Abstract
BACKGROUND: Atrial fibrillation (AF) imposes substantial health care and economic burden on health care systems and patients. Previous studies failed to examine health care resource utilization (HCRU) and costs among patients with incident AF and potential disparity with regard to geographic location.
OBJECTIVES: To examine HCRU and costs among patients with incident AF compared with patients without AF and examine whether a geographic disparity exists.
METHODS: This was a retrospective cohort study. We selected patients with AF and patients without AF from IBM/Watson MarketScan Research Databases 2014-2019. HCRU and costs were collected 12 months following an AF index date. We used 2-part models with bootstrapping to obtain the marginal estimates and CIs. Rural status was identified based on Metropolitan Statistical Area. We adjusted for age, sex, plan type, US region, and comorbidities.
RESULTS: Among 156,732 patients with AF and 3,398,490 patients without AF, patients with AF had 9.04 (95% CI = 8.96-9.12) more outpatient visits, 0.82 (95% CI = 0.81-0.83) more emergency department (ED) visits, 0.33 (95% CI = 0.33-0.34) more inpatient admission, and $15,095 (95% CI = 14,871-15,324) higher total costs, compared with patients without AF. Among patients with AF, rural patients had 1.99 fewer (95% CI = −2.26 to −1.71) outpatient visits and 0.05 (95% CI = 0.02-0.08) more ED visits than urban patients. Overall, rural patients with AF had decreased total costs compared with urban patients (mean = $751; 95% CI = −1,227 to −228).
CONCLUSIONS: Incident AF was associated with substantial burden of health care resources and an economic burden, and the burden was not equally distributed across patients in urban vs rural settings.
DISCLOSURES: Dr Hansen reports grants from the National Science Foundation during the conduct of the study.
Plain language summary
We explored the costs and health care use for people with atrial fibrillation (AF). We found that AF burdens health systems, patients, and payers. We also found gaps in the burden of AF for people in rural areas. Total costs did not differ by location, but rural patients had fewer physician visits, had more urgent care visits, and used more medications.
Implications for managed care pharmacy
We quantified the contemporary burden of AF in the United States. Understanding the costs from different perspectives (payer and patient) will aid decision makers in allocating resources for AF. Finding rural disparities in the burden of AF indicates the necessity for improving access to care to rural areas.
The prevalence of atrial fibrillation (AF) is increasing in the United States. It is estimated that 5.2 million people in the United States had AF in 2010, and this number is expected to increase to 12.1 million by 2030.1 The incidence of AF has been increasing over the past decades from 4.74 cases per 1,000 person-years in 2006 to 6.82 cases per 1,000 person-years in 2018.2 AF is a risk factor for other cardiovascular diseases, including stroke, heart failure, and coronary artery disease.3,4 Among all health conditions, AF alone ranked at 33rd in health care spending with $28.4 billion in the United States in 2016. The projected health care spending by 2030 would be $45.4 billion, given the reported annualized rate of change of 3.4% between 1996 and 2006.1 AF imposes serious health and economic challenges to patients and health care systems.
A few studies have examined the health care resource utilization (HCRU) and costs associated with AF. However, some studies suffer from prevalence-incidence case bias in which prevalent AF cases have been identified along with incident AF cases such that HCRU or costs might have been overestimated or underestimated.5-8 Other studies have primarily focused on HCRU and costs associated with recurring AF or a specific population, such as employees, or did not compare HCRU and costs with that of patients without AF.9-11 In addition, some studies examined the HCRU and yet did not report corresponding costs.12-14 This leaves a knowledge gap in the incremental costs associated with these incremental HCRU, especially by payers and patients. Thus, a study that focuses the incremental HCRU and corresponding costs among the newly diagnosed patients with AF compared with patients without AF is needed.
Many studies have demonstrated the presence of disparities with respect to sex and race and ethnicity in the risk of AF, treatment initiation for AF such as anticoagulation, quality of care, and health outcomes.1,15-18 However, compared with racial and ethnic and sex disparities, rural disparity is not well understood yet in AF. In addition, rural disparity is not uncommon in the area of cardiovascular diseases because managing cardiovascular diseases requires stable and long-term access to resources for disease management, medications, and even surgeries.19-21 As rural disparity is a multifaceted issue associated with socioeconomic status,22,23 assessing the presence and examining the magnitude of rural disparity in HCRU and cost is an important first step to understand the potentially uneven distribution of disease resource and economic burden. Thus, our study aimed to fill this knowledge gap.
The primary aim of this study was to estimate the incremental HCRU and costs associated with AF by comparing newly diagnosed patients with AF with patients without AF. The secondary aim was to examine whether a rural disparity in HCRU and costs exists among patients with newly diagnosed with AF.
This was a retrospective cohort study. Data were collected from IBM/Watson MarketScan Commercial Claims and Encounters Database and Medicare Supplemental and Coordination of Benefits Databases between 2014 and 2019.24 We did not include data in 2020 as we hoped to reduce the impact of COVID-19 on outcomes of interest. Patients with AF were identified by having at least 1 inpatient service or at least 2 outpatient visits with a diagnosis of AF. We used the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) for claims prior to October 1, 2015, and International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) for claims after October 1, 2015 (ICD-9-CM = 427.31; ICD-10-CM = I480-I489). Then, we identified the index date of AF for each patient; defined as the first qualifying date, ie, the date of the first inpatient claim or the date of the second outpatient claim, whichever comes first. We excluded patients with AF who (1) were younger than 18 years on the index date, (2) did not have continuous insurance enrollment for 12 months prior to and after the index date, (3) had an AF diagnosis in any records during 12 months prior to the index date, in order to identify patients with incident AF, and (4) had negative outcome values or missing values for the covariates described below.
Patients without AF were first identified from the rest of the database who did not have any AF diagnosis code in outpatient or inpatient claims between 2014 and 2019. In order to create pseudo-index dates for the non-AF cohort, we randomly selected a patient with AF from the AF cohort for each patient without AF and assigned the index date of AF to the patient without AF. This helped establish the pseudo-index date for patients without AF and match the index date distribution between patients with and without AF. After constructing the pseudo-index date of AF, we excluded patients without AF who (1) were aged younger than 18 years, (2) did not have continuously enrollment for 12 months prior to and after the pseudo-index date, and (3) had negative outcome values or were missing values for the covariates described below. We randomly selected one-third of patients without AF to establish the final non-AF cohort.
We did not match patients with and without AF based on baseline characteristics, as we used regression adjustment to control for imbalances. This is further explained in the Statistical Analysis section.
HCRU outcomes included outpatient visits, inpatient admission, inpatient hospital length of stay (LOS), inpatient services, emergency department (ED) visits, number of outpatient medications prescribed, and total 30-day supply of all medications prescribed. Outpatient visits included physician office visits and other visits, such as laboratory, per the data collection and management methods in the IBM MarketScan database.24 To further differentiate nonemergency and emergency visits, we identified outpatient visits and inpatient admission from ED. This led to the following 2 additional HCRU outcomes: ED visits in outpatient settings and ED visits in inpatient settings. In addition, in the IBM MarketScan database, inpatient admission was derived from inpatient services.24 Both outcomes provided information about unique aspects about the HCRU and cost burden; inpatient services helped inform the number of services used during one admission, whereas inpatient admission helped inform how often a patient with AF was admitted to hospital. As such, we included both as separate outcomes. Costs included the following 3 perspectives: costs for payers, patients (ie, deductibles, copayments, and coinsurance), and total costs. For each perspective, we included costs for aforementioned HCRU. Cost outcomes were adjusted to 2019 US dollars. Costs prior to 2019 were adjusted based on inflation using Consumer Price Index.25 HCRU and costs were collected for 12 months following the index date.
The primary exposure of interest was a binary indicator of incident AF. The secondary exposure of interest was rural or urban status of patients’ geographic location, identified by Metropolitan Statistical Area. Each subject’s geographic location was ascertained on their index date.
We included age as a continuous variable, biological sex (female or male), insurance type (commercial or Medicare), health plans (comprehensive health plan, Health Maintenance Organization, Point of Service, Point of Service with capitation, Preferred Provider Organization, Consumer Directed Health Plans, or High Deductible Health Plan), Charlson Comorbid Index (CCI),26 number of unique generic product identifier, calendar year at the index date, and geographic region (Northwest, Northcentral, South, or West) as covariates. All covariates except CCI were identified on the index date. CCI was calculated based on diagnosis codes in claims in the 12 months prior to the index date; AF was not included in CCI calculation. The generic product identifier is a hierarchical classification system that categorizes medications based on primary therapeutic areas, which helps understand the comorbidities of patients and is commonly used in prior literature to help identify prescribed medicines.27-29 The generic product identifier was collected in the 12 months prior to the index date.
We compared baseline characteristics between the AF and non-AF cohorts by the standardized difference between the groups. A standardized difference close to 0 suggests the mean values of a given baseline characteristic across 2 groups were close to each other.30-32
We constructed 2-part models for analyses. The first model was a logistic regression with a binary outcome defined as having a nonzero HCRU/cost value; the second regression was a generalized linear model with a Poisson distribution for HCRU outcomes and a Gamma distribution for costs, and a log link, in which the outcome was the HCRU/cost value. The second model was applicable to those with nonzero HCRU/cost outcome values. We chose Poisson distribution in the second-part model over negative binomial distribution because the generalized linear model with a negative binomial distribution failed to converge for some outcomes, such as ED visits. Thus, we used Poisson distribution consistently across all HCRU outcomes. We adjusted aforementioned covariates in both parts. The output of the first model was the probability of an individual with nonzero HCRU/cost value. The output of the second model was the expected mean value of HCRU/cost given the outcome of an individual was nonzero. For each individual, we calculated the expected value of a given outcome by multiplying the output from the first model with the output from the second model. As such, we were able to estimate the predicted mean outcome value for patients with and without AF, respectively. Then, we calculated marginal effects (ie, incremental value) by subtracting the expected outcome among patients without AF from the expected outcome among patients with AF. These steps were done in bootstrapping. Bootstrapping with 1,000 iterations and sampling with replacement was used to calculate the mean marginal effect and its 95% CI of the marginal effects. Unlike measures such as odds ratios, the marginal effect produces the incremental cost for each cost outcome or the incremental number of HCRU for each HCRU outcome attributable to AF after adjusting for confounders. Details of the 2-part models and the calculation of marginal effect can be found in prior literature.33,34
In the secondary analysis, 2-part models were applied among patients with AF with a binary indicator for rural vs urban residence. The same method was used to calculate the marginal effect and its 95% CI.
To account for the presence of confounding, we used the technique of regression adjustment. Although some previous studies used matching as a technique to control for confounding, regression adjustment is an equivalent technique to achieve the same goal. We adjusted for all observed confounders in all models.
We used 2-sided statistical tests with a 0.05 level of significance in all analyses.
Data management was performed in SAS 9.4 (SAS Institute Inc.), and analyses were conducted in R 4.1.2 (R Foundation for Statistical Computing).
Only deidentified claims data were used in this work. Therefore, this research is deemed nonhuman subjects research by the Human Subjects Division at the University of Washington.
The final analytic sample consisted of 156,732 patients with AF and 3,398,490 patients without AF (Supplementary Figures 1 and 2, available in online article). HCRU and cost outcomes were heavily right skewed with extremely high values, and thus, we excluded participants who had an outcome above the 99.9th percentile among all participants (N = 5,040 for the AF cohort; N = 15,776 for the non-AF cohort).
Overall, the AF cohort were older than the non-AF cohort (mean age = 65 vs 45), consisted of more males (59% vs 47%), had more patients on Medicare (46% vs 8%), and were sicker (mean CCI = 2 vs 1) (Supplementary Table 1).
Descriptive results for each HCRU and cost outcome are shown in Supplementary Tables 2 and 3. Patients with AF had higher HCRU across all HCRU outcomes compared with patients without AF during the first year following an AF diagnosis. For example, in the first year of AF diagnosis, the mean number of outpatient visits were 27.67 for patients with AF but 9.13 for patients without AF. Patients with AF also had higher out-of-pocket costs and costs for payers. For example, in the first year of AF diagnosis, the mean costs for patients across all services were $2,106 for patients with AF but $877 for patients without AF; the mean costs for payers across all services were $23,306 for patients with AF but $4,839 for patients without AF.
After adjusting for confounders, on average, in the first year of AF diagnosis, patients with AF had 9.04 (95% CI = 8.96-9.12) more outpatient visits, 2.74 (95% CI = 2.71-2.77) more medications prescribed, 0.82 (95% CI = 0.81-0.83) more ED visits, 0.33 (95% CI = 0.33-0.34) more inpatient admission, and 1.47 (95% CI = 1.44-1.49) longer LOS than patients without AF (Table 1).
|
| HCRU outcomes | Part 1a | Part 2a | Marginal effectb Mean (95% CI) | ||
|---|---|---|---|---|---|
| OR: Point estimate (95% CI) | P value | RR: Point estimate (95% CI) | P value | ||
| Outpatient services | 56.85 (49.00-65.95) | < 0.0001 | 1.79 (1.78-1.80) | < 0.0001 | 9.04 (8.96-9.12) |
| Total ED visitsc | 6.03 (5.97-6.10) | < 0.0001 | 1.33 (1.32-1.34) | < 0.0001 | 0.82 (0.81-0.83) |
| ED visits into inpatient admission | 9.29 (9.15-9.43) | < 0.0001 | 1.10 (1.08-1.11) | < 0.0001 | 0.21 (0.21-0.21) |
| ED visits into outpatient visits | 3.93 (3.88-3.97) | < 0.0001 | 1.26 (1.26-1.27) | < 0.0001 | 0.51 (0.51-0.52) |
| Inpatient admission | 8.59 (8.48-8.71) | < 0.0001 | 1.17 (1.15-1.18) | < 0.0001 | 0.33 (0.33-0.34) |
| Inpatient services | 11.12 (10.97-11.26) | < 0.0001 | 1.46 (1.45-1.47) | < 0.0001 | 1.86 (1.83-1.88) |
| Inpatient LOS | 8.59 (8.48-8.71) | < 0.0001 | 1.30 (1.29-1.32) | < 0.0001 | 1.47 (1.44-1.49) |
| All medications | 2.09 (2.04-2.14) | < 0.0001 | 1.39 (1.39-1.40) | < 0.0001 | 2.74 (2.71-2.77) |
| Total 30-day supplyd | 1.96 (1.91-2.00) | < 0.0001 | 1.39 (1.39-1.40) | < 0.0001 | 11.48 (11.33-11.64) |
aAnalyses adjusted for age, sex, insurance type, health plans, CCI, geographic region, calendar year at the index date, and the number of unique generic product identifier.
bMarginal effect was calculated by the difference in expected outcome. Mean marginal effect is the mean value based on 1,000 iteration bootstrapping with sample with replacement, and 95% CI of marginal effect was constructed based on 2.5% and 97.5% percentile of bootstrapped marginal effects.
cTotal number of ED visits was sum of number of ED visits admitted to outpatient settings and ED visits admitted to inpatient settings.
dTotal 30-day supply was calculated based on all medication prescribed within 12 months following the index date.
AF = atrial fibrillation; CCI = Charlson Comorbidity Index; ED = emergency department; HCRU = health care resource utilization; LOS = length of stay; OR = odds ratio; RR = relative ratio.
After adjusting for confounders, on average, in the first year of AF diagnosis, payers of patients with AF spent $13,885 (95% CI = 13,664-14,092) more than payers of patients without AF. Patients with AF had $1,012 (95% CI = 1,002-1,023) higher out-of-pocket costs than patients without AF (Supplementary Table 4). In total, AF was associated with $15,095 (95% CI = 14,871-15,324) higher costs than non-AF (Table 2).
|
| Cost outcomes | Part 1a | Part 2a | Marginal effectb Mean (95% CI) | ||
|---|---|---|---|---|---|
| OR (95% CI) | P value | RR (95% CI) | P value | ||
| Outpatient services | 30.12 (27.08-33.50) | < 0.0001 | 2.74 (2.71-2.77) | < 0.0001 | 6,535.34 (6,443.74-6,627.74) |
| Inpatient services | 11.11 (10.96-11.25) | < 0.0001 | 1.09 (1.07-1.11) | < 0.0001 | 7,918.75 (7,741.23-8,085.63) |
| All medications | 2.09 (2.04-2.14) | < 0.0001 | 1.54 (1.50-1.57) | < 0.0001 | 1,120.52 (1,079.30-1,163.21) |
| Total ED visits | 5.62 (5.56-5.69) | < 0.0001 | 1.33 (1.32-1.35) | < 0.0001 | 1,290.80 (1,269.12-1,313.54) |
| ED visits into outpatient visits | 3.89 (3.85-3.94) | < 0.0001 | 1.34 (1.30-1.35) | < 0.0001 | 924.70 (904.75-943.00) |
| ED visits into inpatient admission | 9.07 (8.92-9.22) | < 0.0001 | 1.03 (1.00-1.10) | 0.033 | 250.84 (242.36-259.55) |
| Total costs for payers and patients across different services | 206.38 (146.69-290.37) | < 0.0001 | 3.32 (3.27-3.37) | < 0.0001 | 15,095.13 (14,871.28-15,324.32) |
aAnalyses adjusted for age, sex, insurance type, health plans, CCI, geographic region as covariates, calendar year at the index date, and the number of unique generic product identifier.
bMarginal effect was calculated by the difference in expected outcome. Mean marginal effect is the mean value based on 1,000 iteration bootstrapping with sample with replacement, and 95% CI of marginal effect was constructed based on 2.5% and 97.5% percentile of bootstrapped marginal effects.
AF = atrial fibrillation; CCI = Charlson Comorbidity Index; ED = emergency department; OR = odds ratio; RR = relative ratio.
Among 156,732 patients with AF, 23,871 patients received care in rural settings and 132,861 patients received care in urban settings (Supplementary Table 5).
After adjusting for confounders, overall, rural patients had significantly fewer outpatient visits (mean = −1.99; 95% CI = −2.26 to −1.71), but a higher number of medications (mean = 0.78; 95% CI = 0.68-0.89) and total 30-day supplies of outpatient prescriptions (mean = 2.24; 95% CI = 1.74-2.71), during the first year following an AF diagnosis (Table 3).
|
| Part 1a | Part 2a | Marginal effectb Mean (95% CI) | |||
|---|---|---|---|---|---|
| OR (95% CI) | P value | RR (95% CI) | P value | ||
| Outpatient services | 0.59 (0.42-0.84) | 0.0031 | 0.93 (0.92-0.94) | < 0.0001 | −1.99 (−2.26 to −1.71) |
| Total ED visitsc | 1.04 (1.02-1.08) | 0.0028 | 1.02 (1.00-1.03) | 0.0093 | 0.05 (0.02-0.08) |
| ED visits into inpatient admission | 0.94 (0.91-0.97) | < 0.0001 | 0.99 (0.97-1.01) | 0.2275 | −0.02 (−0.04 to −0.01) |
| ED visits into outpatient visits | 1.14 (1.11-1.17) | < 0.0001 | 1.01 (1.00-1.03) | 0.1905 | 0.07 (0.05-0.10) |
| Inpatient admission | 1.03 (1.00-1.06) | 0.025 | 1.00 (0.98-1.01) | 0.6024 | 0.01 (0.00-0.02) |
| Inpatient services | 1.02 (0.99-1.04) | 0.3065 | 0.93 (0.91-0.95) | < 0.0001 | −0.24 (−0.34 to −0.14) |
| Inpatient LOS | 1.03 (1.00-1.06) | 0.025 | 0.96 (0.94-0.98) | 0.0006 | −0.05 (−0.15 to 0.05) |
| All medications | 1.08 (1.00-1.16) | 0.0546 | 1.05 (1.05-1.06) | < 0.0001 | 0.78 (0.68-0.89) |
| Total 30-day supplyd | 1.03 (0.97-1.10) | 0.3322 | 1.05 (1.05-1.06) | < 0.0001 | 2.24 (1.74-2.71) |
aAnalyses adjusted for age, sex, insurance type, health plans, CCI, geographic region, calendar year at the index date, and the number of unique generic product identifier.
bMarginal effect was calculated by the difference in expected outcome. Mean marginal effect is the mean value based on 1,000 iteration bootstrapping with sample with replacement, and 95% CI of marginal effect was constructed based on 2.5% and 97.5% percentile of bootstrapped marginal effects.
cTotal number of ED visits was sum of number of ED visits admitted to outpatient settings and ED visits admitted to inpatient settings.
dTotal 30-day supply was calculated based on all medication prescribed within 12 months following the index date.
AF = atrial fibrillation; CCI = Charlson Comorbidity Index; ED = emergency department; HCRU = health care resource utilization; LOS = length of stay; OR = odds ratio; RR = relative ratio.
After adjusting for confounders, during the first year following an AF diagnosis, rural patients had lower costs for payers (mean = −$726; 95% CI = −1,260 to −183) and lower costs for patients (mean = −$28; 95% CI = −50 to −5), and lower total costs (mean = −$751; 95% CI = −1,227 to −228), compared with urban patients (Table 4; Supplementary Table 8). However, they had higher costs for patients in ED visits (mean = $22, 95% CI = 17-28). On the contrary, rural patients with AF had lower payer inpatient costs (mean = −$790; 95% CI = −1,309 to −316). Similarly, rural setting was associated with lower costs for patients in inpatient services (mean = −$33; 95% CI = −44 to −22) but higher costs for patients in outpatient medications (mean = $10; 95% CI = 3-17) and ED visits (mean = $22; 95% CI = 17-28) (Table 4; Supplementary Table 8).
|
| Part 1a | Part 2a | Marginal effectb Mean (95% CI) | |||
|---|---|---|---|---|---|
| OR: Point estimate (95% CI) | P value | RR: Point estimate (95% CI) | P value | ||
| Outpatient services | 0.61 (0.48-0.79) | 0.0002 | 1.00 (0.98-1.00) | 0.9061 | −25.57 (−280.82 to 220.67) |
| Inpatient services | 1.02 (0.99-1.05) | 0.2944 | 0.92 (0.88-0.96) | < 0.0001 | −835.56 (−1,283.56 to −333.59) |
| All medications | 1.08 (1.00-1.16) | 0.0483 | 0.98 (0.95-1.00) | 0.1052 | −84.65 (−189.59 to 33.36) |
| Total ED visits | 1.05 (1.02-1.08) | 0.0005 | 1.02 (0.98-1.05) | 0.3373 | 58.56 (5.60-109.65) |
| ED visits into outpatient visits | 1.12 (1.09-1.15) | < 0.0001 | 0.98 (0.95-1.01) | 0.1667 | 48.83 (9.10-90.63) |
| ED visits into inpatient admission | 1.02 (0.98-1.05) | 0.3537 | 1.05 (0.99-1.12) | 0.0749 | 23.15 (4.22-43.38) |
| Total costs for payers and patients across different services | 0.86 (0.33-2.27) | 0.7631 | 0.97 (0.95-0.99) | 0.0084 | -750.81 (−1,226.82 to −228.64) |
aAnalyses adjusted for age, sex, insurance type, health plans, CCI, geographic region, calendar year at the index date, and the number of unique generic product identifier.
bMarginal effect was calculated by the difference in expected outcome. Mean marginal effect is the mean value based on 1,000 iteration bootstrapping with sample with replacement, and 95% CI of marginal effect was constructed based on 2.5% and 97.5% percentile of bootstrapped marginal effects.
AF = atrial fibrillation; CCI = Charlson Comorbidity Index; ED = emergency department; OR = odds ratio; RR = relative ratio.
In this study, we found that incident AF was associated with substantial burden of health care resources and an economic burden to both payers and patients. In total, an average patient with AF had 9 more physician office visits, 0.8 more ED visits, 0.3 more inpatient admissions, and 3 more prescribed medications than an average patient without AF (Table 1). The additional consumed health care resources for patients with AF entailed in total $15,095 (95% CI = 14,871-15,324) more to both payers and patients (Table 2). Our results were largely consistent with previous findings that AF was associated with significant HCRU and economic burden.5,6,9 However, the magnitude of our estimates is greater than previous literature, potentially because of the increasing health care costs. In addition, we found that the HCRU and economic burden was not equally distributed across patients who received care in different settings. In short, rural patients had on average 2 fewer physician office visits and 0.2 fewer inpatient services than urban patients. However, rural patients tended to have more ED visits and more prescribed medications (Table 3). Furthermore, rural patients had decreased overall economic burden to payers or patients (Table 4).
Our findings have a few implications. First, our results help decision makers better understand the burden of AF to health systems, payers, and patients during the first year of diagnosis. For example, the first-year economic burden for incident AF in the United States from 2021 to 2022 would be estimated to be $34 billion, using the population size in 2021 of 330 million and the incidence of AF 6.82 cases per 1,000 person-years.2,35 In a health plan with 500,000 enrollees, the economic burden to payers would be $47 million. Given the high and increasing prevalence of AF and growing aging population over time, this overall burden will continue to climb in the near future. Furthermore, these aggregate estimates may be helpful in optimizing public health resources allocation in preventing and treating AF when choosing among disease prevention options. For example, as AF is associated with many lifestyle factors and other cardiovascular diseases, there have been public health programs that aim to promote healthy lifestyles and provide access to managing overall cardiovascular health such as Centers for Disease Control and Prevention Heart Disease and Stroke Prevention Programs, which will help prevent AF.36-38 As such, the information on health resources and economic burden to patients and payers will help call for attention to investing on these programs. As mentioned, AF is also commonly associated with other cardiovascular diseases.3,4 The risk of incident AF is associated with both hypertension and obesity, both of which are simultaneously increasing in prevalence in the United States.39,40 Therefore, the current incidence of AF should be expected to increase considerably in the future. Thus, our results may help inform public health decisions on prioritizing disease prevention and management for AF over others.
Moreover, our analyses suggest the presence of geographic disparities in accessing care for AF. It is possible that patients who received care in rural areas had limited access to health care facilities, and/or needed to spend longer time in transportation, and therefore visited their physicians less frequently but refilled prescriptions in larger volumes than patients who received care in urban areas. The access issue may also explain the result that rural patients had fewer inpatient services used compared with urban patients. Although it may be possible that a larger volume of medications in rural patients resulted from a higher adherence of medication of rural patients compared with urban patients, the decreased HCRU for rural patients is concerning and may suggest worse outcomes of rural patients. Our results show that rural patients tend to visit the ED more often than urban patients, which may indicate that the reduced use of outpatient and inpatient services resulted in less effective disease management and unsatisfactory prognostic outcomes for rural patients. Because our analyses had already controlled for comorbidities through CCI and the number of drugs by therapeutic areas (generic product identifier), the higher number of ED visits for rural patients did not likely result from worse health status at the baseline but rather worse health outcomes after AF diagnosis. Our findings are generally consistent with prior literature on rural disparities. For example, a study found that rural pediatric patients with brain injury tend to have fewer visits for physical therapy compared with urban patients.41 Another recent study that focused on rural disparities in postacute care utilization found that older adults in the rural settings tend to have higher mortality after being discharged.42 However, we noticed that evidence of rural disparities especially in HCRU and costs associated with cardiovascular diseases is insufficient. Future studies should examine rural disparities in other cardiovascular diseases or even other disease areas to facilitate the understanding of rural disparities.
Furthermore, the finding that rural patients did not bring sizable economic burden to payers or patients themselves should be interpreted with caution. Based on aforementioned HCRU results by rural status, rural patients did not consume more or less services across all types of services universally. The decreased outpatient visits might offset the increased costs due to more ED visits for rural patients. Moreover, with the financial incentives provided by the Rural Health Clinic program of the Centers for Medicare & Medicaid Services (CMS), rural providers may receive enhanced reimbursement rates from Medicare and Medicaid,43 and therefore, the slightly decreased costs for payers and patients comparing rural to urban patients further confirmed reduced use for rural patients. Because the observed underutilization of services for rural patients may lead to increasing disease burden and economic burden in the future, attention must be paid to reducing the disparities and providing sufficient access to care for rural patients.
Our results on rural-urban disparities have implications for telehealth. It has been suggested that lack of access to specialty providers leads to poorer outcomes in rural settings.44 In the United States, the size of rural population has been increasing over time, and in 2020, there are 57 million people living in rural areas.45 In the most recent report from the CMS in improving health for rural communities, among other proposed solutions, advances in telehealth and telemedicine are warranted because they help overcome standing issues in rural disparities by providing digital support.43 As such, telehealth clinics for AF may be particularly valuable in rural regions, providing patients access to specialty care providers without travel and associated expenses.
Our study has a few strengths. First, we focused on patients with incident AF in that we only selected patients with AF who did not have any AF diagnosis prior to the index date, which allowed us to estimate with appropriate precision HCRU and cost burden during the first year of the disease, helping reduce the prevalent-incidence bias.8 Second, our analyses captured all prescribed medications, instead of a few common medications prescribed to patients with AF, which allowed us to estimate overall disease and economic burden agnostic to specific classes of medications. Third, using IBM/Watson MarketScan Research Databases 2014-2019,24 we were able to obtain a large sample of people. Particularly, we used a careful approach to first select patients without AF in that all patients without AF did not have any AF diagnosis throughout the study period and match index date distributions to patients without AF. Fourth, we controlled for a wide range of covariates, including age, sex, payer, insurance type, US region, CCI, health plans, and the number of unique generic product identifier. Lastly, we not only reported odds ratios and relative ratios from two-part models but also reported the final marginal effects with bootstrapped CIs, aiding in the interpretation of HCRU and costs in their natural units.
Our study also bears with a few limitations. First, patients in the IBM/Watson MarketScan Research Databases may not be a representative sample of the US population. Second, although we accounted for reversals during the study period and excluded records with negative HCRU or cost outcomes, some utilization and cost information might be not precise owing to reversals. Moreover, we controlled for multiple confounders, and yet unmeasured confounders such as race and ethnicity may bias the marginal effects. Lastly, we did not examine long-term HCRU and cost outcomes. We encourage future research such as decision modeling studies to explore long-term outcomes for patients with AF and potential geographic disparities.
AF brings sizable HCRU and cost burden to US health care systems, payers, and patients. The burden is not equally distributed across patients with rural and urban dwellings.
| 1. | Virani SS, Alonso A, Aparicio HJ, et al. Heart disease and stroke statistics-2021 update: A report from the American Heart Association. Circulation. 2021;143(8):e254-743. doi:10.1161/CIR.0000000000000950 Google Scholar |
| 2. | Williams BA, Chamberlain AM, Blankenship JC, Hylek EM, Voyce S. Trends in atrial fibrillation incidence rates within an integrated health care delivery system, 2006 to 2018. JAMA Netw Open. 2020;3(8):e2014874. doi:10.1001/jamanetworkopen.2020.14874 Medline, Google Scholar |
| 3. | Benjamin EJ, Wolf PA, D’Agostino RB, Silbershatz H, Kannel WB, Levy D. Impact of atrial fibrillation on the risk of death: The Framingham Heart Study. Circulation. 1998;98(10):946-52. doi:10.1161/01.cir. 98.10.946 Crossref, Medline, Google Scholar |
| 4. | Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: The Framingham Study. Stroke. 1991;22(8):983-8. doi:10.1161/01.str.22.8.983 Crossref, Medline, Google Scholar |
| 5. | Kim MH, Johnston SS, Chu BC, Dalal MR, Schulman KL. Estimation of total incremental health care costs in patients with atrial fibrillation in the United States. Circ Cardiovasc Qual Outcomes. 2011;4(3):313-20. doi:10.1161/CIRCOUTCOMES.110.958165 Crossref, Medline, Google Scholar |
| 6. | Kim MH, Lin J, Hussein M, Kreilick C, Battleman D. Cost of atrial fibrillation in United States managed care organizations. Adv Ther. 2009;26(9):847-57. doi:10.1007/s12325-009-0066-x Medline, Google Scholar |
| 7. | Coyne KS, Paramore C, Grandy S, Mercader M, Reynolds M, Zimetbaum P. Assessing the direct costs of treating nonvalvular atrial fibrillation in the United States. Value Health. 2006;9(5):348-56. doi:10.1111/j.1524-4733.2006.00124.x Crossref, Medline, Google Scholar |
| 8. | Delgado-Rodriguez M, Llorca J. Bias. J Epidemiol Community Health. 2004;58(8):635-41. doi:10.1136/jech.2003.008466 Medline, Google Scholar |
| 9. | Amin AN, Jhaveri M, Lin J. Incremental cost burden to US healthcare payers of atrial fibrillation/atrial flutter patients with additional risk factors. Adv Ther. 2011;28(10):907-26. doi:10.1007/s12325-011-0065-6 Medline, Google Scholar |
| 10. | Rohrbacker NJ, Kleinman NL, White SA, March JL, Reynolds MR. The burden of atrial fibrillation and other cardiac arrhythmias in an employed population: associated costs, absences, and objective productivity loss. J Occup Environ Med. 2010;52(4):383-91. doi:10.1097/JOM.0b013e3181d967bc Medline, Google Scholar |
| 11. | Reynolds MR, Essebag V, Zimetbaum P, Cohen DJ. Healthcare resource utilization and costs associated with recurrent episodes of atrial fibrillation: The FRACTAL registry. J Cardiovasc Electrophysiol. 2007;18(6):628-33. doi:10.1111/j.1540-8167.2007.00819.x Medline, Google Scholar |
| 12. | Bengtson LG, Lutsey PL, Loehr LR, et al. Impact of atrial fibrillation on healthcare utilization in the community: The Atherosclerosis Risk in Communities study. J Am Heart Assoc. 2014;3(6):e001006. doi:10.1161/JAHA.114.001006 Medline, Google Scholar |
| 13. | Zeitler EP, Ronk CJ, Cockerham A, Huse S, McKindley DS, Kim MH. Healthcare resource utilization in patients with newly diagnosed atrial fibrillation in the United States. Expert Rev Pharmacoecon Outcomes Res. 2022:22(5):763-71. doi:10.1080/14737167.20 22.2045955 Medline, Google Scholar |
| 14. | Vinter N, Cordsen P, Lip GY, et al. Newly diagnosed atrial fibrillation and hospital utilization in heart failure: A nationwide cohort study. ESC Heart Fail. 2021;8(6):4808-19. doi:10.1002/ehf2.13668 Medline, Google Scholar |
| 15. | O’Neal WT, Sandesara PB, Kelli HM, Venkatesh S, Soliman EZ. Urban-rural differences in mortality for atrial fibrillation hospitalizations in the United States. Heart Rhythm. 2018;15(2):175-9. doi:10.1016/j.hrthm.2017.10.019 Medline, Google Scholar |
| 16. | Avgil Tsadok M, Jackevicius CA, Essebag V, Eisenberg MJ, Rahme E, Pilote L. Warfarin treatment and outcomes of patients with atrial fibrillation in rural and urban settings. J Rural Health. 2015;31(3):310-5. doi:10.1111/jrh.12110 Medline, Google Scholar |
| 17. | Sur NB, Wang K, Di Tullio MR, et al. Disparities and temporal trends in the use of anticoagulation in patients with ischemic stroke and atrial fibrillation. Stroke. 2019;50(6):1452-9. doi:10.1161/STROKEAHA.118.023959 Medline, Google Scholar |
| 18. | Meschia JF, Merrill P, Soliman EZ, et al. Racial disparities in awareness and treatment of atrial fibrillation: The Reasons for Geographic and Racial Differences in Stroke (REGARDS) study. Stroke. 2010;41(4):581-7. doi:10.1161/STROKEAHA.109.573907 Medline, Google Scholar |
| 19. | Lawrence E, Hummer RA, Harris KM. The cardiovascular health of young adults: Disparities along the urban-rural continuum. Ann Am Acad Pol Soc Sci. 2017;672(1):257-81. doi:10.1177/0002716217711426 Medline, Google Scholar |
| 20. | Schopfer DW. Rural health disparities in chronic heart disease. Prev Med. 2021;152(Pt 2):106782. doi:10.1016/j.ypmed.2021.106782 Medline, Google Scholar |
| 21. | Loccoh EC, Joynt Maddox KE, Wang Y, Kazi DS, Yeh RW, Wadhera RK. Rural-urban disparities in outcomes of myocardial infarction, heart failure, and stroke in the United States. J Am Coll Cardiol. 2022;79(3):267-79. doi:10.1016/j.jacc.2021.10.045 Medline, Google Scholar |
| 22. | Nuako A, Liu J, Pham G, et al. Quantifying rural disparity in healthcare utilization in the United States: Analysis of a large midwestern healthcare system. PLoS One. 2022;17(2):e0263718. doi:10.1371/journal.pone.0263718 Medline, Google Scholar |
| 23. | Long AS, Hanlon AL, Pellegrin KL. Socioeconomic variables explain rural disparities in US mortality rates: Implications for rural health research and policy. SSM Popul Health. 2018;6:72-4. doi:10.1016/j.ssmph.2018.08.009 Medline, Google Scholar |
| 24. | Adamson DM, Chang S, Hansen LG. WHITE PAPER. Health research data for the real world: The MarketScan databases. Accessed June 30, 2022. http://patientprivacyrights.org/wp-content/uploads/2011/06/Thomson-Medstat-white-paper.pdf Google Scholar |
| 25. | U.S. Bureau of Labor Statistics. Consumer price index. Accessed June 25, 2022. https://www.bls.gov/cpi/ Google Scholar |
| 26. | Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. 2011;173(6):676-82. doi:10.1093/aje/kwq433 Crossref, Medline, Google Scholar |
| 27. | Stewart LK, Sarmiento EJ, Kline JA. Statin use is associated with reduced risk of recurrence in patients with venous thromboembolism. Am J Med. 2020;133(8):930-5.e8. doi:10.1016/j. amjmed.2019.12.032 Google Scholar |
| 28. | Dorjee K, Baxi SM, Reingold AL, Hubbard A. Risk of cardiovascular events from current, recent, and cumulative exposure to abacavir among persons living with HIV who were receiving antiretroviral therapy in the United States: A cohort study. BMC Infect Dis. 2017;17(1):708. doi:10.1186/s12879-017-2808-8 Medline, Google Scholar |
| 29. | Zhang C, Spence O, Reeves G, DosReis S. Cardiovascular risk of concomitant use of atypical antipsychotics and stimulants among commercially insured youth in the United States. Front Psychiatry. 2021;12:640244. doi:10.3389/fpsyt.2021.640244 Medline, Google Scholar |
| 30. | Yang D, Dalton JE. A unified approach to measuring the effect size between two groups using SAS®. Accessed June 30, 2022. https://support.sas.com/resources/papers/proceedings12/335-2012.pdf Google Scholar |
| 31. | Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083-107. doi:10.1002/sim.3697 Crossref, Medline, Google Scholar |
| 32. | Faraone SV. Interpreting estimates of treatment effects: Implications for managed care. P T. 2008;33(12):700-11. Medline, Google Scholar |
| 33. | Tian L, Huang J. A two-part model for censored medical cost data. Stat Med. 2007;26(23):4273-92. doi:10.1002/sim.2847 Medline, Google Scholar |
| 34. | Basu A, Rathouz PJ. Estimating marginal and incremental effects on health outcomes using flexible link and variance function models. Biostatistics. 2005;6(1):93-109. doi:10.1093/biostatistics/kxh020 Medline, Google Scholar |
| 35. | US Census Bureau. QuickFacts: United States. Accessed June 30, 2022. https://www.census.gov/quickfacts/fact/table/US/PST045221 Google Scholar |
| 36. | Lau DH, Nattel S, Kalman JM, Sanders P. Modifiable risk factors and atrial fibrillation. Circulation. 2017;136(6):583-96. doi:10.1161/CIRCULATIONAHA.116.023163 Medline, Google Scholar |
| 37. | Riegel B, Moser DK, Buck HG, et al. Self-care for the prevention and management of cardiovascular disease and stroke: A scientific statement for healthcare professionals from the American Heart Association. J Am Heart Assoc. 2017;6(9):e006997. doi:10.1161/JAHA.117.006997 Medline, Google Scholar |
| 38. | Centers for Disease Control and Prevention. CDC Heart Disease and Stroke Prevention programs. Accessed June 25, 2022. https://www.cdc.gov/dhdsp/programs/index.htm Google Scholar |
| 39. | Vyas V, Lambiase P. Obesity and atrial fibrillation: Epidemiology, pathophysiology and novel therapeutic opportunities. Arrhythm Electrophysiol Rev. 2019;8(1): 28-36. doi:10.15420/aer.2018.76.2 Medline, Google Scholar |
| 40. | Ogunsua AA, Shaikh AY, Ahmed M, McManus DD. Atrial fibrillation and hypertension: Mechanistic, epidemiologic, and treatment parallels. Methodist Debakey Cardiovasc J. 2015;11(4):228-34. doi:10.14797/mdcj-11-4-228 Medline, Google Scholar |
| 41. | Graves JM, Mackelprang JL, Moore M, et al. Rural-urban disparities in health care costs and health service utilization following pediatric mild traumatic brain injury. Health Serv Res. 2019;54(2):337-45. doi:10.1111/1475-6773.13096 Medline, Google Scholar |
| 42. | Kosar CM, Loomer L, Ferdows NB, Trivedi AN, Panagiotou OA, Rahman M. Assessment of rural-urban differences in postacute care utilization and outcomes among older US adults. JAMA Netw Open. 2020;3(1):e1918738. doi:10.1001/jamanetworkopen.2019.18738 Medline, Google Scholar |
| 43. | CMS. Improving health in rural communities. Accessed June 30, 2022. https://www.cms.gov/files/document/fy-21-improving-health-rural-communities508compliant.pdf Google Scholar |
| 44. | Johnston KJ, Wen H, Joynt Maddox KE. Lack of access to specialists associated with mortality and preventable hospitalizations of rural medicare beneficiaries. Health Aff (Millwood). 2019;38(12):1993-2002. doi:10.1377/hlthaff.2019.00838 Medline, Google Scholar |
| 45. | Size of the urban and rural population of the United States from 1960 to 2020. Statista. Accessed June 30, 2022. https://www.statista.com/statistics/985183/size-urban-rural-population-us/ Google Scholar |