Assessment of an Embedded Primary Care-Derived Electronic Health Record Frailty Index (eFI) in Older Adults with Acute Myeloid Leukemia

Introduction

Assessment of frailty is central to medical decision-making for older adults with newly diagnosed acute myeloid leukemia (AML), but validated frailty tools used at the point of care are lacking in this setting. An electronic frailty index (eFI), based on the theory of deficit accumulation, has been developed from routinely collected outpatient primary care data and embedded in the electronic health record (EHR) at Wake Forest Baptist Health (WFBH). The embedded eFI predicts hospitalization and mortality for older adults in a primary care cohort but has yet to be assessed for older adults with AML. The objectives of this analysis were to evaluate this tool among a cohort of newly diagnosed older adults with AML by determining the percent with available data to calculate a score, assessing the distribution of eFI categories (fit, pre-frail, and frail) at the time of treatment by therapy type, and exploring an association of eFI categories with survival.

Methods

Participants in this retrospective cohort study were adults aged ≥ 60 years with newly diagnosed AML from January 2018 - October 2020 who received treatment at WFBH. Patients with acute promyelocytic leukemia (APL) were excluded. Calculation of the EHR embedded eFI requires at least two ambulatory visits over a 2-year period. The eFI comprises demographic information, vital signs, smoking status, ICD-10 diagnosis codes, select outpatient laboratory measurements, and functional information (if available from past Medicare Annual Wellness Visits) during the 2 years prior to diagnosis. For this analysis, eFI was derived using date of treatment initiation as the reference date. Frailty status is categorized as fit (eFI ≤0.10), pre-frail (0.10<eFI≤0.21), and frail (eFI >0.21) based on the proportion of deficits present over the total number evaluated. Intensive therapy is defined as receipt of anthracycline-based chemotherapy. Less-intensive therapy includes hypomethylating agents, low dose cytarabine, and/or venetoclax. The association of eFI category to therapy type was evaluated using a chi square test. Median survival was estimated using the Kaplan-Meier method and compared using log-rank tests.

Results

Among N=163 older adults treated for non-APL AML, 78 (43.3%) had a calculable eFI score. Among those with a calculable score, average age was 74.7 years (range 60 to 90), 68% were male, 90% were white, and average hematopoietic cell transplantation-comorbidity index (HCT-CI) score was 3.5 ± 2.7. Approximately 17% had favorable, 23% had intermediate, and 59% had unfavorable cytogenetics; 54% had secondary AML. Among those with a calculable eFI, 38.5% were classified as fit, 48.7% as pre-frail, and 12.8% as frail. For those who received intensive therapy (N=35), the majority were classified as fit (60%) and the remaining were pre-frail (40%); none were frail. The distribution of eFI categories differed for patients who received less-intensive therapies (N=43), with the majority classified as pre-frail (56%), a smaller proportion classified as fit (21%), and almost one-quarter as frail (23 %) (p<0.05). Median overall survival among those who received intensive therapy (12.1 months) was greater than those who received less intensive therapy (2.9 months) (p <0.01). In exploratory analyses, there was no significant difference in overall survival between eFI fit versus eFI pre-frail among those who received intensive chemotherapy (p=0.33). There was also no significant difference in overall survival among those who received less-intensive therapy by eFI category (p=0.38).

Conclusions

The embedded eFI derived from a longitudinal primary care population was not calculable using available data for most older adults with AML. This may reflect a skewed population referred to a Comprehensive Cancer Center that has higher-risk cytogenetics and/or does not receive primary care through the academic medical center. Among those with a calculable eFI, frailty categories significantly differed by treatment type. There was no association between survival and eFI category within treatment type in exploratory analyses. Our observations highlight the opportunity to adapt the EHR embedded eFI data capture process, including incorporation of inpatient labs, to the AML setting to effectively test the utility of a cancer-adapted passive digital marker for frailty.