Health Literacy and Difference in Current Wealth Among Middle-Aged and Older Adults

Abstract

Numerous studies suggest that health literacy improves health outcomes at older ages. But how, and to what extent, health literacy contributes to improving financial outcomes has not been examined. This study proposed a conceptual framework to explain the mechanisms between health literacy and current wealth. Data from the Health and Retirement Study (HRS) are used to estimate proposed direct and indirect effects between health literacy and current wealth. We found that, for the most part, health literacy is directly associated with wealth rather than indirectly through mediating variables. Alternatively, out of all indirect effects investigated in the model, health literacy affects wealth mainly through the path of chronic condition, work limitation, and income.

Appendix

The binary variable of health literacy is constructed in three steps. First, each individual’s response to an item t was rescaled based on the following formula (Brockett et al. 2002): $B_{ti} = \mathop \sum \limits_{j < i} P_{tj} - \mathop \sum \limits_{j > i} P_{tj} , i = 0, 1, 2, 3..,k_{t}$; where B_ti is an individual’s rescaled response value for the categorical option i (in the case of a binary variable, i = 0, 1) to variable t; P_tj denotes the observed proportion of respondents in options below or above i. This is a linear transformation of categorical responses into numerical values reflecting the relative difficulty of each particular response.

Let us assume two scenarios with five respondents and two different items with a binary response. One item is related to knowledge of colon cancer, denoted as variable c. The first two respondents show the correct answer (=1) while the remaining three provide the wrong answer (=0). In this case, the correct answer is transformed into a numerical value of 0.6 and the wrong answer is converted into − 0.4 by assuming a uniform distribution (i.e. each option i (=1 or 0) takes place with the same probability). That is, B_c1 = (1/5 + 1/5 + 1/5)-0 = 0.6 where there are three responses below i = 1 and zero responses above i = 1; B_c0 = 0 − (1/5 + 1/5) = -0.4 where there are zero responses below i = 0 and two responses above i = 0. The other item is related to word recognition of “Fatigue”, denoted as variable d. In this case, assume that the first four respondents correctly answer but the remaining one respondent does not. In this case, the correct answer is assigned as $B_{d1} = \frac{1}{5} - 0 = 0.2$ where there is one response below i = 1 and there are zero responses above i = 1; $B_{d0} = 0 - \left( {\frac{1}{5} + \frac{1}{5} + \frac{1}{5} + \frac{1}{5}} \right) = - 0.8$ where there are zero responses below i = 0 and four responses above i = 0. As we can see in these two scenarios, the same correct answer (=1) has different values (0.6 vs. 0.2) depending on relative difficulty of the item. This transformed value has desirable characteristics such as bounded in [− 1,1] and has a mean of zero.

The second step is to calculate PRIDIT weights with the rescaled values of the 16 items by using principal components analysis. PRIDIT weights allow more weight to be given to an item with less correlation with the other items because the item is more informative, distinguishing the level of health literacy. More specifically, the first eigenvalue (λ₁) and the corresponding eigenvector (ν₁) is obtained from PCA and used to calculate a weight for each item based on the formula (Lieberthal 2008): $\sqrt {\lambda_{1} } \times \nu_{1}$.

The third step is to classify respondents as either high or low health literacy. A PRIDIT score of an item for an individual is calculated with the formula (Lieberthal 2008): (PRIDIT weight × rescaled values)/first eigenvalue, to obtain a total PRIDIT score for each individual with the sum of 16 PRIDIT scores. Total PRIDIT score has a range of − 1 to + 1 with a mean of 0 and a monotonic relationship with a correct answer. Thus, respondents are classified with a total score of above 0 as high health literacy and of below 0 as low health literacy.