Impact of frailty, malnutrition and socioeconomic status on perioperative outcomes

Abstract

Background

Frailty, malnutrition and low socioeconomic status may mutually perpetuate each other in a self-reinforcing and interdependent manner. The intertwined nature of these factors may be overlooked when investigating impacts on perioperative outcomes. This study aimed to investigate the impact of frailty, malnutrition and socioeconomic status on perioperative outcomes.

Methods

A multicentre cohort study involving six Australian tertiary hospitals was undertaken. All consecutive surgical patients who underwent an operation were included. Frailty was defined by the Hospital Frailty Risk Score, malnutrition by the Malnutrition Universal Screening Tool (MUST) and low socioeconomic status by the Index of Relative Socioeconomic Disadvantage. Linear mixed-effects and binary logistic generalised estimated equation models were performed for the outcomes: inpatient mortality, length of stay, 30-day readmission and re-operation.

Results

A total of 21 976 patients were included. After controlling for confounders, malnutrition and socioeconomic status, patients at high risk of frailty have a mean hospital length of stay 3.46 times longer (mean ratio = 3.46; 95% confidence interval (CI): 3.20, 3.73; P value < .001), odds of 30-day readmission 2.4 times higher (odds ratio = 2.40; 95% CI: 2.19, 2.63; P value < .001) and odds of in-hospital mortality 12.89 times greater than patients with low risk of frailty (odds ratio = 12.89; 95% CI: 4.51, 36.69; P value < .001). Elevated MUST scores were also significantly associated with worse outcomes, but to a lesser extent. Socioeconomic status had no association with outcomes.

Conclusion

Perioperative risk evaluation should consider both frailty and malnutrition as separate, significant risk factors. Despite strong causal links with frailty and malnutrition, socioeconomic disadvantage is not associated with worse postoperative outcomes. Additional studies regarding the prospective identification of these patients with implementation of strategies to mitigate frailty and malnutrition and assessment of perioperative risk are required.

Key Points

• Frailty is a critical consideration in the perioperative patient, the impacts of which are disparate from malnutrition.

• Socioeconomic status is unlikely to have perioperative impacts when frailty and malnutrition are considered.

• Frailty, malnutrition and socioeconomic status are seen as components of a complex system that influence health through multiple mechanisms.

Introduction

Frailty, malnutrition and socioeconomic status are all associated with ageing and reported to contribute to adverse perioperative outcomes [1–6]. However, previous studies that have investigated their effects on surgical outcomes infrequently consider the intimate, co-dependent relationship of this triad [7, 8]. As such, it remains unclear how each variable influences perioperative outcomes, a vital consideration for the ageing population.

Frailty is defined as a clinical phenotype of increased vulnerability, secondary to age-related declines in physiological reserve [9]. This clinical phenotype is operationally defined as weight loss, low energy, sedentary lifestyle and reduced functional capacity (slow gait speed, reduced grip strength) [10]. Frailty has also been operationalised as a risk index which considers the accrual of physical and neurocognitive deficits, psychosocial risk factors, socioeconomic risk factors and geriatric syndromes to accurately predict adverse health outcomes [11, 12]. Malnutrition is defined by the alteration of body composition (decreased fat free mass) secondary to reduced nutritional intake [13] and is driven by a variety of aetiologies including chronic disease, malabsorption, acute illness states and chronic starvation without inflammation (e.g. advanced age, low socioeconomic status) [14].

Frailty, malnutrition and socioeconomic status demonstrate a dependent, mutually perpetuating relationship [15, 16]. The changes to body composition from malnutrition may contribute to impairment in physical and mental function, reducing physiological reserve and impairing clinical outcomes from diseases [17, 18]. Malnutrition is independently related to having more frailty characteristics among geriatric outpatients. This association is driven primarily by functional decline and depression, highlighting the need for assessing nutritional status as part of clinical detection and prevention strategies for frailty [19]. Physical activity and diet quality, in turn, are socially distributed, such that those with more advantage are more likely to exercise regularly and have the health literacy and fiscal ability to access high quality nutrition [20, 21]. Socioeconomic factors, including the inability to afford or access adequate nutrition due to low income or lack of transportation, significantly contribute to malnutrition among older people [8].

Given the interdependent nature of frailty, malnutrition and socioeconomic status, evaluation of these factors together may provide insights for both individual patients and healthcare systems. It is possible in studies where malnourishment and socioeconomic status were more significantly associated with poorer outcomes; they may have instead been acting as surrogates for frailty and exhibiting less direct effects on operative outcomes. Both the direct effects of perioperative frailty on operative outcomes and the delineation of its impact from malnutrition and socioeconomic related factors require more attention. Their effects on operative outcomes are pertinent for both elective surgeries, where it poses a prehabilitation target, emergency surgery cases, where an increased need for a higher degree of perioperative support may be required, and in all surgeries, where informed consent can be better delivered.

This study aimed to investigate the individual impact of perioperative frailty, malnutrition and socioeconomic status on postoperative outcomes in a heterogenous surgical cohort.

Methods

Study design, setting and population

This was a multicentre study of prospectively collected data, entered into electronic medical record by healthcare and administrative staff. All consecutive patients above the age of 18, admitted as an inpatient (>24 h) to a surgical unit in one of six tertiary hospitals in Australia, were included in this study. Patients were excluded if they did not receive an operation or no Malnutrition Universal Screening Tool (MUST) was documented. In the Australian healthcare system, tertiary hospitals represent the highest level of care, providing highly specialised services typically not available at primary or secondary medical facilities. These hospitals often serve as the final referral point for patients requiring advanced diagnostics or complex surgical and medical interventions. Patients can be referred to tertiary hospitals by primary or secondary health professionals, or they may be admitted directly through emergency departments in urgent cases. This system ensures that specialised healthcare is accessible for all critical needs, including those identified in our study population. It is within this context that our study was conducted, allowing for the comprehensive evaluation of frailty, malnutrition and socioeconomic factors in a high-acuity setting. In Australia, Aboriginal and Torres Strait Islander populations face significant health disparities, including higher rates of postoperative morbidity, complications and mortality compared to their non-Indigenous counterparts [22]. These disparities are emblematic of broader systemic issues within healthcare systems that affect indigenous populations globally. For instance, life expectancy for Aboriginal and Torres Strait Islander peoples is markedly lower than that for non-Indigenous Australians, reflecting underlying vulnerabilities including a higher prevalence of chronic diseases [23]. The Central Adelaide Local Health Network Human Research Ethics Committee (reference number: 16860) provided ethics approval for this study, with a waiver of individual consent. The Strengthening the Reporting of Observational Studies in Epidemiology guidelines were followed.

Data collection

The assessment of frailty was undertaken with the validated Hospital Frailty Risk Score (HFRS) and analysed as a categorical and continuous variable [11, 24–26]. The HFRS was chosen as the Clinical Frailty Scale can be influenced by the profession of the assessor and degree of acute critical illness, conferring a significant source of bias in this setting [27]. The HFRS, however, can be easily calculated from routinely collected data and can add value to the prognostic studies of hospitalised patients [28, 29]. Furthermore, E46 (unspecific protein-energy malnutrition), although used in an extensive list of a priori factors to identify potentially frail patients, is not included in the final model (Appendix 1).

The MUST, originally introduced by the British Association for Parenteral and Enteral Nutrition, was chosen due to its institutional integration within SA Health and proven validity and superior sensitivity and specificity [30, 31]. It is a simple tool that consists of five questions, relating to patient’s body mass index (BMI), recent weight loss, appetite and the presence of disease. The scale produces a score of 0, 1 and ≥2, denoting a low, medium and high risk of malnutrition, respectively. Socioeconomic status was ascertained by using each patient’s residential postal code to derive a Socio-Economic Index for Areas (SEIFA) in Australia result. The Index of Relative Socioeconomic Disadvantage (IRSD) includes only variables related to relative disadvantage. This index is therefore appropriate for distinguishing between areas with relative disadvantage (lower deciles) and the relative lack of disadvantage of people in an area (upper deciles).

Study endpoints

Inpatient mortality was the primary endpoint. Secondary endpoints included length of stay (LOS) and 30-day readmission.

Statistical analysis

Baseline characteristics were evaluated using descriptive statistics. Spearman’s and Kendall’s Tau methods were both utilised to assess associations between nominal and scale variables. To deduct potential multicollinearity, due to the binary nature of the dependent variable, a correlation matrix was also obtained from the binary logistic regression model. Variance inflation factors, tolerance, eigenvalue and condition index were all employed for multicollinearity diagnostics. Linear mixed-effects models and binary logistic generalised estimated equations (GEEs) were used for continuous and categorical variables, respectively, against predictors. Clustering on site was adjusted for as a random effect. An unadjusted model was performed first, then an adjusted model controlling for patient-related factors: age (continuous), sex (binary), surgery type (cardiac, general surgery (breast endocrine, upper GI, colorectal, hepatobiliary), orthopaedics and other surgeries), POSSUM operative morbidity and mortality risk, cardiovascular diagnosis, preoperative Glasgow Coma Score, heart rate, systolic blood pressure, creatinine, haemoglobin, potassium, sodium, urea and white cell count. Kaplan–Meier curves and log-rank tests were performed on R. The statistical software used was SAS On Demand for Academics (SAS Institute Inc. 2021). P value ≤.05 was considered to be statistically significant.

Results

There were 39 962 eligible patients. A total of 17 986 patients were excluded due to no operation, resulting in a total inclusion of 21 976 patients. Baseline characteristics are detailed in Table 1. There was no multicollinearity between independent variables HFRS, MUST and SEIFA (Appendix 2). Despite the absence of multicollinearity, there was a significant interaction between independent variables (P < .001) where increasing frailty scores was associated with an increased likelihood of malnutrition and, simultaneously, residence in an area of increasing disadvantage. This relationship was reciprocal.

The median LOS was 95 h (interquartile range (IQR) 44–209). There were 172 inpatient mortalities, 4684 30-day readmissions and 3138 patients requiring re-operation. Kaplan–Meier curves of inpatient survival based on frailty (Appendix 3A), malnutrition (Appendix 3B) and SEIFA (Appendix 3C) are shown the appendix.

Frailty

A total of 578 (2.6%) patients were high risk of frailty (HFRS > 15). In general, increasing risk of frailty was associated with being marginally younger, more comorbid, a relative lack of social disadvantage and not indigenous (Table 2).

In an unadjusted linear mixed-effects model, it was found that there was a statistically significant association between frailty (low, intermediate and high) and hospital LOS (P < .001) when controlling for clustering on site as a random effect. In the adjusted model, hospital LOS continued to have a significant association with frailty where patients at high risk of frailty had a mean hospital LOS 3.46 times that of a patient with low frailty risk (mean ratio = 3.46, 95% confidence interval (CI): 3.20–3.73, P < .001). When considered as a continuous variable, the HFRS remained significantly associated with hospital LOS in an adjusted model, where the mean LOS increased by 9% for every 1-unit increase in HFRS score (mean ratio (MR) = 1.09, 95% CI: 1.08–1.09, P < .001).

In the unadjusted binary logistic GEE models, low frailty risk was significantly associated with fewer 30-day readmissions and less inpatient mortality when compared to intermediate- and high-risk patients (Table 3). In the adjusted model, patients with high risk of frailty had odds of 30-day readmission 2.4 times greater than patients at low risk of frailty (odds ratio (OR) = 2.4, 95% CI: 2.19–2.63, P < .001). Patients at high risk of frailty had odds of dying in hospital 12.9 times greater than patients with low risk of frailty (OR = 12.9, 95% CI: 4.51–36.69, P < .001). When considered as a continuous variable, in an adjusted model, for every 1-unit increase in HFRS, the odds of 30-day readmission increases by 6% (OR = 1.06, 95% CI: 1.05–1.07, P < .001) and odds of inpatient mortality increases by 16% (OR = 1.16, 95% CI: 1.12–1.20, P < .001).

Malnutrition

Of the 21 976 included patients, 79.8% had low risk, 8.6% intermediate and 11.5% high risk (Table 2). Malnutrition was significantly associated with age, comorbidity burden, frailty, indigenous identification and socioeconomic status but not sex.

In an unadjusted linear mixed-effects model, it was found that there was a statistically significant association between malnutrition risk (low, intermediate and high) and hospital LOS (h) (P < .001) when controlling for clustering on site as a random effect. In the adjusted model, hospital LOS continued to have a significant association with malnutrition where patients at high risk of malnutrition had a mean hospital LOS 1.29-fold increase to that of a patient with low malnutrition risk (MR = 1.29, 95% CI: 1.25–1.34, P < .001).

In the unadjusted binary logistic GEE models, low malnutrition risk was associated with fewer 30-day readmissions and fewer inpatient mortalities when compared to intermediate-risk (but not high-risk) patients (Table 3). However, this association for 30-day readmissions was found to be insignificant after adjusting for confounders. In the adjusted model, patients with high risk of malnutrition had increased odds of 30-day readmission (OR = 1.29, 95% CI: 1.16–1.42, P < .001). Similarly, patients at high risk of malnutrition had increase odds of inpatient mortality (OR = 1.55, 95% CI: 1.26–1.91, P < .001).

Socio-Economic Index for Australia IRSD

There were 4602 patients in the lowest SEIFA decile and 700 patients in the highest SEIFA decile (Appendix 4). Compared to patients in the highest decile, those in the lowest decile (with relatively more disadvantage) had mean age that was slightly younger (54.7 years (SD = 19.7) vs 58.5 years (SD = 20.8), P < .001), were slightly less comorbid (Charlson Comorbidity Index: 3.2 (SD = 1.6) vs 3.5 (SD = 1.7), P < .001) and more frequently identified as Indigenous Australian (517 (11.2%) vs 3 (0.4%), P < .001). There was no statistical difference in sex, HFRS or MUST score in patients in the lowest or highest decile.

In an unadjusted linear mixed-effects model, SEIFA IRSD deciles were significantly associated with LOS: for every 1-decile increase in SEIFA IRSD, the mean hospital LOS increases by 1% (MR = 1.01, 95% CI: 1.01–1.02, P < .001). This association, however, loses significance in the adjusted model, where for every 1-decile increase in SEIFA IRSD, the mean hospital LOS did not change (MR = 1.00, 95% CI: 1.00–1.01, P = .49).

In both unadjusted and adjusted models, IRSD was not significantly associated with either 30-day readmission or inpatient mortality. In the adjusted models, for every 1-decile increase in SEIFA IRSD, odds of 30-day readmission decreases by 1% (OR = 0.99, 95% CI: 0.98–1.00, P = .29) and inpatient mortality decreased by 2% (OR = 0.98, 95% CI: 0.91–1.07, P = .68). However, when HFRS is considered a continuous variable, IRSD becomes significantly associated with LOS (OR = 1.02, 95% CI: 1.01–1.03, P < .001), 30-day readmission (OR = 1.01, 95% CI: 1.01–1.03, P < .001) and inpatient mortality (OR = 1.11, 95% CI: 1.03–1.20, P < .001).

Discussion

This study aimed to delineate the individual effects of frailty, malnutrition and socioeconomic status on perioperative outcomes. The results demonstrate that frailty and malnutrition, but not socioeconomic status, are intimately associated with worse postoperative outcomes. Frailty (HFRS > 15) was significantly associated with triple the length of stay, double the rates of hospital readmission and a markedly increased (12-fold) odds of inpatient mortality. A high risk of malnutrition (defined as MUST score >/2) was, however, less consequential but still associated with a significantly longer inpatient admission, higher rate of hospital readmission and inpatient mortality. Socioeconomic status also held no association with length of stay, 30-day readmission or inpatient mortality when HFRS is considered a categorical variable as per the original validation. However, when HFRS is considered as a continuous variable, socioeconomic status is significantly associated with longer hospital LOS, more frequent 30-day readmissions and inpatient mortality.

To place our findings within a broader context, it is essential to consider the operational challenges faced by the Australian healthcare system, particularly in relation to increasing LOS, mortality and health inequalities. Recent data indicate that these issues are of significant concern, with disparities in health outcomes and resource utilisation often reflecting underlying socioeconomic and health-related inequities [32]. Our study’s focus on the intertwined nature of frailty, malnutrition and socioeconomic status provides critical insights that could help mitigate these challenges by informing targeted interventions aimed at improving perioperative care and optimising resource allocation. Recognising the separate and combined effects of these factors on perioperative outcomes could therefore play a crucial role in addressing some of the systemic inefficiencies and inequities in healthcare delivery.

As these findings highlight the necessity of incorporating frailty assessments into surgical care pathways, it becomes essential to consider how these evaluations can be most effectively integrated into clinical practice. The HFRS has emerged as a valuable tool for identifying frail patients using routinely collected hospital data. These findings provide foundational support for developing clinical protocols and workflows that incorporate the utilisation of existing International Classification of Diseases, 10th Revision (ICD-10) codes (indexed to previous admissions). The ability to calculate the HFRS using pre-existing inpatient ICD-10 codes offers a significant advantage, enabling clinicians to promptly identify high-risk patients without the immediate need for a detailed clinical review. It lends credence to the idea of implementing artificial intelligence (AI) into clinical workflows, to aid clinicians in undertaking these higher-level tasks in real time [33–36], particularly in settings where swift decision-making is crucial, such as emergency surgeries.

However, it is important to emphasise that the utilisation of the HFRS does not abrogate the necessity for a Comprehensive Geriatric Assessment (CGA). The CGA remains a cornerstone of perioperative care, offering a multidimensional evaluation that encompasses not only frailty but also other critical domains such as functional status, comorbidities, cognitive function and psychosocial factors [37–39]. Given the complementary strengths of both HFRS and CGA, a hybrid approach may represent the optimal strategy for enhancing perioperative care. Furthermore, by integrating AI-driven algorithms, healthcare systems could ensure that frail patients are promptly identified and appropriately managed, even in the most resource-constrained environments. In this model, a HFRS generated by AI at the point of patient contact with the hospital could serve as a preliminary screening tool, efficiently identifying patients at high risk of frailty. These patients could then be prioritised for a subsequent CGA, ensuring that the detailed, individualised insights provided by CGA are reserved for those most likely to benefit from them. This approach allows for the swift identification and intervention required in urgent surgical settings, while also ensuring that the comprehensive evaluation offered by CGA is not overlooked in cases where it is most needed.

Traditionally, frailty assessments have not been validated for individuals <65 years of age; however, our analysis reveals that younger patients, particularly those from more affluent areas who exhibit signs of malnutrition or other risk factors commonly associated with frailty, might also benefit from these evaluations. This study’s inclusion of a younger demographic presenting with characteristics typically associated with frailty offers a unique perspective. This insight is particularly relevant in the context of changing global demographics and the evolving understanding of health and disease in younger populations. Given the significant overlap observed in our study between malnutrition, socioeconomic factors and frailty indicators in younger adults, there is a compelling argument for the need to further research into this area. Future studies should aim to rigorously characterise the ‘younger frail phenotype’ and explore the validity and reliability of frailty metrics in this group. Such research could fundamentally transform the scope of frailty assessments, making them more universally applicable and significantly enhancing preventive healthcare strategies.

This study has several limitations. The MUST score relies on low BMI and weight loss to identify malnutrition and micronutrients are not directly measured. As such, macronutrient excess with a micronutrient deficit, as observed in the overweight-obese malnourished population, would not [40, 41] be identified by the MUST criteria, and in these cases, other assessments like the Patient-Generated Subjective Global Assessment Short Form (PG-SGA SF) may be more appropriate [42]. However, the PG-SGA still does not directly consider any metric of micronutrient deficiencies. Additionally, while HFRS is advantageous for analysing large datasets and does not require direct clinical evaluation, it is important to note that unlike tools that assess physical and cognitive function directly, HFRS may not fully capture the nuanced aspects of an individual’s day-to-day capabilities. This reliance on diagnostic codes rather than functional assessment represents a limitation in directly correlating clinical outcomes with frailty severity. In this cohort, frailer patients tended to, paradoxically, be marginally younger and live in more affluent areas than their less frail counterparts. This likely reflects a multifactorial selection bias where there are less older and frail patients than older and malnourished patients, due to the increased risk of all-cause mortality associated with the former. Additionally, where frailty represents a cumulative deficit model, with reduced physiological reserve, it is possible that frail individuals of advanced age were not offered surgery, prioritising more conservative or comfort measures. This, however, was outside the scope of this study. While the HFRS and the MUST are widely validated tools for assessing frailty and malnutrition, respectively, it is important to note that their validation in specific populations, including Aboriginal and Torres Strait Islander peoples, may be limited. As such, while HFRS and MUST were applied in this study, we recognise the need for further validation studies to ensure that these tools are appropriately sensitive and specific in identifying frailty and malnutrition within the First Nation Australian communities. The distribution of elective versus emergency surgeries in this study reflects an atypical pattern, heavily skewed towards emergency procedures. This abnormality is primarily due to our focus on inpatient admissions. Additionally, the study period overlaps significantly with the COVID-19 pandemic, during which there was a substantial worldwide reduction in elective surgical activity as resources were redirected to manage the pandemic crisis [43, 44]. These factors should be considered when interpreting the findings, as they may limit the generalisability of the results to typical surgical settings outside of a global health emergency.

Conclusion

Frailty and malnutrition are both strongly associated with worse postoperative outcomes. Despite strong causal links with frailty and malnutrition, socioeconomic disadvantage is not associated with worse postoperative outcomes.

Declaration of Conflicts of Interest

None declared.

Declaration of Sources of Funding

None declared.

Additional Discussion (references in main text)

The absence of multicollinearity between frailty, malnutrition and socioeconomic status is desirable for the regression analysis, to permit a clear interpretation of each variable’s effects [45]; however, it does not negate their co-dependence or interrelatedness. Even if factors are not statistically correlated in a way that multicollinearity would suggest, they can still be interrelated through distinct, indirect, physiological or socio-physiological pathways. Adopting a complex systems perspective can help articulate how these variables interact within a broader socio-ecological model [46]. This approach recognises that health outcomes result from a web of interdependencies rather than simple, linear cause–effect relationships [47]. From this viewpoint, frailty, malnutrition and socioeconomic status are seen as components of a complex system that influences health through multiple, overlapping pathways, which may not always be captured by traditional statistical methods aimed at detecting multicollinearity.

The results are both confirmatory, extends and contends on existing knowledge by providing further granularity on the relationship between frailty, malnutrition and socioeconomic status. The results are concordant with previous evidence that both malnutrition and frailty adversely affect outcomes [16, 48]. Our results provide further clarity, contextualising how the stronger adverse effect of malnutrition on perioperative outcomes that is observed in other studies may be driven by non-considered frailty [49–51]. Our results extend on the existing literature to support the assessment of both frailty and malnutrition in isolation, to provide further clarity around surgical risk and enable more transparent informed consent. Furthermore, our observation that socioeconomic status (in isolation, when HFRS categorical) does not impact postoperative outcomes contradicts previous analysis of Queensland hospitals by De Jager et al., which demonstrated disparities in surgical adverse events for patients of low socioeconomic status [52]. This study, however, notably utilised the Index of Economic Resources which summarises variables related to income and wealth. This differs from the IRSD as the IRSD also considers societal factors such as qualifications, chronic health conditions/disabilities, means of private transport and skilled occupations (Appendix 5). Why socioeconomic status in our cohort demonstrated no significant relationship with a categorical HFRS but significance with a continuous HFRS remains unclear. Overall, it highlights how poorly institutions are at capturing and characterising social determinants of health at an individual level—postcode is a crude metric at best, but there are limited alternatives present unless we better capture data at an individual level.

Recent research underscores the relevance of social determinants such as economic stability, education access and community context, which are deeply intertwined with health outcomes, particularly in surgical patients [6, 53, 54]. Our study acknowledges the critical intersection between frailty, malnutrition and various social determinants such as socioeconomic status, food security and community support systems. These factors are increasingly recognised for their significant impact on health outcomes, particularly in surgical patients. For instance, food insecurity is a pertinent issue that directly correlates with malnutrition and indirectly with frailty, given that both conditions are influenced by an individual’s access to adequate nutrition and stable living conditions [55]. In light of this, our findings not only contribute to the existing literature by mapping the direct impacts of frailty and malnutrition on patient health but also highlight the need for healthcare systems to adapt and address these broader social factors.

References