Clinical and economic burden associated with cardiovascular events among patients with hyperlipidemia: a retrospective cohort study

Background Annual direct costs for cardiovascular (CV) diseases in the United States are approximately $195.6 billion, with many high-risk patients remaining at risk for major cardiovascular events (CVE). This study evaluated the direct clinical and economic burden associated with new CVE up to 3 years post-event among patients with hyperlipidemia. Methods Hyperlipidemic patients with a primary inpatient claim for new CVE (myocardial infarction, unstable angina, ischemic stroke, transient ischemic attack, coronary artery bypass graft, percutaneous coronary intervention and heart failure) were identified using IMS LifeLink PharMetrics Plus data from January 1, 2006 through June 30, 2012. Patients were stratified by CV risk into history of CVE, modified coronary heart disease risk equivalent, moderate- and low-risk cohorts. Of the eligible patients, propensity score matched 243,640 patients with or without new CVE were included to compare healthcare resource utilization and direct costs ranging from the acute (1-month) phase through 3 years post-CVE date (follow-up period). Results Myocardial infarction was the most common CVE in all the risk cohorts. During the acute phase, among patients with new CVE, the average incremental inpatient length of stay and incremental costs ranged from 4.4–6.2 days and $25,666–$30,321, respectively. Acute-phase incremental costs accounted for 61–75 % of first-year costs, but incremental costs also remained high during years 2 and 3 post-CVE. Conclusions Among hyperlipidemic patients with new CVE, healthcare utilization and costs incurred were significantly higher than for those without CVE during the acute phase, and remained higher up to 3 years post-event, across all risk cohorts.


Background
The global cost of cardiovascular disease (CVD) is estimated at $ 863 billion and is estimated to rise to $ 1,044 billion in 2030 [1]. The American Heart Association has estimated the direct costs for CVD in the United States at $195.6 billion, approximately 61 % of the total CVDrelated healthcare costs [2]. Additionally, hyperlipidemia was among the top 10 costliest medical conditions in 2008 in the US adult population [3]. Presence of hyperlipidemia directly correlates with the risk of developing coronary heart disease (CHD) and future cardiovascular (CV) events [4]. Less than half of adults with elevated low density lipoprotein cholesterol (LDL-C) levels receive treatment or are adequately treated [5,6] and as a result, high-risk patients continue to remain at risk for new CV events. Almost 44 % of the US population is projected to be diagnosed with some form of CVD by 2030 [2]. These factors result in a substantial clinical and economic burden in terms of direct healthcare utilization and costs.
While several studies have examined the economic burden of CV events [7][8][9][10][11][12], to our knowledge contemporary and long-term analyses concerning these event costs incurred by hyperlipidemic patients across a range of CVD risk levels is not available. Previous studies focused on short-term healthcare costs due to CV events [13][14][15][16][17] and investigated patient populations diagnosed with acute coronary syndrome [13,14], hypertension [15], atherosclerosis [16] or diabetes [17], but not hyperlipidemia. Furthermore, prior studies focused only on the initial CV event and therefore, limited data are available regarding recurrent and subsequent CV event costs. Prior studies have investigated the economic burden of CV events over various time periods [10]; however, incremental costs among hyperlipidemic patients with and without CV events, and in particular, costs stratified by CVD risk level and associated with myocardial infarction (MI), ischemic stroke (IS) unstable angina (UA), coronary artery bypass graft (CABG), percutaneous coronary intervention (PCI), heart failure (HF) and transient ischemic attack (TIA), all in one study, have not been previously examined. Therefore, the present study is one of the first to estimate the short-term and long-term (up to 3 years) direct clinical and economic burden of new CV events among hyperlipidemic patients at different CVD risk levels and by specific CV event type.

Study design
We conducted a retrospective cohort study including patients with a hyperlipidemia diagnosis who had a new CV event matched to patients without new CV events, using the IMS LifeLink PharMetrics Plus dataset for the study period January 1, 2006 through June 30, 2012. This nationally-representative longitudinal database contains medical and pharmacy claims for over 50 million commercially-insured patients throughout the United States [7,18,19]. All claims data were from a limited dataset with de-identified patient information. No patients were directly involved in the study; therefore, this study was exempt from an Institutional Review Board review.

Study population
Patients (age ≥18) were included in the study if they had ≥1 medical claims for hyperlipidemia (International Classification of Diseases, 9th Revision Clinical Modifications [ICD-9-CM] code 272) [20] from January 1, 2006 through June 30, 2009. The first diagnosis claim date was designated as the hyperlipidemia diagnosis date. As detailed in Appendix 1, patients were required to have at least one inpatient medical claim for a new CV event (MI, IS, UA, TIA, HF, CABG and PCI) after the hyperlipidemia diagnosis date and during the identification period (January 1, 2007 through June 30, 2009). For hyperlipidemic patients with a new CV event, the earliest inpatient claim date was designated as the index date. If a patient had more than one inpatient claim for a new CV event on the index date, only one CV event was selected, according to the following hierarchy: MI, IS, UA, HF, TIA, CABG and then PCI, based on the clinical importance (e.g. acute/urgency) of CV events and CV-related procedures. The comparison group included patients with no new CV event after the hyperlipidemia diagnosis and through the end of the study period (June 30, 2012). Baseline period of the 12 months prior to the index date was utilized to characterize patients' CV risk level (e.g. history of CVE or diabetes) and comorbidity status. Patients were followed from the index date through 3 years post-index date to estimate short-term (first 30 days and 1 year) and longerterm (2 years and 3 years) direct costs.
Matching was completed in a two-step approach. The first step was 1:1 match (age, gender, US region) to assign an index date for patients without a new CV event and define the baseline period for quantifying baseline characteristics (CV risk level, comorbidities). These initially matched patients with no new CV event were then assigned the same index date as that of their matched patients who had a new CV event. This assignment of index date to patients with no new CV event provided the same baseline and follow-up time periods for the comparison of outcomes between patients with and without new CV events.
The second step of matching, propensity score matching (PSM) with 0.01 calipers, was applied to control the differences in baseline clinical and demographic factors between patients with and without new CV events within each risk cohort [21,22]. A standardized difference (STD) of >10 % was used to assess significant practical differences in the case-control comparison [23]. The baseline variables adjusted in the model were age group, gender, US region, Charlson comorbidity index (CCI) score, Chronic Disease Score (CDS), individual comorbid conditions (hypertension, arrhythmias, metabolic syndrome, liver disease, obesity, and chronic kidney disease) and number of inpatient admissions per patient per month (PPPM). The methods used in this study have also been published in prior literature [7,10]. The CCI score is based on ICD-9 codes and CDS uses pharmacy dispensation information for specific drug classes to estimate the burden of comorbidities [24]. The CCI and CDS score have been widely used in many retrospective studies [25][26][27][28].
Based on the risk level during the 12-month pre-index (baseline period), the study sample was subdivided into the following CVD risk cohorts: history of CV event, modified CHD risk equivalent (RE), moderate risk and low risk (Fig. 1). Risk levels were defined based on the National Cholesterol Education Program (NCEP) Adult Treatment Panel (ATP) III guidelines [29] (Appendix 2).
The history of CV event cohort included patients with MI, UA, CABG, PCI or IS, the modified CHD RE cohort included patients with peripheral arterial disease, abdominal aortic aneurysm, coronary artery disease, diabetes or dyslipidemia. Patients with at least two of the following three risk factors a) hypertension and/or pharmacy claim for blood pressure-lowering medication, b) aged ≥45 for men and aged ≥55 for women, c) pre-index high-density lipoprotein cholesterol <40 mg/dl were included in the moderate risk cohort, and patients with zero or one risk factor were included in the low risk cohort [see Appendix 2 for detailed ICD-9-CM codes]. Outcome measures included distribution of CV event type, healthcare utilization, and direct incremental costs (obtained from claims) incurred during the acute (first 1 month post-index date), and long-term (1, 2, 3 years post-index date) follow-up periods for hyperlipidemia patients, stratified by CVD risk level. Healthcare utilization included inpatient, outpatient, outpatient office, emergency room and pharmacy visits and direct costs associated with healthcare utilization were computed from health plan-and patient-paid amounts. Total costs included inpatient, outpatient and pharmacy costs. Costs were adjusted to 2012 US dollars using the annual medical care component of the consumer price index (CPI) to reflect inflation.

Statistical analysis Descriptive analysis
Descriptive analysis was conducted to compare demographic and clinical characteristics between patients with and without a new CV event within each risk cohort. The direct total incremental costs were calculated as the difference in total costs for patients with a new CV event and total costs for patients without a CV event. Negative incremental costs indicate that the total costs were lower for patients with new CV events than for patients without new CV events.

Multivariate analysis
The differences in economic outcomes for each risk cohort were compared among PSM cases and controls. Patients Fig. 1 Patient Selection Flowchart. *Propensity score matching was applied for each cardiovascular disease risk cohort using covariates: age group, gender, US region, baseline Charlson comorbidity index score, Chronic Disease Score, comorbidities and number of inpatient admissions per patient per month. CV: cardiovascular; CHD RE: coronary heart disease risk equivalent; PSM: propensity score matching without new CV events were designated as the reference group (controls). All analyses were performed using SAS® version 9.3 (SAS® Institute Inc., Cary, NC).

Results
Among patients with a new CV event, a large proportion had two or more new CV events (65.8 %) during the 3-year follow-up period. Second and subsequent CV events during follow-up were often the same event type as the first event.

Baseline demographic and clinical characteristics
Baseline demographic and characteristics of unmatched patients with a new CV event (N = 267,165) and patients without a new CV event (N = 184,285) are provided in Appendix 3. Baseline demographic and clinical characteristics for propensity scorematched patients with a new CV event (N = 121,820) and patients without a new CV event (N = 121,820), stratified by CVD risk level, are provided in Table 1. Patients without CV events were well-matched with patients with new CV event within each risk cohort, since the STD was <10 % for all variables included in the PSM. The majority of patients were classified in the modified CHD RE cohort (74.4 %), followed by the history of CV event cohort (8.8 %). Overall, the average age of patients with a new CV event (N = 121,820) ranged from 56 to 72 years; 65-67 % were male; and hypertension was the most common baseline comorbidity (4.7-84.4 %).

Clinical burden
MI was more commonly diagnosed than other CV event types among patients in the low-risk, moderate-risk and modified CHD RE cohorts. Frequency of MI, IS and HF was similar among patients in the history of CV event cohort (Fig. 2).
During the 1 month post-index date, among patients with history of a CV event (n = 10,741), the mean inpatient length of stay (LOS) was significantly longer among hyperlipidemic patients with a new CV event compared to those without (6.4 vs. 0.25 days, p < 0.0001, Table 2). This trend was observed across all risk cohorts. The inpatient LOS remained longer during the shortand long-term follow-up periods among patients with a new CV event, compared to those without, for all risk cohorts (e.g. history of CV event cohort inpatient LOS in year 2 = 4.14 vs. 1.50 days, p < 0.0001, and in year 3 = 3.72 vs. 1.38 days, p < 0.0001) ( Table 3).
During the 1 month post-index date, patients with history of a CV event (n = 10,741) had significantly more outpatient emergency room (ER) visits PPPM compared to patients without a new CV event (0.17 vs. 0.04 visits, p < 0.0001, Table 2). This trend continued across all risk cohorts and during all follow-up periods (Tables 2 & 3). Among hyperlipidemic patients with new CV events, all resource utilization components were highest during the 1-month postindex follow-up phase for all risk cohorts, indicating that the highest healthcare utilization occurred during the first month post-CV event. However, healthcare resource utilization during years 2 and 3 of the follow-up period remained significantly higher for patients with a new CV event than for those without, across all risk cohorts (e.g. history of CV event cohort ER visits PPPM during year 2 =0.05 vs. 0.03 visits, p < 0.0001; and year 3 = 0.05 vs. 0.03 visits, p < 0.0001).

Economic burden
Across all CV event type and risk cohorts, the direct incremental costs ranged from $17,903 to $65,825 in the first year of follow-up period, $474 to $19,617 during the second year post-CV event and $2,598 to $26,982 during the third year post-CV event (Table 4).
Direct incremental costs categorized by CV event type varied in relation to the duration of the follow-up period. The direct incremental costs accrued during the 1-month post-index phase represented approximately 45-90 % of first-year costs (data not shown). During the first year post-CV event, CABG costs were highest ($55,548-$65,825) for all risk cohorts, followed by MI ($47,840-$51,686) and HF ($41,001-$46,890). During years 2 and 3 post-index date, patients diagnosed with HF incurred the highest cost burden (year 2: $11,289-$19,617; year 3: $7,820-$26,982) among all risk cohorts. The direct incremental costs during these years were mainly driven by heart failure. For all CV event types, first-year incremental costs were higher compared to those accrued in the second and third post-CV event years; second-and third-year costs were always higher for hyperlipidemic patients with new CV events than for their matched patients without CV events.

Discussion
Our study showed that the long-term clinical and economic burden associated with CV events among hyperlipidemic patients was substantial across all risk cohorts, but especially among high-risk cohorts (i.e. patients with history of a CV event and prior CHD RE diagnosis). Our mean healthcare resource  [30].
Previous studies have reported that inpatient hospital stays and ER visits are expensive, resource intensive and impose a great clinical burden on patients [31,32]. Higher healthcare resource utilization is a major component of increased healthcare costs. Healthcare costs were higher among hyperlipidemic patients with a new CV event in the acute phase, compared to patients without a new CV event. Our results are similar to the Chapman et al. study, which concluded that patients with new CV events incurred the highest follow-up costs during the acute phase, and acute phase costs were much higher than those in years 2 and 3 [7]. However, our study also determined that incremental costs remained higher through 3 years of follow-up (year 1: $39,869 to $41,648 higher; year 2: $5,900 to $ 9,436 higher; year 3: $4,704 to $11,400 higher), for all risk cohorts of hyperlipidemic patients with a new CV event, compared to those without, emphasizing a sustained economic burden. Compared with the Chapman et al. [7] study with cost estimates from 2001-2006, the present study also provides more recent estimates for healthcare resource utilization and costs across the CVD risk spectrum (history of CV events through low risk) rather than excluding the highest risk cohort (i.e. patients with a history of CV events) as in the Chapman et al. study. Our study also captures the cost of care for multiple CV events thereby providing a more accurate estimate of the direct cost of care for patients experiencing new CV events rather than estimating the cost for each   Refer to Table 3 for length of stay and number of visits per patient per month during years 2 and 3 of the follow-up period PSM propensity score matching, CVD cardiovascular disease, CV cardiovascular, CHD RE coronary heart disease risk equivalent, SD standard deviation, LOS length of stay, PPPM per patient per month, ER emergency room a Chi-square tests were used to evaluate the statistical significance of differences in categorical variables; student t-tests were used for the continuous variables b Outpatient visits included emergency room, laboratory/pathology, radiology, outpatient surgical or diagnostic procedure and office visits    specific CV event type. Setting potentially arbitrary time cut-points to distinguish between different CV events among patients with multiple events may produce artificial cost results, as some CV events may occur with little time gap and the cost of one event is entwined with the cost of the next event.
Our study also brings to light the noteworthy clinical and economic burden among patients in the high-risk cohorts (i.e., history of CV event and modified CHD RE cohorts). Inpatient LOS was, on average, 0.09 to 4.89 days longer among patients with history of a CV event or modified CHD RE, compared to those at moderate or lower risk, during all follow-up time periods, suggesting that high-risk patients have greater healthcare resource needs. During the long-term post-CV event periods (1, 2 and 3 years follow-up), patients with a new CV event in the higher risk cohorts utilized more incremental ER visits PPPM, compared to those in the moderate-and low-risk cohorts, demonstrating the potential for a higher healthcare cost burden during the longer-term post-CV event periods. Future research is warranted to more specifically determine the underlying reasons for the sustained difference in clinical and economic burden between high-risk hyperlipidemic patients with a CV event and those without a CV event.
Our study results were similar to a study done by Karan et al., indicating similarity in findings that CVD had more outpatient and inpatients stays and economic burden of CVD is large [33]. However, this study utilized a national survey of households in India and focused on out-of-pocket spending and non-medical spending for CVD, whereas our present study focused on a patientlevel perspective of direct medical costs for new CV events. Although previous studies provide a general frame of reference, the cost estimates are not directly comparable to the incremental direct costs in the present study since the studies differed in study design (matched cohorts versus survey sample) [34] and composition of the study population (US hyperlipidemia patients versus hypertension or solely acute coronary syndrome patients including those residing in developing countries) [33,34], sample size (n = 10,741 vs. 4,669) [10], CVD risk level (low through high CV risk vs. exclusion of high secondary prevention patients) [7,8] and contemporaneous cost estimates (2009-2013 vs. 2001-2006) [7]. Due to considerable variation in costs by CV event type, the results of our analysis strengthen the importance of evaluating total and individual CV event costs, as this specific information may be essential for secondary prevention and treatment decisions for high-risk patients.
Our present study demonstrated the sustained high economic and clinical burden associated with the occurrence of CV events among hyperlipidemic patients. In patients who have already experienced or who are at high risk for experiencing a CV event, lifestyle intervention strategies alone may not be sufficient to maximally reduce CVD risk [35,36]. Current US treatment guidelines recommend lipidlowering therapy in addition to lifestyle modifications to lower LDL-C levels for primary and secondary prevention for high-risk individuals [37]. Although statins are widely prescribed for elevated LDL-C levels, 9 %-20 % of treated patients, especially highrisk patients, continue to have elevated LDL-C and remain at risk for new CV events [38]. Potential new pharmacological treatments (e.g. anti-proprotein convertase subtilisin/kexin type 9 [PCSK9] monoclonal antibodies) aimed at significantly lowering LDL-C beyond that of current available treatment options [39], could potentially help to reduce the substantial clinical and economic burden.

Limitations
Our study limitations were primarily related to the retrospective use of claims data [7,15]. Misclassification of CV risk, although it cannot be quantified, is likely to be low since the ICD-9-CM codes utilized to capture history of CVD included codes for old MI, stroke sequelae, etc. that would include a history of CVD beyond the baseline period. Similarly, important patient information, including blood pressure, smoking history and family history was not available in the claims data to more accurately classify patients within the CHD RE cohort. Also, administrative claims data do not offer information on whether an elective procedure (CABG, PCI) was planned, thus planned procedures could not be completely excluded from the study. Nevertheless, utilization of the PSM method reduced the differences between patients with and without new CV events and created a balanced study cohort, such that healthcare utilization and incremental costs were more accurately compared.

Conclusion
Substantial incremental costs and healthcare resource utilization 1 month up to 3 years post-CV event highlight the short-and long-term economic and clinical burden especially on high-risk hyperlipidemic patients and the US healthcare system. Interventions used to prevent or reduce the occurrence of CV events among patients with hyperlipidemia may result in substantial cost savings and reduce the clinical burden in the United States. Moderate risk At least two of the following three risk factors identifiable from administrative claims data: a) hypertension (ICD-9-CM code or pharmacy claim for a blood pressure-lowering agent), b) age 45 years or older for men and 55 years or older for womenc) pre-index high-density lipoprotein (HDL) cholesterol below 40 mg/dl.