Improving Adherence Through Data Analysis: An Interview With Kristin Alvarez, PharmD

Kristin Alvarez, PharmD, BCPS, is the director of clinical innovation at Parkland Health & Hospital System in Dallas, Texas. In her current role, she leads the Center for Innovation and Value at Parkland and works with a team of colleagues from diverse disciplines focusing on projects ranging from targeted philanthropic causes to broad organizational strategies. She works closely with other departments within Parkland and University of Texas Southwestern Medical Center to provide academic mentorship and preceptorship for health care professionals pursuing research or quality improvement endeavors.

The Center for Innovation and Value at Parkland is dedicated to finding ways to develop and implement new approaches to providing the highest-quality health care to its patient population in Dallas County. Its mission is to increase value by achieving exemplary outcomes that matter to patients while lowering the per capita cost of care. The center’s guiding principles include empowering patients, developing innovative approaches to patient care, democratizing technology, and advancing diagnostic and therapeutic stewardship.

An editor from The American Journal of Managed Care^® recently conducted a question-and-answer session with Alvarez to discuss a quality improvement initiative designed to help improve medication adherence.

__________________________________________________________________________________________________________

The American Journal of Managed Care^®(AJMC^®): What are some of the barriers patients face regarding medication adherence?

Kristin Alvarez, PharmD, BCPS: Patients face barriers related to many factors, including regimen complexity, transportation, access to care for refills, funding, navigating insurance hoops (prior authorizations, nonformulary authorizations, quantity limits), competing life priorities, adverse events, and mistrust.

AJMC^®: Could you describe your practice setting and the patient population you serve? Are there any barriers to medication adherence that are unique to your setting and/or patient population?

Alvarez: Parkland Health & Hospital System provides care to indigent, uninsured, and underinsured residents of Dallas County, averaging more than 60,000 hospital discharges and 1 million outpatient visits annually. Parkland’s patient payer mix is approximately 30% charity, 30% Medicaid, 20% Medicare, and 10% self-pay. The unique barriers to care facing the populations we serve are related to social determinants of health, which are often overlooked when providing solutions related to medication adherence. For instance, while mail order may be a solution for those with transportation issues or for convenience, it may not work for someone whose housing situation is unstable. Because many Parkland patients are self-pay and charity pay, we have dedicated staff to help them navigate patient assistance programs offered through different pharmaceutical companies or charitable organizations and grant-funded sources. In addition, $4 self-pay prescription drug programs often used by our patients present a unique issue when using claims data to calculate adherence scores.

AJMC^®: How did your team identify a quality gap related to adherence?

Alvarez: Adherence to prescribed therapies is a global barrier to disease state control. The World Health Organization published Adherence to Long-term Therapies: Evidence for Action¹ almost 2 decades ago to bring more attention to this after numerous scientific publications described the issue in the decades prior. Our own anecdotal experience as well as the results of smaller research projects at Parkland confirmed that our patients were not immune to the impact of low medication adherence. Outcomes-based initiatives at Parkland uncovered many opportunities for improving patient care, including medication adherence. We began gathering aggregate data at the population level for many disease states, including patients with diabetes. This started with broad outcomes, such as percentage of patients missing [glycated hemoglobin] A1C results and percentage of patients with A1C values greater than 9%. We parsed this data by clinic and obtained a list of patients who could benefit from intervention. It was clear that prescribing the right therapies was only part of the solution to help patients control their diabetes; the solution would be multifactorial. We also recognized that medication adherence solutions were not only patient-centric—it would also require changes within the system as well.

AJMC^®: Could you describe the quality improvement initiative your team implemented to address medication adherence? What changes did you make, and in what timeframe?

Alvarez: The primary goal of the initiative was to integrate medication adherence information into the health care team workflows. Although this may seem like a simple endeavor, the implementation process was quite complex. The following phases for implementation took 18 months:

• Identify data source for medication fills, including cash pay, and execute data agreement and upload
• Determine which adherence measure would be used: modified proportion of days covered (PDC)
• Execute curation of medications:
  • Exclude medication classes unsuitable for PDC measurement
  • Create custom groupings of medication classes to capture continuous adherence even with medication switches (eg, β-blockers and α/β-blockers, ACE [angiotensin-converting enzyme] inhibitors and ARBs [angiotensin II receptor blockers])
• Validate PDC measurement calculations
• Integrate the adherence scores into the clinical workflow via a diabetes snapshot
• Create a dashboard to describe adherence at the population level
  
  

AJMC^®: Who was involved in developing and implementing the improvement initiative? How was the initiative received by staff and patients?

Alvarez: We assembled a team of physicians, pharmacists, nurses, data analysts, and key stakeholders to create educational materials introducing the new adherence scores along with curated responses to frequently asked questions. During the 10th Annual Diabetes Management Conference, continuing medical education was dedicated to educating 300 local health care professionals on adherence and the implementation of adherence scoring into workflows 10 months after scores were displaying in the EMR [electronic medical record]. Parkland’s executive director of the Global Diabetes Program and the diabetes clinical pharmacy specialist were primarily responsible for in-person education of attending physicians, fellows, and residents as they rotated through clinic or inpatient settings.

From the beginning, staff were excited to have adherence information available in the EMR; however, trust in the score was not fully gained until data on cash pay fills from major retail pharmacy chains with a $4 self-pay prescription drug list were included, which occurred 22 months after the initial rollout. For example, adding cash pay fills data from one major retail chain with the $4 drug list increased adherence scores for metformin among self-pay patients by 12% with only 6 months of data.

AJMC^®: What were the initial results of this quality improvement initiative? Have these results been sustained over time?

Alvarez: Our initiative started 4 years ago, and to date we include 36,343 medications divided into 209 groupings and have removed 6211 medications from adherence scoring. Scores are calculated on 38,000 patients annually. Improvements in monitoring our population adherence metrics have grown to monitor trends related to demographics, such as financial class, diagnosis, sex, ethnicity/race, preferred language, and marital status. Our EMR provider has added an adherence score calculated for a single medication; however, this metric does not accurately provide an overall picture of adherence when there are medication changes within the same class of medication. Therefore, our custom adherence score that has been in place for 4 years is still the primary score used by clinicians.

Medication adherence scoring is now incorporated into other initiatives at Parkland. In 2019, Parkland and Dallas County Health and Human Services completed a Community Health Needs Assessment, which revealed a high rate of diabetes morbidity in certain areas of Dallas. Patients with low adherence scores and poor disease state control are identified by incorporating adherence scores measured from this quality improvement initiative. At baseline, 21% of patients living in vulnerable areas of Dallas had an A1C higher than 9% and medication adherence scores less than 60%. This percentage has favorably decreased from 21% at baseline to 19% of patients after 1 year. Having medication adherence scores with disease outcomes data available at the population level helps care teams identify those in most need of intervention.

Other examples of projects where medication adherence scoring from this quality initiative have been used at Parkland:

• Antiplatelet adherence scores were used to determine which patients need long-term outreach follow-up by cardiology catheterization lab nursing staff after undergoing percutaneous coronary intervention.
• During admission, medication history is collected, and pharmacy teams use adherence scores to identify patients who are having difficulty filling prescriptions or who need refills. This information is also used to help prevent inaccuracies in medication lists upon admission.
• During postdischarge pharmacy follow-ups, pharmacists use adherence scoring to trigger barrier assessments during visits.

There are still some outstanding challenges related to how adherence scores are consumed by the end user. We will need to expand measurements of adherence to not only include persistence scoring, but also measures of therapy initiation and medication therapy gaps. These measures are more appropriate for certain clinical scenarios, and a convenient way to track them in our current EMR display is lacking.

AJMC^®: How might these results be useful to managed care organizations and others interested in improving health at the population level?

Alvarez: Health care professionals cannot act on what they cannot see. Oftentimes, prescription information is stored separately from the medical record, and if the information is not front and center, the team is missing vital information to make clinical decisions. If medication doses are titrated up based on the assumption of adherence, patients are placed at risk for adverse drug events such as hypotension or hypoglycemia.

When using adherence scoring as a payment or quality measure, it is important to understand that if only claims data are used, it may not capture the full picture. On occasion, the cash price is lower than an insurance co-payment, samples may be used, or patients are hospitalized frequently or for prolonged periods of time.

Data storage requirements should also be considered. We currently store around 20 GB worth of information as the result of our weekly analysis of this data. During that weekly analysis, around 10 GB of that data is reprocessed to ensure our patients’ adherence scores are kept up-to-date with any new prescription events from the previous week.

AJMC^®: Is there any other information you would like to share that may be of use to our audience?

Alvarez: There are some things to consider when displaying adherence scoring as part of a workflow:

• Be clear with end users about what is included in the score.
  • Do you have information on all fills, including cash pay?
  • Do you capture medications filled from patient assistance programs?
  • Was the score calculated based on individual medications or a grouping of medications?
• If using pharmaceutical classes or subclasses for groupings, give examples of drugs that fall within that class.
• Keep a record of custom groupings; psychological and neurological medications have complex groupings.
• Provide detailed information, such as the start and end dates of the measurement period, for each calculation and the date the score was last updated.
• Emphasize that adherence scores are conversation starters, and nothing replaces having a meaningful conversation with patients.
• The biggest deterrent to not using the information provided occurs when inaccuracies cause confusion between the patient and provider in the limited time they have together.
• Regular maintenance is required, and users who identify issues with adherence scoring need to have a mechanism to report and resolve them.

AJMC^®: What other quality improvement initiatives is your team currently working on?

Alvarez: Currently, our team is working on a multiple-medication adherence score that will take into account adherence to all medications within a particular diabetes regimen. Adherence will be calculated for the proportion of days in which all medications within a regimen were filled.

In addition, patients are being categorized into 4 main groups. Each grouping will determine the clinical pathway patients will take. The Figure shows examples of pathways for each group.

Finally, further clinical research involving patients who are not adherent but at goal is warranted to assess if certain medications are effective even with low adherence.

Reference

1. World Health Organization. Adherence to Long-term Therapies: Evidence for Action. World Health Organization; 2003. Accessed March 10, 2021. https://www.who.int/chp/knowledge/publications/adherence_report/en/