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


AIDS and Behavior. 2008;12(1):86-94. 

In This Article


Of the 156 participants included in the analyses, 24% were women, mean age was 42 years, 49% were non-white, 94% had finished high school, and 56% reported sex with a man as their principal HIV risk factor. Under 80% reported that it was "definitely worth taking" HIV therapies. With regard to the HIV ARVs reported taken by patients, 48% were NRTIs, 16%NtRTIs, 31% PIs, and 4% NNRTIs. For the monitored ARVs, 45% were NRTIs, 4% NtRTIs, 44% PIs, and 7% NNRTIs.

The mean self-reported adherence for all study visits was 90% for 3 days preceding each follow-up visit, 85% for the 7 days preceding each visit and 79% for the month preceding each visit ( Table 1 ). The corresponding mean MEMS adherence was 72, 71, and 70%, respectively. The correlations between MEMS and self-reported adherence across all study visits for 3, 7, and 30 days were 0.50, 0.49, and 0.55. Mean 1-month self-reported adherence was 81% for the frequency response format, 82% for percent response format, 73% for the rating response format, and 79% for the combined measure (the mean of the previous three measures).

The mean differences between self-reported and MEMS adherence for all study visits, adjusted for clustering, are shown in Fig. 1. For all three time periods (the first three sets of bars), self-report was significantly higher than MEMS [t (145) = 9.67 for 3 days, t (145) = 7.6 for 7 days, and t (146) = 4.76 for 1 month, p < 0.01 for all]. The magnitude of the over-reporting was 17, 14, and 9% for 3, 7 days, and 1 month, respectively. The accuracy of self-report significantly improved with an increasing reporting interval: 1 month was significantly closer to MEMS than 7 days, 7 days significantly closer than 3 days, and 30 days closer than 3 days [t (145) = 4.44, t (145) = 3.26, and t (145) = 6.56, respectively, all p < 0.01].

Mean difference between self-report (SR) and MEMS adherence

To compare the frequency, percent, and rating formats we examined bias, linearity, convergent validity (correlation with means), and predictive validity (relationship to HIV RNA levels), again adjusting for clustering. The frequency and percent response tasks each yielded results that were higher than MEMS, both by 12% (Fig. 1, bottom three sets of bars). The rating response format was not significantly different from MEMS [3% higher, t (146) = 1.44, p > 0.05]. The over-reporting for the frequency and percent response formats was significantly greater than the rating response format [t (146) = 6.32 and t (146) = 5.86, p < 0.01 for both]. Thus, only the rating response showed no bias. To test for linearity we included a squared term in regressions equations to regress the MEMS adherence variable on the self-report adherence variables and their corresponding squared terms, and for all three of the 1-month response formats the squared term was non-significant, suggesting all were linear. The correlations between the frequency, percent, and rating response formats and the corresponding MEMS measures were 0.51, 0.45, and 0.48, respectively ( Table 1 ). To assess efficiency, we examined mean scores for each adherence measure for those with undetectable and detectable HIV RNA ( Table 2 ). For the frequency response format, there was no significant difference after adjustment for clustering ( Table 2 , fourth column, 85.6 vs. 82.2). For the percent and rating response formats, the differences were significant (86.8 vs. 80.7 and 77.6 vs. 71.0). The MEMS measure had the best predictive validity [F (1, 138) = 7.50 and p < 0.01]. The correlations between the 1-month self-report measures and log10 HIV RNA levels ranged from −0.17 to −0.24 ( Table 2 , fifth column), and were not significantly different from one another, and the correlation between the 1-month MEMS measure and log10 HIV RNA was -0.29.

Among the three 1-month self-report items, the rating item was the closest to the 1-month MEMS adherence, with a mean difference that was not significantly different from zero. To understand better why this item performed better than the frequency and percent items, we constructed a series of graphs that show the frequencies of each of the response categories and the mean MEMS adherence for those in each response category (Fig. 2). When comparing the frequency and rating items, the two middle response categories (40 and 60) were more often used for the rating response than for the frequency response. Respondents are more willing to endorse "fair" (13%) and "poor" (16%) than "some of the time" (6%) or "a good bit of the time" (7%) when they have less than optimal adherence. Conversely, respondents were less likely to use the top category "excellent" (30%) with the rating item than with the frequency item ("all of the time," 46%). The percent graph (middle graph) shows results that mirror those of the frequency graph (upper graph).

Distribution of 30-day self-report adherence and corresponding mean MEMS adherence (the height of the bar represents frequency of patients in each self-report category. The dots on each line show the mean MEMS adherence for each self-report category)

The MEMS line in each graph of Fig. 2 shows that respondents with relatively poor adherence (0-60 on the self-report items) tend to underreport this poor adherence. That is, the mean MEMS adherence tends to be higher than the self-report category. Correspondingly, the MEMS line shows a tendency to over-report relatively good adherence. That is, mean MEMS adherence is <100% for those using the highest self-report category. The graphs show that these tendencies to under-report poor adherence and over-report good adherence are greater for the frequency and percent items than for the rating item.


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: