Chapter 8. Study 2: Cognitive function and health literacy decline during ageing
8.3.5 Sensitivity and post-hoc analyses
When memory decline was defined as declines of >2 and >5 points, 1369/5255 (26.1%) and 353/5255 (6.7%) participants were defined as experiencing memory decline. The corresponding values for these re-definitions of verbal fluency decline were 577/5254 (11.0%) and 35/5254 (<1.0%). When the decline variables defined as >2 points decline were added to the final model, the associations between cognitive decline and health literacy decline became slightly stronger in magnitude (OR=1.64; 95% CI: 1.39–1.94 for memory decline and OR=1.52; 95% CI: 1.22–1.90 for verbal fluency decline). When the decline variables defined as >5 points decline were added to the final model, the association between memory decline and health
Chapter 8. Cognitive function and health literacy decline during ageing
112 literacy decline became even stronger (OR=1.86; 95% CI: 1.44–2.41), although the association between verbal fluency decline and health literacy decline was null (OR=1.14; 95% CI: 0.50–2.62). This estimate is imprecise because less than 1% of participants declined by >5 points in verbal fluency. Having a chronic disease diagnosis or depressive symptoms were not associated with health literacy decline and did not affect the ORs for any of the other variables and health literacy decline when added to the final model (not shown). Addition of prospective memory and mental processing speed scores to the cognitive function indices in the model also did not affect any ORs for the other variables and health literacy decline (not shown; see Appendix 8.1 for the results of the published journal article, which are adjusted for prospective memory and mental processing speed).
8.4 Discussion
Consistent with previous research using the same dataset, nearly one third of adults aged 52 years and over had health literacy limitations at wave 2 in the English Longitudinal Study of Ageing (Bostock & Steptoe, 2012). Over the six-year follow-up period, about one-fifth of the sample declined in health literacy skills. Age differences in the likelihood and rate of health literacy decline were pronounced, with adults aged 80 years and over having more than three times greater odds of experiencing health literacy decline than those in their early 50s. Striking social inequalities in health literacy decline were evident, where men and adults from deprived social backgrounds were the most vulnerable to lose the literacy skills required to manage health during ageing. Cognition appears to be a key risk factor explaining health literacy decline. Even subtle, one-point differences in cognitive function affected the likelihood of health literacy decline, and experiencing cognitive decline of any magnitude was strongly associated with health literacy decline.
My finding that cognitive function mostly explained the relationship between older age and health literacy decline was expected based on cross-sectional evidence showing that the constructs of cognition and health literacy overlap to a large degree (Kaphingst, Goodman, Macmillan, Carpenter, & Griffey, 2014; Levinthal et al., 2008; Mõttus et al., 2014; Wolf et al., 2012). Contrary to my findings, three other studies found that the association between age and health literacy was independent of cognitive impairment according to the Mini Mental Status Examination (MMSE) (Armistead-Jehle et al., 2010; D. W. Baker et al., 2002; Gazmararian et al., 1999). However, the MMSE does not detect subtle individual differences in cognitive function, as this study did. An important aspect of this study is that not everyone
Chapter 8. Cognitive function and health literacy decline during ageing
113 who experienced cognitive decline also experienced health literacy decline. The degree to which typical cognitive ageing versus ageing-related cognitive impairments of varying severities affect health literacy skills remains to be elucidated. This study suggests that non-pathological cognitive decline negatively affects health literacy during ageing. Further longitudinal studies that address the fluidity of literacy and cognition during ageing are needed for consideration alongside this study.
8.4.1 Limitations
Although validation data for the individual health literacy measure I used were not available, it was taken from a validated international adult literacy survey (Thorn, 2009). The measure does not capture prose literacy, information navigation, or numeracy, although it is a measure of document literacy that has good face validity. The ability to read and understand a medicine label is crucial to several health outcomes, and has been associated with risk of all-cause mortality among older English adults (Bostock & Steptoe, 2012). The scale had narrow range and a ceiling effect, where over two-thirds of the study sample scored 4/4 on the scale at both time points; this is a common problem in health literacy measures (Davis et al., 1993; Parker et al., 1995). Consequently, few participants declined by more than 1 point on the health literacy measure (316/5256; 6%), preventing me from examining decline of varying magnitudes and from varying starting points. The distribution of change scores among those who declined was: -1 point (716/5256; 14%), -2 points (238/5256; 5%), -3 points (64/5256; 1%), and -4 points (14/5256; <1%). I could not examine non-linear change or change over a period longer than the six-year follow- up. If further follow-up data on health literacy are collected in future waves of the ELSA, this would be an important area for future research.
I observed a degree of health literacy decline among adults aged ≥80 years that was not explained by cognitive variables, which might be because the ELSA dataset does not account for all aspects of cognitive function. For example, inductive reasoning was not measured, but is correlated with both age and health literacy (A Singh-Manoux et al., 2012; Wolf et al., 2012). I also had no measures of the component processes involved with active learning including knowledge integration and text inference, which predict reading comprehension skills among older adults (Hannon & Daneman, 2009). However, the aspects of short-term memory and verbal fluency that were measured are related to these abilities and were taken from established and validated measures that are used in other longitudinal studies of
Chapter 8. Cognitive function and health literacy decline during ageing
114 ageing. This research should therefore be interpreted jointly with other studies that have a greater range of cognitive function measures.
Another important limitation of this study is attrition bias. The prevalence of limited health literacy at baseline was 42% among those who dropped out of the study, but was only 28% among those who remained in the study between waves. Study attrition also increased with age, from approximately 26% among those aged 52-54 years to 71% among those aged ≥80 years. My results may therefore underestimate the true prevalence of limited health literacy among the older English population, particularly in the most elderly age group. Ethnic minorities, participants with no educational qualifications, and those with routine occupations were more also likely to drop out of the study, and were more likely to have limited health literacy at baseline. I may have underestimated the magnitude of associations between these sociodemographic variables and health literacy decline.
8.4.2 Strengths
To the best of my knowledge, this longitudinal study is the first to track health literacy skills over time, particularly among an ageing sample. The large sample size and rich data on sociodemographic and cognitive factors alongside data on fluid health literacy skills in the ELSA made this analysis possible. In addition to adjusting for sociodemographic and cognitive factors, I also accounted for self-reported eyesight. I did this based on the suggestion of a letter to the editor of J Gen Intern Med following the original publication of this study (Matthiesen, Vela, & Press, 2015). It is important to account for this variable because functional health literacy performance partly depends on one’s visual function. The health literacy measure used in the ELSA has good face validity and predictive capability for mortality risk, and was taken from a validated international adult literacy survey. The cognitive function variables that were assessed as predictors of health literacy were those known to be sensitive to change with age that would be minimally affected by literacy skills. As a longitudinal analysis conducted with little prior knowledge on health literacy during ageing, this study provides valuable evidence for future research hypotheses.
8.4.3 Conclusions
The literacy skills required to manage health appear to undergo ageing-related decline among older English adults beginning, on average, around age 65. Rate of
Chapter 8. Cognitive function and health literacy decline during ageing
115 decline increases with age, with adults aged ≥80 years being vulnerable to rapid health literacy decline. Health literacy decline among older adults is marked by social inequalities, whereby men and adults from deprived social backgrounds were the most vulnerable to skill loss during ageing. Cognitive function and even slight cognitive decline during ageing appeared to affect the likelihood of health literacy decline, particularly among the older age groups in the sample. Finally, given that literacy skills are commonly lost during ageing, a time when adults often need health information and services, the current population-wide burden of low health literacy may be substantial. Whether health literacy decline may be prevented through potentially modifiable behavioural influences on health literacy is unknown. For example, Internet use and mentally stimulating social activities may help adults to maintain or improve health literacy through directly stimulating cognitive and literacy skills. These relationships, which have never been investigated longitudinally, will be the focus of the next chapter.
Chapter 9. Internet use, social engagement, and maintaining health literacy
116