Patient Reported Outcomes and Subsequent Management in Atrial Fibrillation Clinical Practice: Results from the Utah mEVAL AF Program

Abstract

Background:

Atrial fibrillation (AF) significantly reduces health-related quality of life (HRQoL), previously measured in clinical trials using patient-reported outcomes (PROs). We examined AF PROs in clinical practice, and their association with subsequent clinical management.

Methods:

The Utah mEVAL program collects the Toronto AF Symptom Severity Scale (AFSS) in AF outpatients at the University of Utah. Baseline factors associated with worse AF symptom score (range 0–35, higher is worse) were identified in univariate and multivariable analyses. Secondary outcomes included AF burden, and AF healthcare utilization. We also compared subsequent clinical management at 6 months between patients with better versus worse AF HRQoL.

Results:

Overall, 1,338 patients completed the AFSS symptom score, which varied by sex (mean 7.26 for males vs. 10.27 for females, p<0.001), age (<65, 9.73; 65–74,7.66; >=75,7.58, p<0.001), heart failure (9.39 with HF vs. 7.67 without, p<0.001), and prior ablation (7.28 with prior ablation vs. 8.84, p<0.001). In multivariable analysis, younger age (mean difference 2.92 for <65 vs. ≥75, p<0.001), female sex (mean difference 2.57, p<0.001), pulmonary disease (mean difference 1.88 p<0.001), and depression (mean difference 2.46, p<0.001) were associated with higher scores. At 6-months, worse baseline symptom score was associated with use of rhythm control (37.1% vs. 24.5%, p<0.001). Similar co-factors and results were associated with increased AF burden and health care utilization scores.

Conclusions:

AF PROs in clinical practice identify highly-symptomatic patients, corroborating findings in more controlled, clinical trials. Increased AFSS score correlates with more aggressive clinical management, supporting the utility of disease-specific PROs guiding clinical practice.

Atrial fibrillation (AF) is the most common sustained arrhythmia and is associated with significant symptom burden and reduction in health-related quality of life (HRQoL), similar to other major cardiac diseases.¹ AF treatment is often targeted at symptom management and improving HRQoL. While AF-related HRQoL is rigorously measured with patient-reported outcomes (PROs) in many major clinical trials, PROs are rarely collected in routine clinical practice, despite a major objective of AF treatment being to improve HRQoL.² To date, reports on PROs from outside clinical trials, are primarily limited to small, selected cohorts sampled once to answer specific questions or assess procedural outcomes for AF.^3–11 Rarely are PROs collected longitudinally, across broad AF cohorts.

In order to implement the routine collection of PROs, and use them to guide AF management, additional data are needed. Specifically, it is not clear to what extent PROs in clinical trials reflect those of routine clinical practice and general AF patient populations. Furthermore, it has not been described how closely AF PROs reflect subjective symptom assessment as currently ascertained by treating physicians. To address these shortfalls, we analyzed AF PRO data from the Utah My Evaluation (mEVAL) program.¹² The objectives of this analysis are to: (1) determine the factors associated with worse AF-specific PROs in routine clinical practice; and (2) examine care patterns of highly-symptomatic AF patients, in order to understand the feasibility and value of using AF-specific PROs to guide treatment. We hypothesize that the factors associated with worse HRQoL in routine clinical practice will be similar to those described most frequently in randomized controlled trials. We also hypothesize that patients with worse AF PRO scores will receive more treatment.

Methods

Data Collection:

In 2016, the University of Utah cardiac electrophysiology clinic began administering the Toronto Atrial Fibrillation Severity Scale (AFSS) through its mEVAL program, as part of an institution-wide commitment to implement PROs in clinical care. Details of the clinic-specific and system-wide implementations of PROs have been previously published.^12–14 The AFSS tool was selected because it is a previously validated, simple to use tool, and was easily incorporated into the Utah mEVAL system. All patients scheduled with a clinical cardiac electrophysiology clinician (physician or advanced practice clinician) in the outpatient setting are eligible to complete PROs. The AFSS is a previously-validated and broadly used AF-specific PRO, which includes 20 items across 4 domains: (1) global well-being; (2) health care utilization; (3) AF burden; and (4) AF symptoms. Global well-being is assessed by one question asking about patient well-being from 1–10, and is not an AF-specific measure; it was thus excluded from this analysis. The health care utilization score is calculated from three questions to report visits to the emergency room, specialists, or hospitalizations because of atrial fibrillation within the last year. Scoring of each visit was described as 0–4 visits corresponding 0–4 points respectively, 5 to 10 visits corresponding to 5 points, 10–15 visits corresponding to 6 points, and greater than 15 visits corresponding to 7 points. Leading to a possible range of scores from 0 to 21. AF burden was calculated from four questions about patient perceived AF episode frequency, duration, and severity with a possible score range of 3–30. Symptom score is calculated from 7 AF-related symptoms questions on 5-point Likert scales for a possible range of scores from 0 to 35, and was the primary outcome of interest for this analysis.¹ Patients in whom an AF burden or healthcare utilization domain score could not be calculated (due to missingness) were included if they did have at least one AFSS symptom score calculable.

Patient Cohort

We included patients who (1) had a clinical diagnosis of AF, defined by ≥2 prior encounters in our health system (and ≥1 as an outpatient) with an International Classification of Disease (ICD) code for AF (427.31 [9^th revision] or I48.0, I48.1, I48.2, I48.91 [10^th revision]), as previously described;¹⁵ (2) had an outpatient visit with a University of Utah adult electrophysiology clinician from October 2016 to January 2020; and (3) completed the entire 7-item symptom score domain of at least one AFSS PRO assessment.

The index visit for the cohort was defined as the first electrophysiology visit where a complete AFSS symptom score was available; completion of domains in AF burden and healthcare utilization were not required. The data sources are derived from the health system’s enterprise data warehouse, and include all administrative billing encounters with diagnosis codes (inpatient, outpatient, procedural), as well as medication orders, laboratory results, electrocardiography (ECG) results, and echocardiography results. Clinical comorbidities were measured using previously validated algorithms for use in administrative data analyses of cardiovascular disease, and includes all health system encounters up to and including the baseline visit.^9,10 Comorbidity rates were calculated based on ICD codes as part of clinical billing encounters, as previously described.¹⁴ Medication rates and laboratory values were based on orders and results between 90 days before to 30 days after the index visit. Echocardiography values were derived from the closest studies between 365 days before and 30 days after the index visit.

Treatments at Follow-up

Among patients with at least 6-months of follow-up, we identified those that received any rhythm control treatment from the day following the index visit to 6 months later. The composite outcome of receiving any rhythm control treatment was defined as (1) any prescribed antiarrhythmic drug therapy, (2) incident cardioversion, and/or (3) incident AF catheter ablation. The relationship between baseline score and subsequent rhythm control intervention was analyzed.

Statistical Methods

We reported descriptive statistics as means (SD) or frequencies (percentages). We used analysis of variance (ANOVA) to describe the cross-sectional association between categorical baseline factors with the AFSS domain scores (symptom score, burden, or healthcare utilization). Baseline factors with ≤25% missing data and a p-value < 0.1 from the univariable ANOVA analsyis were included in a subsequent multivariable analyses to explore the joint association of the baseline factors with the AFSS domain scores. We showed several perspectives on the joint association of baseline factors with AFSS domain scores using a nested sequence of regression models relating each domain score to 1) baseline demographics only, 2) baseline demographics and objective data including vitals and laboratory measurements, and 3) baseline demographics, the objective data, and prior treatment information.

In separate analyses, we examined the association of the AFSS domain scores with subsequent treatment. A subgroup of patients with 6-month data available were separated into groups based on AFSS domain scores above or below the median score value. We used a logistic regression to relate each doain score as a continuous predictor variable to rhythm control, first without covariate adjustment, and with adjustment for 1) baseline demographics, 2) baseline demographics and objective data, and 3) baseline demographics, the objective data, and prior treatment information. All statistical tests were peformed in an exploratory fashion using a 2-sided significance level of 0.05, without adjustment for multiple comparisons. We similarly report 95% confidence intervals without multiple comparison adjustment.

Data processing were performed using R (Version 3.6.3), RStudio (Version 1.2.5033), with appropriate packages.^16–19 Statistical analysis was performed using R (Version 3.4.1), RStudio (Version 1.0.153). Analysis of the data collected as part of routine clinical care, and subsequent reporting of anonymized, aggregate data, was approved by the University of Utah Institutional Review Board.

Results

Baseline Characteristics

From October 4, 2016 to January 17, 2020, 1703 patients completed the AFSS symptom score assessment; 1338 had a pre-existing AF diagnosis and were included in this analysis. Baseline characteristics of the analytic cohort are shown in (Table 1). Overall, 37% of the cohort was female (n=500), had a mean age of 67 years, and a mean CHA₂DS₂-VASc score of 3.7. Five hundred fifty three (41%) had heart failure (HF), 468 (35%) had pulmonary disease, 304 (24%) had a prior stroke, and 380 (29%) had diabetes mellitus. Prior AF treatments included cardioversion (25%), antiarrhythmic drug therapy (22%), and ablation (29%).

Table 1.

Baseline characteristics of AF patients completing AFSS symptom score.

	Overall (n=1338)
Age	67.3 (12.4)
Female Sex	500 (37%)
Race
American Indian and Alaska Native	7 (1%)
Asian	24(2%)
Black or African American	8(1%)
Native Hawaiian / Other Pacific Islander	11(1%)
Other	41(3%)
White or Caucasian	1239(93%)
Medical History
Hypertension	1013 (76%)
Diabetes mellitus	380 (29%)
Myocardial infarction	382 (29%)
Heart failure	553 (41%)
Peripheral vascular disease	566 (42%)
History of stroke	304 (24%)
Dementia	32 (2%)
Pulmonary disease	468 (35%)
Severe liver disease	21 (2%)
Depression	393 (29%)
Medical Therapy
Beta-blocker	488 (36%)
Non-dihydropyridine Calcium-channel blocker	144 (11%)
Oral anticoagulation	568 (42%)
Any antiarrhythmic	289 (22%)
CHA2DS2-VASc Score	3.7 (2.0)
Prior cardioversion	334 (25%)
Prior ablation	394 (29%)
LVEF	57.3 (11.2)
Creatinine	1.2 (0.8)
Hemoglobin	13.9 (2.1)

Baseline characteristics, co-morbidities, medical therapies, and laboratory studies.

Values are presented as n (%) or mean (standard deviation), unless otherwise noted.

AF: atrial fibrillation; AFSS: atrial fibrillation severity score; LVEF: left-ventricular ejection fraction based on echocardiogram.

Factors Associated with AFSS Symptom Score

Baseline univariate analysis showed several differences in mean AFSS symptom scores (Table 2), including by sex (mean 10.27 for females vs. 7.26 for males, p<0.001), age (<65, 9.73; 65–74, 7.66; >=75, 7.58, p<0.001), and patients with a history of HF (9.39 with HF vs. 7.67 without HF, p<0.001). Patients with a prior ablation had significantly lower mean scores (7.28 vs. 8.84, p<0.001) (Figure 1).

Table 2.

Unadjusted mean baseline AFSS symptom scores.

		Baseline AFSS Symptom Score (n=1338), mean (SD)	p-value	Baseline AFSS AF Burden Score (n=888), mean (SD)	p-value	Baseline AFSS Health Care Utilization Score (n=1313), mean (SD)	p-value
Age	<65	9.73 (7.98)	<0.001	16.94 (5.69)	0.013	2.41 (2.83)	<0.001
	65–74	7.66 (6.76)		15.55 (6.29)		1.83 (2.28)
	>=75	7.58 (6.75)		15.98 (6.42)		1.84 (2.41)
Sex	Male	7.26 (6.95)	<0.001	15.98 (6.49)	0.21	1.87 (2.37)	0.001
	Female	10.27 (7.45)		16.50 (5.57)		2.34 (2.78)
Race	White	8.42 (7.24)	0.5	16.20 (6.17)	0.87	2.04 (2.51)	0.95
	Non-White	7.91 (7.79)		16.07 (5.80)		2.06 (2.92)
Heart Failure	No	7.67 (6.99)	<0.001	15.61 (5.99)	<0.001	1.76 (2.19)	<0.001
	Yes	9.39 (7.58)		17.07 (6.26)		2.45 (2.93)
Valvular Heart Disease	No	8.31 (7.21)	0.76	16.00 (6.21)	0.42	2.20 (2.57)	0.078
	Yes	8.43 (7.34)		16.33 (6.08)		1.94 (2.52)
Peripheral Vascular Disease	No	8.02 (7.10)	0.034	15.95 (5.96)	0.16	1.94 (2.34)	0.091
	Yes	8.87 (7.51)		16.54 (6.37)		2.19 (2.79)
Hypertension	No	7.58 (7.31)	0.023	15.60 (6.24)	0.11	1.91 (2.20)	0.28
	Yes	8.64 (7.26)		16.38 (6.10)		2.09 (2.64)
Chronic Pulmonary Disease	No	7.37 (6.88)	<0.001	15.77 (6.32)	0.006	1.92 (2.33)	0.013
	Yes	10.27 (7.63)		16.95 (5.72)		2.28 (2.88)
Diabetes mellitus	No	7.90 (7.05)	<0.001	15.84 (6.12)	0.005	2.01 (2.44)	0.4
	Yes	9.59 (7.72)		17.14 (6.10)		2.14 (2.79)
Hypothyroidism	No	8.06 (7.20)	0.005	15.99 (6.07)	0.1	2.02 (2.51)	0.48
	Yes	9.35 (7.46)		16.78 (6.32)		2.13 (2.63)
Renal Disease	No	8.17 (7.33)	0.032	16.24 (6.07)	0.6	1.99 (2.47)	0.13
	Yes	9.24 (7.03)		15.97 (6.40)		2.26 (2.81)
Depression	No	7.26 (6.75)	<0.001	15.71 (6.23)	<0.001	1.95 (2.36)	0.031
	Yes	11.09 (7.81)		17.30 (5.76)		2.28 (2.92)
Prior Stroke	No	8.13 (7.15)	0.02	16.27 (6.14)	0.45	2.03 (2.47)	0.76
	Yes	9.24 (7.69)		15.91 (6.14)		2.08 (2.77)
Prior MI	No	7.93 (7.07)	<0.001	16.20 (6.17)	0.92	1.88 (2.29)	<0.001
	Yes	9.52 (7.69)		16.16 (6.08)		2.47 (3.05)
Dementia	No	8.34 (7.26)	0.16	16.14 (6.17)	0.14	2.03 (2.52)	0.12
	Yes	10.19 (8.21)		18.00 (4.62)		2.77 (3.22)
Anemia	No	8.22 (7.24)	0.033	16.20 (6.07)	0.88	2.03 (2.52)	0.46
	Yes	9.46 (7.48)		16.12 (6.54)		2.18 (2.69)
Liver Disease	No	8.41 (7.32)	0.23	16.21 (6.12)	0.56	2.05 (2.54)	0.3
	Yes	6.48 (4.41)		15.31 (7.11)		1.48 (2.34)
CHA2DS2-VASc Score	<2	7.71 (7.40)	0.15	15.90 (6.12)	0.53	1.93 (2.11)	0.48
	>=2	8.51 (7.26)		16.25 (6.14)		2.07 (2.61)
LVEF	<50	9.55 (7.43)	0.18	17.94 (6.92)	0.044	3.07 (3.14)	0.006
	>=50	8.58 (7.23)		16.41 (5.95)		2.32 (2.61)
Prior Ablation	No	8.84 (7.48)	<0.001	16.47 (6.30)	0.019	1.92 (2.43)	0.005
	Yes	7.28 (6.68)		15.35 (5.54)		2.35 (2.77)
Prior Cardioversion	No	8.53 (7.38)	0.21	15.80 (6.17)	<0.001	1.88 (2.47)	<0.001
	Yes	7.95 (6.98)		17.55 (5.83)		2.55 (2.70)

AF: atrial fibrillation; AFSS: atrial fibrillation severity score; LVEF: left-ventricular ejection fraction based on echocardiogram.

Figure 1.

Univariate comparisons of AFSS scores among subgroups of interest (higher score is more symptomatic). The inner boxes in the boxplots display 25th, 50th, and 75th AFSS percentiles, with mean AFSS scores designated by diamonds. The upper whisker is the smaller value of the maximum AFSS and 75th percentiles +1.5 IQR and the lower whisker is the larger value of the smallest AFSS and 25th percentiles – 1.5 IQR.

In a multivariable model adjusting for baseline demographics, patients <65 years of age had significantly higher scores (mean adjusted difference 2.92, 95% CI 1.97–3.88), in addition to those with pulmonary disease (mean adjusted difference 1.88, 95% CI 1.06–2.7), depression (2.49, 95% CI 1.6–3.32), and female sex (2.57, 95% CI 1.78–3.37, Figure 2A; Supplemental Material, Table S1).

Figure 2.

Forest plot of multivariable model of baseline AFSS symptom scores (A), AF burden score (B), and health care utilization score (C) (greater, positive score difference is more symptomatic, more burden, and more health care utilization respectively). Shown are mean differences with 95% confidence intervals between the AFSS symptom score for the indicated categories compared to the reference category, adjusted for the other factors included in the model.

AF Burden and Healthcare Utilization

Among patients with calculable AFSS domains of AF burden (n=888) and healthcare utilization (n=1313), similar trends were observed compared with the AFSS symptom score. Younger patients, female patients, and those with pulmonary disease and/or HF tended to have worse scores in these domains (Table 2). Consistently, in multivariable analysis, younger patients, those with HF and/or pulmonary disease tended to have worse AF burden and healthcare utilization (Figures 2B and 2C, Supplemental Material Tables S2 and S3).

Symptom Burden and Treatment

There were 1246 (93%) patients with at least 6-months of follow-up, among which 377 (30%) received a rhythm control intervention during that period. Patients with increased AFSS symptom scores over the median had higher rates of rhythm control interventions (37% vs. 25%, p<0.001), including AF ablation (15% vs. 7.3%, <0.001), antiarrhythmic drug therapy (28% vs. 19%, p<0.001), and cardioversion (12% vs. 6.4%, p=0.002). These findings were consistent for the secondary AFSS domains of AF burden and healthcare utilization (Figure 3).

Figure 3:

Comparison of treatments stratified by AFSS categories symptom score (SS, median=7), AF burden (median=16), and health care utilization (HCU; median=1.0)). Groups represented as score above or below the median and percentage of patients that received treatment. Treatments identified were any antiarrhythmic drug (AAD), AF ablation procedure, or cardioversion from index visit out to 6 months.

In multivariable adjusted analysis, each baseline AFSS domain score (as a continuous co-variate) remained associated with any subsequent rhythm control intervention within 6 months (p<0.001; Supplemental Material, Tables S4, S5, S6).

Discussion

Patient reported outcomes are becoming a mainstay in clinical trials and other controlled circumstances where HRQoL are necessary outcomes, and this is particularly true in AF.^20,21 However, to date, relatively few studies have examined how PROs can be implemented in clinical practice or have sought to ascertain patient characteristics associated with HRQoL, outside of controlled studies. These validated scores lack evidence to support and inform care in routine clinical practice. In this study, we identified factors associated with worse symptom scores using an AF-specific PRO. The factors we identified as associated with worsening AF symptoms were similar to those reported in clinical trials, supporting their external validity to routine clinical care.^22,23 We also assessed how AF patients with worse symptom scores were treated and found that they received more aggressive rhythm control, supporting the concept that AF PROs are capturing similar factors that clinicians use in recommending treatment. Overall, these data lay the groundwork for more systematic implementation of PRO-guided rhythm control management for AF.

Our data show several important clinical variables that were associated with increased PRO-based symptom scores in routine AF clinical care. Patients who were younger, female, and had diabetes and/or pulmonary disease were more likely to have worse AF symptom scores. These findings confirmed our hypothesis and are consistent with large scale clinical trials including CABANA and others.²² They’ve also been observed in other large multicenter registries, such as ORBIT-AF.²⁴ Yet, those are selected cohorts, with devoted HRQoL sub-studies and motivated clinicians and patients. In comparison, our data reflect unselected AF patients in a tertiary-care electrophysiology clinic. This consistency supports the external validity of such studies to accurately test and guide management of unselected AF populations, particularly with respect to HRQoL outcomes. We also found that the AFSS domains of AF burden and healthcare utilization showed consistent findings. This correlation of AF burden, symptoms, and costs provides the potential to improve HRQoL through decreasing AF burden and lower healthcare costs through focused AF care delivery.

We also examined how patients with a baseline PRO measure were treated in practice. We again confirmed our hypothesis and found that patients with worse symptom scores were more likely to receive all forms of rhythm control, including antiarrhythmic therapy, cardioversion, and AF ablation (Figure 3). While we cannot readily determine if clinicians factored the score into rhythm control management, this further support of the clinical utility of PROs is an important finding. For physicians using the AFSS PRO score to guide treatment, it could provide valuable information about patient symptoms and inform a shared-decision making paradigm that provides more discrete data to support an aggressive treatment approach. PROs also may provide a dynamic means to gauge AF therapy effectiveness. Conversely, if physicians are treating patients irrespective of the AFSS PRO score, this suggests physicians may be detecting symptom severity through their own line of questioning, albeit in a more subjective manner. Irrespective of these scenarios, the AFSS broadly patients who received (and likely warranted) more treatment in clinical practice, and additional data are needed to understand appropriate cuff-offs for interventions and expected improvements (Figure 3). Further studies are necessary to understand which AFSS domains can provide the most value to guide and monitor AF management. Such a formal implementation of PRO-guided AF care could provide an evidence-based approach across patients, with more personalized selection of patients for aggressive rhythm control strategies.

Limitations

The data presented are from real-time clinical care and therefore have some variability including irregular data collection intervals and potentially-informative missingness. These data were collected from a single-center within a highly-motivated health care system which may not translate to other institutions. Though validated in other studies, there remain unknown parameters around the AFSS, including the clinically-relevant change in each AFSS domain score and their relationship to clinical outcomes (e.g., hospitalization). Additionally, we cannot ascertain the influence of important lifestyle risk factors (e.g., alcohol intake, exercise) on AF symptom status and interventions.²⁵ Finally, physician approach to treatment is subject to clinical bias and may depend on variables not immediately evident from administrative data.

Conclusions

In summary, data collected in routine clinical practice corroborate HRQoL findings in highly-selected cohorts. Worse AF PRO measures, across domains, were associated with the implementation of rhythm control interventions. These data suggest that implementation of PRO-guided rhythm control management of patients with AF could standardize care and outcomes assessments across heterogeneous patient populations. Future work could (1) help identify optimal intervention candidates, (2) PRO thresholds for intervention, and (3) quantify relative benefits of various treatment approaches, including whether PRO-guided care can improve healthcare value.