Methods

To estimate the per capita incremental cost burden of AF in 2009, we use propensity matching. As patients with undiagnosed AF are not observed in the data, however, we modified the standard propensity matching approach to measure the per capita incremental cost burden for all patients with AF, including both patients who are diagnosed and undiagnosed. A detailed description of this process is mentioned in the following.

To study the cost burden in patients aged 18 to 64 years, we used 2004 to 2010 Touchstone claims data (Optum, Eden Prairie, Minnesota). These commercially available data contain de-identified administrative, claims, and benefit information for employees and dependents insured through 33 Fortune 500 employers. The data contain patient demographics, health status, health service utilization, and cost. The Touchstone data have been used previously to study cardiovascular and other conditions.

To study the burden of undiagnosed AF in patients aged ≥65 years, we used the 2004 to 2010 Medicare Limited Data Set, which is a 5% national sample of fee-for-service Medicare beneficiaries. These data contain information on beneficiary demographics, health status, health service utilization, and cost, but unlike the Touchstone data, the Limited Data Set files do not include prescription drug claims.

To implement a propensity-matching approach, our study identified 2 cohorts of patients: those with a qualifying nonvalvular AF diagnosis and all others. For both groups, we required all patients in our study population to be US residents aged ≥18 years. To identify existing co-morbidities and future economic cost, all individuals also were required to be continuously enrolled for ≥12 months before and for ≥12 months Q3 2009.

Patients were identified as having nonvalvular AF if they had a valid AF diagnosis and no indication of valvular or transient AF. Valid diagnoses are ones with an AF diagnosis if an International Classification of Diseases, Ninth Revision (ICD-9), code of 427.31 appeared on 1 of their inpatient, outpatient, or physician visits claim. Claims in which the providers were not clinicians or in which the place of service was a laboratory were not considered valid for identifying AF. In those with AF, the sample was limited to patients with nontransient and nonvalvular AF as both transient and valvular AF are associated with different levels of stroke risk, and thus, treatment approaches vary. Patients were considered to have valvular AF if they had a diagnosis for a heart valve replacement or mitral valve stenosis (ICD-9-CM codes 42.2, 394.x, 396.1, 396.2, 396.8, or 746.5) or a procedure for a valve replacement (current procedural terminology codes 33405, 33420, 33422, 33425 to 33427, 33430, or 33496; ICD-9 procedure codes 35.0x, 35.1x, or 35.2x). Patients were considered to have transient AF if AF appeared within 12 months after hyperthyroidism (ICD-9-CM code 242.x) or within 30 days after coronary artery bypass surgery (ICD-9 procedure codes 36.10 or 36.19), pericardial surgery (37.10 to 37.12, 37.24, 37.25, 37.31 to 33, 37.35, or 37.40), or structural cardiac repair surgery (35.31 to 35.39, 35.41 to 35.42, 35.50 to 35.54, 35.60 to 35.63, or 35.70 to 36.73).

The primary outcome of interest was direct medical costs (in 2014 USD). Direct medical costs included payments by insurers and patients (e.g., co-insurance, co-payments, deductibles) and were measured over a 12-month period. We divided costs into AF-related, other cardiovascular, and noncardiovascular costs. We defined costs associated with claims with a primary diagnosis code of 427.31 as AF-related cost. Costs appearing on claims with primary diagnosis codes for cardiovascular disease other than AF were defined as other cardiovascular costs, whereas cost for claims with neither AF nor other cardiovascular diagnosis code at the primary position were placed in the noncardiovascular cost category. For the Touchstone data, we considered drug costs associated with anticoagulant prescriptions as AF-related costs, and costs for prescriptions included in RedBook Therapeutic Group 7 (cardiovascular system) as other cardiovascular costs. All other prescription drug costs were included in the noncardiovascular category.

Control variables for the propensity score matching included patient demographics (i.e., age, gender), and co-morbidities (acute myocardial infarction, cancer, cardiomyopathies, chronic pulmonary disease, congenital heart disease, congestive heart failure, diabetes, dyslipidemia, hypertension, ischemic heart disease, obesity, pericarditis, peripheral vascular disorders, renal failure, and sleep apnea). We also measured the annual number of evaluation and management physician visits to predict the likelihood an AF diagnosis was reported in the claims data.

To determine the per capita incremental burden of undiagnosed AF using propensity matching, we estimated the propensity of undiagnosed AF for all patients, matched patients with undiagnosed AF to patients with no AF, and calculated the difference in average costs. As patients with undiagnosed AF are not observed in the data, we implemented a modified propensity-matching approach to measure the per capita incremental cost burden for both patients with diagnosed and undiagnosed AF. First, we calculated the probability of having diagnosed or undiagnosed AF using a constrained maximum likelihood estimator. We used patient demographics and co-morbidities to predict the probability of having AF (diagnosed or undiagnosed) and then used the number of physician visits to model the likelihood a patient is diagnosed. Then, we matched patients with diagnosed AF to those without an AF diagnosis using a 1:1 “nearest neighbor” matching technique. Finally, we reweighted the sample to account for that the propensity score of having AF is applied only to a subset of the full AF population (i.e., those observed to be diagnosed with AF in the data). After the mean per capita incremental cost of AF was estimated, we used estimates of the prevalence of undiagnosed AF estimates from our previous work—based on the same Medicare and commercial data set—to calculate the total US incremental cost burden of undiagnosed AF. See Supplementary data for a more complete description of this approach.

We conducted two sensitivity analyses to test how our conclusions would change various assumption. First, undiagnosed AF may be less costly to treat if undiagnosed patients have less-severe symptoms than those with diagnosed AF; alternatively, the incremental cost burden for these patients may be greater if undiagnosed patients are less likely to receive cost-effective treatments. To address this issue, we tested how the national cost burden of undiagnosed AF would change if the incremental cost burden of undiagnosed AF ranged from 40% less to 40% greater than the measured incremental cost burden for diagnosed AF. Second, we assumed that approximately 10% of US adults with AF are undiagnosed, but some research indicated that rates of undiagnosed AF may fall from 18% to 27% of all patients with AF. We recalculated the US burden of undiagnosed AF using these alternative undiagnosed AF prevalence rates.

Results

Of the initial sample of 2.4 million Medicare beneficiaries and 2.2 million commercially insured people in the Touchstone database in 2009, we identified approximately 913,000 Medicare and 338,000 privately insured beneficiaries who met the age (≥18-year-old), US residency, and continuous enrollment requirements ( Figure 1 ). In the Medicare cohort, we found 81,244 patients with diagnosed AF and in the privately insured cohort, 2,975 patients were found with diagnosed AF. The AF and matched non-AF cohorts were well balanced with regard to gender, age, and co-morbidity covariates after matching (i.e., standardized difference <5%) with the exception of diabetes and dyslipidemia characteristics in the population aged 18 to 64 years ( Table 1 ).

CONSORT diagram. The selection process applied to enrollees from the Medicare Limited Data Set and the Optum Touchstone database. Inclusion criteria of age and US residency requirements and continuous enrollment for 2 years resulted in matched samples of 81,244 and 2,965 enrollees in Medicare and Optum, respectively.

Incremental per capita cost burden was greater for patients aged 18 to 64 years compared to patients aged ≥65 years. For patients aged ≥65 years, mean annual direct costs per patient were $25,322 (95% confidence interval [CI] $25,049 to $25,595) for patients with AF compared to $21,706 (95% CI $21,563 to $21,849) for the matched beneficiaries without AF; an incremental cost difference of $3,616 (95% CI $3,308 to $3,924, p <0.001). For patients aged 18 to 64 years, the per capita annual costs were $38,861 (95% CI $35,781 to $41,950) for patients with AF compared to $28,506 (95% CI $28,409 to $28,603) for matched patients without AF, an incremental cost difference of $10,355 (95% CI $7,050 to $13,659, p <0.001). The incremental AF burden per capita was greater for women compared to men, and inpatient cost was the key driver of additional AF-related cost ( Table 2 ).

Table 2

Per patient incremental cost burden

Cost Breakdown	Individuals <65 years of age			Individuals ≥65 years of age
	AF Cost (n=2975)	Non-AF Matched Cost (n=2975)	Incremental Burden	AF Cost (n=81,244)	Non-AF Matched Cost (n=81,244)	Incremental Burden
All	$38,861	$28,506	$10,355 ∗	$25,322	$21,706	$3,616 ∗
Males	$35,015	$26,982	$8,033 ∗	$24,100	$20,628	$3,472 ∗
Females	$46,654	$31,594	$15,060 ∗	$26,407	$22,663	$3,744 ∗
All	$38,861	$28,506	$10,355 ∗	$25,322	$21,706	$3,616 ∗
AF	$3,649	$85	$3,563 ∗	$1,263	$148	$1,115 ∗
Other CV	$6,778	$4,649	$2,129 ∗	$4,786	$3,908	$877 ∗
Non-CV	$28,434	$23,771	$4,662 ∗	$19,273	$17,649	$1,624 ∗
Inpatient Medical	$16,531	$8,881	$7,650 ∗	$9,139	$7,323	$1,816 ∗
AF	$1,830	$55	$1,775 ∗	$423	$70	$353 ∗
Other CV	$44,180	$2,728	$1,453 ∗	$2,239	$1,813	$426 ∗
Non-CV	$10,521	$6,098	$4,422 ∗	$6,478	$5,441	$1,037 ∗
Outpatient Medical	$15,668	$12,922	$2,746 ∗	$16,183	$14,383	$1,800 ∗
AF	$1,789	$29	$1,760 ∗	$840	$79	$762 ∗
Other CV	$2,458	$1,815	$642 ∗	$2,547	$2,096	$451 ∗
Non-CV	$11,422	$11,078	$344 †	$12,795	$12,208	$587 ∗
Pharmacy	$5,163	$5,438	-$275 †	–	–	–
AF	$524	$72	$452 ∗	–	–	–
Other CV	$1,535	$1,398	$136 †	–	–	–
Non-CV	$3,104	$3,967	-$863 ∗	–	–	–

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Predictors and Outcomes of Routine Versus Optimal Medical Therapy in Stable Coronary Heart Disease
2. Predictive Value of Gamma-Glutamyl Transferase Levels for Contrast-Induced Nephropathy in Patients With ST-Segment Elevation Myocardial Infarction Who Underwent Primary Percutaneous Coronary Intervention
3. Temporal Changes of Platelet Reactivity After Coronary Stenting—A Thing to Think About
4. Vortex Keratopathy: Fabry Related or Amiodarone Induced?

Stay updated, free articles. Join our Telegram channel

Join