Medication Adherence Study Looks at Types of Interventions

Programs use varied techniques, but none have fully shut off the spigot of lost dollars. Greater use of predictive analytics may help.

Patients who don’t take prescribed medications spell trouble not only for their future health, but also for a health plan’s bottom line. A recent research project sought to discover what insurers are doing to promote adherence with prescribed medications, what success they’re achieving, and what other strategies might increase that success.

While nonadherence is a serious problem, its origins are complex and not fully understood. Some of its more commonly accepted causes include complex treatment regimens, side effects, forgetfulness, socioeconomic issues, educational levels, and personal beliefs.

Insurers and other entities that play a pivotal role in patient medication adherence have emphasized novel medication adherence programs and interventions. This is due in large part to poor medication adherence levels in a wide array of patient populations. It is widely understood that many patients do not take their medications properly — including as many as 50% of patients with chronic illnesses.

30 health plans surveyed

This research, conducted in spring 2014, involved 30 health plans. Each was asked about its current medication adherence intervention programs. Interviews involved a questionnaire consisting of open-ended and multiple-choice questions. The questionnaire was divided into three basic parts:

  • Existing medication adherence programs
  • Types of interventions used and their individual effectiveness
  • How patients are selected to receive interventions

Respondents also discussed any unique services their programs provide and any future plans of expansion. All data were captured by five interviewers and then analyzed by the authors.

All 30 health plans interviewed have medication adherence intervention programs that target their member populations. The insurers vary in size and coverage demographics. Nineteen are classified as small plans (enrollment <200,000), while 11 are large (>200,000). The surveyed health plan populations encompass most of the continental United States as well as Puerto Rico.

The coverage demographics of these plans include commercial, Medicare, and/or Medicaid populations. Overall, 53% of health plans cover a commercial population, 80% cover a Medicare population, and 27% cover a Medicaid population. The most prevalent coverage combinations are commercial and Medicare, and Medicare and Medicaid, which account for 23% and 17% of the population, respectively. The authors interviewed employees in charge of their respective medication adherence intervention programs.

Data were gathered about which patients will get interventions and when patients are subject to the interventions (Tables 1 and 2).

TABLE 1 Which patients will get interventions?
Overall Small (enrollment <200,000) Large (enrollment >200,000)
To which of the following diseases do your current intervention programs apply? n=30 n=19 n=11
Diabetes 93% 95% 91%
Hypertension 93% 95% 91%
Hyperlipidemia 90% 89% 91%
Heart failure 50% 47% 55%
Myocardial infarction 17% 21% 9%
Asthma 30% 26% 36%
COPD 30% 37% 18%
Depression 20% 21% 18%
Antiretrovirals 17% 16% 18%
Osteoporosis 10% 5% 18%
Cancer 7% 11% 0%
Are there any disease areas you are considering expanding to? n=17 n=11 n=6
Expansion of disease states (Y/N) 65% 73% 50%
TABLE 2 When do patients get interventions?
Overall Small (enrollment <200,000) Large (enrollment >200,000)
How are patients selected to receive interventions through your intervention programs? n=30 n=19 n=11
Triggers 83% 89% 73%
First fill 30% 37% 18%
New diagnosis 33% 37% 27%
Late to fill 80% 84% 73%
Script coming due 33% 47% 9%
Predictive analytics (prioritizing patients at-risk for nonadherence) 7% 0% 18%
Retrospective adherence measures 40% 47% 27%

Under the heading “Which patients will get interventions?” 93% of respondents targeted patients with diabetes and hypertension, 90% those with hyperlipidemia, and 50% patients with heart failure. Sixty-five percent of the respondents plan on expanding to other disease states including asthma, COPD, and depression (73% small health plans and 50% large ones).

When patients get interventions, the three most common methods are triggers (83%), retrospective adherence measures (40%), and predictive analytics (7%). For trigger interventions, most health plans use pharmacy claims data to determine when to intervene.

The four most popular claims events used are “first fill” (30%), “new diagnosis” (33%), “late to fill” (80%), and “script coming due” (33%). Large MCOs rarely use “first fill” and “script coming due events” (18% and 9% respectively).

Historical levels

Retrospective adherence measures target interventions to patients who have certain historical levels of medication adherence, such as patients who are below a certain threshold of proportion of days covered in the previous year. Health plans were marked as targeting patients using predictive analytics if they used statistical or big-data techniques to forecast individual patients’ future medication adherence and employed those predictions as a factor in determining when to intervene with a member. The four most common intervention channels are telephone outreach (90%), direct mail (77%), provider-centric (73%), and face-to-face visits (50%). Telephone outreach includes automated and live calls and text messages, which are all directed toward patients; 91% of large insurers and 89% of small insurers surveyed use telephone outreach interventions. Direct-mail intervention comprises letters, brochures, and newsletters, with messaging focused on medication management, general education, ways of reducing costs, how to prevent side effects, and reminders to take medications. Eighty-two percent of large health plans and 74% of small health plans surveyed use direct mail-interventions.

In provider-centric interventions, MCOs contacted prescribers and gave them information about patients’ medication adherence levels, refill histories, MPR, or other data. From this information, physicians or their staff members can deliver appropriate interventions to the patients. This category also includes health plans that contacted prescribers to meet specific needs related to medication adherence, such as requesting new scripts when prescriptions have expired or asking for 90-day medication supply scripts. Eighty-two percent of large health plans and 68% of small health plans surveyed currently use provider-centric interventions.

TABLE 3 How do patients get interventions?
Overall Small (enrollment < 200,000) Large (enrollment >200,000)
Which interventions do you utilize for nonadherent patients? n=30 n=19 n=11
Telephone outreach (automated and live calls, text messaging, apps) 90% 89% 91%
Direct mail (e-mails, letters, brochures, newsletters, mail-order interventions) 77% 74% 82%
Provider-centric (physicians and nonphysicians such as nurse practitioners) 73% 68% 82%
Face-to-face visits (in-person pharmacy visits, home visits, support groups) 50% 53% 45%

Face-to-face interventions represent in-person pharmacy visits, home visits, support groups, and information sessions delivering interventions with messaging focused on motivational interviewing, refilling and taking medications, and medication management. Forty-five percent of large health plans and 53% of small health plans surveyed use face-to-face visits.

For where the interventions are conducted, 7% of MCOs outsource all interventions to third parties, 33% do them in-house, and 60% conduct interventions through some combination of in-house and outsourced methods. Many insurers in the study perform telephone outreach interventions in-house and outsource direct mail and provider-centric interventions to contracted PBMs and/or medication adherence vendors.

Health plans measure the effectiveness of their intervention programs through four common parameters: claims-based (59%), quality measures (35%), biological/clinical (24%), and hospitalizations/ER rates (12%). Sixty-seven percent of insurers rate their programs as moderately effective (83% of large health plans). The metrics for each parameter can be seen in Table 4.

Table 4 Where are the interventions designed and are they effective?
Overall Small (enrollment < 200,000) Large (enrollment > 200,000)
Are your interventions designed and conducted in-house or outsourced? n=30 n=19 n=11
In-house 33% 32% 36%
Outsourced 7% 5% 9%
Combination 60% 63% 55%
How do you measure the effectiveness of your medication adherence intervention program? n=17 n=11 n=6
Quality measure (HEDIS and Medicare star ratings) 35% 18% 67%
Hospitalization (ER visits and hospital readmission rates) 12% 9% 17%
Biological/clinical (HbA1c measurement and cholesterol levels) 24% 18% 33%
Claims-based: MPR (medication possession ratio) and PDC (percentage of days covered) 59% 45% 83%
How would you rate the effectiveness of your program? n=15 n=9 n=6
Very effective 33% 44% 17%
Moderately effective 67% 56% 83%

Most MCO medication adherence programs intervene with patients of chronic cardiovascular disease states (diabetes, hypertension, hyperlipidemia, and heart failure). Cardiovascular diseases are a primary concern for MCOs because of patients’ chronic medications/therapy and the strong association of nonadherence with increased hospitalizations.

Most health plans use triggers and retrospective adherence measures to select patients for interventions. Insurers follow a rule-based approach (using specific demographic profiles and predefined events to trigger interventions) rather than treating each patient differently. One way to accomplish a more individualized approach is through predictive analytics. While only 7% of health plans currently use predictive analytics, more than half of the surveyed insurers plan on incorporating predictive analytics into their intervention targeting.

Many are interested in adopting a platform that identifies interventions with a higher probability to engage and influence patient behavior and avoids wasteful spending on less-effective interventions. These enhanced programs are dynamic and can rapidly adapt to new intervention techniques.

Health plans view their current medication adherence programs as only moderately effective because they fail to intervene before patients are nonadherent and personalize interventions. This sense of the ineffectiveness of current interventions that target chronic health problems is in line with the findings of a previous study by Haynes et al.

While most health plans currently have medication adherence programs, many feel these programs aren’t doing the job as well as they might be. This is the driving factor behind growing interest in expanding into platforms that allow for a more personalized approach to address the medication adherence issue troubling our health care system.

Needed: more predictive analytics?

The study results suggest that most MCO medication adherence programs target members with chronic, comorbid cardiovascular diseases through a system of triggers and retrospective adherence measures. Most MCOs intervene using telephone outreach, direct mail, provider-centric, and face-to-face visits through a combination of in-house and outsourced methods. This approach is seen as only moderately effective as it fails to personalize interventions and intervene before a patient becomes nonadherent. As such, predictive analytics platforms can play an increasing role in addressing the needs and shortcomings of existing MCO medication adherence programs.

The study’s limitations

  • Responses were recorded only from managed care organizations that agreed to participate — a potential source of bias
  • Some questions were added midway through the interview process

Clifford Jones is the CEO of AllazoHealth, a predictive analytics company founded in 2011 to address the problem of medication nonadherence for health insurers, PBMs, ACOs, and providers. Its product, AllazoEngine, predicts which patients will not adhere to each of their medications and what interventions will best influence each one to take the medication. Earlier, Jones developed CVS Caremark’s “Pharmacy Advisor” medication adherence program. Other contributors to this article are Ian Sullivan; Kenny Ng, PharmD candidate; Rebecca Elwork; and Fawad Piracha, PharmD candidate (all of AllazoHealth, New York) and Michael Boice; David Coutts, PhD; Stefanie Mazlish, MBA; Aishwarya Nagarian, PharmD, MBA; Slanix Paul Alex, PharmD; Sudarshan Phani, PhD; Kalee Shah; Aneesh Sheth, PhD; Kamilia Sip, PhD; Tracey Van Kempen; and Upal Basu Roy, PhD, MPH (all of the Solutions Lab, a not-for-profit consulting organization in New York).


  • Gurwitz JH, Field TS, Harrold LR, et al. Incidence and preventability of adverse drug events among older persons in the ambulatory setting. JAMA. 2003;289:1107–1116.
  • Haynes RB, Ackloo E, Sohata N, et al. Interventions for enhancing medication adherence. Cochrane Database Syst Rev. 2008 Apr 16;(2):CD000011. doi: 10.1002/14651858.CD000011.pub3.
  • Haynes, RB. National Council on Patient Information and Education. Talk About Prescriptions Month. Oct. 1989.
  • Ho PM, Bryson CL, Rumsfeld JS. Medication adherence: its importance in cardiovascular outcomes. Circulation. 2009;119:3028–3035.
  • Miura T, Kojima R, Mizutani M, et al. Effect of digoxin noncompliance on hospitalization and mortality in patients with heart failure in long-term therapy: a prospective cohort study. Eur J Clin Pharmacol. 2001;57: 77–83.
  • New England Healthcare Institute. Thinking Outside the Pillbox: a System-wide Approach to Improving Patient Medication Adherence for Chronic Disease Aug. 2009. Accessed Aug. 26, 2014.
  • Roebuck MC, Liberman, JN, Gemmill-Toyama M, Brennan TA. Medication adherence leads to lower health care use and costs despite increased drug spending. Health Aff. 2011;30:191–199.
  • Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of medication adherence on hospitalization risk and healthcare cost. Med Care. 2005;43:521–530.