Association of Frailty with 90-Day Postoperative Mortality & Geriatric Comanagement among Older Adults with Cancer

Abstract

Introduction:

Geriatric medical comanagement of older surgical patients with cancer is associated with lower 90-day postoperative mortality. Here, we investigate the hypothesis that this association varies on the basis of the degree of a patient’s frailty.

Methods:

Frailty was assessed using the Memorial Sloan Kettering Frailty Index. Our hypothesis was evaluated using two approaches, both adjusted for age, sex, American Society of Anesthesiologists Physical Status, preoperative albumin, operating room duration, and estimated blood loss. A logistic regression was performed with 90-day mortality as the outcome and geriatric comanagement, continuous Memorial Sloan Kettering Frailty Index, and an interaction term between these two variables as the primary predictors. We then categorized frailty into four levels and, within each level, performed logistic regression with geriatric comanagement as the primary predictor and 90-day mortality as the outcome. Finally, we extracted the effect size and used a meta-analytic approach to test for heterogeneity.

Results:

Of 1687 patients aged ≥75 years who underwent elective cancer surgery with a length of stay of ≥1 day, 931 (55%) received comanagement; 31 patients (3.3%) who received comanagement died within 90 days, compared with 72 (9.5%) who did not. Ninety-day mortality was not statistically significantly different by degree of frailty in either analysis (interaction P=0.4; test of heterogeneity P=0.8).

Conclusion:

Geriatric comanagement is valuable for all older surgical patients, not just the frail, and should be provided to as large a portion of the older surgical population as possible. Further research should examine predictors of response to geriatric comanagement.

1. Introduction

Age-related impairments and decreases in organ function in older patients with cancer may increase the risk of poor surgical outcomes, compared with younger patients.¹ Nonetheless, surgery remains the mainstay of curative treatment for the majority of patients with solid tumors.² As aging is a multifactorial and highly individualized process,³ it is important to assess the specific functional status or physiologic age of a patient, instead of using chronologic age alone, in the decision-making process for surgery.²,⁴ One method of evaluating physiologic age is to assess frailty, a state wherein the body’s ability to tolerate stress is reduced.⁵ Older patients with greater degrees of frailty have a higher risk of poor surgical outcomes, compared with more-fit patients.⁶ Numerous frailty assessments and screening tools have been published. The use of these instruments to identify frail patients provides an opportunity to adopt a risk-stratified approach to the medical treatment of cancer.³ It is also vital to develop interventions to improve the surgical outcomes of frail patients. These efforts can largely be categorized into three groupings: prehabilitation,⁷ enhanced recovery after surgery,⁸ and collaboration between geriatricians and surgeons, also known as geriatric comanagement.⁹

Most of the evidence supporting the benefits of geriatric comanagement has emerged during the last two decades, with the majority of available data from older adults who underwent orthopedic surgery. For example, a systematic review of 17 studies on older adults with hip fracture showed that geriatric comanagement leads to decreases in time to surgery, hospital length of stay, and postoperative mortality.¹⁰ However, data on surgical oncology patients is limited. Two small, single-arm studies of older adults who underwent cytoreductive surgery or cystectomy demonstrated that excellent outcomes can be achieved through geriatric comanagement.¹¹,¹² Spurred by the findings of these two small studies, we retrospectively evaluated the relationship between geriatric comanagement and 90-day postoperative mortality in 1892 patients aged ≥75 years who underwent elective cancer surgery with a hospital length of stay of ≥1 day.¹³ We found that geriatric comanagement was associated with a statistically significant reduction in 90-day postoperative mortality, after adjustment for various confounding factors. In this study, adjusted 90-day postoperative mortality was 4.3% among patients who received geriatric comanagement, compared with 8.9% among patients who did not receive geriatric comanagement.

It is often said that the question facing practicing clinicians is not whether an intervention is of benefit in general but whether it would benefit the patient in front of them. Here, we analyze the data from the cohort mentioned above¹³ to determine whether the relationship between geriatric comanagement and reduction in 90-day postoperative mortality varies on the basis of a patient’s frailty status.

2. Methods

This retrospective study was approved by the Memorial Sloan Kettering Cancer Center (MSK) institutional review board. The need for patient informed consent was waived.

2.1. Patient population

Eligible patients were aged ≥75 years, underwent elective oncologic surgery at MSK between February 2015 and February 2018,¹³ underwent surgery within 60 days of their initial surgery consultation, were evaluated using the Memorial Sloan Kettering Frailty Index (MSK-FI), and had a hospital length of stay of ≥1 day after surgery.

2.2. Frailty

Frailty was assessed using the MSK-FI.¹⁴ The MSK-FI is derived from the modified frailty index¹⁵ and is composed of 10 items related to comorbid conditions and 1 item related to functional activity. Comorbid conditions were retrieved using submitted ICD9/10 codes in the first 24 h of the hospital stay. All patients completed a questionnaire that assessed their functional activity. Patients had impaired functional activity if they were dependent on others for any of the following activities: bathing, dressing, grooming, walking outside of home, or preparing meals. A patient’s score ranges from 0 to 11, with a higher score corresponding to a greater number of geriatric assessment impairments. In validating the MSK-FI, we have shown that MSK-FI score is associated with overall postoperative mortality¹⁴ and with surgical decision-making and poor overall survival among women with advanced ovarian cancer.¹⁶

2.3. Geriatric comanagement

Geriatric comanagement at MSK has two phases: preoperative and postoperative. Because of the large volume of patients with cancer treated at our institution, patients aged ≥75 years are referred to the Geriatrics Service for preoperative clearance at the discretion of the surgery services. Surgeons also have the option of referring any patient to the MSK Cardiology, Pulmonary Medicine, or General Medicine Services or the patient’s outside primary care provider for preoperative clearance. The Pulmonary Medicine Service routinely sees patients again as inpatients after surgery. The Cardiology Service sees patients after surgery on an as-needed basis.

During geriatric preoperative evaluation, patients complete a geriatric assessment^17,18 and meet with a registered geriatric nurse or nurse practitioner. Subsequently, a geriatrician evaluates the patient and provides various instructions on managing perioperative medications, engaging in more-vigorous physical activity, and optimizing comorbid conditions before surgery. Concerns and questions are shared with the surgery and anesthesiology teams via e-mail. In the postoperative period, patients are seen by the Geriatrics Service during their hospital stay, with the focus of improving mobility, achieving proper pain control, improving respiratory function, implementing delirium risk reduction interventions, managing comorbid conditions, and engaging in a hospital disposition plan.

2.4. Surgical service management

Patients in this group were not referred to the Geriatrics Service for preoperative evaluation. As a result, the Geriatrics Service did not perform a geriatric assessment and did not provide any recommendations during the preoperative phase. Most of these patients were managed only by the surgical service teams in the postoperative period. The surgery team may have requested a geriatric consultation in the postoperative period. Surgeons also had the option of referring any preoperative patient to the MSK Cardiology, Pulmonary Medicine, or General Medicine Services or the patient’s outside primary care provider for preoperative clearance. Whereas the Pulmonary Medicine Service routinely sees patient again as inpatients after surgery, the Cardiology and General Medicine Services see patients after surgery on an as-needed basis.

2.5. Statistical analysis

In our initial study,¹³ 205 patients with missing MSK-FI scores were retained in the analysis, as we used multiple imputation by chained equations to impute missing data. However, as MSK-FI score is a critical factor for our current hypothesis, which requires a defined level of frailty for each patient, we excluded patients without an MSK-FI score, leaving 1687 patients for analysis.

We evaluated whether there was a difference in the association between comanagement status and 90-day mortality on the basis of frailty by using two approaches. The first was to perform an interaction analysis. We used a multivariable logistic regression with 90-day mortality as the outcome; geriatric comanagement, continuous MSK-FI score, and an interaction term between geriatric comanagement and continuous MSK-FI score as the primary predictors; and age, sex, American Society of Anesthesiologists (ASA) Physical Status, preoperative albumin, operating room duration, and estimated blood loss as covariates. For our second approach, we categorized frailty into four levels on the basis of MSK-FI score: 0 vs. 1-2 vs. 3 vs. ≥4. We specifically defined the levels to differentiate between 2 and 3 because the MSK-FI is an adapted version of the modified frailty index, where a score of ≥3 is considered to indicate frail status. For each of the different levels of frailty, we performed separate multivariable logistic regressions with 90-day mortality as the outcome, geriatric comanagement as the primary predictor, and adjustment for age, sex, ASA score, preoperative albumin, operating room duration, and estimated blood loss. We then extracted the effect size for geriatric comanagement from each of the models and calculated Cochran’s Q. We additionally visualized the association between 90-day mortality and MSK-FI score by performing locally weighted regression (lowess) of 90-day mortality on MSK-FI score for patients who were comanaged by the Geriatrics Service and, separately, for patients who were not comanaged by the Geriatrics Service and plotted estimates for both groups on the same figure.

Among our cohort, 37% of patients were missing at least 1 of the covariates included in our models. The most frequent missing covariate was estimated blood loss (unknown in 25%), followed by preoperative albumin (17%) and ASA score (4.1%). Multiple imputation by chained equations was used to impute missing data. All analyses were performed using the measured and imputed values combined across 20 imputations using Rubin’s method. All statistical analyses were performed using Stata 15.0 (StataCorp, College Station, TX).

3. Results

Among our cohort of 1687 patients, 931 (55%) were comanaged by the Geriatrics Service. Table 1 depicts patient characteristics stratified by geriatric comanagement status. Compared with patients who did not receive geriatric comanagement, patients comanaged by the Geriatrics Service were older (mean, 81 vs. 80 years), had a longer length of stay (mean, 7 vs. 6 days), and underwent longer procedures (mean, 206 vs. 139 min) (all P<0.0001). Supplemental Table 1 presents procedure types by geriatric comanagement status.

Table 1.

Patient and Perioperative Characteristics (N=1687)

Characteristic	No Geriatric Comanagement (N=756; 45%)	Geriatric Comanagement (N=931; 55%)	P
Age at surgery, yearsa	80 (4)	81 (4)	<0.0001
Male	392 (52)	446 (48)	0.12
Length of stay, daysa	6 (8)	7 (8)	<0.0001
Operating room duration, mina	139 (109)	206 (149)	<0.0001
ASA score ≥3	707 (94)	880 (95)	0.4
Preoperative albumin, g/dLa	3.9 (0.5)	4.0 (0.4)	0.0003
Estimated blood loss, mLa	130 (249)	204 (304)	<0.0001
MSK-FI score			0.6
0	85 (11)	99 (11)
1	209 (28)	271 (29)
2	190 (25)	260 (28)
3	131 (17)	152 (16)
4	83 (11)	84 (9.0)
5	34 (4.5)	42 (4.5)
≥6	24 (3.2)	23 (2.5)
Major complication	32 (4.2)	69 (7.4)	0.007

Data are no. (%) unless otherwise noted. Data for characteristics with incomplete data are shown combined across 20 imputations. P values were assessed using univariable regression and combined across 20 imputations using Rubin’s method. ASA, American Society of Anesthesiologists; MSK-FI, Memorial Sloan Kettering Frailty Index.

^aMean (standard deviation).

In total, 103 patients died within 90 days of surgery: 31 (3.3%) who received geriatric comanagement and 72 (9.5%) who did not receive geriatric comanagement. In our multivariable model, the association between geriatric comanagement and MSK-FI score was not statistically significant (P=0.4). Figure 1 shows the association between geriatric comanagement and 90-day mortality in each of the different frailty groups—again, no evidence of heterogeneity was observed (P=0.8). A post hoc sensitivity analysis that investigated frailty groups defined instead as 0-1 vs. 2 vs. 3 vs. ≥4 yielded similar results (test of heterogeneity P=0.9). Because of the low number of events among patients with an MSK-FI score of 0, we performed another post hoc sensitivity analysis, this one investigating the unadjusted association between geriatric comanagement and 90-day mortality in the different frailty groups—this yielded similar results (test of heterogeneity P=0.6). Results remained similar in a final post hoc sensitivity analysis that additionally adjusted for procedure type. To perform this analysis, we first created a multivariable model with 90-day mortality as the outcome and all of the procedure types as the predictor. From this model, we derived the linear predictor and then included this linear predictor as an additional covariate in the models of our primary analyses.

Figure 1.

Forest plot of the association between geriatric comanagement and 90-day mortality within different frailty groups. Test of heterogeneity P=0.8. CI, confidence interval; MSK-FI, Memorial Sloan Kettering Frailty Index; OR, odds ratio.

These analyses failed to reject the hypothesis that the relative risk reduction of comanagement is independent of the degree of frailty. However, an intervention that has a constant relative risk across different levels may have very different levels of benefit if the baseline risk is subject to large variation. For example, halving of risk is of greater benefit to a patient with a 50% risk, compared with a patient with a 5% risk. Figure 2 shows that there is little variation in risk among MSK-FI scores, and, consequently, the absolute benefit of geriatric comanagement is similar irrespective of MSK-FI score.

Figure 2.

Risk of 90-day mortality on the basis of Memorial Sloan Kettering Frailty Index (MSK-FI) score among patients who did (solid blue line) and did not (solid red line) receive geriatric comanagement. Dashed lines indicate 95% confidence intervals.

4. Discussion

With the increase in the aging population and the prevalence of frailty as high as 42% in the geriatric population,¹⁹ there is a need to implement interventions aimed at improving the outcomes of these patients. Previously, we showed that geriatric comanagement was associated with a statistically significant reduction in 90-day postoperative mortality among adults aged ≥75 years who underwent elective oncologic surgery with a hospital length of stay of ≥1 day at our institution.¹³ The possible mechanisms underlying the reduction in 90-day postoperative mortality seen with geriatric comanagement have been discussed in detail.^10,13 In the present study, we assessed whether geriatric comanagement is more beneficial for frail older patients than for fit older patients. The results show that the benefits of geriatric comanagement are independent of the degree of patient frailty. For example, 90-day postoperative mortality among patients with an MSK-FI score of 1 was 3.0% among those who received geriatric comanagement and 8.1% among those who did not. For frailer patients, with an MSK-FI score of 4, 90-day postoperative mortality was 4.8% among comanaged patients and 12% among patients who did not receive comanagement. Our findings have clear clinical implications, as more institutions have become interested in developing geriatric surgery and geriatric surgical oncology programs.^20–23 With the emergence of more data on the benefits of geriatric comanagement,^9,13,24–26 it is critical to provide guidance on the characteristics of patients who may benefit the most from geriatric comanagement. In medical oncology practice, a stepwise approach starting with a very short frailty screening tool, such as the G8,²⁷ has been proposed and evaluated. Screening would be followed by a more in-depth geriatric assessment for patients who are positive on screening and would include interventions by a geriatrician either directly or indirectly.^28,29,30 However, data are not available on the use of such an approach in geriatric surgery and especially geriatric surgical oncology. As there are many frailty screening tools and instruments available, comparing different iterations of this approach and determining the best frailty screening tool overall are problematic. One systematic review assessed the relationship between frailty and surgical outcomes and found that, in 23 studies, 21 different instruments were used to measure frailty.⁶ Another systematic review, including 20 studies that investigated the association between frailty and surgical and medical outcomes in patients with cancer, also found a substantial variation in frailty instruments used.³¹ Such a pronounced heterogeneity in frailty screening tools may contribute to a negative attitude toward preoperative frailty assessment among surgeons. One study showed that only 48% of surgeons consider preoperative frailty screening a necessary component of presurgical assessment.³² These two issues—the wide variety of frailty screening tools and the lack of surgeon interest in implementing frailty screening preoperatively—may hamper the much needed efforts and interventions aimed at improving perioperative care and outcomes in older patients with cancer. The results of our study suggest that, for all patients aged ≥75 years, a referral to the Geriatrics Service for geriatric comanagement whenever logistically feasible could be associated with improved 90-day postoperative mortality. As a result, the logical next step in establishing geriatric surgical oncology as a vital component of cancer care is to develop the infrastructure and resources needed to comanage all (or nearly all) patients aged ≥75 years with cancer and to formally engage geriatric experts in multidisciplinary cancer care teams.

The results of the present study warrant further discussion of frailty screening methods. We assessed frailty by use of the MSK-FI, which has been shown to be associated with geriatric assessment and surgical outcomes in our patient population.¹⁴ However, the MSK-FI is based on 10 comorbid conditions and 1 item related to functional activity. The comorbid components are simply the presence or lack of certain conditions, such as diabetes or peripheral vascular disease. Thus, the tool does not consider the severity or overall control of these comorbid conditions. It is possible that the severity of comorbidities differed between patients with the same MSK-FI score. Moreover, the MSK-FI is mainly composed of comorbidities, with only 1 of 11 items related to functional activity. It is possible that a frailty index with a more balanced mix of comorbidities and functional impairments may have shown that the relationship between geriatric comanagement and improvement in 90-day postoperative mortality differs on the basis of degree of frailty. Future studies aimed at exploring the role of geriatric comanagement in improving perioperative outcomes among older adults on the basis of frailty status should aim to collect more-granular data on comorbidity burden and aging-related impairments among patients.

Our study has several limitations. Our primary analysis of the relationship between geriatric comanagement and 90-day postoperative mortality was nonrandomized. Although there are good reasons to believe that the association is causal, it may be only correlative. In addition, not all eligible patients at our institution are referred to the Geriatrics Service, and it may be that referral patterns reduced the predictiveness of the MSK-FI. For example, there is likely a rationale underlying why some patients with lower MSK-FI scores were nonetheless referred to the Geriatrics Service and thus underwent geriatric comanagement. Likewise, there may be other underlying characteristics prompting patients with higher MSK-FI scores to not be referred to the Geriatrics Service. Additionally, our study was conducted in a single tertiary-care cancer center, and, as a result, the findings may not be generalizable to other institutions. For example, it is possible that, in other institutions, patients of even younger age may benefit from geriatric comanagement. We also did not elaborate on every component of geriatric comanagement, its duration, and its relationship with surgical outcomes. In the future, we intend to do so. We also did not perform an analysis of the costs associated with geriatric comanagement. Future studies should assess various models of geriatric comanagement in limited or extended versions and the cost-effectiveness of each model. Our study may have been strengthened by focusing on a more homogeneous cohort of patients and surgical procedures in a prospective manner. Previously, we described the differences in characteristics between those who were and were not referred for geriatric comanagement.³³ Ongoing prospective studies in our institution are addressing this issue. As mentioned above, although our frailty screening tool has been shown to have a strong relationship with geriatric assessment and surgical outcomes,¹⁴ different results may have been found if another frailty assessment tool had been used. When feasible, future studies should consider the use of a battery of frailty screening tools to stratify the relationship between geriatric comanagement and surgical outcomes on the basis of frailty.

Our study has several strengths. It remains the largest study on geriatric comanagement and surgical outcomes among adults aged ≥75 years with cancer. Frailty was assessed retrospectively, and, as a result, health care providers, especially the surgery teams, were not aware of frailty status. Consequently, it is unlikely that perioperative care was modified by the surgery teams on the basis of frailty status. Finally, included patients were not required to give consent for the study, which therefore limited healthy selection bias.

5. Conclusions

In our cohort of patients aged ≥75 years who underwent elective oncologic surgery with a hospital length of stay of ≥1 day, geriatric comanagement was associated with lower 90-day postoperative mortality regardless of degree of frailty. Efforts should be directed to develop infrastructure capable of providing geriatric comanagement for as large a portion of the older surgical population as possible.

Abbreviations:

ASA

American Society of Anesthesiologists

CI

confidence interval

MSK

Memorial Sloan Kettering Cancer Center

MSK-FI

Memorial Sloan Kettering Frailty Index

OR

odds ratio