“…However, many studies have shown self-report measures of adherence to be relatively accurate for predicting clinically meaningful outcomes (e.g., meta-analysis by Nieuwkerk & Oort, 2005) and their accuracy is improved by controlling for social desirability bias (Simoni et al, 2002), as we did in the present study. Although the reliability of the adherence measure was low (a = 0.58 at baseline and a = 0.60 at follow-up), we considered the measure acceptable because it has: (1) performed similarly in other studies (Jerant et al, 2008); (2) been widely used for a variety of medical conditions (Murray et al, 2007;Parienti et al, 2004;Roth et al, 2009); and (3) been shown to have predictive validity (Morisky et al, 1986) and concordance with pharmacy refill data (Krousel-Wood et al, 2009). Moreover, we obtained significant results despite the ''random noise'' of unreliability, which mitigates against a measure yielding results.…”