Treatment outcomes in older patients presenting to a radiation oncology clinic based on an electronic health record-based frailty index

Introduction

For older adults with cancer, assessments of frailty, which take into account physical and functional performance, medical comorbidity, cognitive function, psychological and nutritional status, inform treatment decisions more reliably than age alone [1–4]. Among studies of frailty in patients undergoing radiotherapy (RT), few have demonstrated an association between pre-treatment comprehensive geriatric assessment (CGA) and morbidity or mortality after RT [5]. For CGA studies that demonstrated an association, several domains were analyzed and assessments were performed by geriatricians and/or geriatric oncologists using multiple tests and instruments [6,7]. More simplified pragmatic tools are needed to standardize and facilitate frailty assessment in the radiation oncology clinic.

The electronic frailty index (eFI) is a readily available metric embedded within our institutional electronic health record (EHR) that is calculated at the time of initial consultation [8]. Frailty determined by the eFI has been significantly associated with overall mortality, hospitalization, emergency department (ED) visits, and falls in older patients [8]. In patients over 65 undergoing unselected elective surgery or lower extremity revascularization surgery, eFI was associated with discharge to acute care facilities, 30-day readmission, and six-month mortality [9,10]. eFI has also been associated with the intensity of systemic therapy delivered in patients with acute myeloid leukemia and with survival in patients with solid tumors receiving chemotherapy [11,12]. To our knowledge, EHR-based metrics that elucidate fitness – or frailty – have not been studied in older patients planned for RT. In this study, we aim to describe overall survival, treatment completion, emergency visits, and hospitalizations by frailty status in patients consulting for RT.

Materials and Methods

In this IRB-approved (IRB00102365) retrospective study, we performed a systematic query of the EHR and identified consecutive patients with an eFI automatically calculated at the time of initial consultation. Patients were seen within our radiation oncology network (one tertiary academic center, two community affiliate sites) between August 2019 and July 2022. The eFI, which has been available in the institutional EHR (Epic) since 2019, is an EHR-integrated metric that compiles diagnosis codes, vital signs, laboratory measurements, medication, functional status, and smoking data over a two-year period [8,9].The eFI score is calculated automatically at the point of contact as the unweighted sum of 54 candidate factors already present in the EHR (i.e., no information is required to be inputted by the encounter provider to calculate it), divided by the sum of the total number of non-missing factors. For a complete list of candidate factors, see Supplemental Table 1. At least two outpatient visits with a measured blood pressure and at least 30 of the 54 factors (including at least 9 of the 20 laboratory measurements, smoking status, and body mass index) over a two-year period are required for eFI score calculation. The eFI ranges from 0 to 1 with higher scores indicating a greater level of frailty. Based on eFI score, patients are categorized as fit (eFI <0.10), pre-frail (eFI between 0.10 and 0.21) and frail (eFI >0.21) [8]. Additional data were abstracted from the EHR including date of encounter, sex, race, ethnicity, primary malignancy, stage (where applicable), and intent of therapy (curative versus palliative). For those treated with RT, dates of RT, prescribed dose, fraction number, concurrent chemotherapy (for diagnoses where concurrent chemotherapy is a consideration in the curative setting), and whether the patient completed the course of prescribed RT were also abstracted. Acute care outcomes were assessed in the RT group and included any ED visit or hospitalization at our facility within 30-, 90-, and 180-days from the date of RT initiation.

Data were summarized using count, frequency, and 95% Clopper-Pearson confidence intervals for categorical variables and median (interquartile range, IQR) for continuous variables. Overall survival (OS) was calculated from the date of consultation to the date of death from any cause or last follow-up visit in the EHR, estimated using the Kaplan Meier method, and compared between groups using the log-rank test. Cox proportional hazards modeling was used to assess for an association between frailty and OS while controlling for clinically important factors including age, sex, race, primary cancer site, and stage. After assessing each variable for proportionality using the Schoenfeld test, a multivariable model including eFI category, sex, and stage (0 through IV) was generated. All analyses were performed using R version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

A total of 399 patients were seen in consultation and had an eFI calculated. Demographics of included patients are listed in Table 1. The median eFI score was 0.14 (IQR, 0.09–0.21). According to eFI score, 119 (29.8%) patients were categorized as fit, 194 (48.6%) as pre-frail, and 86 (21.6%) as frail. Baseline patient characteristics did not differ by frailty status (Supplemental Table 2).

Table 1:

Prescribed radiotherapy and treatment outcomes by frailty category in patients treated with curative or palliative radiotherapy

Curative Intent (n=183)
	Fit (n=50)	Pre-Frail (n=95)	Frail (n=38)
Prescribed RT Dose (Gy)	60.0 (50.1–66.0)	52.6 (50.0–60.0)	51.8 (48.5–60.0)
Number of Prescribed Fractions	28 (20–30)	25 (16–30)	25 (11–28)
Elapsed Treatment Duration	38 (26–44)	32 (21–42)	24 (15–41)
Concurrent Chemotherapy*	15 (68.2)	18 (51.4)	9 (60.0)
Incomplete Treatment	0 (0)	4 (4.2)	4 (10.5)
Emergency Visit or Hospitalization
30 days	2 (4.0)	9 (9.5)	5 (13.2)
90 days	4 (8.0)	12 (12.6)	6 (15.8)
180 days	5 (10.0)	12 (12.6)	6 (15.8)
Palliative Intent (n=69)
	Fit (n=19)	Pre-Frail (n=31)	Frail (n=19)
Prescribed RT Dose (Gy)	30.0 (21.5–30)	30.0 (22.0–30.0)	30.0 (30.0, 30.0)
Number of Prescribed Fractions	10 (5–10)	10 (5–10)	10 (10–10)
Elapsed Treatment Duration	13 (8–15)	13 (6–15)	13 (11–14)
Incomplete Treatment	1 (5.3)	1 (3.2)	1 (5.3)
Emergency Visit or Hospitalization
30 days	2 (10.5)	4 (12.9)	1 (5.3)
90 days	3 (15.8)	5 (16.1)	3 (15.8)
180 days	3 (15.8)	6 (19.4)	3 (15.8)

All data are presented as count (frequency) and median (interquartile range).

^*Including 72 patients with brain, gastrointestinal, genitourinary, gynecologic, head and neck, lung, and skin cancers treated with curative intent where concurrent chemotherapy is a consideration.

In total, 252 of 399 patients were treated with RT. Of these 252 patients, this included 69 of 119 (58%) fit patients, 126 of 194 (65%) pre-frail patients, and 57 of 86 (66%) frail patients. Overall, the intent of treatment was curative in 73% and palliative in 27%. Table 1 displays the RT details, treatment completion and acute care use in patients treated with curative or palliative intent. For patients treated with curative RT, median prescribed dose was 52–60 Gy in 25–28 fractions over an elapsed duration of 24–38 days. Concurrent chemotherapy was provided in 51% to 68% of patients. For patients treated with palliative intent, median RT dose was 30 Gy over 10 fractions over an elapsed duration of 13 days. The rates of incomplete treatment by frailty category for patients treated with curative intent were: 0% fit (95% CI: 0–7.1), 4.2% pre-frail (95% CI: 1.2–10.4), and 10.3% frail (95% CI: 2.9–24.8). The corresponding rates of incomplete treatment for patients treated with palliative RT were: 5.3% (95% CI: 0.1–26.0), 3.2% (0.1–16.7), and 5.3% (0.1–26.0), respectively. The rates of ED visits or hospitalization within 90 days for patients treated with curative intent were 10%, 13%, and 16% for fit, pre-frail, and frail patients, respectively. Corresponding 90-day rates of ED visit/hospitalization for patients treated with palliative RT were 16%, 19%, and 16%, respectively (Table 1).

Median follow-up was 19.3 months. eFI was associated with OS in all patients seen in consultation (p=0.002) and in patients treated with curative RT (p=0.01, Figure 1A), but not in patients treated with palliative RT (p=0.38, Figure 1B) or patients seen in consultation but not ultimately treated with RT (p=0.08, Figure 1C). Two-year OS rates for fit, pre-frail and frail patients were 71.7%, 58.4%, and 51.1%, respectively. Corresponding two-year OS rates for patients treated with curative RT were 89.4%, 73.6%, and 61.2%. After adjusting for sex and stage, the hazard of death was higher for pre-frail (hazard ratio [HR] 3.59, 95% CI: 1.50–8.55) and frail (HR 5.78, 95% CI: 2.19–15.23) patients treated with curative RT, compared to fit patients (Supplemental Table 3).

Figure 1:

Overall survival in patients treated with curative RT (A), palliative RT (B) or not treated with RT (C) by eFI frailty category.

Discussion

In older patients with cancer, CGA provides a valuable opportunity to identify the various deficits that comprise frailty and to intervene on these deficits to improve the quality of life and medical outcomes [1]. However, currently available methods of CGA have not been associated with meaningful outcomes after RT and can be complex and logistically cumbersome to obtain [1,5–7]. Given the intersection of physical, cognitive, nutritional, psychological status, and available support systems involved when assessing older patients for cancer treatment, more efficient, standardized methods to evaluate frailty are needed to guide decision-making. The eFI has been developed to identify patients most at-risk for all-cause mortality, acute care usage (emergency department visits and hospitalizations), and falls [8]. Now embedded into the EHR and available at the point of care for patients with sufficient component medical data available, it is routinely included as part of our departmental consultation notes.

In this study, we describe prescribed treatments, duration of therapy, treatment incompletion, acute care visits, and survival in patients based on eFI-derived frailty category in older patients (≥65 years) with cancer presenting to a radiation oncology clinic in consultation for curative or palliative RT. Though additional studies are warranted to more clearly elucidate the prognostic characteristics of the eFI and whether its use in the clinic meaningfully impacts patient care or outcomes, the eFI is a unique and efficient metric that warrants further study. Future prospective studies should determine whether incorporating the eFI at the time of consultation assist radiation oncologists with counseling and selection of appropriate fractionation schedules and meaningfully impacts patient care or outcomes. For example, shorter courses could be selected or additional on-treatment monitoring could be employed in eFI pre-frail or frail patients in an effort to minimize treatment incompletion and acute care visits. Acute and late toxicity of RT could also be considered an important endpoint in a prospective study, where the grading and monitoring of adverse events is more standardized and reproducible.

The eFI has been shown to predict postoperative outcomes in patients undergoing outpatient elective surgeries and open revascularization procedures [9,10]. It has also been studied in older patients with cancer planned for systemic therapy [11,12]. Cheng et al. evaluated 166 patients with acute myeloid leukemia and found that patients with increasing levels of frailty were more frequently treated with less intensive systemic therapy regimens [11]. These findings are similar to ours, in which fewer fractions and shorter treatment times were prescribed in frail patients treated with curative RT, while the rate of RT incompletion was still highest among frail patients. This suggests that additional on-treatment support or more systematic implementation of hypofractionation (where appropriate) may be informed by the eFI [13]. In contrast to patients with leukemia, eFI was strongly associated with OS in patients treated with curative RT, which may be related to the better prognosis of unselected patients treated with RT compared to those with acute leukemia.

As with more formal CGA methods, careful consideration of the eFI frailty status by the treating oncologist could serve as an opportunity to intervene on potentially reversible sources of frailty. Although the actual mathematical calculation itself is hidden to the provider within the EHR and the ultimate score and frailty category is all that is displayed, a general understanding of the component factors could aid oncology providers in directing referrals. For instance, for patients identified as frail or pre-frail, a review of the patient’s medical, social, and functional history could identify factors such as unmanaged comorbidities, vital sign abnormalities, laboratory measurements, and functional deficits for which referrals for medical management, physical therapy, counseling, smoking cessation, and pharmacy consultation for polypharmacy evaluation could be considered. While standardized processes are not currently available in our system, investigations into eFI-based frailty as a trigger for additional supportive services or referrals for various patient populations are ongoing.

The study is limited by its retrospective nature, small sample size, and the focused patient population studied (those with eFI available at consultation encounter). The study design and sample size limit this study to primarily descriptive, hypothesis-generating analyses. A prospective study with clearly defined clinically relevant endpoints would be needed to comprehensively evaluate the prognostic capability and clinical utility of the eFI. Due to the focused nature of our methodology for patient identification for this cohort, only patients with an eFI calculated were identified. Therefore, we did not collect encounters without the eFI, and do not have data on the total number of patients seen in clinic during the study period, or for those without an eFI calculated, which limits our understanding of the availability and applicability of eFI to all-comers. This also precludes a more in-depth analysis of reasons for inability to calculate eFI; this may be in part related to the lack of two outpatient encounters within the last two years from which the eFI is calculated. As a tertiary academic referral center, patients presenting to our institution may seek their outpatient primary care elsewhere, and the candidate factors to calculate the eFI may not be available in our her; the inherent requirements (intended to create a robust frailty index per standard procedure) are at least 30 of 54 candidate factors including smoking status, body mass index, blood pressure, and 9 of 20 laboratory measurements are required for eFI calculation [14]. Patients presenting to our radiation oncology clinic who are established elsewhere for routine care may not satisfy these requirements, as many candidate factors are closely aligned with values obtained at routine primary care visits. Additionally, since we do not have data on acute care visits outside our hospital system, our observed rates of acute care utilization may under-represent the true value and limit our ability to distinguish this metric between frailty categories. While the total number of outside acute care visits is unknown, as a referral center with over 50 hospitals in our catchment area, the total number of patients seeking acute care at outside, unaffiliated hospitals during our study period is likely higher than our estimates herein.

In conclusion, we describe RT treatment and acute care visits based on frailty determined by the eFI and identify eFI frailty category to be associated with survival for patients treated with curative-intent RT. Further study is warranted to better understand the therapeutic and prognostic impacts of this metric, and should focus on availability, generalizability, prognostic performance, impact on clinical care, and subsequent clinical outcomes.

Abbreviations:

eFI

electronic frailty index

EHR

electronic health record

RT

radiotherapy