Organizational Costs and Benefits of a Health System Quality Improvement Intervention to Increase Aspirin Use for Primary Prevention of Heart Attack and Stroke

Abstract

Performance improvement on clinical quality outcomes typically requires significant effort by personnel in health care organizations. Understanding the cost of quality improvement is important given diffusion of value-based contracting. This study investigates the organizational costs and benefits associated with planning and implementing the Ask about Aspirin intervention to increase use of low-dose aspirin in clinically recommended patient populations. Data from 4 health systems in Minnesota were used to estimate personnel effort and labor resource costs as well as corresponding benefits, measured as the change in aspirin use among eligible candidates during the study period. Overall personnel effort across the 4 systems was approximately 3900 hours with corresponding resource costs estimated to be $214,385. Aspirin use increased 4.7% overall, corresponding to roughly 1530 new users in the aspirin candidate population. Significant variation was observed by system in total hours reported, distribution of effort by activity type, and in benefits realized from the intervention.

Cardiovascular disease (CVD), including acute myocardial infarction (AMI) and stroke, are leading causes of death and disability in the United States.¹ The economic effects of CVD are significant as well. In 2014–2015, the estimated annual total cost of CVD was more than $350 billion, including $213.8 billion in direct medical spending.² Given these large health and economic costs, identifying cost-effective strategies to reduce CVD mortality and morbidity may confer significant societal benefits.

Effective approaches for reducing CVD mortality and morbidity include lifestyle interventions such as tobacco cessation; improved control of hypertension and cholesterol; and selective use of revascularization strategies following an index CVD event.³ One well-studied CVD prevention approach includes the use of daily low-dose aspirin. Research has demonstrated that aspirin is a cost-effective strategy for preventing a first AMI or stroke among individuals at high risk and that it decreases risk of subsequent events among those who already have had an AMI or stroke.^4–7 In current US Preventive Services Task Force Primary Prevention recommendations, low-dose aspirin is advised for adults aged 50 to 69 years who have a 10% or greater 10-year CVD risk, are not at increased risk for gastrointestinal bleeding, have a life expectancy of at least 10 years, and are willing to take aspirin daily for at least 10 years (B and C recommendation).⁸ Despite the clinical recommendation of this cost-effective prevention strategy, rates of low-dose aspirin use remain suboptimal.⁹

Launched in 2015, the Ask about Aspirin study (https://askaboutaspirin.umn.edu) is a pragmatic clinical trial conducted by a research team from the University of Minnesota to increase guideline-based use of low-dose aspirin for primary prevention of CVD in health system–based patient populations across the state of Minnesota. In the context of a mass and social media campaign, the main intervention included health professional education to disseminate accurate, evidence-based cardiovascular health information to prepare clinical professionals to rapidly evaluate the individual benefit and risk of aspirin use in the target patient population. Practice facilitators (PFs) employed by the research team also worked with health system and primary care clinic personnel to implement aspirin prescriptive behaviors.

For care delivery organizations, improving performance on clinical quality outcomes typically requires significant planning and implementation of practice changes, including changes in staff responsibilities and workflows, continuing education, and modifications to electronic health records systems. This study investigates the organizational costs and clinical benefits associated with planning and implementing the Ask about Aspirin quality improvement (QI) intervention to increase use of low-dose aspirin in patient populations for whom it is clinically recommended. As value-based payment models diffuse more broadly, health systems will face even stronger incentives to improve performance. To do this, they will need to make strategic investments around QI activities, given limited human and financial resources. Additionally, this work can inform investment decisions by payers to support practice transformation and QI efforts.

Methods

Study Setting and Target Patient Population

Fifteen health systems participated in the Ask about Aspirin study overall. Of these, 6 were invited to participate in this supplemental analysis. Four health systems (Systems 1–4) agreed to participate and provided both organizational data as well as changes in aspirin use for the target patient population over the study period.

In this context, the target patient population includes individuals between 50 and 69 years of age, who had at least 2 face-to-face visits with an eligible clinician (doctor of medicine, doctor of osteopathy, physician assistant, or advanced practice registered nurse) in the past 2 years, and had been seen in a family or internal medicine clinic affiliated with a participating health system. Primary prevention aspirin candidates were restricted to those who were estimated to have a 10% or greater 10-year CVD risk, who did not have an existing CVD diagnosis (AMI, stroke, peripheral arterial disease), a bleeding ulcer from the stomach or intestines, current anticoagulant use, or an aspirin allergy. All participating systems and clinics signed a Memorandum of Understanding, and the University of Minnesota’s Institutional Review Board approved and provided ethical oversight.

Table 1 provides summary information about each participating health system’s attributes, including number of clinics, number of physicians, time period (months) of engagement and data collection, specific QI tactics adopted by each system, and the average number of eligible aspirin candidates in each system’s target patient population. The 4 participating health systems are structurally diverse in size (eg, number of clinics and physicians) as well as their geographic service areas (eg, urban, suburban, and rural markets). The systems also vary with respect to their target patient population sizes, and range from slightly more than 3000 aspirin candidates, on average, in System 3 to almost 13 500 candidates in System 2.

Table 1.

Participating Health Systems, Implementation Strategy, and Aspirin Candidates.

	System 1	System 2	System 3	System 4
Number of clinics	15	9	5	8
Number of MDs	92	56	35	50
Time frame of data collection	36 months	30 months	30 months	24 months
Primary implementation strategy	• Best Practice Alert during clinic visit identifying primary prevention ASA candidates in the EMR	• Combination of pre-visit identification of ASA candidates by rooming staff and EMR-based algorithm to calculate patient’s ASCVD risk score	• Best Practice Alert during clinic visit identifying primary prevention ASA candidates in the EMR	• Best Practice Alert during clinic visit identifying primary prevention ASA candidates in the EMR
	• CME-accredited webinar training of all clinic health professionals	• CME-accredited webinar training of all clinic health professionals	• CME-accredited webinar training of all clinic health professionals	• CME-accredited webinar training of all clinic health professionals
	• Patient educational materials	• Patient educational materials	• Patient educational materials	• Patient educational materials
Average size of eligible targeted population	7398	13 499	3005	8928

Abbreviations: ASA, acetylsalicylic acid; ASCVD, atherosclerotic cardiovascular disease; CME, continuing medical education; EMR, electronic medical record; MD, doctor of medicine.

For the QI intervention, all 4 health systems made use of a continuing medical education–accredited webinar training for their clinical professionals and were provided educational materials on low-dose aspirin use for clinically recommended candidates. All systems leveraged their electronic medical records to facilitate identification of primary prevention aspirin candidates, although these were operationalized in slightly different ways across organizations.

Data Collection and Measurement

Research team members created an electronic data collection tool for organizations to systematically report employees’ effort, measured in hours, on the Ask about Aspirin QI intervention. Information was reported on each individual employee who participated in the QI project, job title, whether the employee’s position location was at the system or clinic level, and hours spent on the following 4 categories of QI-related activities: (1) Planning and Managing QI Implementation, including strategic and operational planning to execute the QI initiative; (2) Delivering QI Activities, including implementation of the strategic plan within the clinic flow; (3) Providing information technology support, including planning for electronic medical record-related data capture, extracting data to identify the at-risk population, and reporting of outcomes; (4) Managing QI Dissemination, including assurance that activities are implemented appropriately, internal goals are met, and all oversight and reporting functions are completed.

Data collection spanned 2 broad phases. The planning phase captured each employee’s total number of hours expended on activities corresponding to the initial strategic and operational planning of the project and varied based on each system’s time line (eg, months needed to prepare to launch). Once each system officially launched its QI intervention in its clinics, data were collected and reported to the research team on a quarterly basis to measure effort for maintenance-related activities. For the maintenance phase, respondents were asked to report on each individual employee’s effort corresponding to their average number of hours per week spent on the project. The research team then scaled these values accordingly to arrive at the reported effort levels.

As part of the Ask about Aspirin intervention, up to 2 PFs from the research team worked periodically with personnel from each health system to implement guideline recommendations through staff development and implementation of a workflow and/or through facilitation of clinician-patient discussions. PFs used standardized onboarding processes to support system-specific QI project teams in identifying primary aspirin candidates through the use of a CVD risk assessment calculator, offered patient education regarding use of low dose aspirin for primary prevention, and documented outcomes of discussions between provider and patients (ie, recommended, contraindicated, patient refusal). These data, including PF hours of engagement, were captured in internal information systems and are reported separately.

Organizational Costs.

Although organizations may incur both labor and capital-related costs associated with QI efforts, the former represent the predominant type of cost incurred by organizations for this particular QI intervention. To estimate the labor costs of personnel, the research team used reported hours as well as information on hourly wage rates by occupation in Minnesota for 2018 from the US Bureau of Labor Statistics Occupational Employment Statistics.¹⁰ Labor costs were estimated for 3 assumed wage rate levels corresponding to the 25th, 50th, and 75th percentiles of the wage rate distribution for each occupation category. Job titles reported by health system respondents were mapped to standardized occupation categories and included chief executives, medical and health services managers, business operations specialists, database administrators, family and general practitioner physicians, registered nurses, and pharmacists. Supplemental Appendix Table 1 (available with the article online) provides a crosswalk between organizations’ reported occupation titles and standardized occupation categories as well as the hourly wage rates used in this analysis. All costs were estimated in 2018 dollars.

Organizational Benefits.

From an organization’s perspective, the goal of the intervention is to reduce the number of AMI and stroke events in the target population as a result of increased aspirin use. In this analysis, the research team measured the intermediate clinical benefit of increased aspirin use and then used this information in conjunction with existing scholarly evidence on the associated risk reduction of aspirin to predict the number of events averted. Throughout the data collection process, in addition to reporting personnel effort, each system also generated quarterly reports of the total number of eligible aspirin candidates based on the aforementioned criteria and the number of patients using low-dose aspirin as documented in the electronic medical record.

Data Analysis

Descriptive analyses were performed to summarize both organizational costs and benefits overall and by system. The research team estimated how effort and costs were distributed between system- versus clinic-based personnel; by phase of the QI project (planning vs maintenance phases); and by the activity types performed in the organizations. To summarize clinical benefits, the team calculated the average size of the target patient population (eg, aspirin candidates) for each system. The team then calculated the system-level aspirin use rate at baseline when the QI intervention launched and again at the final point of data collection (24 months post launch for Systems 1–3 and 12 months for System 4). Finally, the team estimated the expected number of new aspirin users resulting from the QI intervention based on the percentage point change in aspirin use and the target patient population size.

Results

Table 2 summarizes personnel hours. Across the 4 health systems, total personnel effort was approximately 3900 hours. Individual systems varied considerably, with reported personnel effort ranging from 633.5 to 1388.5 hours. When considering effort relative to each system’s total number of aspirin candidates, Systems 1 and 3 expended considerably more effort relative to Systems 2 and 4 (Table 2). Organizations also varied in how they distributed effort between system- and clinic-level personnel. Notably, 3 of the 4 systems (1, 3, and 4) reported more than half of all personnel effort at the system level.

Table 2.

Health Care System and Clinic-Level Personnel Hours and Costs Associated With Planning, Implementation, and Maintenance of the Ask about Aspirin QI Intervention.

	System 1	System 2	System 3	System 4
Total organization personnel hours	1388.5	909.05	968.9	633.5
Target population size	7398	13 499	3005	8928
Hours per aspirin candidate	0.19	0.07	0.32	0.07
Total hours by system versus clinic personnela
System level	814.5	408.5	968.9	383.5
Clinic level	574	500.6	0	250
Total hours by stage of Ql implementationa
Planning	316	352	94	120
Maintenance—Year 1	617.5	244.4	536.9	513.5
Maintenance—Year 2	455	312.7	338	N/A
Total hours by activity typea
Managing Ql implementation (strategic and operational planning)	560	302.6	332.5	447.5
Providing IT support	144.5	186.2	281.8	147
Delivery activities (implementing the strategic plan in clinic flow)	494	264.3	247.1	156
Managing Ql dissemination and reporting (assuring activities are implemented and goals are met; oversight and reporting)	190	156	7.5	0
Total hours of research team-based practice facilitators	6.5	1 1.5	9.5	5.5

Abbreviations: IT, information technology; Ql, quality improvement.

^aCalculation excludes practice facilitator data, which were not collected by phase or activity.

Three of the 4 systems (1, 3, and 4) reported the highest share of effort during the first year following the launch of the QI intervention, whereas System 2 reported a plurality of effort during the planning phase. Differences across systems also were observed in the distribution of effort by activity type. Planning and management activities required the largest share of personnel effort across all systems. Delivery activities reflected the second most effort-consuming category for 3 of 4 systems. Providing information technology support related to both the electronic medical record best practice alert creation as well as data extraction constituted between roughly 10% and 30% of total effort. Finally, managing QI dissemination and reporting was the most variable category by system, ranging from 0 hours reported for System 4 to 190 hours for System 1. External to the organizations, research team-based PFs provided modest levels of direct support (Table 2).

Table 3 summarizes the estimated labor costs by system. Costs vary as a result of systems expending different levels of effort as well as their use of different types of personnel in the planning and operationalization of the QI intervention. Altogether, the 4 systems dedicated approximately $214 385 in labor resources over the study period. Again, wide variation was observed across systems, and ranged from $37 829 for System 2 to $78 818 for System 1 (assuming median wage rates). For Systems 1, 2, and 3, labor costs were relatively balanced between the planning and management of implementation and QI delivery activities. System 4 was an exception; it reported the vast majority of expenses for the latter category only. Information technology support ranged from 8% (System 1) to 20% of total costs (Systems 2, 3, and 4).

Table 3.

Health System Estimated Labor Costs (2018 Dollars).

	System 1	System 2	System 3	System 4
Estimated total costs across study period (exclusive of practice facilitator costs)
25th percentile of wage distribution	63 474	30 023	46 953	30 965
50th percentile (median) wage distribution	78 818	37 829	58 700	39 038
75th percentile of wage distribution	91 642	47 398	67 748	45 531
Total costs by system versus clinic personnela
System level	44 318	15 422	58 700	24 047
Clinic level	34 500	22 407	0	14 991
Total costs by stage of Ql implementationa
Planning	15 992	14 785	4991	8736
Maintenance—Year 1	35 037	1031 1	32 132	30 302
Maintenance—Year 2	27 789	12 733	21 577	0
Total costs by activity typea
Managing QI implementation (strategic and operational planning)	33 742	1 1 745	21 742	29 110
Providing IT support	6260	7706	12 208	8052
Delivery activities (implementing the strategic plan in clinic flow)	30 312	1 1 410	24 390	1877
Managing Ql dissemination and reporting (assuring activities are implemented and goals are met; oversight and reporting)	8504	6968	361	0

Abbreviations: IT, information technology; Ql, quality improvement.

^aCosts estimated using median wage rate assumption.

Table 4 summarizes each system’s benefits, measured as the change in rate of aspirin use for its respective target population. The 4 systems started at very different baseline levels of aspirin use at the time of the QI intervention launch. System 1 had the lowest rate (22.29%), whereas System 3 had the highest (52.5%). Column 3 reports the percentage point change in aspirin use between the QI launch and the final data collection point during the maintenance phase (24 months for Systems 1–3 and 12 months for System 4). Systems 1 and 3 yielded the largest gains in absolute terms, whereas Systems 2 and 4 reported minimal changes. Using each system’s average target population size, the research team next estimated the total number of new aspirin users and the cost per user. The largest gains in terms of numbers of new users were experienced by Systems 1 and 3. Additionally, System 1 had the lowest average cost per new user relative to the other systems.

Table 4.

Change in Aspirin Use Among the At-Risk Population Over Time.

System	Average at-risk population size	Baseline rate of aspirin use at Ql launch (%)	Aspirin use rate at final data collection point (%)	Estimated percentage point change in the number of at- risk patients taking aspirin at 2 years post implementation	Estimated # of new aspirin users	Estimated cost per new aspirin usera
System 1b	7397	22.29	38.26	15.97	1182.46	$67
System 2	13 499	44.78	45.59	0.81	109.34	$346
System 3	3005	52.5	61.26	8.76	263.24	$223
System 4	8928	36.34	36.07	−0.27	−24.11	NA

Abbreviations: CVD, cardiovascular disease; QI, quality improvement.

^aCalculation uses median hourly wage rates in computation of labor costs.

^bAverage determined for time periods during which target patient population information was accurately measured (eg, incorporated the 10-year CVD risk information into determination of eligibility).

Discussion

Measuring the cost of QI can inform health care delivery organizations’ efforts to prioritize where they choose to allocate resources for performance improvement, given existing resource constraints. This study characterizes the experiences of 4 health systems in Minnesota that participated in the Ask about Aspirin study, a QI intervention to increase low-dose aspirin use for primary prevention in clinically-recommended patient populations.

Study results suggest that the combined personnel effort across the 4 systems to plan and implement the intervention was approximately 3900 hours with an estimated resource cost of $214 385. Correspondingly, the estimated number of new aspirin users was 1530 among an aspirin candidate population size of approximately 32 830 persons (4.7% increase). At 10% risk, this would translate into 153 cases in 10 years. Given an average risk reduction of 12%, this suggests that approximately 18 cases would be averted if they took aspirin for 10 years. Although it is not possible to estimate the economic benefits associated with averted events for this specific target population precisely, one can approximate these benefits using published estimates of average episodic spending for patients with AMI and stroke diagnoses. Recent estimates from the Centers for Medicare & Medicaid Services Bundled Payments for Care Improvement Initiative Year 5 evaluation report average total allowed amounts for 90-day AMI and stroke episodes (including an inpatient stay) to be $25 970 and $31 205, respectively.¹¹ Under value-based purchasing models in which a health system takes financial risk for attributed enrollees’ utilization and spending, the prevention of AMI or stroke events could confer meaningful savings to the organization.

Study findings reveal significant variation across systems with respect to personnel effort, allocation between system- and clinic-level personnel, and across the 4 activity types. Although multiple factors may contribute to a given system’s effectiveness with respect to increasing aspirin use, the 2 systems that deployed higher levels of resources overall relative to the size of their target patient populations experienced larger intermediate clinical benefits. Observations by research team members also suggest that the 2 systems demonstrating greater improvement exhibited strong leadership support throughout the initiative as well as somewhat greater investment during the initial year of the rollout. In contrast, the other systems did not exhibit the same degree of leadership support and/or were experiencing significant organizational change, given recent merger and acquisition activities. The experiences of these 4 health systems illustrate how factors including leadership support for QI, organizational culture and change management, and variability in implementation processes (eg, coordination, communication, information technology capacity, effective teamwork) can influence the likelihood of success.

Limitations

This study is subject to several potential limitations. First, although the participating health systems are organizationally and geographically diverse, one cannot easily generalize these results to broader health system populations. Second, health care delivery organizations can pursue a range of strategies to improve the quality of care delivered to their patient populations. This study analyzed the costs and benefits of a specific intervention—one focused on health professional education to enhance guideline adherence for aspirin use. The research team acknowledges that other interventions and approaches could be used to address the clinical quality issue at hand and that these alternatives may generate different costs or outcomes. Third, various assumptions had to be employed to estimate the economic costs. Notable among these are the classification of position titles reported by health system respondents to standardized occupation categories in the Occupational Employment Statistics data. As positions in clinics and systems may not be defined uniformly, this may lead to some measurement error. Moreover, the wage rates used from the occupational employment data may not correlate perfectly with the actual labor costs incurred by the organization. However, by estimating costs using different points in the occupation-specific wage distributions, it is possible to have a range of plausible values. Finally, while the Ask about Aspirin intervention is a group randomized trial, the economic analysis reported here is based solely on observational data from a subset of health systems invited to participate. Thus, one cannot easily isolate the effect of this intervention relative to other secular changes that may have been occurring in these health systems or the geographic markets they serve.

Conclusion

Improving the quality of care can require significant investment of monetary and human resources in delivery organizations.¹² Yet this information often is not available to organizations when they are formulating strategic priorities around quality improvement. This study contributes to a growing literature examining the organizational costs associated with QI activities, including participation in diabetes QI collaboratives,¹³ depression care model implementation,¹⁴ primary care–based behavior change interventions,¹⁵ and quality reporting requirements.¹⁶ Such information can facilitate better planning by health system leaders as they assess the cost-benefit trade-offs of different QI investments, and a clearer understanding of the cost of QI can inform negotiations between providers and payers, particularly as value-based payment models diffuse more broadly.