Measuring health literacy: Interviewer-dependent variation in literacy assessments in the REALM

Health literacy is an important component of health communication but has not been well studied among women using the maternal-child care system. In a study of 1,034 low-income women in pregnancy and postpartum the Rapid Estimate of Adult Literacy in Medicine (REALM) was used to determine literacy. The rates of low literacy in this sample (4.8%) was approximately one quarter that expected based on demographic characteristics. We wished to determine if the commonly utilized REALM had failed.

Methods: Inter-interviewer variability was assessed in the initial sample by test of equality of variance. A sample of women was randomly selected for re-assessment of literacy using the REALM. Additionally we delivered the Short Test of Functional Health Literacy in Adults (STOFHLA) for validation.

Results: In the initial sample of 1,034 women literacy scores varied significantly with interviewer (F test, P<0.05). In the re-assessment (n=237) the distribution of literacy scores corresponded to that expected by demographic variables; 13% of the sample had low literacy (< 6th grade), 36% had marginal literacy (<7th-8th grade), and 51% had higher literacy (>9th grade). In the re-assessment the REALM and STOFHLA scores were highly correlated (74%) as expected.

Discussion: We found that in a sample of low-income women in pregnancy and postpartum the REALM assessment was unreliable. This test is widely used in the developing field of health literacy and health communication but accurate scoring is dependent on interviewer interpretation of responses. Further research is needed to determine if interviewer characteristics contribute to this variability.