Heart Health: A Cardiovascular Prevention Choice Tool

Notes on its development

Submitted by Sandra Hartasanchez

Introduction

Cardiovascular disease (CVD) continues to be a leading cause of mortality and disease burden worldwide. There are several approaches to prevent CVD and new ones continue to emerge.[1] The American College of Cardiology/American Heart Association guidelines for primary cardiovascular (CV) prevention recommend shared decision making (SDM) to co-create individualized plans for preventive care.[2]

As part of an NIH-funded project which aims to identify implementation approaches that promote high-quality SDM about CV prevention for patients in primary care settings, designers and researchers at the KER Unit created Heart Health: A CV Prevention Choice Tool, an SDM tool for use by patients and clinicians during the clinical encounter to co-create a plan of care. The tool:

• Offers an individualized 10-year estimate of the patient’s atherosclerotic cardiovascular disease (ASCVD) risk.
• Offers lifestyle interventions – Mediterranean diet, physical activity, smoking cessation – and medication – e.g., statins, aspirin, other lipid, blood pressure, glucose lowering drugs –  options and demonstrates how starting/continuing/stopping them can affect jointly and cumulatively the patient’s ASCVD risk.
• Supports a conversation by which patient and clinician create an evidence-based plan that is desirable, useful, and feasible for both.

This document describes the development process we followed to produce Heart Health.

Steps for the design, evaluation, and prototyping of the tool

A. Literature review

Clinicians and researchers thoroughly searched the literature for relative risk reduction estimates of major adverse cardiovascular events for the interventions of interest. The relative risk (RR) is the probability of an outcome in the exposed group compared to the probability of an outcome in the unexposed group.[3] The RR estimates for smoking cessation and exercise come from observational cohort studies. All other estimates of the effect of these interventions on CV risk come from randomized trials. When possible, we sought high quality systematic reviews of these primary studies. For each intervention we also sought evidence on other (nonCV) benefits, medication administration routine, adverse effects, and patient out-of-pocket costs. 

Among the interventions of interest were nutrition and diet, antiplatelet medication, glucose-lowering medications, lipid-lowering therapy, blood pressure lowering medications, smoking cessation, and weight loss.

B. Summarizing the evidence

In addition to the factors needed to compute a 10-year ASCVD risk, we also included 2 additional parameters: lipoprotein (a) and coronary calcium score. We assumed these could be helpful in patients who were unsure as to how intensely to pursue preventive care (e.g., patients at so-called intermediate CV risk). When evaluating the use of the tool in clinic, it was evident that these parameters were rarely available in primary care settings and rarely used in discussions. Thus, we decided to exclude them from the current version.  

Another parameter that was added after discussing with experts was a question on family history of premature (males <55 years, females <65 years) myocardial infarction, stroke, or sudden death in a first degree relative. This question did not affect the risk calculations per se but, if selected, a disclaimer would be displayed when calculating risk: “Your family history of heart disease means that your risk may be higher than shown. Consider further discussion with a preventive cardiologist.”

For the activities included (smoking cessation, Mediterranean diet, and exercise), it was decided to include links to patient education websites created by the Mayo Clinic for each activity, where patients and clinicians could obtain more detailed information and suggestions on how to make these changes to their lifestyle.

For Diabetes medications, we included GLP-1 agonists and SGLT-2 inhibitors. If the patient has diabetes, these two medications are part of the medications table. If the patient does not have diabetes, they are not initially included. However, the option of adding them to the table is available.

• Risk calculators

 This tool uses the ASCVD risk calculator to estimate the patient’s 10-year ASCVD risk and a 100-person pictograph to display this risk. Then we use best estimates of risk reduction against this risk estimation to propose a revised ASCVD risk given the interventions chosen, assuming independence. This is based on the approach used in the highly popular and effective Statin Choice tool.

To calculate the current risk of having a coronary event (described as “heart attack” in the tool) in the next 10 years, the tool uses the ASCVD risk calculator equation and data such as: age, sex at birth (M/F), African American (Y/N), smoker (Y/N), Diabetes (Y/N), treated blood pressure (Y/N), total cholesterol (100-350 mg/dL), HDL cholesterol (10-120 mg/dL), and systolic blood pressure (90-250 mmHg) that has to be completed by the clinician or auto-populated from the electronic health record.  For further detail on how to calculate the current risk using the ASCVD risk calculator, please refer to pages 32-34 on the 2013 Report on the Assessment of Cardiovascular Risk: Full Work Group Report Supplement. [4]

For each intervention, we used their RR estimates to calculate the future risk of having a coronary event in the next 10 years if the patient started, or stopped using, that intervention. The future risk is calculated by multiplying the current risk by the RR overall. The RR overall is calculated using the RR estimates shown in the following table, which are different depending on the use status of the intervention (i.e. if a medicine and/or an activity is started or stopped). A patient’s RR overall is calculated by multiplying together the RRs of each intervention that the patient chooses to start using, and then further multiplying by the RRs of each intervention that the patient is currently using that they and their clinician elect to stop using. The RR of an intervention that is stopped is the inverse of the RR of an intervention that is started

RR overall = RR interventions started * RR interventions stopped

This means for example, that if a patient switches from medium to high dose statins, then:
 RR overall = RR_{statins high} * 1/RR_{statins medium}

Any interventions that the patient is currently using and will continue to use are not included in the calculation of RR overall and therefore do not contribute to the estimate of future risk. This reflects the difficulty of determining any further risk reduction that may be achieved by continuing current interventions beyond that which is reflected in the patient’s current risk estimate. (For example, for a patient currently taking a blood pressure lowering medicine, the benefits of the intervention are reflected in the estimate of the patient’s current risk and it is unknown to what extent continued use of the medication will lower that risk further.)

Future risk= current risk * RR overall.

Activity or medicine option	RR activity/med started	RR activity/med stopped
Not smoking	0.61[5]	1.64 (1/0.61)
Heart-Healthy Diet	0.7[6]	1.42 (1/0.7)
Exercise	0.75[7]	1.33 (1/0.75)
Statins medium	0.75[8]	1.33 (1/0.75)
Statins high	0.6[8]	1.66 (1/0.6)
Ezetimibe	0.94[9]	1.06 (1/0.94)
Aspirin	0.91[10]	1.09 (1/0.91)
Blood Pressure Lowering Medications	0.88[11]	1.13 (1/0.88)
PCSK-9 inhibitors	0.86[12]	1.16 (1/0.86)
GLP-1 Agonists	0.88[13]	1.13 (1/0.88)
SGLT-2 Inhibitors	0.86[14]	1.16 (1/0.86)

As described above, in order to overcome evidence limitations, a number of methodological compromises were made in calculating future risk. This reflects principles that are used in developing all our shared decision-making tools:

• Patients and clinicians need support in making decisions even when optimal evidence does not exist.
• Risk is only a device that in some circumstances may be helpful in decision making.
• It is more important that any risk presented is a useful approximation that can help people make reasonable decisions than that it is precise, particularly when imprecision is unlikely to affect the final decision.
• The most appropriate method for calculating risk should be based on the quality of the reasonably applicable evidence and its ability to contribute usefully and feasibly to patient and clinician decision making.

• Reporting of other outcomes

For each medication stated above, we created a table with key information, the most common adverse effects, and other benefits of the medication. We selected the most discussed in practice and the most relevant to the clinical context of primary prevention. Also, we gathered information on average cost of these medications per month according to the GoodRx service, recognizing that these estimates vary greatly depending on the patient’s insurance.

	Other benefits	Side effects
Statins medium dose	Prevents strokes by 29%[15]	Muscle aches (0-5 in 100)[8]
Statins high dose	Prevents strokes up to 48%[16]	Muscle aches (0-10 in 100)[8]
Ezetimibe	Prevents strokes up to 14%  alone or in combination with statins.[17]	Muscle/joint aches, flu-like symptoms.[18]
Aspirin	Prevents colorectal cancer by 20%.[19]	Easy bruising, bleeding (3 in 1000)[10]
Blood pressure medications	Prevents strokes up to 40%; other benefits depend on medicine used.[11]	Depends on medicine used
PCSK9-inhibitors	Prevents strokes up to 20%. [20]	Flu-like symptoms.[12]
GLP1-agonists	Prevents death by 11%, loss of >5% of body weight, prevents kidney failure by 20%. [21, 22]	Nausea-vomiting (2-3 in 10), diarrhea (1 in 10).[23]
SGLT2-inhibitors	Prevents death by 20%, loss of 2% of body weight, prevents kidney failure by 30%.[14]	Urinary and genital infections (200 in 1000 over 5 years), DKA (4 in 1000 over 5 years).[24, 25]

• Design process: Prototyping and refinement of the tool

Designers at the KER Unit start the process of designing the tool by reviewing how relevant conversations take place in medical encounters in usual practice. After obtaining patient and clinician informed consent, we video recorded 5 preventive cardiology visits which were then reviewed by members of the team. These observations were fundamental to creating the first version of the Heart Health tool.

We invited 4 clinicians working in preventive cardiology, consultative internal medicine, and primary care at Mayo Clinic to test this tool prototype with their patients. We tested two versions of the tool with a total of 8 patients. All encounters were video recorded and reviewed. Each version of the tool was improved considering user experience, observed misuses, and clinician recommendations for adding, removing, or modifying the tool’s content.

One example of the kind of refinements done to the tool (that came from the observation of its use in clinical practice and from expert feedback) is the change to the order of display of the intervention screens. Initially, our tool showed the medications tab first and the activities tab second. The order of these seemed to imply that the best approach to discussing CV risk prevention was by focusing on medications first.

However, when observing the videos, we saw that the conversation always started with changes in lifestyle given their large impact on CV health. We also recognized that the issues presented in the tool that pertained to medicines weren’t particularly informative when considering lifestyle changes (e.g. side effects). We wanted our tool to support as much as possible the usual conversation that patients have with their clinicians. For this reason, we made the very simple but relevant decision to order the tool so that activities were available for discussion before medicines.

The third version of the prototype tool was considered the final one and was sent to a software development company that later released the tool to be implemented within the electronic workflow and online as a standalone webapp. The team at KER Unit was in constant communication with the software development team, and together we worked on improving the tool’s visual display while making sure the purpose and logic of each screen was maintained.

For more examples of SDM tools designed by researchers at the KER Unit, please refer to http://www.carethatfits.org/tools

References