This study sought to evaluate the clinical epidemiology of fatigue in a large cohort of newly diagnosed HF patients from a non-clinical trial, community-based setting. The extensive EMR data repository of the Geisinger Health System was utilized to provide a detailed perspective on the prevalence, predictors, and prognostic value of clinically documented fatigue in a large HF population receiving primary care and other health care services through a single health care system. We found that 39% of incident HF patients had clinically documented fatigue at the time of diagnosis. Fatigue was often part of a symptom cluster, as individuals with fatigue were also significantly more likely to have documented dyspnea, chest pain, edema, syncope, and palpitations. The variable showing the strongest association with fatigue was depression, suggesting a strong psychological component to fatigue in HF. Fatigue was also strongly associated with other factors possibly indicative of fluid homeostasis, cachexia, and/or frailty, such as volume depletion, lower body mass index, and abnormal weight loss. Fatigue was not independently associated with either short- or long-term mortality after considering the strongest correlates of fatigue.
Heart failure is often a very symptomatic condition, and the presence and severity of symptoms often contribute to its diagnosis . Fatigue in particular has been noted as a very common symptom in HF, but estimating its true prevalence is challenging as the fatigue construct is intrinsically difficult to define in an objective manner and no fatigue measurement tool exists that has gained widespread popularity in either the clinical or research settings [2, 5,6,7]. Prior studies have reported that the prevalence of fatigue in HF ranges from 28 to 59% with variation likely attributable to the method of measurement, specific characteristics of the HF population studied, timing of the measurement with respect to disease exacerbations, and other factors [10, 13,14,15,16,17]. Our estimate of fatigue prevalence in a newly diagnosed HF population (39%) falls within the range observed in previous studies. The definition of fatigue employed in the current study involved observing any one of three fatigue-related ICD-9 codes either near, or prior to, the HF diagnosis date. Of note, fatigue had to be either directly coded into the EMR by clinical staff or abstracted from a clinical note reviewed by a billing coding specialist. Given the ubiquity of fatigue, it seems likely that fatigue documented in the clinical environment resulted from the fatigue being greater than what is typical or expected with respect to frequency, severity, duration, and/or ease of provocation, but we are unable to validate this supposition. Evaluating prevalence estimates from prior studies in concert with the employed fatigue definition may help clarify the meaning of clinically documented fatigue in our study population. For instance, Ingle et al. reported that 27% of HF patients had some, a lot, or very much fatigue at rest, while 38% reported fatigue limited their daily activity some, a lot, or very much . Perez-Moreno et al. noted that 43% of HF patients reported fatigue at rest or with slight exertion . Ekman et al. found that 50% of HF patients reported fatigue either walking at normal pace on a flat surface, walking slowly on a flat surface or during washing or dressing, or at rest . Finally, Barnes et al. found that 59% of chronic HF patients were moderately to extremely troubled by fatigue .
Identifying the origin(s) of fatigue in HF can also be challenging as fatigue may be a consequence of one or more underlying etiologies. In the context of HF, fatigue may be attributable to peripheral sequelae of the cardiac dysfunction itself, psychological morbidity which often accompanies HF, any of several additional comorbidities often coexisting with HF, or aging [2, 4]. Our analysis of a large number of candidate predictors of fatigue serves not only to identify potential etiologic origins of fatigue but also suggests possible therapeutic opportunities for its alleviation. The variable with the strongest association with fatigue in our analysis was depression, suggesting a strong psychological component to fatigue in HF. Depression is common in HF (26% in our study), and depressive symptoms have significant overlap with several HF-related symptoms, including fatigue [3, 5, 18]. Several previous studies have noted an association between depression and fatigue in HF, but the nature of the causal relationship is not clear and may be bidirectional [6, 18, 19]. In the current study, depression was documented prior to fatigue in 59% of patients with both conditions documented, while fatigue was documented first in 36% (5% on same day). Our results also revealed that fatigue was often part of a symptom cluster, as dyspnea, chest pain, edema, syncope, and palpitations were all independently associated with fatigue. Recognition of symptom clusters in HF is increasing, along with their potential to impact adversely both life quality and expectancy [20,21,22,23]. Symptom clusters may reflect a common underlying etiology (e.g. volume overload) causing multiple somatic manifestations and/or an increased awareness of additional symptoms by patient or provider once an initial problematic symptom has been identified [20,21,22,23,24].
We also found volume depletion, lower body mass index, and abnormal weight loss to be among the strongest predictors of fatigue. These factors may reflect some combination of volume dysregulation, cachexia, and/or frailty, all of which are not infrequent in the HF population and signal an advanced stage of disease and an adverse prognosis . These traits may be a direct result of catabolic processes in the skeletal musculature believed to be activated in HF, leading to structural and functional alterations which impede oxygen delivery and utilization in the working muscle, causing muscle weakness, atrophy, and wasting [7, 9, 25, 26]. These aberrations have been hypothesized as a direct cause of fatigue and other symptoms in HF, and importantly, both the muscle abnormalities and fatigue have been shown to improve with exercise training [4, 9, 25,26,27]. Beyond depression and markers of volume regulation and frailty, other notable independent predictors of fatigue were female gender, anemia, sleep apnea, vitamin D deficiency, and higher age – female gender and anemia in particular have shown to be predictors of fatigue in prior studies [4, 6, 17, 28, 29].
The logic behind evaluating the independent prognostic effect of fatigue (or any symptom) lies in its potential as a signal of some prognostically relevant pathophysiologic process not captured by usual clinical measures. In this regard, our results are largely consistent with prior studies showing that fatigue has at most a small, but perhaps no, independent association with all-cause mortality after adjusting for the strongest correlates of fatigue [10, 11, 15, 17, 29, 30]. Though our results showed fatigue was associated with a 49% increased risk of short-term death within 6 months following HF diagnosis, the effect diminished greatly (HR = 1.12, p = 0.16) after controlling for the strongest correlates of fatigue. Notably, this small effect diminished further when evaluating 12-month (HR = 1.07, p = 0.26) and overall (HR = 1.00, p = 0.97) mortality, suggesting that any prognostic signal provided by fatigue wanes over time. Importantly, our results suggest that the presence of dyspnea (often a marker of congestion), anemia, volume depletion, and a lower body mass index explain much of the increased risk associated with fatigue in unadjusted models. Multiple prior studies have observed no to modest increased mortality risk with various measures of fatigue. In a recent secondary analysis of the CORONA trial (Controlled Rosuvastatin Multinational Trial in Heart Failure), Perez-Moreno et al. showed that the highest level of fatigue was associated with a non-significant 17% increased risk of all-cause mortality (p = 0.26) in patients with systolic HF of ischemic origin, while Ekman et al. reported no independent prognostic effect of fatigue among systolic HF patients enrolled in COMET (Carvedilol or Metoprolol European Trial) [10, 17]. Notably, these two studies did show that fatigue was associated with an increased risk of HF hospitalization and worsening HF, respectively [10, 17]. Our work confirms and extends the mortality findings to a community-based cohort with HF of various origins and ejection fractions.
Study limitations derive largely from the retrospective nature of this study originating from EMRs. Study data is derived from usual clinical practice, thus there was no standardized assessment of fatigue or other study variables and variability in inter-coder practices is inevitable. The patient had to either spontaneously report fatigue as a concern during a clinical encounter or clinical staff had to specifically query about fatigue in the absence of any stimulus. Furthermore, clinical staff had to either explicitly document fatigue with an appropriate ICD-9 code or within a clinical note with subsequent abstraction by a coding specialist. These factors may have led to an underestimation of fatigue prevalence. We also had no information on fatigue severity, frequency, duration, or ease of provocation, and thus were unable to evaluate whether predictors or the prognostic relevance of these fatigue attributes differ from what was observed for our binary definition of fatigue. Furthermore, data elements necessarily were restricted to diagnostic codes and other fixed-field elements which restricted data availability. For instance, New York Heart Association class is a popular metric for assessing the symptomatic impact of HF and likely overlaps with fatigue, but was not available in the current study. Nonetheless, this study was able to include a large, real-world community-based cohort with newly diagnosed HF and evaluate a large number of potential correlates of fatigue well in excess of prior studies.