Optimal Recall Period and Response Task for Self-Reported HIV Medication Adherence

Minyi Lu; Steven A. Safren; Paul R. Skolnik; William H. Rogers; William Coady; Helene Hardy; Ira B. Wilson

Disclosures

AIDS and Behavior. 2008;12(1):86-94. 

In This Article

Methods

We used data from a randomized, crossover trial of a physician-focused intervention to improve adherence with HIV ARVs. We enrolled subjects from five sites in the greater Boston area including two academic medical centers, a community health center, a general medicine practice based in an academic medical center, and a private infectious diseases practice. Eligibility requirements included detectable HIV RNA at the most recent clinical visit, currently taking ARVs, willing to use an electronic pill bottle cap for one of their ARV medications for the duration of the study, and fluency in English. Of the 282 persons referred to study coordinators, 220 (70%) agreed to discuss participation with study enrollment coordinators, 197 (90%) of those agreed to participate, and 156 (79%) completed the run-in period which tested their willingness and ability to use the MEMS caps.

Enrollment occurred between November 18, 2002 and January 31, 2005. The protocol included an enrollment study visit and five follow-up study visits at 2-3-month intervals. At the enrollment visit, patients were instructed in the use of MEMS caps, and a single medication, either a once-a-day or a twice-a-day medication, was chosen for monitoring. At each follow-up visit we downloaded the MEMS data and noted medication changes. In addition, patients completed adherence self-report surveys. The number of participants and the number of follow-up visits completed were as follows: 156 (one visit), 141 (two visits), 126 (three visits), 114 (four visits), and 106 (five visits). To obtain HIV RNA data we reviewed participants' medical records. Of the 643 study visits, we found HIV RNA data from the previous 30 days in 488 (76%).

Medication Event Monitoring System. We used the MEMS Smart Caps (AARDEX Ltd., Union City, CA, USA). MEMS caps have a pressure-activated microprocessor that records the date and time of each bottle opening and closure. There are a number of possible ways to summarize MEMS data. To optimize the conceptual alignment between the self-reports (which ask about whether specific doses were taken), and the MEMS summary measure, we used the percent of doses taken correctly as the MEMS summary measure. To operationalize this, we defined a day as the period starting at 3a.m. and continuing for 24 h. We did not use 12a.m. because many people's "day" ends after mid-night. That is, they may take the "evening" dose of a medication after mid-night. For those taking a twice-a-day medication we divided the day in half at 3p.m. Thus, for those on once-a-day dosing, any dose between 3a.m. 1 day and 3a.m. the next day was classified as a "correct dose." Similarly, for those on twice-a-day dosing, any dose taken between 3a.m. and 3p.m. was a "correct" morning dose, and any dose taken between 3p.m. and 3a.m. the next day was a "correct" evening dose. The adherence summary measure was the percent of correct doses taken during a given interval (3, 7, or 30 days prior to the study visit at which the self-report was done).

Three Days. We asked patients whether they had taken each of the prescribed doses of the monitored medication on each of the 3 days prior to the study visit. There were separate sets of items for those on once-a-day and twice-a-day regimens. For once-a-day regimen, the items ask a patient: "Did you take your medication yesterday?", "... 2 days ago?" and "...3 days ago?" For twice-a-day regimen, the same questions were asked for morning (first) dose and evening (second) dose for each of the previous 3 days. The 3-day adherence measure was the proportion of the sum of the doses taken for the 3 days divided by the total possible number of doses. Other items in the series (not reported on here) ask patients what time of day they try to take their medication, and whether they were more than 3 h early or late with their dose.

Seven Days. We asked how many doses they had missed during the 7 days prior to the study visit. Again, there were separate sets of items for those on once-a-day and twice-a-day regimens. For once-a-day regimen, the items ask a patient: "Of the seven doses you were supposed to take during the last week, how many doses did you miss?". For twice-a-day regimen, the items ask a patient: "Of the seven morning doses you were supposed to take during the last week, how many doses did you miss?" An identical item asked about the evening dose. The 7-day adherence was calculated as 1 minus the proportion of the sum of the missed doses for the 7 days divided by the total possible number of doses.

One Month. We asked participants to assess their adherence with their ARV medications during the previous month with three different response formats: (1) frequency, (2) percent, and (3) rating response. The frequency item was "Did you take all your medications all the time?" (six response categories: None of the time, a little of the time, some of the time, a good bit of the time, most of the time, and all of the time). The percent item was, "What percent of the time were you able to take your medications exactly as your doctor prescribed them?" (11 categories, 0, 10, 20, ...100%). The rating item was, "Rate your ability to take all your medications as prescribed" (six categories: Very poor, poor, fair, good, very good, and excellent). Hereafter these response tasks are referred to as frequency, percent, and rating tasks. For analysis we assigned the six categories scores of 0, 20, 40, 60, 80, and 100, with 100 being the best adherence (Preston and Colman 2000). All self-reported adherence measures at all study visits are presented as percentages, with 100% considered perfect adherence. We averaged the three 1-month scores to generate a combined 1-month adherence that we refer to as the combined 1-month self-report outcome measure.

Demographic and background characteristics assessed included age, gender, education, racial and ethnic background, housing accommodations, employment status, marital status, sexual orientation, HIV risk factors, and attitude toward HIV therapies. We use the Medical Outcome Study 12-item short-form health survey to assess physical and mental health (Ware et al. 1996), and the PC-SAD to assess depression (Rogers et al. 2002). The PC-SAD is a 37-item DSM-IV compatible depression screener. The scoring algorithm creates a "Yes" or "No" for each of the nine DSM-IV depression symptoms. As per DSM-IV, five or more symptoms was considered major depression. Plasma HIV RNA levels of <75 copies/ml were classified as undetectable.

All analyses used STATA. Our data included 156 patients who made 643 study visits. We treated MEMS and self-reported ARV adherence as continuous variables. The visit was the unit of analysis. To compare MEMS adherence with 3-, 7-day, and 1-month self-reports, we subtracted the self-report from the MEMS value for each visit, and reported the mean of the difference scores. For the 1-month self-report we used the combined measure described above. Because each individual contributed between one and five observations to the dataset, the data for MEMS and self-reported adherence as well as the difference scores between the two measures at each study visit may be correlated for the same patient. Therefore, we adjusted the intra-patient correlation for the mean differences using the clustering procedure in linear regression in STATA. The clustering also accounts for missing data on different study visits by the same patient.

To determine whether over-reporting was greater for one type of self-report compared with another, we subtracted one difference score from the other [e.g. (1-month self-report - 1-month MEMS) - (3-day self-report - 3-day MEMS)], again using clustering. We considered several methods to compare two different measures with MEMS, including parametric and non-parametric correlations. We chose this difference of differences approach because it preserved the magnitude of the associations (i.e., points on a 0-100 scale).

To compare accuracy of the three different 1-month self-reports, we examined bias, linearity, convergent validity, and predictive validity. We examined bias (the tendency to over or under report) using methods described in the preceding paragraph. To assess linearity, we constructed an equation that included a squared term for the self-report adherence variable, and regressed MEMS adherence variable on both the self-report adherence variable and its square term. HIV RNA was treated both continuously (as log10 HIV RNA copy number) and dichotomously (detectable or undetectable). We used Pearson correlation coefficients to assess the relationships between each self-reported adherence measure and log10 HIV RNA (convergent validity). To assess predictive validity, we compared mean adherence for those with detectable and undetectable HIV RNA. P-values, as above, were adjusted for clustering.

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as:

processing....