Table 2) between the two groups were also observed. During surgery, the two groups showed differences in the proportion of emergency surgeries, surgical location, operative duration, stratification of operative risk, duration of mechanical ventilation, intraoperative crystalloid fluid infusions, estimated blood loss, and blood transfusions. After surgery, patients with PPCs were observed to have a higher ICU admission rate, elevated all-cause in-hospital mortality, and prolonged hospital stay (all |SMD| ≥ 0.10; Table 3). Other baseline and perioperative data are presented in Tables 2 and 3 and Supplementary Table 4.
Association between the mFI-5 score and PPCs
In the univariate logistic analysis, an mFI-5 score ≥ 1 compared with a score of 0, was separately associated with an increased risk of PPCs (all P < 0.001; Table 4). After adjusting for confounding factors, higher mFI-5 scores remained robustly correlated with an elevated risk of PPCs in comparison with a score of 0 (all P < 0.05; Table 4 and Supplementary Table 5).
In the female subgroup, an mFI-5 score ≥ 2 compared to a score of 0 was independently associated with an increased risk of PPCs (all P < 0.001). Similar findings were observed in the moderate stress level surgery subgroup (all P < 0.001). In the other subgroups, the results were consistent with those of the entire cohort (Supplementary Table 6).
Among the five items of the mFI-5, COPD or current pneumonia (OR: 2.518, 95% CI [1.529–4.146], P < 0.001), diabetes mellitus (OR: 1.353, 95% CI [1.018–1.799], P = 0.037), and dependent functional status (OR: 3.865, 95% CI [2.911–5.131], P < 0.001), not including congestive heart failure (OR: 1.387, 95% CI [0.595–3.238], P = 0.449) and hypertension (OR: 0.966, 95% CI [0.747–1.249], P = 0.793), were independently associated with PPCs (Supplementary Table 7).
We further performed a multivariate-adjusted restricted cubic spline analysis to explore the dose-response relationship between the mFI-5 score and the OR of PPCs. As shown in Fig. 2, the OR of PPCs increased with an increase in the mFI-5 score (P for nonlinearity = 0.7337).
Additive value of the mFI-5 score in PPC risk prediction
Adding the mFI-5 score significantly increased the discriminative power of each basic PPC risk model. Specifically, whereas basic model 2 demonstrated weak discrimination for PPCs (AUROC: 0.676, 95% CI [0.660–0.692]), after adding the mFI-5 score, discrimination for PPCs was fair (AUROC: 0.741, 95% CI [0.725–0.756]; DeLong’s test: P < 0.001). Similarly, whereas basic model 4 exhibited fair discrimination for PPCs (AUROC: 0.782, 95% CI [0.768–0.796]), after incorporating the mFI-5 score, discrimination for PPC was classified as good (AUROC: 0.812, 95% CI [0.799–0.826]; DeLong’s test: P < 0.001) (Table 5).
The incorporation of the mFI-5 score also enhanced or maintained the calibration of the basic models. In basic model 1, we observed poor calibration (Hosmer-Lemeshow test: P = 0.001); however, after incorporating the mFI-5 score, the model was well calibrated (Hosmer-Lemeshow test: P = 0.649). The mFI-5 score also improved the goodness of fit (likelihood ratio χ2 test: all P < 0.001) and the explained variance for each model (Table 5).
Reclassification analysis showed that the categorical NRI values in those patients with and those without PPCs were all > 0 after the mFI-5 score was added to the four basic models (Supplementary Table 8). Overall, the total categorical NRI showed that adding the mFI-5 score into basic models 1, 2, 3, and 4 separately improved the risk reclassification of PPCs by 24.9% (95% CI [17.4%–32.3%]), 22.1% (95% CI [15.0%–29.2%]), 16.4% (95% CI [9.9%–23.0%]), and 13.2% (95% CI [7.1%–19.2%]), respectively (all P < 0.001; Table 6). Continuous NRI analysis demonstrated that integrating the mFI-5 score into basic models 1, 2, 3, and 4 separately showed an improved reclassification of 69.3% (95% CI [58.6%–80.0%]), 50.0% (95% CI [39.1%–60.9%]), 50.8% (95% CI [40.0%–61.7%]), and 63.9% (95% CI [53.1%–74.6%]) of the population, respectively (all P < 0.001; Table 6). The IDI analysis showed that all IDI values were above 0, further confirming the additive prognostic value of the mFI-5 score for PPCs beyond that of the basic models (all P < 0.001; Table 6).
Discussion
In this retrospective study investigating elderly patients who underwent major abdominal surgery, our primary analyses determined that preoperative frailty, identified by an mFI-5 score ≥ 1, was independently associated with an increased risk of PPCs. We further found that incorporating the mFI-5 score into the basic PPC risk models significantly improved the predictive performance.
In the current study, the incidence of PPCs was 10.6%, which is slightly lower than previously reported PPC rates (11%–45%) [4–6]. This discrepancy may be explained by a few factors. First, we focused on in-hospital PPCs rather than 30-day PPCs, as reported in other studies [5,6], resulting in a lower rate of PPCs. Second, we only analyzed PPCs classified as Clavien-Dindo grade II or higher and did not cover all PPC events, which might have further underestimated the rate of PPCs. Third, severely frail patients with a high risk of PPCs are typically excluded as candidates for major abdominal surgeries during preoperative decision-making. As observed in this study, only 30 patients had an mFI-5 score of 4, and none had a score of 5 (i.e., most severe level of frailty). Finally, considering the retrospective nature of this study, some PPC events may have been missed due to incomplete medical records.
Frailty, considered “the most problematic expression of population ageing” [25], has emerged as an important risk factor for healthcare outcomes beyond geriatric medicine [26]. The relationship between frailty and PPCs has been examined in several surgical settings. For example, Fan et al. [15] found that the risk of PPCs in frail patients (defined as a Comprehensive Assessment of Frailty [CAF] scale score ≥ 11) was approximately four times that in non-frail patients after cardiac surgery. In another cohort study, Chen et al. [16] reported that preoperative frailty (defined as a Fatigue, Resistance, Ambulation, Illnesses, and Loss of weight [FRAIL] scale score ≥ 3) was associated with a 6-fold increase in PPC risk in older patients undergoing thoracoscopic pulmonary resections. Notably, these studies measured frailty using tools based on physical frailty phenotypes or self-reported questionnaires. Although the physical frailty phenotype is well studied, it incorporates performance-based grip strength and gait speed criteria that make its application in clinical practice challenging as clinicians may lack the time and busy perioperative settings may have limited space or lack the availability of dynamometers [27]. Furthermore, disability status may weaken the accuracy and reliability of the assessment results due to “ceiling effects” [18]. Self-reported questionnaire instruments may also present issues such as recall bias, communication barriers, or gender bias (e.g., males may be less “willing” to report difficulty in physical function than females) [28].
Given the above, we chose to use an accumulated deficits frailty model and selected the mFI-5, which is derived from the original Canadian Study of Health and Aging Frailty Index (CSHA-FI) [19,29]. In addition to its clinician-friendly accessibility (readily captured in < 1 min without special equipment or trained personnel during routine clinical practice) and patient-friendly simplicity (the total score directly reflects the number of potential risk factors and severity of frailty), the mFI-5 score particularly appealed to us because most of its components are potential risk factors for PPCs. Hence, we reasonably inferred that the mFI-5 may be useful for predicting PPCs. More importantly, to the best of our knowledge, no previous study has specifically explored the relationship between the mFI-5 score and PPCs in older surgical patients. In the multivariate logistic analyses and restricted cubic spline analysis across the entire cohort, we revealed that an mFI-5 score ≥ 1 was independently correlated with an increased risk of PPCs in a dose-response manner compared to a score of 0. Despite the null association between an mFI-5 score of 1 and PPCs in the female cohort and moderate operative stress level subgroup, the consistent directional trend between the primary analysis and subgroup analyses provided robust support for the positive association between frailty and PPCs in older patients (i.e., accumulating health deficits confer increasing adverse effects on the postoperative recovery of respiratory function). According to these findings, assessing frailty using the mFI-5 could help clinicians identify high-risk patients and aid in clinical decision-making before surgery, allowing for designing individualized perioperative management strategies for older patients to improve their postoperative outcomes.
Among the five components of the mFI-5, we found that congestive heart failure was not independently related to PPCs, which seems to contradict clinical practice. This phenomenon may have a few explanations. First, the low number of patients with chronic heart failure (n = 41) in this cohort might have reduced statistical power and masked the true effect of chronic heart failure on PPCs, thereby increasing the risk of type II errors. Second, by carefully reviewing the medical records of these patients, we found that almost all the surgical procedures were strategically scheduled during the remission stage of chronic heart failure. Third, these patients had access to targeted care management during hospitalization, including preoperative cardiology consultation and, if necessary, cardiopulmonary function optimization, close cardiovascular and respiratory monitoring and aggressive management during surgery, and postoperative ICU admission, all of which likely further reduced the occurrence of PPCs. Similar findings are clinically plausible regarding hypertension. However, hypertension may theoretically be involved in the development of PPCs through mechanisms such as cardiac dysfunction, vascular endothelial injury [30,31], and inflammatory responses [32]. This theoretical basis, combined with our main findings that a greater deficit burden is associated with an elevated risk of PPCs, may lead to a hypothesis that other coexisting health deficits, such as COPD, diabetes mellitus, or functional impairment, exert a potential interactive effect on the association between hypertension and PPCs. Further studies are warranted to explore this hypothesis and elucidate the potential pathophysiological interaction mechanisms and clinical implications of our findings.
To the best of our knowledge, this is the first study to evaluate the additive predictive value of the mFI-5 score for PPC risk prediction. Based on our results, adding the mFI-5 score to the basic PPC risk models led to notable improvements in discrimination. Among the existing PPC prediction tools [23,33–35], the ARISCAT score is widely used and has demonstrated good predictive performance in Western European cohorts [23,36]. However, its discriminative ability was found to be inadequate in older Asian surgical populations [37], as observed in our cohort. After integrating the mFI-5 score, the discrimination of the ARISCAT significantly improved, which may be used to help clinicians better stratify the risk of PPCs before surgery. The mFI-5 also considerably improved the AUROC of the full model (i.e., basic model 4) such that it exceeded 0.8, which is considered the threshold for strong discrimination in clinical practice [38]. As simplified PPC risk models, basic models 2 and 3 demonstrated improved discrimination after incorporating the mFI-5 score. Particularly in basic model 2, the AUROC increased from 0.68 to 0.74 (an AUROC > 0.7 is considered potentially useful for clinical decision-making) [38], further suggesting the meaningful clinical utility of the mFI-5 score. In busy perioperative contexts, combining such reduced models with the mFI-5 score may serve as a feasible and efficient approach for the real-time screening of PPC risk.
Based on our results, adding the mFI-5 score improved or maintained the calibration of the basic PPC risk models. Calibration reflects the degree to which a model correctly predicts the absolute risk and is critical for clinical decision-making [39]. Our findings indicate that measuring preoperative frailty using the mFI-5 score may enable clinicians to accurately estimate the risk of PPCs, thereby improving clinical decision-making. Although the Nagelkerke R2 value is controversial as a measure of the proportion of explained variation in logistic regression (unlike linear regression) [40], the increase in Nagelkerke R2 value still provides meaningful evidence for the predictive role of the mFI-5 score.
According to the categorical NRI analysis results, some patients’ PPC risk stratifications meaningfully changed after the mFI-5 score was integrated, with improvements observed among cases reclassified into higher-risk categories and non-cases reclassified into lower-risk categories. Accordingly, these correctly reclassified patients may be offered intervention programs that correspond to their actual risk and achieve better postoperative outcomes. The continuous NRI and IDI analysis results further confirmed the additive predictive value of the mFI-5 score. These findings suggest that integrating the mFI-5 score with conventional PPC risk factors can provide more accurate prognostic information and help guide clinicians to provide more effective care for high-risk patients.
Considering our results, we recommend conducting a frailty assessment using the mFI-5 before surgery to identify older patients at high risk for PPCs. Once a frail patient is identified, tailored interventions targeting the modifiable components of frailty can be implemented to improve the patient’s health status to withstand surgical stress, such as preoperative medication optimization to ameliorate relevant comorbidities and functional rehabilitation exercises to improve functional status. To determine whether these interventions reduce the risk of PPCs requires further investigation.
In addition to its retrospective nature, our study had a few limitations. First, patients with the most severe frailty (i.e., mFI-5 score = 5) were not included in our cohort because they were excluded as candidates for major abdominal surgery. This may have introduced a selection bias, potentially resulting in an underestimation of the impact of frailty on PPCs. Second, important data related to PPCs, such as preoperative pulmonary function test results and postoperative pain information, were not acquired due to the lack of documentation in most patients’ medical records, which might have confounded the effects of frailty on PPCs. Finally, our study was based on clinical datasets from 2018 to 2021; thus, any subsequent changes in treatment protocols or revisions to guidelines may affect the generalizability of our findings to current clinical settings. Additionally, the Coronavirus Disease 2019 (COVID-19) pandemic’s disruption of routine healthcare services during the observation period might have introduced unmeasured confounding factors into the results. Therefore, caution should be exercised when extrapolating these results to post-pandemic healthcare settings without external validation using more recent data. Given the above limitations and our single-center retrospective design, multicenter prospective cohort or external validation studies are warranted to further validate our findings. Despite these limitations, our results demonstrate clinical significance and provide insights for further investigation.
In conclusion, our study demonstrated that frailty, as measured using the mFI-5 score, was independently associated with an increased risk of PPCs in older patients undergoing major abdominal surgery. Integrating the mFI-5 score into conventional PPC risk factors significantly improved the predictive performance. Based on our findings, preoperative frailty assessment should be considered in perioperative settings. Further studies are warranted to investigate whether preoperative frailty interventions reduce the risk of PPCs in older patients.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
