Setting and subjects
This study was conducted at three family practice clinics affiliated with an urban academic teaching hospital in Toronto, Ontario. These clinics provide primary care to a large patient population living in the central area of the city. From January to September 2001, consecutive patients presenting to the clinic during regular office hours on selected weekdays were approached and asked to participate in a study of comprehension of prescription labels. Days on which patients were enrolled were selected on the basis of availability of a member of the research team. A total of 130 participants were enrolled. Patients were excluded if they were too ill to participate or were unable to communicate in English. Patients who participated in the study gave written informed consent and received a $5 payment. The St. Michael's Hospital Research Ethics Board approved this study.
Data collection
One of the investigators (CQNT) conducted face-to-face interviews with participants that obtained information on demographic characteristics, native language, and education. Participants were then presented with five instruction labels regarding how certain prescription medications should be taken or their potential side effects (Figure 1). Pharmacies usually affix these labels to the bottles or packaging of certain medications when they are dispensed to the patient. We selected five labels from among those in common use by pharmacies in Toronto. Labels were deliberately chosen to provide a wide range in terms of our assessment of the complexity of both the written information and the accompanying illustration. The labels were presented as black-and-white images that were enlarged from the original label size of 4 × 1 cm to a final size of 8 × 2 cm to enhance readability. The five labels were presented on a single sheet of letter-sized paper.
Participants were first shown five labels with text only (Figure 1, left column), then a separate sheet of paper with the labels with identical text and the addition of visual illustrations (Figure 1, right column). In each case, participants were asked, "If this label were on your pill bottle, how would you take this medication?" Participants were allowed unlimited time to reply. The interviewer wrote down participants' verbatim responses on a survey form; interviews were not audiotaped.
Following presentation of all labels, participants were given the Rapid Estimate of Adult Literacy in Medicine (REALM) test. The REALM test is a previously validated instrument that uses an objective scoring system to assign a grade-range estimate of literacy (grade 0 to 6, grade 7 to 8, or grade 9 and above) [8]. This test is a simple and widely used research instrument that is highly correlated with other measures of health literacy such as the Test of Functional Health Literacy in Adults (TOFHLA) [9].
Data coding
Two researchers independently coded participants' interpretations of medication labels as incorrect, partially correct, or completely correct. Disagreements in coding were resolved by consensus after further discussion. Labels A, C, and D were deemed to each convey a single main informational component, and interpretations of these labels were coded as either incorrect or completely correct. Labels B and E were deemed to each convey three main informational components. For example, for label B the informational components were: (1) possibility of drowsiness, (2) alcohol may worsen effect, and (3) caution if driving or handling machinery. If none of these components could be identified in the participant's verbatim response, their answer was coded as incorrect. If one or two of these components was found, the response was coded as partially correct. If all three components were found, the response was coded as completely correct. This method of coding, although not previously validated, offered useful detail regarding the completeness of the participants' comprehension. At the time of coding, investigators were blinded to participants' literacy level. The effect of the addition of illustrations on participants' performance was classified as improved, worse, or unchanged, based on the categorization of their first and second responses as incorrect, partially correct, or completely correct. When improvement or worsening was noted, responses were examined to characterize the nature of the change.
Statistical analyses
The sign test was used to assess whether there was significant improvement or worsening in the interpretation of the label with the addition of the illustration. Chi-square analyses were used to assess whether the effect of the addition of illustrations on participants' performance on label interpretation (improved, worse, or no change) was significantly associated (p < 0.05) with sex, age (under 25 years, 25 to 39 years, 40 to 64 years, or 65 years and over), native language (English or other language), or grade-range estimate of health literacy (assessed using the REALM). All significance tests were two-sided. Statistical analyses were performed using SPSS 10.0 (SPSS, Inc., Chicago, IL).
The power of this study was dependent on both the number of individuals whose performance on label interpretation changed with the addition of illustrations, and the anticipated magnitude of this effect. Assuming that about 25% of subjects or 30 individuals would demonstrate a change in performance, this study would have had 80% power to detect a ratio of 3:1 or greater in terms of the proportion of individuals with improved performance vs. worse performance.