Background
Health literacy is defined by the Institute of Medicine of the National Academies, USA as the degree to which individuals have the capacity to obtain, process and understand basic health information and services needed to make appropriate health decisions [
1]. There are two ways of conceptualising health literacy: a risk factor or an asset. Health literacy as a risk factor fits best in clinical settings. It focuses on improved communication between doctors and patients [
2]. Assets are a set of capabilities needed for everyday life in order to make decisions that affect ones’ health.
Low heath literacy is common in Australia. The 2006 Adult Literacy and Life Skills Survey (Australia) found 60 % of adults to be performing at the lowest levels of health literacy when assessed for prose literacy, document literacy, numeracy and problem solving [
3]. The latest available data show that 41 % of Australians aged 15–74 had a level of health literacy that was adequate or above [
4]. Health literacy (HL) deficits affect half of the overall American patient population, especially the elderly [
5]. Low HL can make it difficult for patients to function effectively in the health care system [
6]. Low health literacy has been consistently associated with poor health outcomes, including poorer health status [
7‐
9], lack of knowledge about medical conditions and related care [
10], lack of engagement with health care providers [
11], decreased comprehension of medical information [
10], mortality [
12], and poorer use of preventive health services [
10,
12], poorer self-reported health [
10], and increased hospitalizations [
10,
12] and higher health care costs [
13,
14].
Health-related quality of life (HRQoL) refers to how individuals subjectively assess their own well-being and their ability to perform physical, psychological, and social functions [
15]. Although there are many studies examining the relationship between HRQoL and heath literacy (HL) among patients with chronic diseases [
10,
16], less is known about this relationship among patients without vascular disease or diabetes or who only have risk factors for these conditions.
The SF-36 and SF-12 are widely used measures of HRQoL. Investigators from numerous countries representing diverse cultures have determined that both measures are sensitive to differences in a number of socio-demographic variables, including gender [
17,
18], age [
17,
19], income [
18,
20,
21], employment [
18‐
20], education [
20,
21] and country of birth [
19]. In addition to patient demographic variables, patient lifestyle risk factors have been found to be associated with HRQoL. Other than low HL [
7,
10], several studies found that smoking [
22], poor diet [
23,
24], alcohol risk [
25], insufficient physical activity [
26‐
28] and overweight [
29‐
31] were negatively associated with HRQoL.
In this study, we investigated the relationship between HL and HRQoL in a large sample of adults without vascular disease or diabetes from four Australian states, using SF-12 version 2 after adjustment for both patient demographic and lifestyle risk factors. We are unaware of previous studies investigating the association between HL and HRQoL after simultaneous adjustment for both patient demographic and lifestyle risk factors in adults without vascular disease or diabetes. Based on the findings and some identified gaps in the literature, the following research questions were posed:
(1)
What are the differences in HL according to patients’ gender, age, place of birth, socio-economic status (home ownership, education and employment) and patient lifestyle risk factors (smoking, diet, alcohol, physical activity and BMI)?
(2)
What is the magnitude of the association between HL and HRQoL after adjustment for patient gender, age, place of birth, socio-economic status and lifestyle risk factors and are there any clinically significant differences in HRQoL among patient subgroups?
(3)
Are there any clinically significant differences in HRQoL between levels of each categorical independent variable within low or high HL patients?
Results
The mean age was 55.5 years (interquartile range (IQR) = 49–62) and 69 % were females.
The low and high HL comparison for demographic characteristics and risk factors of patients and practice characteristics is presented in Table
1. The HL score was available for more than 98 % of patients. Low HL patients tended to be younger (40–54 years) than high HL patients (50 % vs 38 %,
P = 0.001). High HL patients tended to be retired (24 % vs 14 %,
P = 0.003) and 5 % more owner-occupiers (
P = 0.030) than low HL patients. Low HL patients were more likely to be smokers (12 % vs 6 %,
P = 0.005), to have insufficient physical activity (63 % vs 47 %,
P < 0.001) and be overweight (68 % vs 52 %,
P < 0.001). The distribution of practice characteristics over the low and high HL groups was similar (Table
1).
The overall means of PCS-12 scores for low and high HL were 46.2 (SD = 12.2) and 51.9 (SD = 7.9) respectively (Table
2). Similarly, overall mean MCS-12 scores were 47.3 (SD = 11.0) for low HL and 54.4 (SD = 6.9) for high HL. The two summary scores were available for 94 % of patients. Table
2 shows the differences between PCS-12 or MCS-12 scores of low and high HL patients for the subcategories of patient and practice characteristics and patient risk factors (Table
2). Low HL patients reported poorer physical health than high HL for all categories of independent variables. Similarly, low HL patients reported poorer mental health than high HL for all categories of independent variables except for alcohol risk group (Table
2). Among low HL patients, those who were female, older, not well educated, unemployed, a smoker, overweight or had insufficient physical activity were likely to have lower physical health. The differences in physical health between less well educated and well educated (43.0 vs. 48.9, effect size = 0.50) and unemployed and employed (38.5 vs 48.7, effect size = 0.92) were the only clinically significant differences for low HL patients (Table
2). The education and employment remained significant in predicting physical health after adjustment for confounding effects with multilevel analysis for low HL patients (Table
5). Similarly, among high HL patients those who were less well educated, unemployed, retired, overweight or had insufficient physical activity were likely to have lower physical health. Only the difference in physical health between unemployed and employed (49.2 vs. 53.3, effect size = 0.55) was clinically significant for high HL patients (Table
2). The employment remained significant in predicting physical health after adjustment for confounding effects for high HL patients (Table
5). Among low HL patients, those who were younger, unemployed, and had insufficient physical activity tended to have lower mental health (Table
5). Similarly among high HL patients those who were younger tended to have lower mental health (Table
2). All the differences in mental health were not clinically significant. The above effects were not adjusted for confounding effects.
The associations between low and high health literacy and PCS-12 and MCS-12 are presented in the multivariate multilevel regression analyses in Tables
3 and
4 and separately for patients with high and low health literacy in Table
5.
Patients with low health literacy were likely to have lower physical health with a large clinically significant effect size of ≥ 0.56 (B (regression coefficient) ≤ - 5.4,
P < 0.001) and lower mental health with a large clinically significant effect size of ≥ 0.78 (B ≤ −6.4,
P < 0.001) compared to those with higher HL after adjustment for confounding factors (Tables
3 and
4). After accounting for confounding factors, regression coefficients for lower HL in the main model were only marginally increased for PCS-12 (from −5.57 to −5.35) and for MCS-12 (from −7.02 to −6.43). This shows that other patient factors had negligible influence on the strong association between HL and PCS-12 or HL and MCS-12. MCS-12 scores increased with age (
P < 0.05). Gender or home ownership were not significantly associated with either component score. Patients who were employed (effect size = 0.75, B = 6.0,
P < 0.001) or retired (effect size = 0.35, B = 3.2,
P < 0.05) were likely to have higher PCS-12 scores. Similarly, patients who were employed (effect size = 0.23, B = 2.4,
P < 0.05) or retired (effect size = 0.55, B = 2.9,
P < 0.05) tended to have higher MCS-12 scores than unemployed. Employment interacted with HL for PCS-12 and MCS-12 (Tables
3 and
4). Education was positively associated with PCS-12 (effect size = 0.34,
P < 0.05). Older patients (55–69 years) tended to have higher MCS-12 scores than younger patients (40–54, years) (effect size = 0.33,
P < 0.05). However retirement and age were no longer significant in the separate analysis of high HL patients (Table
5).
Patients with insufficient physical activity were more likely to have a lower physical health (effect size = 0.42, B = -3.1,
P < 0.001) and lower mental health (effect size = 0.37, B = −2.6,
P < 0.001) than those with sufficient physical activity. Being overweight had a negative effect on PCS-12 (effect size = 0.41, B = −1.8,
P < 0.05). Poor diet or alcohol risk were not associated with PCS-12 scores or MCS-12 scores (Tables
3 and
4). There was an interaction between smokers and HL for PCS-12 and physical activity and HL for MCS_12. Smokers had a negative effect on physical health (Tables
3 and
5) and physical activity had a negative effect on mental health (Tables
4 and
5) for low HL patients. Practice characteristics were also not associated with either PCS-12 or MCS-12 scores.
Percentage of variance explained
At the practice level (level 2), 72 % (of which HL explained 16 %) and 100 % (of which HL explained 84 %) of the practice variances in PCS-12 and MCS-12 were explained respectively by the variables used in the main model (Tables
3 and
4). At the patient level (level 1) the variance in PCS-12 explained was 20 % of which 9 % was explained by HL and the remaining 11 % was due to all other independent variables (Table
3). Similarly, at the patient level (level 1) the variance in MCS-12 scores explained was 19 % of which 13 % was explained by HL and the remaining 6 % was due to all other independent variables (Table
4). Remarkably, HL accounted for 45 and 70 % of the total between patient variance explained in PCS-12 and MCS-12 respectively.
Discussion
To the best of our knowledge, this is the first study to investigate the impact of HL on HRQoL after simultaneous adjustment for both patient demographics and lifestyle risk factors in adults without chronic vascular disease or diabetes. We found two studies which examined the effect of HL on PCS-12 and MCS-12 after adjustment for confounding factors in patients with chronic conditions. The first study [
50] comprised of 3260 elderly (≥65) patients with ten chronic health conditions (Heart attack, Stroke, Asthma, Cancer, Diabetes etc.). The effects of inadequate HL (Reference: adequate) on PCS-12 (B = −2.53,
P < 0.001) and MCS-12 (B = 1.41,
P < 0.001) were weaker than those of our study. The second study [
51] included 1581 men with newly diagnosed clinically localized prostate cancer from a population based study. In this study, patients with higher health literacy had better MCS-12 (B = 2,
P < 0.04) than low HL. However, the association between HL and PCS-12 was significant only in the bivariate model. Again the difference of HRQoL between low and high HL patients was weaker than those of our study. This may be due to the lower HRQoL of chronically ill patients and that they have less variation when compared to better HRQoL of patients without vascular disease or diabetes. For example, the overall average of PCS-12 in the current study was 49.0 compared to 42.4 in Australian patients with diabetes or cardiovascular disease [
19].
This study provides comprehensive data on the association between the HL and self-rated physical and mental health of Australian adults without previously diagnosed vascular disease or diabetes. Our findings demonstrated that lower HL patients reported clinically significant poorer physical health and mental health than higher HL patients with HL accounting for 45 and 70 % of the total between patient variance explained in PCS-12 and MCS-12 respectively. A potential explanation of these negative associations between low health literacy and physical and mental domains of HRQoL may be that low health literacy can make it difficult for patients to effectively navigate the health care system [
6]. Patients with lower HL may have difficulties with complex health tasks and ability to seek and understand health information [
40], have limited access to the health care [
40], lack engagement with health care providers [
11], and have poorer uptake of preventive health care [
10,
12]. Patients with lower health literacy tend to have difficulties with communication, which prevent them from asking questions, clearly expressing their concerns, emotions, and needs to providers and seeking additional services such as support for mental health [
10,
52]. Being able to recognise low health literacy is important in general practice as there is good [
53] or mixed [
54] evidence that tailoring health related communication to those with low health literacy can improve health outcomes.
Few studies have examined the impact of HL and its interaction with socio-demographics factors on HRQoL. For PCS-12, our study showed some significant interaction effects between HL and age, HL and education (borderline with
P = 0.06), HL and employed or retired, and HL and smokers (Table
3). Similarly for MCS-12, interaction effects between HL and employment, HL and age (borderline), HL and physical activity were significant (Table
4). These interaction effects for PCS-12 and MCS-12 were explored in Table
5. Among lower HL patients, educational attainment was positively associated with physical heath. For high HL patients, this association between education and physical health was not significant after adjustment for confounding factors (Table
5). This suggests that the combination of lower HL and lower educational attainment is particularly important in physical health. Practitioners need to be alert to problems with communication and adherence to treatment plans and access to services as these patients encounter double disadvantage. Education had no association with the mental health of low or high HL patients (Tables
2,
4 and
5).
There are no other studies we are aware of that examined the impact of the association between health literacy and employment in predicting HRQoL. The interaction between HL and employment status showed that the negative impact of unemployment was greater in low HL patients than high HL. We found that the negative impact of low health literacy in unemployed patients with 8.7 lower PCS-12 than employed and 6.0 lower PCS -12 than retired (Tables
3 and
5). Similarly, unemployed patients had 4.2 lower MCS-12 than employed for low HL patients (Tables
4 and
5). Consistent with other research, lower socio-economic groups reported lower PCS-12 and MCS-12 [
18,
20,
55].
The finding that mental health was higher in the older age group is consistent with our previous research [
19]. The older age had a negative effect on PCS-12 and positive effect on MCS-12 for low HL patients.
We found that almost half the patients in this study met the criteria for insufficient health literacy. This is consistent with studies in primary care in other developed countries [
56] and the prevalence reported in the Australian community [
57]. Patients with low HL were more likely to be smokers, report being overweight or obese, and exercise inadequately. In multivariate analysis, inadequate physical activity tended to have a negative effect on physical health (Table
5). Being overweight also had a negative effect on physical health (Table
5). This supports findings from previous studies demonstrating associations between HRQoL and physical activity [
31,
58] or BMI [
31,
59]. The analyses showed that life style risk factors interacted with HL (Tables
3 and
4). Low HL smokers were likely to have 6.4 lower PCS-12 than non-smokers and low HL patients with insufficient physical activity tended to have 5.1 lower MCS-12 after adjustment for confounding factors (Tables
3,
4 and
5). Our results extend findings from previous studies by demonstrating the possible beneficial association of regular physical activity, non-smoking or normal weight with HRQoL is more relevant to low HL patients.
There are a number of limitations to this study. Patients identified by the practice as being unable to read English, with psychosis, cognitive impairment, diagnosed substance abuse problems or severe mental illness were excluded from the study. Others may have self excluded in their response to the written invitation to participate, particularly patients with low health literacy who may not have understood why it was important to be involved, or what might have been required of them. It is possible that non-respondents might have assessed their physical and mental health differently from those who responded and had lower health literacy. The response rate was also low (15 % compared to 30 % in our previous study [
60], possibly due to the recruitment method or because patients no longer wished to attend the practice). The patients responding to the survey were older and more were female compared to patients from the clinical audit in the same practices. Participants of this study were predominantly not from low socio-economic groups. The majority of them were Australian-born, had a degree or diploma, were in paid employment and owned their home. It is possible that due to selection bias more high health literacy patients responded and this could have diluted the effects of this study.
The strengths of the study include a broad sample of 739 patients from 30 general practices in four states of Australia, the adjustment for confounding for both patient (demographic and lifestyle risk factors) and practice variables and the correction for practice level clustering with multilevel modelling.
Acknowledgements
The following are members of the PEP (Preventive Evidence into Practice) Partnership Group:
Professor Mark Harris (Chief Investigator and Study Lead) University of NSW Australia;
Professor Nicholas Zwar (Chief Investigator) University of NSW Australia;
Associate Professor John Litt (Chief Investigator) Flinders University Australia;
Professor Danielle Mazza (Chief investigator) Monash University Australia;
Professor Mieke van Driel (Chief Investigator) University of Queensland, Australia;
Professor Richard Taylor (Chief Investigator) University of NSW Australia;
Professor Grant Russell (Chief investigator) Monash University Australia;
Professor Chris Del Mar (Associate Investigator) Bond University Australia;
Dr Jane Lloyd (Associate Investigator) University of NSW Australia;
Associate Professor Jane Smith (Associate Investigator) Bond University Australia;
Associate Professor Elizabeth Denney-Wilson (Associate Investigator) University of Technology Australia;
Dr Rachel Laws (Associate Investigator) Deakin University Australia;
Dr Teri Snowdon, Ms Helen Bolger-Harris and Mr Stephan Groombridge (Supporting Partner) Royal Australian College of General practitioners (RACGP);
Dr Stan Goldstein and Ms Teresa Howarth (Supporting Partner) Bupa Health Foundation Australia;
Dr Nancy Huang and Ms Jinty Wilson (Supporting Partner) National Heart Foundation of Australia.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
UJ designed the paper, conducted the data analysis and drafted the manuscript. MH supervised each process of the study. SP coordinated the patient survey, collated responses, compiled and cleaned the patient data set. MH, SP, JL, MvD, DM, CDM, JL, JS, NZ and RT contributed to the revision of the manuscript. All authors read and approved the final manuscript.