Impact of Frailty on Outcomes in Surgical Patients: A Systematic Review and Meta-analysis

Abstract

Importance:

Age has historically been used to predict negative post-surgical outcomes. The concept of frailty was introduced to explain the discrepancies that exist between patients’ chronological and physiological age. The efficacy of the modified frailty index (mFI) to predict surgical risk is not clear.

Objective:

We sought to synthesize the current literature to quantify the impact of frailty as a prognostic indicator across all surgical specialties.

Data Sources:

Pubmed and Cochrane databases were screened from inception to 1 January 2018.

Study Selection:

Studies utilizing the modified Frailty Index (mFI) as a post-operative indicator of any type of surgery. The mFI was selected based on a preliminary search showing it to be the most commonly applied index in surgical cohorts.

Data Extraction and Synthesis:

Articles were selected via a two-stage process undertaken by two reviewers (AP and DS). Statistical analysis was performed in Revman (Review manager V5.3). The random-effects model was used to calculate the Risk Ratios (RR).

Main Outcome(s) and Measure(s):

The primary outcomes: post-operative complications, re-admission, reoperation, discharge to a skilled care facility, and mortality.

Results:

This meta-analysis of 16 studies randomizes 683,487 patients, 444,885 frail, from gastrointestinal, vascular, orthopedic, urogenital, head and neck, emergency, neurological, oncological, cardiothoracic, as well as general surgery cohorts. Frail patients were more likely to experience complications (RR 1.48, 95%CI 1.35–1.61; p<0.001), major complications (RR 2.03, 95%CI 1.26–3.29; p=0.004), and wound complications (RR 1.52, 95%CI 1.47–1.57; p<0.001). Furthermore, frail patients had higher risk of readmission (RR 1.61, 95%CI 1.44–1.80; p<0.001) and discharge to skilled care (RR 2.15, 95%CI 1.92–2.40; p<0.001). Notably, the risk of mortality was 4.19 times more likely in frail patients (95% CI 2.96—5.92; p< 0.001).

Conclusions and Relevance:

This study is the first to synthesize the evidence across multiple surgical specialties and demonstrates that the mFI is an underappreciated prognostic indicator that strongly correlates with the risk of post-surgical morbidity and mortality. This supports that formal incorporation of pre-operative frailty assessment improves surgical decision-making.

Introduction

The US population is aging with those over 65 years old projected to reach 88 million by 2050, representing a 105 percent increase from 2015.¹ As the elderly patient demographic grows, surgeons will more commonly operate on these patients, with studies showing that up to 30 percent of invasive procedures performed in Medicare beneficiaries occurring in the final year of life.² Although chronological age has historically been used to predict poor post-surgical outcomes, in recent years, the concept of frailty was introduced to reconcile the mismatch between a patient’s chronological and physiological age.

The precise definition of frailty is an evolving one, but is generally recognized as a state of decreased physiologic reserve coupled with an increased vulnerability to stressors which leads to higher susceptibility to adverse health outcomes, disability, and death.³ In general, frail individuals are those clinically characterized by an impairment in their nutrition, endurance, mobility, physical strength and muscle power, balance, and cognitive function.^4,5

A frailty model, termed the Frailty Index (FI), proposed by Rockwood et al. supports that frailty is a multidimensional risk state characterized by the accumulation of various health-related deficits, whose severity is dependent on the number rather than the nature of each individual health deficit. ^6,7 The model’s variables capture multiple domains of health including cognitive dysfunction, functional limitation, psychosocial risk factors, geriatric syndromes, and medical co-morbidities.^8,9.10

In 2011 Obeid et al. mapped the variables included in the FI against the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database to propose the modified frailty index (mFI).¹¹ The mFI consists of 11 variables (Figure 1). Each of these criteria corresponds to one point. The index is then calculated by adding all the applicable points and dividing by the total, that is 11. The majority of studies analyzing the impact of frailty in post-surgical outcomes utilize the mFI because it is based on easily identifiable patient characteristics that are relevant to surgical patients and which can be extracted using simple history taking and physical examination.

Fig. 1.

The eleven mFI variables.

Studies have supported that frailty indices are an efficient and accurate way of identifying patients who are at higher risk of negative post-operative outcomes.¹² Recognizing the necessity for frailty assessment in surgical care, in 2012 the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) and the American Geriatrics Society (AGS) conjointly proposed best practice preoperative guidelines that recommended formal assessment of frailty.¹³

In this review and meta-analysis we seek to summarize the currently available literature in order to quantify the strength of the mFI, and indirectly frailty, as a prognostic indicator of postoperative outcomes across all surgical specialties.

Patients and Methods

Selection Criteria

The Cochrane Handbook for Systematic Reviews of Interventions version 5.1.0 was consulted to establish the review strategy.¹⁴ The review was registered on the Research Registry UIN: reviewregistry472 (http://www.researchregistry.com). The PUBMED and Cochrane electronic databases were screened from inception to 1 January 2018 using the keywords: frailty, frail, mFI, Modified Frailty Index, surgery, and surgical (Supplemental Table 1). The syntax of each database determined the search format.

The inclusion criteria were limited to any prospective or retrospective studies including case-control studies, cohort studies, case series, and randomized controlled trials, reporting on surgical outcomes in frail patients, classified as such using the mFI. This index was selected as it is the most commonly investigated frailty index in surgical populations, enabling us to include the highest number of articles in our review. All surgical interventions and specialties were considered. Inclusion of generic terms like surgery increases the likelihood of the search strategy being comprehensive (Supplemental Table 2).

Articles were selected via a two-stage process undertaken by two reviewers (AP and DS). All data of identified studies were extracted into Microsoft Excel® 2017 (Microsoft, Redmond, WA, USA), followed by screening of the title and abstracts to select eligible studies. In the second stage the manuscript of eligible studies was assessed for inclusion. The last author was consulted to resolved inconsistency between the two reviewers (IS).

Quality Assessment

The methodological quality of the studies was independently assessed by two reviewers (DS, YE) using the Grading of Recommendation Assessment, Development and Evaluation (GRADE) guidelines.¹⁵

Statistical Analysis

The Cochrane Collaboration and the Quality of Reporting of Meta-analyses (QUOROM) guidelines were consulted and the meta-analysis was performed in RevMan (Review manager V5.3).¹⁶ The random-effects model was used to calculate the Risk Ratios (RR). Statistical heterogeneity was quantified using the I² and X² statistics with their corresponding P-values. Publication bias was assessed using funnel plots.

Subgroup Analysis

Subgroup analyses were performed to investigate the difference in complication occurrence and mortality in non-frail and frail patients with mFI scores of 0.09, 0.18, 0.27, and >0.36.

Sensitivity Analysis

Sensitivity analysis was performed to assess whether outcomes were altered when the analysis was restricted to higher quality studies and whether limiting the analysis to studies utilizing data prior to 2012 had any effect on results.

Results

Primary Studies Included in the Literature Review

A search of the PUBMED database resulted in 192 relevant articles and further search in the Cochrane database yielded no additional articles (Figure 2). Of the 192 studies 88 studies were excluded based on the type of study, according to our exclusion criteria, and of the remaining 104 studies, seven were excluded based on their title, and 63 based on their abstract as they stratified frailty using indices other than the mFI. Full manuscripts were evaluated for 34 publications but only 16 fulfilled the entry criteria for the meta-analysis.^11,17–31 The 18 papers excluded from the meta-analysis did not provide the appropriate numerical data necessary for statistical analysis, but were included in the literature review.^32–49

Fig. 2.

The selection process of the studies included in the literature review and meta-analysis.

Main Study Characteristics and Methodological Quality Assessment

Of the 34 included studies, only three were prospective (Table 1). All studies were case series and had a Level of Evidence (LoE) of 4 as defined by the Oxford Centre for Evidence-Based Medicine.⁵⁰ No randomized controlled trials were found. Numerical data for meta-analysis could be extracted from 16 studies, three of which focused on orthopedic, three on vascular, and six on gastrointestinal surgery. Seven of 16 studies were of low and six of very low quality on the GRADE scale. Three studies were deemed moderate quality based on their large effect according to GRADE guidelines (Supplemental Table 3).

Table 1:

Studies included in the literature review, organized by year of publication.

Study	Publication Year	Study Period	Country*	Type	Type of Surgery	Patients	Outcomes
Youngerman et al	2018	2008-	USA	R	Neurosur	9149	Mort, Comp, Major Comp, LOS, DSC
Ali et al	2017	2010	USA	R	Vascular	4704	Mort, Comp
Gani et al	2017	2014	USA	R	Gastroint	2714	Mort
Konstantinidis et	2017	2005-	USA	P	Gastroint	1171	Mort, Major Comp
Levy et al	2017	2008-	USA	R	Urogenit	23104	Mort, Major Comp
Mazzola et al	2017	2015-	Italy	P	Gastroint	76	Mort, Comp, Major Comp, LOS, RA, DSC
Mogal et al	2017	2005-	USA	R	Gastroint	9986	Mort, Comp, Major Comp
Park et al	2017	2007-	USA	R	Oncolog	846	Mort, Comp
Runner et al	2017	2005-	USA	R	Orthoped	90260	Mort, Major Comp, Wound Comp, RA
Seib et al	2017	2016-	USA	R	General	140828	Comp, Major Comp
Vermillion et al	2017	2005-	USA	R	Gastroint	41455	Mort, Comp, Major Comp, LOS
Wachal et al	2017	2006-	USA	R	Head and	343	Comp, Major Comp, LOS, DSC
Wahl et al	2017	2007-	USA	R	General	236957	Mort, Comp, RA
Wen et al	2017	2006-	USA	R	Gastroint	272	Comp, RA
Abt et al	2016	2006-	USA	R	Head and	1193	Mort, Major Comp, RA, RO
Ali et al	2016	2006-	USA	R	Orthoped	18294	Mort, Major Comp, Wound Comp
Arya et al	2016	2011-	USA	R	Vascular	15843	Comp, DSC
Cloney et al	2016	2000-	USA	R	Neurosur	243	Mort, Comp, LOS
Ehlert et al	2016	2006-	USA	R	Vascular	72106	Mort, Major Comp
Flexman et al	2016	2006-	Canada	R	Orthoped	53080	Mort, Major Comp, LOS, DSC
Leven et al	2016	2005-	USA	R	Orthoped	1001	Mort, Comp, Wound Comp, RO
Mosquera et al	2016	2005-	USA	R	Gastroint	94811	Mort, Comp
Phan et al	2016	2010-	USA	R	Orthoped	3920	Mort, Comp, Wound Comp, RO
Brahmbhatt et al	2015	2005-	USA	R	Vascular	24624	Mort, Comp, Major Comp, LOS
George et al	2015	2008-	USA	R	Urogenit	66105	Mort, Major Comp, Wound Comp
Lascano et al	2015	2005-	USA	R	Urogenit	41681	Mort, Major Comp
Tsiouris et al	2015	2005-	Canada	P	Cardioth	1940	Mort, Comp, Major Comp
Keller et al	2014	2009-	USA	R	Gastroint	859	Mort, Comp, RA, RO, LOS
Kolbe et al	2014	2005-	USA	R	Gastroint	104952	Mort, Major Comp
Adams et al	2013	2005-	USA	R	Head and	6727	Mort, Comp, Major Comp
Karam et al	2013	2005-	USA	R	Vascular	67308	Mort, Comp, Wound Comp
Obeid et al	2012	2005-	USA	R	Gastroint	58448	Mort, Comp, Major Comp, Wound Comp
Farhat et al	2011	2005-	USA	R	Emergen	35344	Mort, Comp, Wound Comp
Louwers et al	2007	2005-	USA	R	Gastroint	10300	Mort, Major Comp, LOS

^*:Institution of lead author; R: retrospective; P: prospective; Mort: Mortality; Comp: Complications; RA: Readmission; RO: Reoperation; DSC: Discharge to skilled care; LOS: Length of hospital stay.

Definitions of Frailty

Although the mFI was not proposed to be used as a dichotomous variable, for the purpose of this meta-analysis, a score of 0 was considered non-frail and a score greater than 0 was considered frail. This selection was made in order to include the highest number of studies possible, as there was large variation in the use of the mFI between studies.

Definitions of Outcome Measures

Complications were reported by 24 studies and were defined as any 30-day negative outcome other than death.^{11,17–20,22–27,29–31,33,36,38,40,41,43,46–49} Complications were further subclassified into major complications and wound complications, with some studies focusing solely on these complications and not providing information on minor complications. Major complications were reported in 21 studies.^{11,17–19,21,23–25,29,32–35,37,39,40,42,44–49} The majority of the studies defined major complications as Clavien-Dindo class IV complications, that is, complications that are life-threatening or require critical care management. This includes, but is not limited to, postsurgical myocardial infarction, pulmonary embolism, stroke, or coma.^{11,18,19,21,25,29,32–35,39,40,42,44,45,47,48} Three studies defined major complications as Clavien-Dindo class >III complications^24,46,49 and two as class > II.^23,37 For the purpose of this review and meta-analysis, major complications were considered any Clavien-Dindo class >III complications. Wound complications were reported in eight studies.^{11,22,26,27,34,36,39,40} Four of these defined wound complications as the occurrence of any superficial, deep, or organ space wound infection, or wound infections with dehiscence.^26,36,39 One study focused on wound infections with dehiscence.²⁷ Two papers looked at wound infections without further defining them^11,40 and one paper analyzed wound complications but did not specify what this outcome included.²² Length of postoperative hospital stay was reported in nine studies which defined the length of stay in days.^{19,20,23,29,37,41,45,48,49} Reoperation was reported in four studies and was defined as an unplanned return to the operating room.^22,26,32,41 Six studies looked at readmission defined as an unplanned re-admission to hospital within 30 days post-surgery.^{23,27,30–32,41} Discharge to skilled care was reported in five studies, four of which defined it as discharge to a care facility rather than home^18,23,29,37 and one termed it unfavorable disposition, that is a discharge destination with a higher level of care than before admission.⁴⁸ Mortality was defined by 30 studies as death within 30 days of surgery.^{11,17,19–22,24–27,29,30,33–49}

Results for the Overall Meta-Analysis

Sixteen studies were included in the meta-analysis providing a total of 683,487 patients, of which 444,885 were frail. The results of the primary outcomes analysis are shown in Table 2. Frail patients were more likely to experience complications, including major and wound complications. Frail patients had a higher risk of readmission and reoperation and were more likely to be discharged to a skilled care facility. Most notably, the risk of mortality was 4.19 times more likely in frail patients.

Table 2:

Forest plots for primary outcomes in frail versus non-frail patients. (see figure 2b)

Study	Frail		Non-frail		Weight %	RR M-H Random,
	No.	Total	No.	Total
Complications
Mogal et al (2017)	2381	6157	1292	3829	10.3	1.15[1.09, 1.21]
Runner et al (2017)	10444	66847	3051	23413	10.5	1.20[1.15, 1.24]
Mazzola et al (2017)	11	41	10	35	1.3	0.94[0.45, 1.94]
Wen et al (2017)	23	147	16	125	1.9	1.22[0.68, 2.21]
Wachal et al (2017)	79	241	17	102	2.7	1.97[1.23, 3.15]
Ali et al (2017)	376	4019	47	685	4.9	1.36[1.02, 1.83]
Seib et al (2017)	1465	67793	992	73033	9.9	1.59[1.47, 1.72]
Mosquera et al (2016)	15032	52304	8273	42507	10.7	1.48[1.44, 1.51]
Cloney et al (2016)	49	198	3	45	0.6	3.71[1.21, 11.38]
Leven et al (2016)	292	612	137	389	8.0	1.35[1.16, 1.59]
Phan et al (2016)	347	1895	217	2025	8.0	1.71[1.46, 2.00]
Arya et al (2016)	1626	5914	1638	9929	10.2	1.67[1.57, 1.77]
Brahmbatt et al (2015)	4276	12738	1916	9886	10.4	1.73[1.65, 1.82]
Obeid et al (2012)	11225	35818	4492	22630	10.6	1.58[0.53, 1.63]
Total (95% CI) I2 = 92%	47626	25472	2210	18863	100.0	1.48[1.35, 1.61]
Mortality
Mogal et al (2017)	238	6157	51	3829	12.8	2.90[2.15, 3.92]
Runner et al (2017)	115	66847	16	23413	10.7	2.53[1.50, 4.27]
Wahl et al (2016)	1863	18970	122	47251	13.6	3.80[3.17, 4.57]
Ali et al (2017)	96	4019	4	685	6.5	4.09[1.51, 11.80]
Konstantinidis et al (2017)	17	455	9	716	8.1	2.97[1.34, 6.61]
Mazzola et al (2017)	15	41	0	35	1.4	26.57[1.65, 428.63]
Phan et al (2016)	5	1895	2	2025	3.4	2.67[0.52, 13.75]
Leven et al (2016)	10	612	1	389	2.4	6.36[0.82, 49.46]
Mosquera et al (2016)	1530	52304	236	42507	13.9	5.27[4.60, 6.04]
Brahmbatt et al (2015)	433	12738	98	9886	13.4	3.43[2.76, 4.26]
Obeid et al (2012)	2547	35818	164	22630	13.8	9.81[8.39, 11.48]
Total (95% CI) I2 = 91%	6869	37023	703	15336	100.0	4.19[2.96, 5.92]
Major Complications
Mazzola et al (2017)	15	41	7	35	11.4	1.83[0.84, 3.97]
Mogal et al (2017)	1869	6157	983	3829		1.18[1.11, 1.26]
Wachal et al (2017)	25	241	4	102		2.65[0.94, 7.41]
Seib et al (2017)	622	67793	349	73033		1.92[1.68, 2.19]
Konstantinidis et al (2017)	51	455	48	716		1.67[1.15, 2.44]
Brahmbatt et al (2015)	2894	12738	1239	9886		1.81[1.71, 1.93]
Obeid et al (2012)	5195	35818	718	22630		4.57[4.24, 4.93]
Total (95% CI) I2 = 99%	10671	12324	3348	11023	2.03[1.26, 3.29]
Wound Complications
Runner et al (2017)	236	66847	49	23413	1.69[1.24, 2.29]
Leven et al (2016)	25	612	10	389	1.59[0.77, 3.27]
Phan et al (2016)	45	1895	32	2025	1.50[0.96, 2.35]
Obeid et al (2012)	8924	35818	3718	22630	1.52[1.47, 1.57]
Total (95% CI) I2 = 0%	9230	10517	3809	48457	1.52[1.47, 1.57]
Reoperation
Leven et al (2016)	50	612	19	389		1.67[1.00, 2.79]
Phan et al (2016)	103	1895	47	2025		2.34[1.67, 3.29]
Total (95% CI) I2 = 13%	153	2507`	66	2414		2.10[1.54, 2.86]
Discharge to Skilled Care
Mazzola et al (2017)	7	41	2	35		2.99[0.66, 13.46]
Wachal et al (2017)	18	241	1	102		7.62[1.03, 56.31]
Arya et al (2016)	659	5914	518	9929		2.14[1.91, 2.39]
Total (95% CI) I2 = 0%	684	6196`	521	10066		2.15[1.92, 2.40]
Re-admission
Mazzola et al (2017)	11	41	4	35		2.35[0.82, 6.72]
Runner et al (2017)	3485	66847	714	23413		1.71[1.58, 1.85]
Wahl et al (2016)	22540	18970	3722	47251		1.51[1.46, 1.56]
Wen et al (2017)	11	147	4	125		2.34[0.76, 7.16]
Total (95% CI) I2 = 68%	26047	25674	4444	70824	100.0	1.61[1.44, 1.80]

Complications, mortality and major complications had high heterogeneity, readmission had moderate heterogeneity, and wound complications, reoperation and discharge to skilled care had low heterogeneity. Funnel plots were asymmetric suggesting publication bias (Supplemental Figure 1).

Subgroup Analysis

Subgroup analysis for complication occurrence in the different mFI score classes was based on four studies,^11,22,27,28 three of which looked at more than 20,000 patients.^11,27,28 The occurrence of complications increased proportionally to the mFI score. Frail patients with a score of 0.36 or more, indicating a score of 4 points out of 11 (RR 2.08, 95 percent CI 1.50 to 2.87; p<0.001) were more likely to develop surgical complications than patients with an mFI of 0.27, or 3 points, (RR 1.90, 95 percent CI 1.46 to 2.48; p<0.001), 0.18, 2 points, (RR 1.52, 95 percent CI 1.29 to 1.80; p<0.001), and 0.09, 1 point (RR 1.25, 95 percent CI 1.16 to 1.34; p<0.001) undergoing surgery. Subgroup analysis for mortality was based on four studies.^11,22,27,31 Mortality was more likely to occur in frail patients with a score of 0.36 or more (RR 33.87, 95% CI 27.94 to 38.67; p<0.001) than patients with an mFI of 0.27 (RR 14.84, 95% CI 6.84 to 32.20; p<0.001), 0.18 (RR 5.70, 95% CI 2.36 to 13.75; p<0.001), and 0.09 undergoing surgery (RR 2.88, 95% CI 1.52 to 5.45; p = 0.001; Supplemental Table 4). Subgroup analysis for the other primary outcomes could not be performed as the majority of papers did not provide data for each individual mFI score.

Sensitivity Analysis

When performing the meta-analysis with data from the three studies deemed to be of moderate quality, again frail patients were more likely to experience complications (RR 1.41, 95% CI 1.20 to 1.65; p < 0.0001) and had a higher mortality rate (RR 3.82, 95% CI 1.87 to 7.79; p = 0.0002) (Supplemental Table 5) than non-frail patients.

Sensitivity analysis focusing on pre-2012 data could be performed for six of the primary outcomes (Supplemental Table 6). Frail patients were again more likely to experience complications (RR 1.50, 95% CI 1.37 to 1.65; p<0.001),^{11,17–20,22,24,25,29} major complications (RR 2.22, 95% CI 1.11 to 4.42; p=0.02),^11,19,24,29 wound complications (RR 1.52, 95% CI 1.47 to 1.57; p<0.001),^11,22 mortality (RR 4.78, 95% CI 3.00 to 7.62; p<0.001),^{11,17,19,22,24,25} and discharge to skilled care (RR 2.69, 95% CI 1.03 to 7.03; p=0.04).^18,29

Discussion

In this meta-analysis of 16 studies that used the mFI in surgical patients we found that frailty was associated with a higher rate of all-cause complications and mortality. This result held true across multiple surgical specialties. Studies have suggested that frail patients have reduced physical reserves and hence less ability to respond well to surgery.⁵¹ In addition, frail patients often have multiple comorbidities which increase the risk for post-operative complications such as myocardial infarction and pulmonary embolism.⁵² It should be noted that frailty is not simply a count of comorbidities and the mFI provides a more general picture of physical condition by including social, functional, and cognitive factors as well as comorbidities. Subgroup analysis showed that as the mFI increased so did the relative risk for complications. The higher rate of wound complications may be due to the fact that frail patients are more likely to have a low physiological reserve that can predispose to surgical wound complications such as infection.³⁷ In addition, studies have shown that frail patients have increased inflammation, possibly owing to higher levels of acute phase reactants and coagulation factors, including CRP, factor VIII and fibrinogen.⁵³ These factors may increase further following a major inflammatory stressor such as surgery and could account for the higher level of complications, particularly those associated with wound healing.⁵³

The rate of reoperation was higher in frail patients as well, with only one of the four studies reporting a non-statistically significant increase.²⁶ By combining the data of this study in the meta-analysis a statistically significant difference between frail and non-frail patients was found. A higher rate of reoperation may be due to the higher rate of complications seen in frail patients, with one study reporting that a large percentage of reoperations was due to higher rates of wound infections seen with increasing mFI scores.²²

Four of six studies reporting on readmission rates found that although this was more likely to occur in frail patients it was not found to be statistically significant.^23,30,32,41 The two studies with the highest number of patients showed that, in their analysis, readmission was strongly associated with mFI. As this studies were both included in the meta-analysis, along with two studies with a small number of patients, a statistically significant positive correlation was found between frailty and readmission.^27,31

Three of four studies included in this review found that frail patients are statistically more likely to be discharged to a care facility.^18,29,37 The fourth study was limited by a small number of patients but was included in the meta-analysis which showed that frailty is a strong predictor for non-home discharge.¹⁶ Flexman et al. analyzed the rate of discharge to a care facility as it correlates to the mFI using both an unadjusted and adjusted analysis. A higher mFI was first found to be a strong predictor of discharge to skilled care in an unadjusted analysis, which was followed by confirmation with an analysis which adjusted for pre-specified procedure and patient specific variables.³⁷ This increased rate of discharge to a care facility may be explained by the higher rate of complications seen in frail patients. Arya et al. used a stratified analysis to show that frailty was a strong positive predictor for non-home discharge, both independently but also in association with the occurrence of postoperative complications.¹⁸ Specifically, they found that frail patients without complications were more likely to be discharged to a care facility than non-frail patients with no complications, with the lack of functional reserve in frail patients being the likely reason. Likewise, frail patients with complications had a higher risk of non-home discharge than non-frail patients with complications.¹⁸

Eight of the nine studies reporting on the length of stay found that a longer length of stay correlates with pre-operative frailty status.^{19,20,29,37,41,45,48,49} The increased length of stay could be due to several factors, including the higher rate of complications but also the increased likelihood of discharge to a skilled care facility which can often delay discharge.

Interestingly, Farhat et al found that post-operative mortality was strongly associated with frailty and using a multivariate analysis showed that mFI was better at predicting mortality than age or the ASA score.³⁶ Aging does not independently increase mortality, however, frailty with its associated comorbidities and deterioration of physical condition is more likely to be the cause of increase in surgical risk.⁵⁴

Given the introduction of the NSQIP/AGS preoperative guidelines in 2012 we sought to perform a sensitivity analysis investigating whether the effect we were seeing from the retrospective studies was due to a secular trend caused by changes in preoperative selection of patients from 2012 to 2017. Elimination of all studies post-2012 did not affect our conclusions, highlighting that frail patients were still more likely to experience complications, discharge to skilled care and mortality.

It should be emphasized that frailty should not be considered an absolute contraindication for surgery. Rather, risk stratification through frailty assessment based on strict criteria provides the opportunity to optimize patients’ physiology prior to surgery through pre-habilitation, such as nutrition and exercise regimes, as well as post-surgery through proper rehabilitation.

In addition, frailty assessment can enhance identification of patients who may require reoperation and discharge to a skilled care facility. This would allow for better planning, identification of adequate resources, and engagement of patient support systems, which may decrease the length of stay post-surgery.⁵² The decision making of the patient would also be improved as a surgeon would be better able to inform the patient of their individual potential for surgical risk and benefit.

This review and meta-analyses was the first to synthesize the evidence of the effect of frailty on postoperative outcomes across surgical specialties, analyzing the most recent studies including 12 studies from 2017. A study protocol was not published a priori, however, this study follows the PRISMA guidelines.⁵⁵ The study is non-commercial and set strict inclusion and exclusion criteria. In addition, the GRADE criteria were used to assess the methodological quality of the studies, and a sensitivity analysis was performed to provide more robust conclusions.

This study carries limitations commonly seen with systematic reviews and meta-analyses. First, the quality of this meta-analysis is dependent on the quality of the studies it analyzes, all of which are non-randomized. Non-randomized studies carry inherent biases, including selection bias. All included studies are case series which tend to be perceived as lighter surgical research evidence because of several short-comings including having vague objectives and exaggerated conclusions. However, they carry relevance to a certain degree as proof of a potential cause-effect relationship.⁵⁶ Accepting the limitations of case series allows surgeons to learn from such evidence.⁵⁷ Furthermore, according to the GRADE criteria, 12 of 14 studies included in this meta-analyses were of low or very low quality. We sought to overcome this limitation by repeating the analysis using data solely from studies deemed to be of moderate quality. Last, the study is subject to publication bias as all but one study were conducted in the US and Canada. Adding to the publication bias is the fact that our criteria excluded studies that were unpublished and included only studies published in the English language. In addition, the funnel plots indicate a lack of studies showing no effect.

Based on this meta-analysis, the mFI is an underappreciated prognostic indicator that strongly correlates with higher risks of post-operative complications, longer hospitalization periods, greater rates of readmission, reoperation, and discharge to skilled care, as well as higher mortality. Despite not being a perfect measure, the mFI is easy to calculate and uses patient characteristics extracted during history taking and examination. This meta-analysis serves to highlight that the mFI is a powerful prognostic tool. Overall, identifying patients at risk prior to surgery can improve patient outcomes, discharge planning and post-operative support, and have long lasting financial benefits for the healthcare system. Prospective actions such as prehabilitation can improve the frailty status of surgical patients leading to better postoperative outcomes. This supports the formal incorporation of pre-operative frailty assessment using the mFI which has the potential to improve surgical risk stratification.

Keypoints.

Question:

Is the Modified Frailty Index (mFI) an effective tool for risk-stratifying surgical patients?

Findings:

Based on this meta-analysis, the mFI strongly correlates with higher rates of post-operative complications, readmission, reoperation, discharge to skilled care, longer hospitalization periods, as well as greater rates of mortality.

Meaning:

The mFI is an underappreciated prognostic indicator and formal incorporation of pre-operative frailty assessment using the mFI has the potential to improve surgical risk stratification.

Highlights.

The Modified Frailty Index (mFI) is an effective tool for risk-stratifying surgical patients.

Based on this meta-analysis, the mFI strongly correlates with higher rates of post-operative complications, readmission, reoperation, discharge to skilled care, longer hospitalization periods, as well as greater rates of mortality.

The mFI is an underappreciated prognostic indicator and formal incorporation of pre-operative frailty assessment using the mFI has the potential to improve surgical risk stratification.