Influence of Frailty and Health Status on Outcomes in Patients with Coronary Disease Undergoing Percutaneous Revascularization

Abstract

Background

While older patients frequently undergo percutaneous coronary interventions (PCI), frailty, comorbidity, and quality of life (QOL) are seldom part of risk prediction approaches. We assessed their incremental prognostic value over and above the risk factors in the Mayo Clinic risk score (MCRS).

Methods and Results

Patients ≥ 65 years who underwent PCI were assessed for frailty (Fried criteria), comorbidity (Charlson index), and QOL [SF-36]. Of the 628 discharged [median follow-up of 35.0 months (IQR, 22.7-42.9)], 78 died and 72 had an MI. Three year mortality was 28% for frail patients, 6% for non-frail patients. The respective 3-year rates of death or MI were 41% and 17%. Following adjustment, frailty [hazard ratio (HR) 4.19 [95% confidence interval (CI), 1.85, 9.51], physical component score of the SF-36 (HR, 1.59; 95% CI, 1.24-2.02), and comorbidity, (HR, 1.10; 95% CI, 1.05, 1.16) were associated with mortality. Frailty was associated with mortality/MI (HR, 2.61, 1.52, 4.50). Models with conventional MCRS had C-statistics of 0.628, 0.573 for mortality and mortality/MI respectively. Adding frailty, QOL, and comorbidity, the C statistic was (0.675, 0.694, 0.671) for mortality, and (0.607, 0.587, 0.576) for mortality/MI respectively. Including frailty, comorbidities, and SF-36, conferred a discernible improvement to predict death and death/MI (integrated discrimination improvement 0.027 and 0.016 and net reclassification improvement of 43% and 18% respectively).

Conclusions

Following PCI, frailty, comorbidity and poor QOL are prevalent and are associated with adverse long-term outcomes. Their inclusion improves the discriminatory ability of the MCRS derived from the routine cardiovascular risk factors.

The 2000 US census data reported that 12.4% people are above 65 years, and 5.9% over the age of 75 years.¹ Coronary artery disease (CAD) is common in the elderly and these elders represent an increasing proportion of the >1 million percutaneous coronary interventions (PCI) performed annually in the United States.² Assessing risk in CAD is the foundation of current cardiovascular practice. It is required to identify optimal treatments and to communicate the risks associated with alternative treatments to patients. Predicting risk in the elderly, however, is complex given their highly variable health status. For example, among geriatric patients, syndromes associated with impaired functional reserves are known to increase the risk of subsequent disability and death.^3-5 With the availability of better assessment tools comorbidity, frailty, and quality of life [QOL], the opportunity to leverage these tools to improve risk stratification is a potentially important and novel method to better understand elderly patients' prognosis and in turn provide a foundation for better medical decision-making.

To optimize risk-stratification in the general population of patients undergoing coronary revascularization, we have previously described the Mayo Clinic Risk Score (MCRS), a model based on traditional cardiovascular risk factors (e.g., age, myocardial infarction, serum creatinine etc.). The MCRS accurately predicts in-hospital mortality and major adverse cardiovascular complications following PCI.⁶ The model has been validated and demonstrates good calibration and discrimination for in-hospital outcomes⁷ but has only modest discrimination to predict long-term mortality.⁸ We reported that addition of comorbidity improves the discriminatory ability of the MCRS. ⁸ Given the growing population of elderly patients treated with PCI, and the known prognostic importance of frailty, comorbidity, and QOL in the elderly^8-10, we hypothesize the addition of these measures can improve the prediction accuracy of the MCRS for long-term outcomes in the elderly. Thus, the objective of this study was to evaluate the role of frailty, comorbidity, and QOL variables on long-term outcomes following PCI and to explicitly test whether they provide incremental improvements in risk-stratification over and above the MCRS alone.

Methods

This prospective cohort study consisted of patients undergoing PCI at Mayo Clinic, a tertiary care center, in Rochester, MN and at Franciscan Skemp Hospital, a community hospital, in La Crosse, WI. Patients who were 65 years or older, underwent PCI from October 2005 to September 2008 and survived until hospital discharge, were included. Pre discharge, standardized questionnaires and tests were administered to each patient to ascertain frailty, comorbid conditions, and their QOL. Informed consent was obtained from each participant and the Mayo Clinic Institutional Review Board approved the study. Patients with a history of stroke with residual neurological deficits, severe Parkinson disease, or severe dementia were excluded from the study.

Data Collection and Follow-Up

At the time of PCI, patient-specific data including clinical, procedural, and angiographic information were recorded and later entered into a prospective database. A clinical research nurse contacted all patients at 6 months, 12 months, and annually thereafter. Medical records of all patients requiring hospitalization at Mayo Clinic or elsewhere were reviewed to further characterize any clinical events during follow-up. Deaths were confirmed using patient records and/or death certificates.

Mayo Clinic Risk Score

This score was previously reported and externally validated for both in-hospital mortality and major adverse cardiovascular events. ⁶This model is based solely on clinical and non-invasive laboratory, and procedural parameters. Five clinical and 3 angiographic variables were significantly correlated with procedural complications: cardiogenic shock, left main coronary artery disease, severe renal disease, urgent or emergent procedure, congestive heart failure class III or higher, thrombus, multivessel disease and older age.

Frailty Assessment

Frailty was assessed during the index hospitalization, using the Fried and Walston definition of frailty.¹¹ The five criteria measured were unintended weight loss (>10 lb in the preceding year), exhaustion, physical activity, time required to walk 15 feet, and grip strength by Jamar® handgrip dynamometer. Exhaustion was measured by the subscale of the Center of Epidemiologic Studies-Depression subscale in which subjects were asked two questions: how often in the past week did they feel the following (a) I felt that everything I did was an effort, and (b) I could not get going. Subjects who answered “a moderate amount of time (3-4 days)” or “most of the time” to either of the statements were categorized as meeting the exhaustion criteria for frailty. Physical activity was measured by the short version of the Minnesota Leisure Time Activity questionnaire. Table 1 details the thresholds for each of the five frailty criteria. Frailty was defined as present if the subjects had 3 or more core elements, and intermediate frailty, if they had 1- 2 core elements.

Table 1. Measurement thresholds for frailty criteria.

Criteria	Threshold for meeting core frailty element		Study prevalence
Exhaustion	Responds “3 or more days in the past week” to how often they felt either of the following: I felt that everything I did was an effort I could not get “going”		211/595 (35%)
Weight loss	Unintentional weight loss of ≥10 lbs in past year		47/625 (7.5%)
	Women	Men
Physical activity	AMI<270 kcal/week	AMI<383 kcal/week	169/521 (32%)
Grip strength	≤17.0 kg if BMI<=23 kg/m2	≤29.0 kg if BMI<=24 kg/m2	202/617 (33%)
	≤17.3 kg if BMI 23-26 kg/m2	≤30.0 kg if BMI 24-26 kg/m2
	≤18.0 kg if BMI 26-29 kg/m2	≤31.0 kg if BMI 26-28 kg/m2
	≤21.0 kg if BMI>29 kg/m2	≤32.0 kg if BMI>28 kg/m2
Walk time	≥7 seconds if ≤159 cm tall	≥7 seconds if ≤173 cm tall	249/606 (41%)
	≥6 seconds if >159 cm tall	≥6 seconds if >173 cm tall

AMI, Activity Metabolic Index; BMI, body mass index

Comorbid conditions

Comorbidity was assessed by the Charlson index, with higher values indicative of greater comorbidity burden.

Quality of life

Assessment was performed with the Short Form 36 (SF36). The scales used in these analyses range from 0-100. The SF36 is a generic instrument widely used to measure health status with physical (PCS) and mental (MCS) component summary measures- higher scores indicating better QOL. A score of 50 is used to denote the mean of the US population and each 10-point deviation represents 1 standard deviation from that mean.

Outcomes of the study

The outcome variable of interest was mortality defined as all-cause mortality during follow-up. The second main outcome was MI defined as presence of 2 of 3 following criteria: prolonged (≥20 minutes) ischemic chest pain and elevation of cardiac biomarkers (creatinine kinase -MB or relative index) more than 2 times upper limit of normal, or electrocardiographic changes (ST/T wave changes or new Q waves).

Statistical analyses

Continuous variables are presented as mean ± standard deviation; categorical variables as frequency (percentage). Trends were tested for patient characteristics across the three frailty classifications. For continuous variables, these tests were conducted using linear models with contrasts of group means. Age and sex were added to the models to get age-sex adjusted p-values. For binary variables, the Armitage trend test was used for unadjusted comparisons. For age-sex adjusted p-value a logistic regression model was used and the frailty variable scored as 0, 1 and 2 for the three groups tested. All hypothesis tests were two-tailed with a 0.05 significance level.

Because frailty classification was unavailable in 84 patients due to missing data, multiple imputation methods were employed to handle missing data for follow-up event analyses. Five multiple imputation data sets were created using the aregImpute function in S-Plus (available from the Hmisc library by Frank Harrell). Proportional hazards regression models were then estimated on each data set and the resulting estimates combined according to Rubin's rules. ¹²

The Net Reclassification Improvement (NRI) and Integrated Discrimination Index (IDI) for the follow-up event models were calculated at the 1 year follow-up point using equations involving predicted event rates from the original article by Pencina et al.¹³ Predicted survival rates from Cox proportional hazards models were used in equations 8 and 9 of the Pencina paper for NRI, as well as equation A9 of the appendix for IDI. The c-statistic (AUC) for the follow-up models was calculated using the modified version of the c-statistic proposed by Harrell et al.¹⁴

Results

We screened 1545 patients of whom 629 (508 at Rochester and 121 at La Crosse) consented. More men (69% vs. 60%) and younger patients (74.3±6.4 years vs. 76.0±6.9 years) consented as compared with non-consented patients. Six hundred twenty (99%) of the patients were white. Of the 629 study participants, 117 (18.6%) were frail, 298 (47.4%) had intermediate frailty, and only 130 (20.6%) were not frail. Frailty classification could not be assessed in 84 patients (13.3%) due to incomplete or inaccurate Minnesota Leisure Time Activity Questionnaire form completion. The most prevalent frailty measure was the 15-feet walk test (41%) and the least common measure was the weight loss of more than 10 lb in the preceding year

As compared to those with and without intermediate frailty, frail patients were older, with higher body mass index, more likely to be female, and had greater number of comorbid conditions, including diabetes mellitus, hypertension, and chronic kidney disease (Table 2). Other comorbidities including prior heart failure, coronary artery bypass surgery, and MI were also more prevalent in frail patients as compared with intermediate frailty and patients not determined to be frail. Multivessel disease was seen more frequently in frail patients. Patients with frailty/intermediate frailty had a longer mean in-hospital stay following PCI (3.9 ± 3.3 days versus 2.5 ± 1.9 days, P<0.001) (data not shown).

Table 2. Baseline characteristics of patients with and without frailty.

Patient characteristics by frailty class
Variable	Not Frail (N=130)		Intermediate Frailty (N=298)		Frail (N=117)		P-value	Adj. P-value†
Age, mean ± SD	72.6	± 5.8	74.6	± 6.0	77.4	± 6.8	<.001
Men, No. (%)	106	(82%)	205	(69%)	65	(56%)	<.001
Body mass index	29.0	± 4.2	30.1	± 5.4	30.7	± 6.6	0.014	<.001
Hypertension, No. (%)	91	(72%)	244	(82%)	103	(88%)	0.001	0.008
Current smoker, No. (%)	4	(3%)	18	(6%)	7	(6%)	0.30	0.047
Diabetes mellitus, No. (%)	20	(16%)	93	(31%)	47	(40%)	<.001	<.001
Chronic kidney disease, No. (%)	5	(4%)	39	(13%)	28	(25%)	<.001	<.001
Peripheral Arterial Disease, No. (%)	7	(6%)	29	(10%)	25	(22%)	<.001	<.001
Congestive Heart Failure, No. (%)	14	(11%)	47	(16%)	32	(28%)	<.001	0.013
Atrial Fibrillation, No. (%)	9	(7%)	41	(14%)	30	(26%)	<.001	<.001
History of CVA/TIA, No. (%)	11	(9%)	41	(14%)	23	(20%)	0.011	0.004
History of vascular surgery, No. (%)	11	(9%)	42	(14%)	25	(22%)	0.003	0.005
History of congestive heart failure, No. (%)	13	(10%)	40	(14%)	33	(29%)	<.001	0.004
History of MI, No. (%)	29	(23%)	94	(32%)	41	(38%)	0.013	0.043
Prior PCI, No. (%)	38	(29%)	117	(39%)	41	(35%)	0.31	0.28
Previous CABG, No. (%)	21	(16%)	79	(27%)	39	(33%)	0.002	0.012
Multivessel disease, No. (%)	70	(56%)	182	(67%)	79	(71%)	0.017	0.005
Left main 70% stenosis, No. (%)	11	(9%)	26	(10%)	18	(17%)	0.072	0.14
Chronic lung disease/COPD, No. (%)	9	(7%)	35	(12%)	31	(26%)	<.001	<.001
Rheumatologic disease, No. (%)	45	(35%)	135	(46%)	68	(58%)	<.001	0.005
Any tumor, No. (%)	32	(25%)	64	(22%)	25	(22%)	0.53	0.44
Metastatic solid tumor, No. (%)	13	(10%)	39	(13%)	14	(12%)	0.68	0.89
Lymphoma, No. (%)	0	(0%)	2	(1%)	2	(2%)	0.12	0.22
Unexpected fall within 6 months, No. (%)	5	(4%)	46	(15%)	27	(23%)	<.001	<.001
Fracture, No. (%)	42	(33%)	105	(36%)	36	(31%)	0.76	0.84
Sleep apnea, No. (%)	16	(13%)	50	(17%)	27	(24%)	0.022	0.002
Charlson Index, Median (Q1, Q3)	2.0	(1.0, 3.0)	2.0	(1.0, 4.0)	3.0	(2.0, 5.0)	<.001	<.001
Sachdev, Median (Q1, Q3)	1.0	(1.0, 3.0)	3.0	(1.0, 4.0)	3.0	(2.0, 6.0)	<.001	<.001

^†P-values adjusted for age and sex;

CABG, coronary artery bypass surgery; COPD, chronic obstructive airway disease; CVA/TIA, cerebrovascular accident/transient ischemic attack; MI, myocardial infarction; PCI, percutaneous coronary intervention.

Follow-up events

Ninety-eight percent of patients had a follow-up contact within two years prior to analyzing the data. The median follow-up was 35.0 months (inter quartile range, 22.7-42.9 months). Seventy eight (12%) patients died and 70 had non-fatal MI, and 140 (22%) had death/MI.

Long-term outcomes

Univariate associations with long-term mortality and mortality/MI are shown in Table 3. The MCRS, frailty, Charlson index and SF-36 physical component scores were significantly associated with death and death/MI.

Table 3. Unadjusted associations with the three follow-up endpoints.

	Death			Death/MI
	HR	95% CI	P-value	HR	95% CI	P-value
Mayo Clinic Risk Score	1.15	(1.08, 1.22)	<0.001	1.10	(1.04, 1.15)	<0.001
Comorbidities
Charlson Index	1.12	(1.06, 1.18)	<0.001	1.05	(1.01, 1.10)	0.024
Frailty group			<0.001			<0.001
Intermediate Frailty	1.90	(0.85, 4.25)	0.120	1.40	(0.84, 2.33)	0.192
Frail	5.36	(2.41, 11.9)	<0.001	3.04	(1.80, 5.15)	<0.001
Health status variables			<0.001			0.032
SF-36 Mental Comp. (per 10 point decrease)	1.02	(0.81, 1.27)	0.893	1.09	(0.92, 1.29)	0.326
SF-36 Physical Comp. (per 10 point decrease)	1.72	(1.36, 2.18)	<0.001	1.24	(1.04, 1.47)	0.015

SF-36, Short-Form 36

Multivariate Predictors of Late Outcome

In models containing both the MCRS and frailty, comorbidity, and QOL variables, MCRS was an independent predictor of long-term mortality and mortality/MI, Table 4. Frailty [hazard ratio (HR) 2.74, 95% confidence interval (CI), 1.12, 6.71], comorbidity (HR, 1.09, 1.03, 1.15), and a ten point decrease in the physical component of the SF-36 for QOL (HR, 1.32, 1.02, 1.71) were significant risk factors for long-term mortality. When added to the MCRS, all variable sets produced substantial gains in prognostic information for follow-up mortality, increasing the c-statistic from 0.628 to 0.675 with the addition of frailty, to 0.671 with comorbidity, and 0.694 with SF-36 for QOL. The addition of frailty and comorbidity was associated with (22% and 34% respectively) net reclassification improvement (NRI). For long-term mortality or MI, only frailty (HR, 2.45, 1.33, 4.53) was a risk factor with significant NRI to the existing MCRS. Adding all three variables to the mortality model, i.e. frailty, comorbidity, and SF-36, increased the C-index by 0.097 to 0.724 and NRI by 43%, p=0.007. Similar addition to a model for mortality/MI resulted in modest improvement in the C-statistic, NRI, and IDI of the risk model based on the MCRS.

Table 4. Multivariate model results for the three primary follow-up endpoints.

Variable	C-statistic (AUC)	C-statistic Increase	NRI %a	IDIa
DEATH
MCRS	0.628
+ Frailty/Intermediate frailty	0.675	0.047	22% (p=0.13)	0.007
+ Charlson index for comorbidity	0.671	0.043	34% (p<.001)	0.016
+ Short-form-36 (Mental and Physical)	0.694	0.066	23% (p=0.010)	0.010
+ Frailty, Charlson, and Short form-36	0.724	0.097	43% (p=0.007)	0.027
DEATH/myocardial infarction
MCRS	0.573
+ Frailty/Intermediate frailty	0.607	0.035	16% (p=0.038)	0.013
+ Charlson index for comorbidity	0.576	0.003	5.5% (p=0.23)	0.004
+ Short form-36 (Mental and Physical)	0.587	0.015	2.8% (p=0.62)	0.001
+ Frailty, Charlson, and Short form-36	0.608	0.035	18% (p=0.027)	0.016

IDI, integrated discrimination index; NRI, net reclassification index

^a– Based on event rates at one year.

Table 5 shows the number of participants reclassified with the addition of the frailty, comorbidity, and QOL variables during 2 years of follow-up as compared to the risk categories assigned by the MCRS-only model. For example, for mortality, there were 12 patients who died in the first year who were classified by the MCRS only model as 0.04-0.10 risk. Of those 12 patients, 6 remained in that classification in the MCRS+ frailty, comorbidity, and QOL model, 1 was incorrectly moved down to 0.02-0.04, and 5 were correctly moved up to 0.10-0.20 risk. So the net reclassification improvement was 4 patients out of 12. Similarly, with the addition of frailty, comorbidity, and QOL variables, participants were moved into higher risk categories in the first and second years of follow-up.

Table 5. Reclassification of Predicted Mortality with the Addition of Frailty, Comorbidity, and QOL variables*.

MCRS only	MCRS + Frailty, Comorbidity, Poor QOL					Total
n (row %)	<0.02	0.02-0.04	0.04-0.10	0.10-0.20	0.20+
Patients with event by 365 days
<0.02	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0
0.02 - 0.04	1 (8.3)	7 (58.3)	3 (25.0)	1 (8.3)	0 (0.0)	12
0.04 - 0.10	0 (0.0)	1 (8.3)	6 (50.0)	5 (41.7)	0 (0.0)	12
0.10 - 0.20	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (100.0)	1
0.20+	0 (0.0)	0 (0.0)	0 (0.0)	1 (100.0)	0 (0.0)	1
Total	1	8	9	7	1	26
Patients with event after 365 days
<0.02	2 (100.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	2
0.02 - 0.04	3 (16.7)	6 (33.3)	8 (44.4)	1 (5.6)	0 (0.0)	18
0.04 - 0.10	1 (3.4)	4 (13.8)	15 (51.7)	7 (24.1)	2 (6.9)	29
0.10 - 0.20	0 (0.0)	0 (0.0)	1 (33.3)	2 (66.7)	0 (0.0)	3
0.20+	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0
Total	6	10	24	10	2	52
Patients censored before 365 days
<0.02	0 (0.0)	1 (100.0)	0 (0.0)	0 (0.0)	0 (0.0)	1
0.02 - 0.04	18 (36.0)	16 (32.0)	16 (32.0)	0 (0.0)	0 (0.0)	50
0.04 - 0.10	2 (10.0)	8 (40.0)	8 (40.0)	2 (10.0)	0 (0.0)	20
0.10 - 0.20	0 (0.0)	0 (0.0)	0 (0.0)	1 (50.0)	1 (50.0)	2
0.20+	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0
Total	20	25	24	3	1	73
Patients event-free at 365 days
<0.02	14 (70.0)	3 (15.0)	3 (15.0)	0 (0.0)	0 (0.0)	20
0.02 - 0.04	123 (41.6)	122 (41.2)	47 (15.9)	2 (0.7)	2 (0.7)	296
0.04 - 0.10	17 (11.2)	53 (34.9)	70 (46.1)	11 (7.2)	1 (0.7)	152
0.10 - 0.20	0 (0.0)	0 (0.0)	4 (44.4)	5 (55.6)	0 (0.0)	9
0.20+	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0
Total	154	178	124	18	3	477

Frailty (Fried classification), comorbidity (Charlson index), and Short-Form 36 MCRS, Mayo Clinic risk score

Discussion

The key points from this study on patients undergoing PCI are that, 1) sixty-six percent of patients ≥ 65 years undergoing PCI were either frail or had intermediate frailty; 2) frailty, comorbid conditions, and the physical component of SF-36 were independent risk factors for long-term mortality; 3) long-term mortality/MI were significantly higher in patients determined to be frail ; and 4) the addition of frailty, comorbidity, and QOL significantly improve the prognostic ability of the Mayo Clinic risk assessment mortality model based on traditionally-assessed cardiovascular risk factors. These findings highlight important characteristics of the geriatric population that warrant consideration for prognosis and follow-up of elderly patients undergoing PCI.

These findings substantially expand previous insights into the importance of frailty among the elderly. The prevalence of frailty and comorbidity increases with age and varies across studies, reflecting differences in population characteristics, inclusion criteria, and definitions used.^{9,11, 15-17} The results of the current study further illuminates the clinical consequences of the frailty phenotype in patients deemed healthy enough to undergo PCI for the treatment of CAD. Frail patients in the CAD population had a higher body mass index, and weight loss may be a less important criterion for identifying frailty. Compared with non-frail patients, prior studies noted higher prevalence of frailty in patients with CAD and the findings of the present study are congruent with the prior literature.¹⁸ One small prospective study, using Fried criteria, observed frailty in 27% of patients older than 70 years with significant CAD at cardiac catheterization.¹⁹ Given the high prevalence of frailty in patients with CAD, its assessment will further classify the at-risk elderly patients along a spectrum of functional status with the traditionally-measured coronary risk factors.

Frailty, Comorbidity, Quality of Life and long-term outcomes following PCI

There are several reasons why frailty, comorbidity, and QOL should be included in the risk assessment of an elderly patient undergoing a PCI procedure. First, the presence of frailty, worse health status, and a higher number of comorbid conditions identify a subset of elderly patients who are at-risk of dying or having an MI during follow-up after a successful PCI procedure. The prognostic influence of health status on long-term outcomes has been reported in patients with CAD or patients undergoing CABG but comprehensive measurement of other components, for example, frailty, is lacking in patients at the time of PCI.^20-22 Second, the magnitude of risk associated with the presence of frailty, comorbidity, and QOL is greater than predicted from the MCRS. For example, the presence of frailty in a patient undergoing PCI increased the risk of mortality 5 fold and mortality/MI about 2.5 fold than a patient not determined frail at the time of the coronary intervention. Third, these tests are simple to administer and are entirely non-invasive. Assessment of frailty and comorbidity has not usually been included for long-term outcome assessment despite several population-based studies demonstrating their prognostic significance.^{10, 23,24} Reasons for this are unknown but could relate to limited familiarity with the data and tests, perceived concerns about the complexity of obtaining such measurements in the clinic or hospital setting, or skepticism on relative utility of frailty, comorbidity, and QOL to assess mortality and other risks. Fourth, determination of these variables may be useful when counseling patients regarding the risk of worse long-term outcomes in conjunction with traditional risks included in the current PCI models.²⁵ For example, if a patient is at moderate risk for long-term worse outcomes, he/she may decide against the procedure if he/she knows the incremental risk from associated frailty, comorbidity, and poor QOL. Alternatively, a higher risk person may be considered a better candidate who is not frail with good functional status, even if his MCRS is high. This is relevant as it resulted in reclassifying individuals to new and clinically meaningful risk categories. Our data indicate that by the addition of frailty, comorbidity, and QOL variables, 33% and 42% of participants who died in the first year of follow-up were correctly moved to higher risk categories assigned by the MCRS risk model. Identification of higher risk can prompt comprehensive geriatric evaluation or therapy targeting factors that may be contributing to frailty of poor health status. We previously demonstrated that comorbidity improves prediction of long-term mortality in patients undergoing PCI.^8,10 The present study amplifies and extends these results and underscores the need to routinely measure frailty and health status in these patients with significant prognostic implications.

Frailty, Comorbidity, QOL and risk-stratification models for PCI

There are several parsimonious PCI risk prediction models that provide information on in-hospital events, yet which lack information on long-term prognosis.^6,25,26 Additionally, the covariates that predict in-hospital outcomes may not predict follow-up events underscoring the need to develop models that accurately predict long-term outcomes. The C-statistic of the MCRS model based on traditional cardiovascular risk factors was 0.659 for long-term mortality. Inclusion of health status, burden of comorbid conditions, and measures of frailty, currently not included in any of the contemporary PCI risk models, improved the C-statistic to 0.753 along with significant changes in the NRI and IDI. We previously demonstrated improvement in the risk stratification of community cohorts presenting with MI and patients undergoing PCI with the addition of comorbid conditions, underscoring the need for their inclusion in estimation of long-term prognosis.¹⁰ The present study extends that observation with the inclusion of measures of frailty and poor health status that predicted long-term mortality above and beyond the traditional cardiovascular risk factors and improved the discriminatory ability of the existing MCRS.

Limitations

The principal limitation of this study is that it represents the experience of two centers on patients undergoing PCI. The race, ethnic composition and other characteristics of our population may limit generalizability to other populations. There is likely some selection bias as only 41% of the eligible cohort was consented. The timing of assessment following a recent invasive procedure may increase the prevalence of problems with the walk test and increase the prevalence of intermediate and/or frailty. Additionally, the applicability of this model to pre-procedure timeframe and the influence of PCI on the self-reporting of measures of quality of life can't be answered from the present study. Furthermore, we used the MCRS as our base predictive model, though it was not intended for use as a predictor of follow-up events. It may be that more sophisticated follow-up models based on clinical and angiographic parameters would not see such improvement with the addition of frailty, comorbidity, and QOL variables.

Strengths of the Study

Whether frailty, comorbidity, or QOL add useful information for risk prediction is unknown. The present study was undertaken to address these shortcomings. With the use of these variables, it is possible to define groups with 2- to 5-fold difference in mortality or mortality/MI. Adding frailty, comorbidity, and QOL to conventional risk factors improves the C statistics significantly. As the C statistic has been criticized as insensitive to small changes in predictive accuracy, we also calculated a newer measure called the NRI. This metric improves when frailty, comorbidity, and QOL correctly assign an individual to higher or lower risk categories. The addition of these variables led to a significant change in the NRI for long-term mortality. Our data indicate that 43% are moved to a higher risk category by the addition of frailty, comorbidity, and poor QOL for long-term mortality following PCI. When applied to the existing MCRS model, these data may further improve the risk stratification for the patients undergoing high-risk coronary revascularization procedures. Health status is a recognized priority for patient-oriented research yet clinicians do not use it. Thus, fostering its purposeful use can improve management by moving the focus of clinical care from the disease to the patient

Conclusions

This study demonstrates that frailty, comorbidity, and poor QOL are highly prevalent in elderly patients undergoing PCI. Presence of these variables is associated with worse long-term outcomes and inclusion of these measures improves the discriminating ability of the risk model derived from routine cardiovascular risk factors. Frailty, comorbidity, and QOL variables need to be included in future risk stratification models that predict long-term mortality after PCI.

What Is Known

• Frailty, comorbidity, and quality of life are important variables that predict long-term outcomes in older adults.

• Current risk-prediction models do not routinely incorporate these measures in the risk assessment of the patients undergoing percutaneous coronary interventions.

What The Study Adds

• Adding frailty, comorbidity, and quality of life to the traditional cardiovascular risk factors significantly improves risk prediction of older patients undergoing percutaneous coronary interventions.

• About 40% patients were moved to the higher risk category with the addition of these variables.