Rx for Getting Patients to Follow Doctors’ Orders

Medication adherence improves patient outcomes, but U.S. patients generally have low rates of compliance. To help patients stay on top of their dosing regimens, physicians need to identify those most likely to be nonadherent and the reasons for their recalcitrance to implement interventions. But adherence itself bears costs, as do interventions. Payers want proof that strategies are both cost-effective and feasible.

Patients with chronic diseases such as hypertension, dyslipidemia or congestive heart failure (CHF) benefit from medical therapy but only if they actually take their drugs as prescribed. Conditions such as acute MI require that patients receive and adhere to medication plans ordered by their physicians at discharge. But the PARIS Registry, which is tracking more than 5,000 patients for two years after stent implantation, found that the incidence of nonadherence to dual-antiplatelet therapy was 2 percent at 30-day follow-up, putting nonadherent patients at risk of stent thrombosis. According to the World Health Organization, only half of patients with chronic diseases in developed countries take their medications as prescribed, to the detriment of their health and the healthcare systems that may absorb the costs of subsequent emergent care and hospitalizations.

Someone—whether a patient, his or her insurance company, Medicare, other government payers or providers—has to foot the bill for these pharmaceuticals. “Medication adherence is a measure of prescription drug utilization, so increasing adherence generally means you spend more on medication,” says M. Christopher Roebuck, MBA, a health economist and principal of RxEconomics in Hunt Valley, Md.

But do the savings from reduced hospitalizations offset the costs of adherence? Numerous studies have shown that adherence leads to lower overall healthcare costs, but the research is observational and consequently cannot establish a causal link.  

As the former director of strategic research at CVS Caremark, Roebuck collaborated with colleagues at the retail pharmacy as well as at the Centers for Medicare & Medicaid Services to answer that question using two powerful tools: the pharmacy benefits manager’s database of prescription drug claims from sponsors of health insurance plans and an econometrics fixed-effects method to tease out confounding variables that hamper observational studies (Health Affairs 2011;30[1]:91-99). The researchers focused on four chronic vascular conditions that are both costly and prevalent: CHF, hypertension, diabetes and dyslipidemia.

They extracted pharmacy data on people with continuous health insurance between 2005 and 2008 for a sample that included 16,353 patients with CHF, 112,757 with hypertension, 42,080 with diabetes and 53,041 with dyslipidemia. For costs, they looked at annual pharmacy, medical and total healthcare costs.

They found that adherence carried a price. Annual pharmacy spending for adherent patients with CHF was $1,058 more; for hypertension, $429 more; for diabetes, $656 more; and for dyslipidemia, $601 more. But annual medical spending was lower, with reduced average spending for CHF, hypertension, diabetes and dyslipidemia, totaling $8,881, $4,337, $4,413 and $1,860, respectively. The adherence effects were more pronounced for patients who were 65 years or older.
 
“Those savings certainly speak to the value of medication adherence,” Roebuck says. Of course, cost savings is a critical part of the equation for payers, who want reassurance that the money doled out for medications leads to lower healthcare use and consequently fewer and lower payouts. But that is not the same as improving adherence. “How you move the needle on adherence is another ball of wax.”

Nonadherent's portrait

Adherence rates can vacillate over time, according to Barbara J. Riegel, DNSc, RN, who studies nonadherence in HF patients. As director of the Biobehavioral Research Center at the University of Pennsylvania School of Nursing in Philadelphia, she and colleagues test disease management approaches for these patients.

“Even the patients who were serious about taking their medicine regularly weren’t 100 percent perfect,” says Riegel in reference to one nonadherence study involving 202 chronic HF patients. “They were good most of the time.”

Compounding the problem, many patients with cardiovascular disease struggle with comorbidities that make simultaneous control of multiple conditions difficult to achieve. In a study involving two different patient groups who received medications to control hypertension, diabetes and dyslipidemia for a median of four years, achieving simultaneous control of all three risk factors was rare and transient (Circ Cardiovasc Qual Outcomes, online July 31).

Of the 5,269 patients at Denver Health, an inner city healthcare system in Colorado, only 16.2 percent achieved simultaneous control of all three risk factors and only 13 percent maintained control over 90 days. Of the 23,458 patients at Kaiser Permanente Colorado, a managed care organization, 30.3 percent achieved simultaneous control but only 5 percent maintained control of all three risk factors.

Researchers who study adherence issues agree that in many cases the first step in improving adherence is identifying those patients who likely will be nonadherent. Janet Shin, PharmD, PhD, a research scientist in the pharmacy analytical services research group at Kaiser Permanente Southern California (KPSC) in Downey, Calif., and colleagues targeted patients who fail to pick up a newly issued prescription to determine the factors associated with these primary nonadherents (Am J Manag Care 2012;18[8]:426-434).

“We wanted to use our data in a real-world setting and a large sample size to provide evidence-based results, so clinicians can make better decisions and improve care,” Shin says. To do that, she and colleagues designed a retrospective cohort study that took advantage of KPSC’s resources: 3.3 million members at 14 medical centers and an EMR that captures all prescription entries, patient characteristics and prescribing physician characteristics.

They selected 10 therapeutic drug groups, including cardiovascular agents and antihyperlipidemics, for a total of 874 individual drugs prescribed between Dec. 1, 2009 and Feb. 28, 2010, to nearly 400,000 patients. Their primary outcome was failure to pick up a prescription at 14 days after the index date.

Cardiovascular agents made up 8.6 percent and antihyperlipidemics 3.9 percent of the nearly 570,000 new prescriptions written in the three-month period. The overall primary nonadherence rate was 9.8 percent, but it varied by drug group. Antihyperlipidemics had a nonadherence rate of 22.3 percent while cardiovascular agents showed a higher compliance rate of 7.8 percent.

Adherent patients tended to have a history of having filled a prescription previously and have a prescription for a symptomatic disease.
Primary nonadherents were more likely to be:

• Prescribed a greater number of prescriptions on the index date;
• Prescribed chronic medications;
• Have more comorbidities;
• Be black;
• Have prescriptions for brand name drugs and treatments for asymptomatic disease; and
• Be “treatment naïve,” that is, they had not been prescribed a drug in that therapeutic class previously.

“Most of these patient characteristics are not modifiable like age, gender or race,” Shin says. “The goal is to use these characteristics to help clinicians identify patients who are primary nonadherent and then modify the medication behavior itself.”

Find it and fix it

Riegel’s research on HF patients looks beyond “who” to also assess why they may be nonadherent. By ferreting out the reasons, they can pinpoint modifiable factors and then apply evidence-based interventions designed to overcome barriers to compliance.
The American Heart Association estimates that 5.7 million people in the U.S. have HF, which accounted for 1.1 million hospitalizations in 2004 at a cost of almost $29 billion. Medication adherence may reduce those costs, though. In one analysis of Medicaid beneficiaries, researchers calculated that costs per year for adherent HF patients was 23 percent less than for nonadherent beneficiaries (Am J Manag Care 2009;15:437-445).

Many studies of nonadherent HF patients rely on self-reports to measure adherence. Riegel and colleagues instead designed a study using an electronic monitoring system that incorporated a small device into the cap of a pill vial to objectively track real-time medication-taking in 202 Stage C patients with HF (Circ Heart Fail, online May 30). Prescribed medications included angiotensin-converting enzyme inhibitors (ACEIs), angiotensin II receptor blockers (ARBs) and beta-blockers.

The majority—57.9 percent—of the patients were on a twice-daily dose regimen while 39.1 percent had a once-daily regimen and 3 percent took medications three or more times per day. They found taking, dosing and timing adherence decreased from the first three-month period to the second three-month period.

They identified two distinct trajectories of nonadherence: a persistent adherence group and a steep decline subgroup. Further analyses showed that lapses in attention, excessive daytime sleepiness and regimens that required two or more medication dosings per day contributed to steep declines in adherence. While the data apply to HF, the three factors associated with medication nonadherence may occur in other chronically ill patient populations.

“We know that medication adherence is a problem in a lot of different patient groups,” Riegel says.

Based on their findings, she offers a number of strategies for physicians, hospitals and payers to consider for improving adherence. For instance, other research has shown that HF patients take on average eight medications per day, and those under a multiple-dosing regimen may take 18 pills a day. Generics, while less expensive, often require a multiple-dosing regimen that contributes to the decline in adherence.

“If we spend more money [on branded drugs] and get the once-a-day dosing, then these patients may stay out of the hospital,” she says. “That has a cost implication.”

Physicians should be on the lookout for attention lapses and excessive daytime sleepiness. “What we are recognizing in heart failure patients is that a large number of them—up to 50 percent—have some subtle cognitive impairment and that is probably associated with this lack of vigilance,” she says. Lack of attentiveness also may be a response to poor sleep quality, though.

In HF patients, daytime sleepiness may result from sleep-disordered breathing, which is common with this disease. One study examining 53 stable HF outpatients found that 63 percent suffered from sleep-disordered breathing (CHEST 2005;128[4]:2116-222). And the prevalence of sleep disorders such as moderate to severe obstructive sleep apnea has not changed in HF patients, despite the increased use of beta-blockers and spironolactone (J Card Fail 2009;15[4]:279-285).

Physicians gauge potential sleep-disordered breathing by patients’ complaints about daytime sleepiness. They then will refer the patient for testing.  However, HF patients tend to not complain about their daytime sleepiness, Riegel says, which means their physicians need to be extra vigilant.

“My message for the providers is that they need to have a very low threshold for referring patients for testing,” she advises.

Connecting the dots

Riegel’s study group expected cost to percolate into the top nonadherence factors, based on the literature. In one study, nonadherent HF patients pointed to cost as the reason they ignored prescriptions for statins (Mayo Clin Proc 2011;86[4]:273-281). The population-based study monitored adherence of beta-blockers, ACEIs, ARBs and statins between 2007 and 2009. Of the 209 patients studied, 10 percent reported in a questionnaire that they decided not to fill a prescription because of cost; 8 percent admitted that they stopped taking medications because of cost; and 4 percent said they skipped doses to reduce costs. Across all drug classes, a higher percentage of patients with poor adherence blamed cost for their noncompliance compared with patients with good adherence.

One possible difference between that study and Riegel’s? The population-based study involved a county with a small number of medical providers, which includes the Mayo Clinic in Rochester, Minn. Riegel et al enrolled many Veterans Affairs patients, who may receive medications for free.  

Many studies have examined the effect of cost on nonadherence in various disease states. Others added insights on the effect of adherence on outcomes. Niteesh K. Choudhry, MD, PhD, an associate physician at Brigham and Women’s Hospital in Boston, took research a step further by evaluating adherence, outcomes and cost in MI FREEE. The policy study enrolled insured patients who had been discharged with an MI, randomized them by their insurance plan sponsor and measured clinical and cost outcomes (N Engl J Med 2011;365:2088-2097).

“We designed a study to answer the ‘so what’ question upfront,” Choudhry says.

One study group consisted of 1,494 plans with 2,845 patients who had fully paid prescription coverage while a second group consisted of 1,486 plans with 3,010 patients who had standard prescription coverage. Drugs for both groups included statins, beta-blockers, ACEIs and ARBs. The primary outcome was a first major vascular event or revascularization. Secondary outcomes were rates of medication adherence, total major vascular events or revascularization, first major vascular event and pharmacy and medical expenditures. They used health service claims to assess outcomes for a median follow-up of 394 days.

There was no significant difference between the groups for the primary outcome, but the intervention reduced the rates of the secondary clinical outcomes. Adherence in the fully covered group was 4 to 6 percentage points higher and expenditures were $5,770 lower. Patients in the fully covered group paid 26 percent less in overall out-of-pocket costs compared with the control group, as well.

Choudhry says that eliminating the co-payment nudged adherence up only modestly, but they deliberately designed the study to be simple and scalable with the goal that payers could easily put the intervention into practice. “All we did is change co-payments for these patients,” he says. “Keeping it simple allowed for wider dissemination and generalizability, which was as important as demonstrating that the concept worked.”

In a later study involving Pitney Bowes, Choudhry et al found that a change in policy that led to lower copayments for statins and clopidogrel was cost neutral (J Am Coll Cardiol online Oct. 3).

Nonetheless, modest increases in adherence may lower spending. In a study that analyzed the relationship between medications commonly used to treat CHF and Medicare expenditures, researchers found that a 10 percent increase in the average daily pill count for prescribed ACEIs or ARBs, beta-blockers, diuretics or cardiac glycosides might trim Medicare spending by $508, $608, $250 and $1,244, respectively, over three years (Am J Manag Care 2012;18[9]:556-563).

“Even small savings over a large population can have a significant impact on Medicare expenditure,” says the study’s lead author, Ruth Lopert, MD, a visiting professor in the department of health policy at George Washington University in Washington, D.C. She and colleagues also found that 17 percent of the study population was not taking medications in any of the four classes of drugs. “What are the possibilities of much larger savings, both through better adherence among the poorer adherers and initiating and adhering to treatment among those who are not being treated at all?”

Making payers believers

Choudhry suggests that the MI FREEE type of intervention may be effective for a variety of chronic conditions, but only those that are common, bear high costs and are supported by strong evidence showing the efficacy of medications. Patients recently discharged with HF may present an appropriate opportunity, because they have a high risk of readmission or recurrent cardiovascular or cardiac-oriented events that is reduced if they adhere to their prescribed drugs.

“As a whole, this would work well for are higher risk chronic disease management, such as hypertension, diabetes, hyperlidipidemia, coronary artery disease, heart failure and maybe COPD [chronic obstructive pulmonary disease],” he says. “This is not the strategy for lower-risk primary prevention or for situations where the drug treatments are less well evaluated.”

Interventions don’t come free though. The MI FREEE model requires payers to invest resources beyond expenditures for covering medication costs. Some value-based insurance designs involve a starting date for eliminating co-payments, which is fairly easy to implement. With the MI FREEE design, payers must identify patients and change authorization codes soon after the MI event to allow them to get medications at no cost.

“There is a bit of administrative exercise here,” Choudhry admits. “Most large insurers have the capacity to do this. For smaller insurers, this may be more of a challenge.”

Aetna, which sponsored MI-FREEE, announced that it planned to make available in January 2013 a program that eliminates co-pays on generic cardiac drugs and reduces co-pays on branded cardiac drugs for members who experience an MI.

KPSC researchers are developing and testing interventions using technology to target primary nonadherence to cholesterol and osteoporosis medications.  These drug categories had the highest nonadherence rates found in the study.

“Technology can be helpful for primary nonadherence,” Shin says. “The next step is to look at how other technologies can be used.” In her ongoing research, she is analyzing clinical data available through the database to track if and to what degree primary nonadherence affects clinical outcomes.

Riegel also is applying the results from their study to evaluate an intervention to reduce the daytime sleepiness burden in HF patients. She is collaborating with colleagues at the University of Pennsylvania to test the effectiveness of cognitive behavioral therapy to improve sleep quality among HF patients. Physicians may not be able to remove all of the attentiveness problems in patients who also have cognitive deficits, she says. “But we can get people thinking better if we can get them sleeping better. They are less sleepy during the day and more alert.”

Given the complexity and challenges of nonadherence, no one intervention likely will have a dramatic impact on patients’ medication-taking behavior. Instead, Choudhry sees providers considering a mix-and-match approach that allows them to select several interventions that accumulatively push adherence rates to a level that significantly improves outcomes.

“We think cost is a simple and scalable intervention but there could be other things like reminder systems, education, outreach, easy access, simplifying treatment and dosing regimens,” Choudhry observes. “Those are where we need to go from here. If eliminating copayments improves adherence by 5 percent on average, then we could do the next thing and push it up another 5 percent. Maybe with a suite of interventions—four, five, six different interventions—we may get to a number that we think is relatively acceptable, closer to 80 percent.”

That is an ambitious goal, given the state of medication adherence today. Adherence in the usual-coverage group in MI-FREEE, for instance, ranged from a low of 36 percent for ACEIs and a high of 49 percent for statins. But Choudhry sees efforts such as his and others who try to lift these rates from the abyss as having an impact beyond their clinical and financial findings.

These initiatives raise awareness among policymakers and patients that nonadherence is a problem. And they give practicing cardiologists comfort in knowing that through these interventions some patients—even if it is a modest 5 percent increase—are following their doctors’ orders.

The Prescription Gap

For every 100 prescriptions written…

• 50–70 are received by a pharmacy

• 48–66 are picked up

• 25–30 are taken properly

• 15–20 are refilled as prescribed

Source: National Association of Chain Drug Stores, Pharmacies: Improving Health, Reducing Costs, July 2010. Based on IMS Health data

Measuring adherence

Researchers apply a variety of methods to determine adherence. The cutoff threshold varies study to study, but many define 80 percent and higher compliance as adherent.

Roebuck et al measured adherence as the proportion of time a patient had medication on hand, as seen in the claims data. They used a metric known as the medication possession ratio (MPR), which is based on the number of days a patient possessed a medication divided by the number of days in a year. They set a condition-level MPR below 0.80 as nonadherent.

Choudhry et al also used an MPR to evaluate medication adherence. They multiplied MPRs by 100 for absolute adherence percentages. They also calculated the proportion of patients who had complete adherence to each and to all three medication classes. They defined full adherence as 80 percent or greater.

Shin et al defined primary nonadherence (PNA) as failing to fill a prescription within 14 days of it being written. They calculated PNA rates overall and by drug class. Riegel et al used a tracking device within a pill vial to track medication taking. Their study measured medication taking, dosing and timing adherence. 

Lopert et al gauged adherence based on average daily pill counts over a three-year period.

Some interventions work, but how well?

The Agency for Healthcare Research and Quality reviewed scientific literature to identify studies on the efficacy and effectiveness of interventions to improve medication adherence across an array of chronic conditions. Reviewers also graded the evidence linking adherence to improved outcomes. Based on an analysis, researchers determined that a variety of interventions improved adherence for hyperlipidemia, hypertension, heart failure and MI. But while they found many pathways to improve adherence, they reported little strong evidence that improved adherence led to better outcomes, such as improved systolic or diastolic blood pressure.