Introduction
Prominent geographic disparities in life expectancy (LE) among older adults are present in the United States (US) with the highest 2017 LE observed in Hawaii and the lowest in Mississippi [
1]. The etiologies underlying these disparities are complex, and potential causes may include human biology and genetic risk, behavioral, mental health and socio-environmental factors, as well as variations in access to health care and healthcare utilization. Nonetheless, the reasons for the disparities between the states with highest (leading states) and lowest (lagging states) LE are not fully understood. Understanding how disease-specific mortality contributes to geographic disparities in LE is important for optimization of health policy and interventions aimed at mitigating the LE gap.
As the leading contributor (explaining approximately 40% of the total differences in LE) to geographic disparities in LE in the US [
2], heart failure (HF) accounted for approximately one in eight deaths in the U.S. in 2017 [
3]. About 6.2 million adults were living with HF in 2013–2016 [
4], and a projected 71% of all HF cases will be among adults aged 65+ in 2030 [
5]. While sex and racial disparities in HF risks and mortality are well studied [
6,
7], the substantial geographic disparities in HF mortality [
8‐
10] received less attention. This presents a potential future problem as the prevalence of HF driven primarily by population aging [
11] is expected to increase substantially within the next decades and will likely surpass the prevalence of other cardiovascular diseases [
12]. Furthermore, the gradual decline of HF mortality observed over the past few decades [
10,
13] has been reversing since 2012 [
9,
10].
In this study, we investigate several scenarios to explain the variations of HF mortality across the U.S. We hypothesize that regions with higher HF-specific and total mortality have: (a) a higher HF incidence; (b) poorer survival of patients with HF; (c) higher pre-existing prevalence of HF at the time of entring (age 65) the Medicare – the primary payer for health service in older U.S. adults.
Discussion
Our study found substantial geographic disparities in HF outcomes across the US: older adults aged 65+ in the lagging states had higher HF mortality, incidence, prevalence, and lower survival, with most of the disparities in mortality originating from differences in HF incidence, pre-existing prevalence of HF at age 65, and survival after HF diagnosis, accompanied with increasing incidence as well as declining prevalence and survival in HF patients in both leading and lagging states. The findings in this study are consistent with previous works on HF mortality [
8,
9,
18] and incidence [
19].
Our study has a significant advantage: it is based on the analysis of over 5 million Medicare beneficiaries which provides sufficient power for the analysis of relatively small geographic regions and is nationally representative sample of older U.S. adults aged 65+, covering all geographic regions in the U.S. and allowing for evaluation of both morbidity and mortality. Furthermore, this Medicare-based analysis is combined with death certificate data from CDC-WONDER to reduce the impact of the limitations associated with administrative data. Previous studies on geographic patterns of HF mortality were based on death certificate data only and did not investigate the associated epidemiological measures (e.g., incidence, prevalence) [
8,
9]. Furthermore, most such studies were community-based cohort studies [
13,
20] with poor generalizability to the total U.S. population. The single existing Medicare-based study [
19] identified by the authors investigated the geographic disparities of HF in four U.S. census regions (i.e., Midwest, Northeast, South, and West) and only examined HF prevalence and incidence.
Our study showed that all hypothesized scenarios (higher incidence, poorer survival and higher prevalence at age 65) contributed to the geographic differences in HF mortality. Study results showed substantially higher pre-existing prevalence of HF in lagging states prior to Medicare eligibility. This phenomenon could be explained by the earlier onset of HF in the lagging states. Data from the Behavioral Risk Factor Surveillance System showed that the prevalence of coronary heart disease and myocardial infarction at ages younger than 65 was higher in the lagging states [
21], which suggest a higher incidence and prevalence of HF before age 65 in the lagging states. The gap of prevalence at age 65 narrowed after 2008, which explained part of the observed decline in the gap in HF mortality. Early primary prevention efforts targeting HF risk factors in young and middle-aged adults in the lagging states are desirable.
Disparities of HF incidence - another important contributor to the geographic disparity - are often attributed to the differences in the distribution of associated risk factors (e.g., hypertension, diabetes) [
19,
22], and their impacts [
23]. We found that the incidence gap between the leading and lagging states was decreasing over the time period available for our study. However, this did not represent a beneficial trend as it was caused by relatively quicker incidence growth in the leading states rather than incidence reduction in the lagging states. This suggests the need for intensive prevention and awareness programs even in those states with relatively good epidemiological profiles.
Finally, HF survival was also lower in the lagging states, likely due to between-the-state differences in stage at diagnosis, access to/quality of healthcare, behavioral habits, and the prevailing comorbidity profiles [
8‐
10,
18]. For example, data from the CDC showed that patients in lagging states had greater nonadherence to arterial hypertension treatment and cholesterol-lowering medication intake, as well as higher eligibility for cardiac rehabilitation coupled with low participation rates [
24]. This suggests that tertiary prevention work targeting the treatment and management after HF onset is needed to improve the survival in the lagging states.
In addition to the substantial geographic disparities, we observed that the CBMs showed recent increases while the IBM sustained relatively stable, that may be associated with the administrative nature of Medicare data, where patients are followed up till the end of their enrollment instead of their deaths. Study results showed a decade-long plateau stage in HF IBM and decline in prevalence in both the leading and lagging states that may be attributable to joined effects from the increases in incidence and declines in prevalence at age 65 and survival. The recent increase in incidence may be attributed to the adoption of better testing methods [
25], the increasing rates of obesity and diabetes [
10], and some negative lifestyle changes (e.g., low physical activity) [
4,
26]. The decline in prevalence at age 65 may be related to the declines in survival while the survival decline may be related to increasing levels of multimorbidity in the elderly as well as the increasing proportion of HF with preserved ejection fraction (HFpEF), a common subtype of HF among older patients that does not have a specific treatment [
4,
26‐
29]. Although HFpEF is not ascertainable in our data, the overall trends in mortality are supportive of these findings.
Another explanation of the declining survival is the Hospital Readmissions Reduction Program (HRRP) that was discussed in 2007–2009, announced in 2010 and implemented in 2012. The HRRP aims to encourage hospitals to improve the quality of health care by imposing Medicare payment penalties on hospitals with higher-than-expected readmission rate, with three diseases initially covered, including HF, acute myocardial infarction, and pneumonia [
30]. The penalties may lead the hospitals to take inappropriate care strategies, such as delaying patients’ readmission beyond day 30, increasing observation stays without admission, shifting inpatient care to outpatient/emergency care [
31], increasing the coding disease severity [
32], that may adversely affect the health outcomes in HF patients [
31]. Previous studies based on the data from the Medicare Beneficiaries showed that the 30-day and 1-year HF mortality rates were higher after the implementation of the HRRP [
31,
33,
34], despite the successful reduction of the hospital readmission rates [
31,
33,
35,
36]. Further studies are needed to investigate this issue.
The study results showed greater geographic disparities in HF outcomes among Whites than Blacks, which might be related to the differences in risk factor distribution [
37] and the impacts of these factors [
38]. On the contrary, Hispanics in the leading states had higher incidence, that can be partially explained by associated higher prevalence of risk factors in the leading states than the lagging states [
37,
39] as well as lower access to health care, lower coverage of health insurance [
40] and hospitalization rate [
41] in the lagging states, that may lead to an underdiagnosis of HF. More data are needed to investigate the causes.
In our study, patients aged 80+ had less pronounced between-the-state differences in HF incidence compared to patients aged 65–79 years old. Possible explanations could be that more people with HF risk factors do not survive to age 80 in the lagging states, that may lead to a smaller between-the-state differences for the prevalence of obesity, diabetes, and arterial hypertension among patients aged 80+ than patients aged 65–79 years old [
21] and respective age-specific impacts of these factors on HF incidence [
42].
This study has an important limitation: information on specific subtypes of HF is limited in Medicare claims. This reduces the generalizability of our findings to patient groups with well-defined HF subtypes and suggests the need for more granular studies.
Based on the results of this study, future studies will apply trend decomposition analyses (such as partitioning [
43‐
46]) to verify the causes of these trends and the relative changes in the magnitude of their effects over time as well as the role of complementary trends in related comorbid conditions (e.g., diabetes, coronary heart disease, and myocardial infarction). Furthermore, the quantification of differences in treatment and medication prescription/utilization patterns (which can be derived from Medicare Part D data) can provide further insight into the mechanisms generating these disparities.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.