FormalPara Key Points for Decision Makers

Healthcare cost burden of acute myocardial infarction (AMI) is high and hospital services accounted for the largest proportion of healthcare costs for AMI patients.

Compared with other MI groups, Non-ST segment elevation MI is the biggest cost driver.

Healthcare costs for AMI decreased during the study period in the province of Alberta, Canada despite an increase in health service utilization among AMI patients. Further investigation is needed to better understand the trends and to explore replication of the factors that contributed to a decrease in AMI healthcare costs in other jurisdictions.

1 Introduction

Acute myocardial infarction (AMI) is an acute condition requiring expedited diagnosis and intervention. It is the leading cause of morbidity and mortality in many parts of the world [1] and accounts for half of the 17 million worldwide annual deaths from cardiovascular disease (CVD) [2]. Hospital readmission is common among patients with AMI [3]. It is estimated that 37% of the US population were living with CVD in 2010 and the direct medical costs for CVD are predicted to triple from US$272.5 billion in 2010 to US$818.1 billion in 2030, while the productivity losses are projected to increase 61% during the same period [4]. In Canada, CVD is the second most expensive healthcare burden, costing Can$22 billion in 2000 (~ Can$30 billion in 2016 dollars) [5]. The higher prevalence of AMI associated with an aging population [6] and improved survival among patients with AMI [7] will likely increase its cost burden over time. However, little is known about the current cost burden of AMI on the healthcare systems, and how this cost burden is changing over time. In this study, we examined the healthcare costs associated with AMI between 2004 and 2013 in the province of Alberta, Canada.

2 Methods

2.1 Data Source and Study Population

We conducted a retrospective cohort study using five linked administrative databases including Ambulatory Care Database, acute care inpatient records (Discharge Abstract Database [DAD]), Practitioner Claims, Pharmaceutical Information Network (PIN), and Population Registry in the province of Alberta, Canada [8]. Ambulatory Care Database contains all ambulatory care utilization (same-day surgery, day procedures, emergency department visits, and community rehabilitation services at publicly funded facilities). It provides information on patient demographics, diagnoses, procedures, and Comprehensive Ambulatory Classification System (CACS) grouping identifying homogenous patient clusters [9]. DAD tracks all acute-care hospitalizations in the province. It contains patient demographics, diagnoses, and Case Mix Group (CMG) classification (equivalent to CACS for hospitalized patients) [10]. Practitioner Claims contains fee-for-service claims information for physicians and other providers for insured health services. PIN records information on drugs dispensed for prescribed medications at almost all pharmacies in the provinces. Finally, Population Registry provide demographic and vital statistics for all inhabitants of Alberta [8]. The annual population of Alberta during 2004–2013 was used to calculate AMI prevalence over time [11].

We first selected all ambulatory care visits and hospitalizations from April 1, 2004 to March 31, 2014 (fiscal years [FY] 2004–2013) in which AMI (International Classification of Diseases [ICD], 10th revision, codes I21 or I22) was recorded as the primary diagnosis to create the study cohort. All practitioner claims relating to cohort members with a primary diagnosis of AMI (ICD, 9th revision, code 410) and AMI-related drug dispensing events to cohort members that occurred during the study period were then retrieved. AMI-related drugs were defined as (1) antihypertensive agents, (2) diuretics, (3) peripheral vasodilators, (4) β-blockers, (5) calcium channel blockers, (6) agents acting on the renin–angiotensin system, (7) lipid-modifying agents, and (8) antithrombotic agents. Anatomical Therapeutic Chemical (ATC) drug codes are presented in Supplemental Table S1 in the electronic supplementary material.

2.2 Main Outcomes

The main outcomes were total and annual costs for AMI. From a healthcare payer perspective, we defined annual costs as the annual summation of hospitalization, ambulatory care, practitioner claims, and drug costs incurred for all patients with AMI as the primary diagnosis. Other outcomes were distribution of healthcare costs by MI subtypes (ST-segment elevation MI [STEMI], non-ST-segment elevation MI [NSTEMI], and undefined/recurrent MI [Other MI]), and by sex and age (< 50, 50–59, 60–69, 70–79, and ≥ 80 years). We also examined distribution of hospitalization costs by revascularization procedures (percutaneous coronary intervention [PCI]) and coronary artery bypass grafting [CABG]).

2.3 Hospitalization Costs

We used the Alberta Interactive Health Data Application (AIHDA) Hospital Inpatient Care Case Costs version 2013 to provide dollar values for each hospitalization based on its CMG classification [12]. The AIHDA CMG costs provide functional center direct (e.g., nursing, diagnostic, and therapeutic costs) and indirect costs (e.g., general administration and support services) [13]. As the CMG codes were only available for FYs 2006–2012 in the DAD, we developed a generalized linear model (GLM) with gamma distribution and log link to estimate the average costs per hospitalization during this time period using a previously described algorithm [14]. The variables included in the model were FY, MI subtypes, patient sex and age group. We used the model’s coefficients to calculate average annual costs per hospitalization for FYs 2004, 2005, and 2013 for which CMG codes were not available. Annual AMI hospitalization costs for each combination of sex, age group, and MI subtype were derived by multiplying number of hospitalizations in each group by the costs per hospitalization in that group.

2.4 Ambulatory Care Costs

We used the AIHDA Ambulatory Care Case Costs version 2013 to provide dollar values for each ambulatory care visit based on its CACS code [12]. Similar to AIHDA CMG costs, the AIHDA CACS costs provide functional center direct and indirect costs [15]. The CACS codes were only available for FYs 2006–2012 in the ambulatory care database. Therefore, we also developed a GLM model with gamma distribution and log link with FY, MI subtypes, patient sex and age group being independent variables to estimate the average costs per ambulatory visit during this time period and used the model’s coefficients to calculate average costs per ambulatory care visits for FYs 2004, 2005, and 2013, for which the CACS codes were not available. Annual AMI ambulatory care costs for each combination of sex, age group, and MI subtype were derived by multiplying the number of visits in each group by the costs per visit in that group.

2.5 Practitioner Claims Costs

The practitioner claims database provides a paid amount and a system assessed amount to each claim. We used the paid amounts as costs for fee-for-service claims. For alternative relationship plan (ARP) claims (~ 10% of all claims) where the paid amounts were unavailable, the system assessed amounts were used instead. We summed all claim costs in a year to determine annual practitioner claims costs.

2.6 Drug Costs

The drug data were only available from FY 2008–2013. We used the Alberta Drug Benefit List (ADBL) [16] to provide drug unit prices. If a drug price was not listed in the ADBL, we used the market price at Canada Drugs [17]. Drug costs of a dispensing event were then derived by multiplying the price of a dispensed unit by the number of units dispensed. We summed the costs of individual dispensing events per year to determine annual drug costs.

2.7 Statistical Analysis

We summarized patient characteristics using means (± SD), medians (interquartile ranges), counts and percentages, as appropriate. A patient was counted each year if s/he had either a hospitalization or an ambulatory care visit or a practitioner claim during that year. As there is no indication of MI subtype in the practitioner claims and PIN databases, we assumed the MI subtype of the previous ambulatory care visit or hospitalization, whichever was closest, for a claim or drug dispensing event. We used previously validated ICD codes to identify patient comorbidities [18]. Comorbidities were considered to be present if they were recorded in any hospitalization or ambulatory care visit in each year of the study period. We used the Canadian Classification of Health Interventions [19] to identify PCI (codes of 1.IJ.50** and 1.IJ.57.GQ**) and CABG (code of 1.IJ.76**) in any of 20 intervention fields in the DAD record. Univariate GLM regression was used to test for trends of costs over time. A non-parametric trend test was used for trend tests of medians, and univariate linear regression was used for trend tests of means. All costs were converted to Canadian 2016 dollar values using the Bank of Canada inflation calculator [20].

All analyses were performed using Stata version 14 (Stata Corporation, College Station, TX, USA); Two-sided p values < 0.05 were considered statistically significant. The study was approved by the Health Research Ethics Board at the University of Alberta (IRB number: Pro00066424).

3 Results

3.1 Study Cohort Descriptive Statistics

Between FY 2004 and 2013, there were 55,384 hospitalizations, 75,309 ambulatory-care visits, 524,238 practitioner claims, and 4,798,869 drug-dispensing events involving 52,912 unique patients with AMI (Table 1). Patient characteristics by year are presented in Table 2. The number of patients increased from 6031 in 2004 to 7455 in 2013 (p < 0.001), as did the number of hospitalizations (2004: 4935; 2013: 6071; p < 0.001). However, population prevalence rates remained stable (p = 0.782). The mean age of patients decreased over time (p < 0.001), as did the proportion of females (p < 0.001). The median hospital length of stay (LOS) decreased from 7 days in 2004 to 5 days in 2013 (p < 0.001). Heart failure and diabetes mellitus were the most common comorbidities. The rates of heart failure decreased (2004: 16.4%; 2013: 12.7%; p < 0.001) while the rates of diabetes mellitus increased over time (2004: 20.1%; 2013: 29.0%; p < 0.001). However, the median Charlson score remained unchanged during the study period (p = 0.263).

Table 1 Health services utilization for patients with acute myocardial infarction (AMI) in Alberta, Canada, 2004–2013
Full size table
Table 2 Characteristics of study population
Full size table

3.2 Total Healthcare Costs

Overall, AMI cost the Alberta healthcare system Can$1033 million between 2004 and 2013. Of this, hospitalization costs accounted for the majority (Can$716.4 million), followed by drug costs (Can$147.2 million), ambulatory care costs (Can$94.5 million), and practitioner claims costs (Can$74.9 million). Proportionally, hospitalization, drug, ambulatory care, and practitioner claims costs, respectively, accounted for 63.1, 21.1, 8.8, and 7.0% of the total healthcare costs for AMI between FY 2008 and 2013 where data of all four service components were available (Fig. 1).

Fig. 1
figure 1

Healthcare costs for AMI in Alberta, Canada in 2016 Canadian dollars, 2004–2013

Full size image

Both hospitalization (annual average Can$71.6 million; p = 0.141) and ambulatory care (annual average Can$9.5 million; p = 0.888) costs remained unchanged despite a decrease in costs per hospitalization during the study period (2004 Can$14,116; 2013 Can$11,792; p < 0.001). Practitioner claims costs increased from Can$5.9 million in 2004 to Can$9.1 million in 2013 (p < 0.001), primarily due to a 7.6% annual increase in the number of claims during the study period. In contrast, drug costs dropped from Can$28.6 million in 2008 to Can$18.3 million in 2013 (p < 0.001), despite a constant increase of annual drug dispensing events (Table 1). As a result, annual healthcare costs for AMI decreased from Can$124.6 million in 2008 to Can$108.6 million in 2013 (p = 0.002).

3.3 Distribution of Costs by Myocardial Infarction Subtype, Sex and Age, and Drug Class

Overall, care for STEMI was more costly than it was for other MI subtypes in both hospital and ambulatory care settings (Table 3). However, since NSTEMI occurred more frequently, it accounted for more than half of the total healthcare costs. While the proportion of STEMI costs (34.2%) remained stable during the study period (p = 0.921), the NSTEMI contribution increased from 44.2% in 2004 to 54.7% in 2013 (p = 0.006) and other MI costs decreased from 21.5% in 2004 to 10.8% in 2013 (p = 0.006) (Fig. 2). A detailed distribution of the costs by MI subtype in hospitalization, ambulatory care, practitioner claims, and drug domains are presented in Supplemental Figure 1 (see electronic supplementary material).

Table 3 Hospitalization and ambulatory care costs for acute myocardial infarction (AMI) in Alberta, Canada in 2016 Canadian dollars, 2004–2013
Full size table
Fig. 2
figure 2

Distribution of healthcare costs by AMI subtype in Alberta, Canada, 2004–2013. MI myocardial infarction, NSTEMI non-ST-segment elevation MI, STEMI ST-segment elevation MI

Full size image

The total AMI healthcare cost for males was double that for females. Elderly groups (aged ≥ 70 years) accounted for more than half of total costs for females, but only 34.1% for males. In contrast, the youngest group (aged < 50 years) accounted for 14.1% of total costs for males, about two times higher than their female counterparts (Table 4).

Table 4 Healthcare costs for acute myocardial infarction (AMI) by sex and age in Alberta in 2016 Canadian dollars (millions), 2004–2013
Full size table

Lipid-modifying agents, antithrombotic agents, and agents acting on the renin–angiotensin system accounted for the highest drug costs, while diuretics and peripheral vasodilators accounted for the lowest drug costs during the study period (Fig. 3). The annual costs of lipid-modifying agents (2012: Can$12.7 million; 2013: Can$5.7 million; p < 0.001) and agents acting on the renin–angiotensin system (2012: Can$5.4 million; 2013: Can$3.9 million; p < 0.001) dropped gradually over time whereas the annual costs of antithrombotic agents reached a peak of Can$9.5 million in 2011 and then dropped by half in the following years (Table 5).

Fig. 3
figure 3

Drug costs distribution by main Anatomical Therapeutic Classification (ATC) groups

Full size image
Table 5 Drug costs for acute myocardial infarction (AMI) by Anatomical Therapeutic Classification (ATC) groups in Alberta in 2016 Canadian dollars (thousand), 2004–2013
Full size table

3.4 Distribution of Hospitalization Costs by Revascularization Procedure

The proportion of hospitalizations with PCI increased from 33.6% in 2004 to 37.8% in 2013 (p < 0.001). In contrast, the proportion of hospitalizations with CABG decreased from 3.0% in 2004 to 1.1 % in 2013 (p < 0.001). The cost per hospitalization with CABG (Can$32,880) was twice as high as the cost per hospitalization with PCI (Can$14,147; p < 0.001). Accordingly, the annual costs for hospitalizations with PCI increased from Can$24.2 million (34.7%) in 2004 to Can$27.8 million (38.9%) in 2013 (p = 0.012), whereas the annual costs for hospitalizations with CABG remained stable during the study period (Can$2.4 million/year; p = 0.356).

4 Discussion

Our study of healthcare costs for patients with AMI in the province of Alberta, Canada showed that AMI cost the province approximately Can$1 billion between 2004 and 2013 (Can$100 million per year, equaling 0.4% of provincial health expenditure in 2013 [21]). Hospital services accounted for the greatest share and practitioner claims accounted for the smallest share of total AMI healthcare costs. During the study period, annual AMI healthcare costs decreased due to a reduction in drug costs and stable hospitalization and ambulatory care costs. NSTEMI was the main cost driver for patients with AMI in the province of Alberta.

In response to ever-increasing healthcare costs, cost containment efforts have been implemented in many countries, including Canada [22]. In addition, there have been significant advances in diagnosis and treatment of AMI in the last several decades, which may lead to productivity improvements in AMI care [23]. For example, there has been a substantial increase in the use of PCI and a decrease in the use of CABG, which is far more costly and resource-intensive than PCI in AMI patients [24,25,26]. Furthermore, the Canadian Institution for Health Information reports a substantial increase in the generic share of prescriptions in Alberta from 56.6% in 2008 to 72.1% in 2013, and this could contribute to the decreased annual drug costs as observed in this study [27]. The combination effect of these factors might explain the decrease in annual healthcare costs for AMI in Alberta. However, further investigation is needed to better understand this effect.

We found that the per capita cost of AMI in Alberta was Can$31.7 per capita in 2010 (Can$117.4 million for a population of 3.7 million [11]). This is lower than the cost of Can$131.2 per capita in the US [4, 28] (converted to Can$, 2016 values using the Federal reserve system exchange rate [29] and Bank of Canada inflation calculator [20]). In addition, we observed a downward trend in healthcare costs while Heidenreich et al. forecasted the opposite to happen in the US [4]. This may be due to the fact that the US study included a broader population (AMI plus angina pectoris and other forms of chronic ischemic heart disease). Furthermore, variations in patterns of practice between the two countries may have an impact on medical costs [30, 31], and overall drug prices in the US are approximately 2.9 times higher than in Canada [32].

Our study confirmed that hospital services and drugs account for the majority of total healthcare costs for patients with AMI. While Seo et al. reported that hospitalization costs could account for 73% of total medical costs for AMI in Korea [33], Mantovani et al. reported a 77 and 15% attribution of inpatient care and drug costs, respectively, to the total medical costs for AMI in Italy [34]. In addition to opportunities to improve quality of care while controlling costs by preventative strategies [35], these findings emphasize the importance of cost-saving strategies in hospital (i.e., utilizing the right procedure for the right patient and reduction of LOS). During the study period, we observed decreased costs per hospitalization which may be a composite effect of the reduction in costs for patients who underwent PCI [36] and decreased LOS during the study period. It was reported previously that differences in procedure and LOS could contribute 34 and 53% to the variation in total hospital costs, respectively [31].

Even though STEMI is an acute condition requiring expedited diagnosis and intervention and was more costly to treat, we found that NSTEMI was the biggest resource consumer and cost driver. Our findings are consistent with previously reported results which show a significant increase of the NSTEMI proportion from 52.8% in 2002 to 68.6% in 2011 in the US [37]. These findings suggest NSTEMI patients could be a potential target for future cost-containment efforts. Recent studies have shown a high proportion of NSTEMI patients in Canada could be admitted to telemetry wards rather than resource-intensive and costly critical care units with no difference in clinical outcomes [38, 39].

Although our study provides novel data on the healthcare costs of AMI in the province of Alberta, it has some limitations. First, we did not include productivity losses due to mortality and morbidity. Indirect costs for AMI could be equal to or higher than direct medical costs [4, 33]. Therefore, the true cost burden of AMI on Albertans is substantially higher than what we reported. Second, we did not have drug data for the years 2004–2007, so the drug costs could be significantly higher. Third, we did not include the healthcare costs incurred at privately funded services (e.g., privately funded rehabilitation services). However, these costs should be minor in a universally covered and publicly funded health system like in Alberta. Finally, we could not account for any diagnostic errors and misclassification (in terms of MI subtype). Although Canadian administrative data have been shown to be valid relative to chart abstraction, these data may be affected by data entry errors, omissions, and inconsistencies [40, 41].

5 Conclusions

Healthcare costs for AMI are significant; however, they decreased slightly during the study period in the province of Alberta. Not surprisingly, hospital services accounted for the largest proportion of the AMI care costs. There are still opportunities to further improve cost-saving efforts in AMI care. Our study suggests that further investigation is needed to better understand cost-saving efforts in the province of Alberta and to explore the possibility of replication to other jurisdictions.