key: cord-0842157-f56clhvb
authors: Owodunni, Oluwafemi P.; Mostales, Joshua C.; Qin, Caroline Xu; Gabre-Kidan, Alodia; Magnuson, Thomas; Gearhart, Susan L.
title: Preoperative Frailty Assessment, Operative Severity Score, and Early Postoperative Loss of Independence in Surgical Patients Age 65 Years or Older
date: 2020-12-29
journal: J Am Coll Surg
DOI: 10.1016/j.jamcollsurg.2020.11.026
sha: bd0edbb8bac9bde046b3aab1f8edd4fea223fa99
doc_id: 842157
cord_uid: f56clhvb

INTRODUCTION: Preoperative discussion around postoperative discharge planning have been amplified by the COVID pandemic. We wish to determine if our preoperative frailty screen would predict postoperative LOI. METHODS: This single-institutional study included demographical, procedural, and outcome data from patients ≥65 years who underwent frailty screening prior to a surgical procedure. Frailty was assessed using the Edmonton Frailty Score (EFS). The Operative Severity Score was used to categorize procedures. The Hierarchical Condition Category (HCC) risk-adjustment score as calculated by CMS was included. LOI was defined as an increase in support outside of the home following discharge. Univariable, multivariable logistic regressions, and adjusted post-estimation analyses for predictive probabilities of best fit were performed. RESULTS: 535 patients met inclusion and LOI was seen in 38 (7%) patients. Patients with LOI were older, had a lower BMI, a higher EFS score (7vs.3.0, p<0.001), and a higher HCC score than patients without LOI. Being frail and undergoing a procedure with an OSS ≥ 3 was independently associated with an increased risk of LOI. In addition, social dependency, depression, and limited mobility was associated with an increased risk for LOI. On multivariable modeling, frailty status, undergoing a surgery with an OSS ≥ 3, and having an HCC score ≥ 1 was the most predictive of LOI (OR 12.72, 95% CI: 12.04, 13.44, p<0.001). In addition, self-reported depression, weight loss, and limited mobility was associated with a nearly 11-fold increase risk in postoperative LOI. CONCLUSIONS: This study is novel as it identifies clear, generalizable risk factors for LOI. In addition, our findings support the implementation of preoperative assessments to aid in care coordination and provide specific targets for intervention.

It is currently estimated that more than 50% of surgeries performed in the United States are performed in patients over 65 years of age and the prevalence of frailty in this older patient population ranges from 10% to 37%. [1] [2] [3] Frailty is an indicator of a patient's vulnerability to the physiological stress of surgery and the potential for long-term postoperative effects. 4 It has become a well-established predictor of poor postoperative outcomes including an increased risk for complications and a longer length of stay in many of the surgical sub-specialties. [5] [6] [7] Recently, Berian et al demonstrated that the increase in complications seen in our older patient population was significantly associated with an increased rate of loss of independence and early mortality. 8 Frail surgical patients who suffer loss of independence often are placed in skilled nursing facilities upon discharge from the hospital. This has been particularly challenging given the uncertainty around SARS-CoV-2 infections within skilled nursing facilities. Skilled nursing facilities are particularly vulnerable to SARS-CoV-2 infection and residents are at risk for severe outcomes. 9 In 2013, the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP), in collaboration with the American Geriatrics Society (AGS), issued best practices guidelines recommending the incorporation of frailty assessment into clinical practice to address the growing trend of a rapidly aging U.S. population. 10 Dr. Ko, in a systematic review identified 67 different frailty instruments frequently cited for their ability to identify this vulnerable population. 5 The instrument types are characterized into two unique models, the Physical Frailty Phenotype characterized by the 5-clinical features of a decline in lean body mass, grip strength, endurance, walking speed, and physical activity, and the Deficit Accumulations Model characterized as a model of multi-morbidity. Several of the instruments J o u r n a l P r e -p r o o f lack a cognitive assessment and fail to acknowledge psychosocial determinants of health. In 2018, we implemented the Edmonton Frail Scale in our clinical setting to screen patients prior to a surgical intervention for the risk of frailty. This scale is an 11-item scale in which 9 items are self reported. It has been reported to take on average 5 minutes to complete and does not require someone with geriatric expertise to administer. [11] [12] The assessment does incorporate a cognitive screen and screens for psychosocial determinants and has been validated as a tool to evaluate frailty. [11] [12] [13] Understanding the interaction between frailty and surgical outcomes has been hindered by the high degree of heterogeneity of the frailty assessments. However, more importantly, is the limitation imposed by the variability in the types of procedures performed on these patients.

Most studies examining frailty have limited their patient population to a select disease state (cancer) or procedure types which greatly limits the generalizability of their findings. Recently, to address this issue, Shinall et al developed and validated an "Operative Severity Score" which assigns an ordinal number (1 -5) relative to the degree of physiologic stress experienced by patients to a large variety of surgical procedures. 4 We wished to evaluate the use of the Edmonton Frail Scale to predict postoperative outcomes in our surgical population. To aid in the generalizability of our findings, we wished to examine the utility of the operative severity score to categorize the elective surgical interventions in our older population. In particular, we wanted to understand the effect of our frailty screen and the physiological severity of the planned procedure on postoperative loss of independence and mortality.

J o u r n a l P r e -p r o o f Between June 2019 and June 2020, patients ≥ 65 years who underwent frailty screening prior to a surgical intervention were included in this study. All patient data were abstracted from the hospital's electronic medical record (EMR). Abstracted data included patients' demographic characteristics, including age, sex, race, and lifestyle determinants of health, including body mass index (BMI), alcohol use, and smoking status. In addition, access to advanced care directives within the EMR prior to the planned procedure was assessed. Procedural data included Current Procedural Terminology (CPT) coding designation and procedure duration. The patient class designation of elective surgical inpatient or outpatient procedure was utilized.

To evaluate for the presence of frailty in patients ≥ 65 years, deviations from normal physiological functions were considered across several frailty domains, including independent capacity, cognition, nutrition, disease chronicity, and the presence of geriatric syndromes.

Individual scores were calculated as the proportion of deficits out of the total number of the domain-specific items included in the scale to compare frailty levels. EFS was first dichotomized, with a score of ≥ 6 considered frail. To understand the impact of the individual domains of the EFS on outcomes, each domain was tallied independently using the following algorithm:

Cognitive impairment: complete failure of the clock draw or partial failure and self-reporting forgetting to take medications (2 points) Social Dependence: self-reported failure to perform 2 or more ADLs, self-reported lack of any social support, or self-reported failure to perform 1 or more ADL and self-reported lack of some 

The OSS has been previously described, and the techniques for the score evaluation are reviewed. 4 Briefly, modified Delphi consensus methods using CPT codes were used to develop OSS. OSS is assigned using an ordinal scale of 1 to 5 relative to the degree of physiologic stress experienced by patients, with larger scores registering more physiologic stress levels. The research team assigned scores based on CPT and procedural description. Only procedures with an established OSS score were included in this study.

The Centers for Medicare & Medicaid Services (CMS) utilizes patient demographic characteristics and diagnosis-based clinical measures to generate a summary risk-adjustment scoring system. 14 The CMS-HCC model adjusts for patients' insurance capitation fees and predicts healthcare charges for which payment plans are subject. The CMS-HCC model is prospective as the mean expenditures collected in one year provides predictive data for the following year. Moreover, beneficiaries rely on several parameters for coding diagnoses and, consequently, a comprehensive clinical profile prediction is generated. Five domains of clinical and diagnostic categories are used to assign CMS-HCC and include the following sources: 1).

Primary hospital inpatient diagnosis, 2). Secondary hospital inpatient diagnosis, 3). Hospital outpatient, 4). Physician visit 5). Clinically trained non-physician. For the individual HCC scores, we evaluated the most updated reported predictive ratio for each patient prior to the J o u r n a l P r e -p r o o f surgical intervention. Ratios approaching one are characteristic of an accurate prediction, ratios less than one under-predict expenditure, and ratios greater than one over-predict expenditure.

The main outcomes of interest evaluated included loss of independence (LOI) and mortality. LOI was defined as discharge destination other than home. Patients discharged to skilled care facilities (SNF/Rehab), or other locations that differed from their point of origin were considered dependent and met the LOI definition. Mortality is defined as patients who died within a 30-day period from the primary procedure. Other outcomes of interest included length of hospital stay (days) following the surgical procedure and readmissions occurring within 30 days following discharge.

Descriptive analyses were performed for the demographic and perioperative clinical 

Overall, 535 surgical patients ≥ 65 years with a preoperative EFS score participated in this study. The median age was 72 years (IQR 68, 77), with 59.8% being male and 74.6% white ( Table 1 ). The median EFS score for the study population was 3.0 (2.0 -5.0). Within the domains of the EFS assessment there was a 15.1% rate of cognitive impairment, an 8.2% rate of J o u r n a l P r e -p r o o f social dependence, a 16.0% rate of depression, 20.4% of patients suffered weight loss, and 34.4% of patients had limited mobility. Most procedures were classified as inpatient (55.0%) and median procedure duration was 160 min (105 -230 min). Using the OSS to define the case type, the majority of procedures performed were an OSS of 2 with 33.5% of the total population having a procedure with an OSS ≥3 performed. With regards to the domains of the EFS score, having self-reported weight loss and measured poor mobility was associated with an 8.7 fold increase in mortality (OR 8.78, 95% CI: 7.82, 9.87, p<0.001). This finding was confirmed with post-estimation predictive margins which demonstrated a 12% increased risk for mortality in patients with self-reported weight loss and measured limited mobility (PM 0.12, 95% CI: 0.04, 0.21, P=0.004, table 4).

Loss of independence is part of the conceptual pathway known as "the disablement process" that elderly persons often experience. Verbrugge and Jette define the disablement process as a description of how chronic and acute conditions affect functioning in specific body systems and how personal and environmental factors speed or slow down the disablement process. 18 One of the environmental factors that has been shown to facilitate the disablement process is subjecting an elderly person to the physiological stress of surgery and the complications often associated with an operative intervention. It therefore is not surprising that LOI occurs in the elderly following many surgical interventions for vascular disease, colorectal cancer, and general abdominal emergencies. 6, 8, 18, 19 Frailty is an established risk factor for poor postoperative outcomes including an increased risk of discharge to a skilled nursing facility. In addition, LOI identified on postoperative discharge has been shown to be predictive of long-term LOI. In patients ≥ 80 years of age undergoing surgery for colorectal cancer, nearly half of patients with LOI at discharge continued to have LOI at 30 days. 18 Similarly, De Roo and associates demonstrated that 33% of older adults undergoing major colorectal surgery suffered a functional decline as measured by the activities of daily living (ADL) scale at 1 year. 20 We therefore wished to determine the ability of our frailty assessment tool, the EFS, to predict postoperative LOI at our institution.

We implemented the EFS into our multi-specialty surgical clinic workflow through integration with our EMR. An EFS ≥ 6 is an indicator of vulnerability and this score has been utilized to trigger interventions including a referral to geriatric medicine for a comprehensive geriatric assessment. As our surgical population is heterogeneous, each surgical procedure was assigned an OSS. Our results demonstrated that frailty status (EFS ≥ 6) was significantly associated with LOI both in the inpatient and outpatient surgical population. This finding J o u r n a l P r e -p r o o f validates the independent use of this screening system to predict this outcome and supports the use of existing resources and workflow to complete the screening assessment and the incorporation of the screening results into shared-decision making with the patient during the clinic visit. In addition, this finding supports the allocation of appropriate resources needed for coordination of perioperative care. In a similar study, Donald et al utilized the validated Clinical Frailty Score (CFS) in the preoperative setting to successfully predict LOI in patients only undergoing major vascular surgery. Our findings are more generalizable as they apply to a variety of surgical procedures. In this data set, the CMS-HCC score was not associated with a risk for LOI, however, was significantly associated with the risk of 30-day mortality. Our findings are corroborated by Kumar et al 21 who recently reported that current risk adjustment and comorbidity index including the CMS-HCC were poor at predicting post-acute care SNF utilization and readmission in patients undergoing joint replacement. This suggests that automated scoring systems that rely heavily on coding data and multi-morbidity may not provide the granularity necessary to evaluate some of our important surgical outcomes.

In this study, we chose to examine all variables for an interaction to develop the best predictive model for LOI. After examining all potential variables for possible interactions, we did note that utilizing all three scoring systems, frailty (EFS ≥ 6), OSS ≥ 3, and HCC score ≥ 1 was associated with a 13-fold increase risk for LOI. As predictive models vary in outcomes, we also chose to examine our findings using post-estimation predictive margins. Using this modeling, we determined that frail (EFS ≥ 6) older (age ≥ 80) patients undergoing a more physiological stressful procedure (OSS ≥ 3) were at a nearly 40% risk of postoperative LOI. Our combined results suggest that knowledge of the preoperative frailty status and application of validated global score of operative stress provides the surgeon with valuable information to J o u r n a l P r e -p r o o f incorporate into the process of shared-decision making with the patient and should assist in the allocation of appropriate resources to coordinate perioperative care.

When examining the domains of the EFS to determine for possible areas of intervention, we noted that self-reported depression and weight loss and limited mobility was associated with LOI using both predictive modeling methods. This finding is similar to a recent study from Pederson et al which demonstrated that a delay in mobilization defined as out of bed ≥ 36 hours after surgery was independently associated 2-fold higher risk of 30-day readmission or mortality. 22 In their study, 24% of patients undergoing emergent abdominal surgery (the most common procedure was cholecystectomy) had delayed mobilization. In addition, Hirvensalo et al demonstrated similar findings in community dwelling older adults. 23 In their study, mobility impairments independently predicted 5-fold increase in loss of independence among older men and a 3-fold increase among older women. It is also important to note that LaCroix et al in a similar study found that a higher level of physical activity was associated with better functioning in those with chronic disease. Moreover, those individual with chronic disease who had a higher level of physical functioning were less likely to lose their mobility than those individuals with chronic disease that were sedentary. 24 In aggregate, this work supports the implementation of a multimodal prehabilitation protocol such as the Michigan Surgical Home and Optimization Program (MSHOP) 25 to optimize preoperative mobility and nutrition in a targeted population of frail patients undergoing a higher physiologically stressful procedure.

It is interesting to note that our data did not show that cognitive impairment assessed with the clock-draw was independently associated with LOI or mortality. Previous work using National Surgery Quality Improvement Project (NSQIP) data has shown cognitive impairment as well as postoperative delirium to be a predictor of early postoperative LOI. 18 As we have J o u r n a l P r e -p r o o f recently implemented our delirium screening tool in our institution, we currently lack the ability to identify the rate of postoperative delirium, which is more likely to occur in patients with underlying cognitive impairment and is a known risk factor for poor surgical outcomes. 26 It is also plausible that early identification of these at risk individuals with cognitive impairment in our patient population has prevented the outcome of LOI and further work to examine the effect of cognitive impairment are necessary.

This study has several limitations. First, we utilized the EMR to obtain our data and therefore, our data is dependent upon correct documentation in the appropriate location within the EMR. Our EFS data may have not reflected the absolute condition of the patient at the time of surgery, we have allowed for a 6-month gap in performance of the assessment. In addition, we internally validated administration of the EFS by our medical assistants, and there is always change in staffing allowing for some newer medical assistants to be less familiar with EFS scoring. We did develop an EFS learning tool and have performed several re-education sessions, however, we currently do not have a method of assessing the performance of the medical assistants prior to them administering the EFS assessment. Second, although we were able to assess a variety of elective surgical procedures, some procedure may have been more urgent than others given our current OR release policy and the ability of some procedures in inpatients to be posted to the next day schedule as an elective case. In addition, nearly half of our cases were outpatient procedures. As we noted that nearly 20% of our patients with LOI were scheduled for outpatient procedures, we felt it was important to include these cases into our study as unplanned admissions can utilize a large amount of hospital resources. Finally, we were concerned about our low number of patients with LOI which is most likely a result of including outpatient procedures. Low numbers may have made predictive models less accurate; however, we used 2 J o u r n a l P r e -p r o o f separate models to investigate our population. Regardless, further work in this area is necessary to better define measures to identify vulnerable individuals and implement strategies to prevent poor outcomes in this patient population. 

Elderly patients in surgical workloads: a population based analysis

An aging population and growing disease burden will required a large and specialized health care workforce by 2025

Association of Preoperative Patient Frailty and Operative Stress with Postoperative Mortality

Preoperative Frailty Evaluation: A promising Risk Stratification Tool in Older Adults Undergoing General Surgery

Frailty screening and a frailty pathway decrease length of stay, loss of independence, and 30-day readmission rates in frail geriatric trauma and emergency general surgery patients

Association fo Frailty and Postoperative Complications with Unplanned Readmissions after Elective Outpatient Surgery

Association of Loss of Independence with Readmission and Death after Discharge in Older Patients after Surgical Procedures

Asymptomatic SARS-CoV-2 Infection and COVID-19 Mortality During an Outbreak Investigation in a Skilled Nursing Facility

Optimal Perioperative Management of the Geriatric Patient: A Best Practices Guideline from the American College of Surgeons NSQIP and the American Geriatrics Society

Preoperative frailty assessment: a review

The Gap Between Clinical Research and Standard of Care: A Review of Fraillty Assessment Scales in Perioperative Surgical Settings

The prevalence and impact of undiagnosed cognitive impairment in older vascular surgical patients

Risk adjustment of Medicare capitation payments using the CMS-HCC model

Computing Interaction Effects and Standard Errors in Logit and Probit Models

Using the Margins Command to Estimate and Interpret Adjusted Predictions and Marginal Effects

Simulation-Based Sensitivity Analysis for Matching Estimators

Loss of Independence in Older Patients after Operation for Colorectal Cancer

Preoperative frailty assessment predicts loss of independence after vascular surgery

Long-term functional decline after high-risk elective colorectal surgery in older adults

Current Risk Adjustment and Comorbidity Index Underperforms in Predicting Post-Acute Utilization and Hospital Readmissions after Joint Replacement: Implications for Comprehensive Care for Joint Replacement Model

The impact of delayed mobilization on postdischarge outcomes after emergency surgery: a prospective cohort study in older patients

Mobility Difficulties and Physical Activity as Predictors of Mortality and Loss of Independence in the Community-Living Older Population

Maintaining mobility in late life II. Smoking, alcohol consumption, physical activity and body mass index

The Michigan Surgical Home and Optimization Program is a scalable model to improve care and reduce costs