METHODS OF MEASURING MEDICATION USE/COMPLIANCE IN THE COMMUNITY DWELLING POPULATION OF OLDER PEOPLE
Patient Self Report
In many studies, information obtained using the patient self-report method has correlated positively to information obtained from random pill counts, biochemical measures, and electronic medication monitors. In several studies reviewed by APhA (2003), querying patients about their compliance resulted in the detection of over 50 percent of patients with poor compliance, with a specificity of 87 percent. Inaccuracies in the patient self-report method arise when patients overreport actual compliance, either because they don’t want to displease the provider or because they are unaware of their actual compliance. Vik et al. (2004) report that the consensus of the literature is that subjects who report nonadherence are in fact nonadherent, but other methods are required to detect those who report that they are adherent, but in fact are not. Vik et al. note that scaled questionnaires have been employed to assess self-reported adherence with success. For example, Morisky, Green, and Levine’s (1986) four-item scale for assessing antihypertensive treatment demonstrated acceptable psychometric properties, and scaled scores correlated with blood pressure control.
Kaplan et al. (2004) assessed patient compliance with lipid lowering medications using the question, “In the last month, how often did you take your cholesterol medication in the way your doctor prescribed,” with a five-point scale ranging from “None of the time,” to “All of the time.” Subjects who reported that they took their medication in the manner prescribed “some of the time,” “a little of the time,” or “none of the time” over the last month were defined as noncompliant. Overall, 12 percent of the subjects were defined as noncompliant, and there was a significant association between achieved cholesterol level (serum cholesterol level was measured) and self-reported noncompliance.
De Klerk et al. (2003) reported on the use of the Compliance-Questionnaire-Rheumatology (CQR) in a validation study comparing the CQR to electronic medication event monitoring, which they refer to as “the gold standard” in compliance measurement. The CQR is a 19-item instrument that measures patient compliance to drug regimens, identifies factors that contribute to suboptimal patient compliance, and may be used to measure future compliance in patients with rheumatoid arthritis (RA), polymyalgia rheumatics (PMR), and gout.39 In a prior study, the CQR had good test-retest reliability and moderate internal consistency, and validation using discriminant analyses against an overall patient self-report compliance measure showed a sensitivity of 98 percent, a specificity of 67 percent, and an estimated kappa of 0.78 to detect low compliance (de Klerk et al. (1999). Treharne, Lyons, and Kitas (2004) used the CQR scale, and found good internal consistency (Chronbach’s Alpha = 0.81). In their study to investigate the effects of specific psychosocial factors on adherence to medication in RA patients, CQR adherence score was used as the main dependent variable.
In the de Klerk et al. (2003) study, 85 patients visiting the outpatient rheumatology wards at three hospitals in The Netherlands completed the CQR and were given their medications in bottles fitted with Medication Event Monitoring System (MEMS) caps (Aardex Ltd., Zug, Switzerland). Multiple linear regression analyses showed that the total, weighted CQR score significantly and adequately predicted medication-taking compliance (r 2=0.46, p=0.001) and correct dosing (r 2=0.42, p=0.004). Discriminant analyses showed that specificity and sensitivity to detect good taking compliance were 95 percent and 62 percent, respectively. The predictive value to detect unsatisfactory taking compliance was 86 percent and to detect good taking compliance was 83 percent. Four items were especially predictive of taking compliance, and explained 35 percent of the variance: fear of forgetting to take the drug (“I definitely don’t dare to miss my antirheumatic medications”); being able to function well (“The most important reason to take my antirheumatic medicines is that I can still do what I want to do”); routines in daily life (“My medicines are always stored in the same place, and that’s why I don’t forget them”); and side effects (“If you can’t stand the medicines, you might say ‘throw it away, no matter what’”). The predictive values for these items are somewhat lower than the predictive value of the full CQR-19, with a sensitivity of 51 percent and a specificity of 87.5 percent.
The high predictive values of the CQR-19 make it an attractive tool as a screening instrument in future studies of polypharmacy and driving, to identify subjects who are compliant with their medication regime.
Another self-report method—the “brown bag method” of medication review—entails participants collecting containers of their current prescription and over-the-counter medications in a brown paper bag (usually provided by the organization or agency conducting the review) and bringing the bag with the medications to their physicians or a pharmacists during scheduled medication review appointments. As noted by Caskie and Willis (2004), the usefulness of self-reports of medication use depends on the willingness and the ability of the individual to volunteer such information. They further state that the congruence of pharmacy records and self-reported medication use is of importance because self-reports of medications are often used as a surrogate for health status or the presence of chronic diseases, and are important when studying medication compliance, polypharmacy, and drug interactions. Self-reports of medication are frequently used in population-based studies where pharmacy records are lacking or are expensive to obtain.
In England, brown bag reviews are a component of the medicine review process, and supplement information provided by the patient’s medical record and pharmacy reports. The Task Force on Medicines Partnership and the National Collaborative Medicines Management Services Programme (2002) indicates that a review of scripts in the absence of the medical record or a review of the record in the absence of the patient are screens for significant prescribing error. A medication review that does not take account of what the patient actually takes—rather than what is on the prescription or in the record—is incomplete. Patient information leaflets describing medication reviews include a request for the patients to bring all their medications to the review clinic, including herbal remedies, medicines that are bought from a pharmacy, and medicines that are bought over the counter. Medicines are described as anything the patient takes including tablets, liquids, inhalers, creams and ointments, and any medicines that the patient has but no longer takes (Pharmacy Practice and Medicines Management Group, 2002).
Furthermore, in England the National Service Framework (NSF) for Older People (Department of Health, 2001) recommends that an in-depth evaluation of all a patient’s medications (prescribed and nonprescribed) be targeted at those older people known to be at higher risk of medicine-related problems, including: those prescribed four or more medicines (polypharmacy); those discharged from a hospital, those residing in in-care homes; those for whom medicine-related problems have been identified through routine monitoring; patients 75 and older, as a part of their annual health check; and for those who have had an adverse change in health (dizziness, confusion). An NHS milestone by the year 2002 was for all people over age 75 to have their medications reviewed at least annually, and for those taking four or more medications to have a 6-month review. A 2004 milestone set by the NSF is for every Primary Care Group (PCG) or Primary Care Trust (PCT) to have schemes in place so that older people get more help from pharmacists in using their medications.
The study by Caskie and Willis (2004) included 294 members of a State pharmaceutical assistance program (PACE – Pharmaceutical Assistance Contract for the Elderly) who were also participants in a clinical trial on cognitive training at the Pennsylvania State University (ACTIVE – Advanced Cognitive Training for Independent and Vital Elderly). Subjects ranged in age from 65 to 91, with a mean age of 74.5. In this study, self-reported medication data were collected by use of the brown bag method; however, over-the-counter medications were excluded from analysis. Computerized pharmacy claims for prescription fills and refills were obtained from PACE for a time period that included the date of the brown bag data collection. Medications in the PACE pharmacy data whose supply was estimated to be depleted more than 5 days before the brown bag assessment were also excluded. Data were examined for 10 major therapeutic drug classes: antihistamines; anti-infective agents; autonomic drugs; blood formation and coagulation; cardiovascular drugs; CNS agents; electrolytic, caloric, and water balance; ear, eyes, nose, and throat preparations; gastrointestinal drugs; and hormones and synthetic substitutes. Specific drug classes within the class of cardiovascular drugs and CNS agents were examined in greater detail because of the frequency with which they are prescribed to older people and because of their possible side effects. The following specific cardiovascular drug classes were examined: ACE inhibitors, cardiac glycosides, beta blockers, calcium channel blockers, antilipemic drugs, hypotensive agents, and vasodilating agents. Specific CNS drug classes examined included: NSAIDs, opiate agonists, and benzodiazepines. For both data sets, a code was entered to indicate whether the participant had at least one medication in that class or had no medication in the class. The three most commonly occurring major therapeutic drug classes in both data sets were cardiovascular drugs (PACE = 59% of participants; self-report = 67% of participants); hormones and synthetic substances (PACE = 32% of participants; self-report = 37% of participants) and CNS agents (PACE = 28% of participants; self-report = 34% of participants).
Across the 10 major drug classes, there was an average of 91 percent agreement (9 of 10 classes) between the self-report and pharmacy records. Agreement scores ranged from 50 to 100 percent, and almost half (49%) had perfect agreement between their self-report data and their pharmacy data. Within the specific cardiovascular drug classes, agreement scores ranged from 43 to 100 percent, with an average of 96.4 percent (6.7 of 7 classes). Within the specific CNS drug classes, agreement scores ranged from 33 to 100 percent, with an average agreement of 92.7 percent (2.8 of 3 classes).
Sources of discrepancy between self-reported medications and the pharmacy database in the Caskie and Willis (2004) study most often occurred because the pharmacy records excluded medications that were contained in the self-reports. The pharmacy records did not include an average of 24 percent of the major drug classes that were self-reported, whereas on average, the study participants did not report 7 percent of the major drug classes contained in their pharmacy reports. Similarly, for the specific cardiovascular drugs, the average rate of omissions was 1 percent for self-reports, and the average rate of omissions for the pharmacy records was 13 percent. For the specific CNS drug classes, self-reports showed on average omissions for 5 percent of the drug classes, whereas the pharmacy records reported an average of 13 percent fewer CNS classes than the self-report data.
The study authors examined the variables that predicted agreement and disagreement between the self-report and the pharmacy data. They found the following significant predictors for the major drug classes:
For the cardiovascular drug classes, the following significant relationships were found:
For the CNS drug classes, the following significant relationships were found:
Caskie and Willis’s study findings support the conclusion that the brown bag method provides a reasonable substitute for pharmacy records as a measure of current medications.
In another study utilizing the brown bag method, Freml, Farris, Fang, and Currie (2004) performed an analysis of 50 randomly selected brown bag reviews performed by pharmacy students and 100 randomly selected brown bag reviews performed by pharmacists in the Iowa brown bag review to compare types of recommendations (e.g., cost savings versus therapeutic) made by pharmacists and students. The Iowa brown bag reviews were performed by community pharmacists, as well as by doctoral pharmacy students during their final year of pharmacy education. The mean age of the 150 subjects in the Freml et al. study was 77, and more than 70 percent were female. Interestingly, the number of prescription medications the subjects brought to the pharmacy in the brown bags for review was higher than the number of medications they self-reported using over the past month. There were no differences in the demographics, disease variables, or medication use variables between the pharmacist-reviewed and the student-reviewed groups. Results found that each subject used an average of 5.5 prescription drugs of which 2.2 (41%) were generic, and an average of 2.5 nonprescription drugs of which 1.6 (64%) were generic. There were no differences between pharmacist and pharmacy student groups in the rate of substitution and interchange of medications. However, there were statistically significant differences in pharmacist and student recommendations. In general, pharmacy students were more likely than pharmacists to make any recommendation. Specifically, students made significantly more “stop prescription drug,” “switch to alternate brand,” and substitution recommendations. Grouping recommendations into cost-saving and therapeutic categories, pharmacy students made significantly more cost-saving recommendations than pharmacists. Students also made more therapeutic recommendations, but the difference was not significant.
Freml et al. (2004) provide a few reasons why pharmacy students may have provided more recommendations, including possibly having more time than pharmacists to do reviews (as pharmacists had to time-share with other on-the-job tasks), having better interviewing skills as a result of recent clinically oriented education, unfamiliarity with patients’ wishes, and lack of concern about damaging provider relations at the pharmacy site. Freml et al. (2004) concluded that students can be valuable members of the health care team by providing cost-saving and therapeutic recommendations, and by decreasing pharmacist workload. Prior research has documented that physicians frequently accept recommendations from students at a rate similar to which they accept recommendations from pharmacists (Chisholm and Hawkins, 1996; Chisholm, Taylor, and Hawkins, 1997; Slaughter, Ericson, and Thomson, 1994; Taylor, Church, and Byrd, 2000; Briceland, Kane, and Hamilton, 1993).
Nathan, Goodyer, Lovejoy, and Rashid (1999) reported on medication use of 205 patients who volunteered to participate in a brown bag review. The mean age of the participating patients was 64.45. The number of drugs reviewed per patient ranged from 1 to 14, with an average of 6.2. Pharmacists made interventions in 87 percent of the reviews. Interventions included: providing information about the purpose of at least one medication (65% of the reviews); improving or correcting usage of at least one medication (46%); providing knowledge on common or important adverse drug reactions or side effects (52% of reviews). Fifty-eight percent of patients admitted to or were suspected of either not using at least one of their medications at all or not using them according to prescribed directions. Interactions between medications (sometimes between prescribed and over-the-counter medicines) were identified in 4 percent of the reviews.
39 The 19 items were derived from a series of patient interviews and a focus group interview, and reflects statements made by individual patients regarding their drug-taking behavior. It requires approximately 12 minutes to complete. Patients are asked to indicate how much they agree with each statement on a 4-point Likert scale (1=don’t agree at all; 2= don’t agree; 3=agree; and 4=agree very much). One item, for example, states, “If the rheumatologist tells me to take the medicines, I do so.” Six items are stated negatively, and their scoring is reversed. For example, “If I can help myself with alternative therapies, I prefer to do that to what my rheumatologist prescribes.” The CQR total score is calculated by summing the items, subtracting 19, and dividing by 0.57, enabling the CQR total score to vary from 0 (complete noncompliance) to 100 (perfect compliance).