A Multivariate Analysis of Factors Associated With Depression:  Evaluating the Role of Health Literacy as a Potential Contributor FREE

MORE THAN one third of Medicare managed care enrollees have inadequate health literacy skills.¹ Although it is likely that health literacy skills are related to general literacy skills, an individual's functional health literacy—the ability to read and comprehend prescription bottles, appointment slips, and other essential health-related materials required to function well in the health care environment—may be significantly worse than their general literacy, because health literacy is context specific.² Although many patients with inadequate health literacy are unaware of their deficiency,³ others feel significant shame and are unwilling to disclose their reading problem.^3- 4 One study found that among individuals who admitted they had trouble reading, the majority had never told their spouse or family, and 19% had never disclosed their difficulty reading to anyone, including their health care provider.³ Poor self-esteem, along with the shame and embarrassment often felt by individuals with low health literacy skills, may lead to social isolation and pose an important psychological barrier to asking for help.³ It is possible that these feelings could also contribute to a higher prevalence of depression among individuals with low health literacy skills.

Among community residents older than 60 years, rates of significant depressive symptoms are approximately 15%.^5- 9 Although no study has specifically examined the relation between health literacy and depression, data generally indicate that higher education levels are associated with fewer depressive symptoms.^10- 15 Some studies propose that the benefits from education may be the result of greater financial success, improved lifestyle behaviors, and problem-solving capacity.¹⁶ Duncan¹⁷ suggested that there may be something about the skills acquired in school, or possibly other characteristics (eg, perseverance), of those who receive more schooling that helps better-educated individuals avoid undesirable life events. Because years of school completed is an inaccurate indicator of literacy,^1,18- 21 previous studies of the relation between educational attainment and depression may significantly underestimate the true association.

The potential relation between health literacy and depression may also be mediated by health status. There are some data that suggest that inadequate health literacy is associated with poorer self-reported health status,²² and there is extensive literature that documents the strong predictive power of health status on depression.^5,8,23- 31 These results are of particular concern among elderly patients: their higher rate of low health literacy,^1,19 coupled with a high prevalence of chronic conditions, may lead to higher rates of depression. In addition to health status, there is some research that has shown that demographic factors,^5,15,24 social support,^{5,15,24,32- 33} and health behaviors^34- 37 are also related to depression. However, these other factors are often not examined together in multivariate analyses or in an elderly population.

We conducted this analysis to determine whether older adults with inadequate health literacy skills were more likely to report depressive symptoms and whether health literacy was an independent predictor of depression after accounting for other known determinants of depression.

This analysis is a part of a larger study that examined the prevalence of low functional health literacy among community-dwelling Medicare enrollees in a national managed care organization. Details of the study population and locations are described elsewhere.¹ Briefly, we selected 4 of the 8 locations where a national managed care organization had a large population of Medicare managed care enrollees (Cleveland, Ohio, Houston, Tex, South Florida, including Fort Lauderdale and Miami, and Tampa, Fla). One project coordinator and 8 interviewers were hired at each site and participated in intensive training. Several interviewers at the Houston, South Florida, and Tampa locations spoke both English and Spanish.

A letter of introduction describing the study and confidentiality policies and protocols for the study was sent to each member 65 years or older 3 months after they enrolled in the managed care organization (n = 8409). One week after the letters were sent, an interviewer called each enrollee to determine eligibility for the study. Of the 7471 individuals contacted, 3247 refused to participate. Individuals who indicated that they were not comfortable speaking either English or Spanish, were blind, had severely impaired vision that could not be corrected with glasses, or were living in a nursing home were not eligible to participate in the study (n = 737). We also excluded enrollees who missed 1 or more screening questions for severe cognitive impairment (not able to correctly identify year, month, state, year they were born, or their address). Of the 3487 enrollees who were eligible and agreed to participate, 143 people did not show for their interview, and 84 did not complete the survey, the literacy testing, or both. A total of 3260 individuals completed a 1-hour in-person interview between June and December 1997. Written informed consent was obtained from all participants before beginning the interview. For this analysis, we further excluded individuals who scored in the "severe" category of the Mini-Mental State Examination (n = 68), since our measure of depression symptomatology (Geriatric Depression Scale [GDS]) is less valid in assessing cognitively impaired individuals.³⁸ We also excluded individuals with missing data for 5 or more questions on the 15-item depression scale (n = 21). Our final sample size for this analysis included 3171 Medicare managed care enrollees.

The dependent variable we examined was depression symptomatology, measured by the abbreviated GDS.³⁹ This instrument is the only instrument that has been developed with the elderly specifically in mind. It has a high sensitivity and specificity. Because it is widely used, we could compare our results to other studies in an elderly population.^40- 42 The abbreviated GDS is composed of 15 questions and takes about 5 to 7 minutes to complete. One point is counted for each depressive answer, and the points are added to form a total score ranging from 0 to 15. For the small number of respondents who did not complete up to 4 of the questions (n = 21), we imputed values for missing questions based on a proportion of the positive responses to the other answered questions (eg, 0.5 was assigned to the one missed question if 7 of the 14 answered questions were positive). Individuals who scored from 0 to 4 are classified as not being depressed, 5 to 9 as having mild depression, and 10 to 15 as having moderate to severe depression.⁴³ Preliminary analyses examined predictors for all 3 categories and separately compared the mild and moderate-to-severe groups to the not depressed group. Because of the similar results these comparisons yielded and the small sample size for the moderate-to-severe group (n = 77), we collapsed the mild and moderate-to-severe groups to form the depressed group of respondents.

The primary independent variable we examined was health literacy as measured by the Short Test of Functional Health Literacy in Adults (S-TOFHLA).⁴⁴ The S-TOFHLA uses actual materials that patients might encounter in the health care setting and consists of 2 parts. The reading comprehension section is a 36-item test that uses the modified Cloze procedure; that is, every fifth to seventh word in a passage is omitted, and 4 multiple choice options are provided.⁴⁵ This section measures patients' ability to read and understand prose passages selected from instructions for preparation for an upper gastrointestinal tract radiograph series and the patient "Rights and Responsibilities" section of a Medicaid application. Readability levels of these passages on the Gunning-Fog index⁴⁶ are grades 4.3 and 10.4, respectively. The numeracy section is a 4-item test that uses actual hospital forms and labeled prescription vials. It tests a patient's ability to comprehend directions for taking medicines, monitoring blood glucose levels, and keeping clinic appointments. Correct responses for each question in the reading and numeracy sections receive one point, and incorrect responses do not receive any points. Each item in the reading comprehension is multiplied by 2 (for 36 items) to create a score from 0 to 72, and each numeracy question is multiplied by 7 (for 4 items%) to create a score from 0 to 28. The sum of the 2 sections yields the S-TOFHLA score, which ranges from 0 to 100.

Scores on the S-TOFHLA are classified and interpreted as follows: inadequate health literacy (scores 0 to 53)—individuals will often misread the simplest materials, including prescription bottles, appointment slips, and the instructions for an upper gastrointestinal tract radiograph series; marginal health literacy (scores 54 to 66)—individuals will perform better on the simplest tasks but will have difficulty comprehending the Medicaid "Rights and Responsibilities" passage; adequate health literacy (scores 67 to 100)—individuals will successfully complete most of the tasks required to function in the health care setting, although many still have difficulty comprehending more difficult information (ie, materials written above a 10th-grade reading level).

We also included other variables that have been shown to be related to depression in elderly populations, including sociodemographics (race and language, sex, age, education, income), social support (marital status, social support scale), health behaviors (exercise, body mass index [BMI], drinking alcohol), and health status (health conditions, instrumental activities of daily living [IADL], activities of daily living [ADL], self-reported health status).^5- 8

We measured social support as defined by the tangible social support subscale of the Medical Outcomes Study.⁴⁷ This index is composed of 4 questions that assess how often each of the following kinds of support are available: someone to help you if you were confined to bed, someone to take you to the physician if you needed it, someone to prepare your meals if you were unable to do it yourself, and someone to help with daily chores if you were sick. For each question, the response options (and assigned points) were none of the time (1 point), a little of the time (2 points), some of the time (3 points), most of the time (4 points), or all of the time (5 points). The responses to each of these questions were summed and then grouped according to how they corresponded to the original categories: none or a little of the time (4-12), some of the time (13-16), most of the time (17-19), and all of the time (20).

Exercise habits were based on responses to the following question: "In an average week, how many times do you engage in physical activity for at least 20 minutes? Specifically, exercise or work which is hard enough to make you breathe heavier and your heart beat faster?" Respondents were coded as exercising 4 or more times a week, 3 times a week, 1 or 2 times a week, or less than 1 time per week.

We calculated BMI based on self-reported weight and height (weight in kilograms divided by the square of height in meters). Respondents were grouped into 1 of 4 categories based on criteria recommended in the consensus statement of the 1985 National Institutes of Health Development Conference on the Health Implications of Obesity.⁴⁸ For men, underweight was defined as a BMI of 20.7 or less, midweight as 20.8 to 27.7, overweight as 27.8 to 31.0, and severely overweight as 31.1 or more. For women, underweight was defined as a BMI of 19.1 or less, midweight as 19.2 to 27.2, overweight as 27.3 to 32.2, and severely overweight as 32.3 or more.

Drinking patterns were defined based on recommendations for safe limits of drinking by the US Department of Agriculture and the National Institute of Alcohol Abuse and Alcoholism.^49- 51 For women, a heavy drinker was defined as having more than 7 drinks per week, moderate drinker as between 1 and 6 drinks per week, and nondrinker as no drinks per week. For men, a heavy drinker was defined as having more than 14 drinks per week, moderate drinker as between 1 and 13 drinks per week, and nondrinker as no drinks per week.⁵¹

Both ADL and IADL were measured using standard instruments adapted from the Medical Outcomes Study.⁵² The ADL functions are essential for an individual's direct self-care⁵³ and are based on 5 questions assessing any limitations with eating, dressing, mobility (getting in and out of bed), bathing, and toileting. The IADL functions are more concerned with self-reliant functioning,⁵³ for example, housework, meal preparation, shopping, transportation, and money management. Respondents who answered that they need some or complete help on any of the questions were coded as having a limitation.

Health conditions was a summed index of any of the following conditions: coronary heart disease or acute myocardial infarction (heart attack, coronary heart disease, heart bypass surgery or angioplasty, angina), chronic obstructive pulmonary disease (chronic bronchitis, emphysema), chronic heart failure (heart failure, enlarged heart, fluid in the lungs), stroke, asthma, high blood pressure or hypertension, diabetes, arthritis, cancer, cataracts, deafness, hearing trouble, and hip fracture. Respondents were coded as having either no conditions, 1 condition, 2 conditions, 3 to 4 conditions, or 5 or more conditions.

Analysis of survey data consisted of examining the overall distribution of selected enrollee characteristics and comparing the distribution of characteristics by GDS categories. χ² Analyses were conducted to determine significant differences between selected characteristics and GDS. A P value of .05 was used to determine statistical significance.

We conducted several logistic regression analyses to determine the relation between health literacy and depressive symptoms. We first calculated the unadjusted models and then conducted multiple logistic regression analysis using a hierarchical approach. All multivariate models included depression as the dependent variable and health literacy as an independent variable. The first model also included sociodemographic factors (age, sex, education, and income), the second model added social support measures (marital status and social support scale), the third model added health behaviors (exercise, BMI, and drinking), and the fourth model added health status (health conditions, ADL, IADL, and self-reported health status). We also assessed for potential interactions (ie, cognitive impairment, age, marital status, education, and health status). All analyses were conducted using the Statistical Analysis Software package.⁵⁴

The majority of the respondents were white (76%), female (57%), and between 65 and 74 years old (64%), had at least a high school education (64%), and currently earned more than $10,000 per year (66%).¹ Almost 35% of respondents had low health literacy skills (inadequate or marginal%) (Table 1). One third of respondents were widowed, and 42% had access to excellent social support. More than 40% of the respondents indicated they exercised at least 4 times a week, 57% had a midweight BMI, and 3.3% were heavy drinkers. Almost 90% had at least 1 health condition, 4.3% had an ADL limitation, 30% had an IADL limitation, and 27% of respondents described their health as fair or poor.

Rates of depression symptomatology, as measured by the GDS, varied by levels of health literacy (Figure 1). For example, individuals with inadequate health literacy skills were more likely to have mild depression symptoms (18%) than individuals with adequate health literacy skills (8%). Overall, 13% of respondents were classified as depressed according to the GDS (Table 2). Most of the characteristics differed by GDS category (P<.001); for example, 23% of individuals with inadequate health literacy skills were considered depressed compared with 10% of individuals with adequate health literacy. Individuals with at least a high school education or higher income had lower rates of depression compared with individuals with less than a high school education or lower income, respectively. Moreover, individuals who were currently married or had better social support had lower rates of depression. Individuals who exercised at least once a week, were either midweight or overweight, or were moderate or nondrinkers also had lower levels of depression. Individuals who had at least 2 health conditions, had an ADL or IADL limitation, or rated their health as fair or poor were all more likely to be depressed.

The unadjusted logistic regression analysis indicated that individuals with inadequate health literacy had 2.7 times the odds (95% confidence interval, 2.2-3.4) of being depressed compared with individuals with adequate health literacy skills (Table 3). The relation between health literacy remained significant, although diminished, even after adjusting for demographic (model 1), social support (model 2), and health behavior (model 3) factors. However, controlling for health status characteristics (model 4) eliminated the significant relation between health literacy and depression (odds ratio, 1.2; 95% confidence interval, 0.9-1.7). In the full model (model 4), individuals who had less social support, exercised less than 3 times a week, drank heavily, or had poor health status (at least 3 health conditions, ADL limitation, IADL limitation, or fair or poor self-rated health) had significantly higher odds of depression. Assessment of several interactions (cognitive impairment, age, marital status, education, and health status) did not alter these findings.

In our population of Medicare managed care enrollees, the overall rate (13%) of depression was comparable to other studies^5- 8,30,55 of older community residents. Although individuals with inadequate health literacy were more than twice as likely to report depressive symptoms, this was mostly explained by their worse health status. Similarly, although age, education, and income were strongly related to depressive symptoms in bivariate analyses, these variables were not significantly associated after adjusting for social support, health behaviors, and health status. Thus, although low health literacy, advancing age, low education, and low income are strongly associated with depression, the causal pathways for these relations are likely to be through inadequate social support, worse health behaviors, and functional limitations.

Similar to other studies,^{5- 6,15,26,32- 33,56} we found that available social support was a strong predictor of depression. However, unlike other studies,^5- 6 once we controlled for other factors, widowhood was not related to depression. Thus, the relation between widowhood and depression appears to be mediated by lack of tangible social support. This finding is also consistent with other research³² that shows subjective social support as the most important dimension of support related to depression.

Even after controlling for other factors, we found that individuals who are inactive are twice as likely to have symptoms of depression than more active persons. Although this finding has been shown by others,^34- 37 the other study populations were not restricted to older adults (≥65 years). Although we do not know the exact causal relation between exercise and depression in our study, limited evidence suggests that physical activity may protect against the development of depression.^34,36- 37 Similar to 2 other studies,^34,36 our data suggested a possible dose-response relation between levels of physical activity and measures of depression symptomatology. These results are promising in terms of possibly targeting specific factors, such as social support and exercise, that may be amenable to interventions.

To our knowledge, this is the first study that has empirically examined the relation between health literacy and depression symptomatology. Moreover, the rich database of information from community-dwelling, older adults enhances previous research. Despite the strengths of this study, we only have an estimate of depression symptoms at one point in time. It would be important to examine the temporal relation between the characteristics of interest and depression in future prospective studies. For example, we do not know if depression leads to less exercise or if individuals who do not exercise are more likely to have depressive symptoms. We also do not know the exact relation between the GDS measure of depression symptomatology and clinical diagnosis of depression; however, another study⁵⁵ among elderly managed care enrollees indicated poor agreement between these 2 measures. Finally, we cannot determine whether the outcomes of depression differ for those with inadequate health literacy. Research is needed to determine if patients with low health literacy skills are less likely to report depression to their physicians or less likely to comply with treatment regimens.

The managed care environment provides a good setting for developing and implementing interventions to reduce depression among elderly patients. Although approximately 85% of older adults living in the community are seen by a primary care physician at least once a year, there is considerable unrecognized, undiagnosed, and untreated depression among these individuals.^5,30,55,57 The high use of the health care system by elderly patients provides an opportunity for early diagnosis and appropriate treatment by physicians.

Managed care organizations can educate their physicians to assist them in recognizing mental health conditions, increasing the likelihood that patients with these conditions will obtain appropriate treatment. The strong relation between depression and poor health status suggests the need to target interventions to improve the care of mental health and medical illness concurrently.⁵⁷ Plans and physicians should be aware of the prevalence of low health literacy and take steps to assist patients, especially patients with chronic disease, to understand the important aspects of managing their health care. Previous studies^58- 62 with public hospital patients have shown that, compared with literate patients, patients with low health literacy are less likely to know their discharge diagnosis, reasons for taking medications, plans for follow-up diagnostic tests and appointments, and elements of self-management. For patients with poor health literacy skills, alternative methods of education may be necessary to improve their health care knowledge. Treatment plans must be clearly articulated, and physicians should have patients repeat back instructions to be sure they have adequate comprehension. The influence of particular interventions on depression, such as referral to community support contacts and recommendations for an exercise program, suggest promising areas to further evaluate.

Accepted for publication April 28, 2000.

This work was partially supported by a grant from The Robert Wood Johnson Foundation, Princeton, NJ.

Reprints: Julie Gazmararian, MPH, PhD, USQA Center for Health Care Research, 2859 Paces Ferry Rd, Suite 820, Atlanta, GA 30339 (e-mail: julie.gazmararian@aetna.com).