The authors have declared that no competing interests exist.
Conceived and designed the experiments: SS JW CvW MW. Analyzed the data: SS CvW LC MW. Wrote the paper: SS LC JW CvW MW. Data acquisition: LC MW. Data interpretation: SS LC JW CvW MW.
There is ongoing debate on whether health literacy represents a skill-based construct for health self-management, or if it also more broadly captures personal ‘activation’ or motivation to manage health. This research examines 1) the association between patient activation and health literacy as they are most commonly measured and 2) the independent and combined associations of patient activation and health literacy skills with physical and mental health.
A secondary analysis of baseline cross-sectional data from the LitCog cohort of older adults was used. Participants (n = 697) were recruited from multiple US-based health centers. During structured face-to-face interviews, participants completed the Test of Functional Health Literacy in Adults (TOFHLA), the Patient Activation Measure (PAM), the SF-36 physical health summary subscale, and Patient Reported Outcomes Measurement Information Service (PROMIS) short form subscales for depression and anxiety.
The relationship between health literacy and patient activation was weak, but significant (r = 0.11, p<0.01). In models adjusted for participant characteristics, lower health literacy was associated with worse physical health (β = 0.13, p<0.001) and depression (β = −0.16, p<0.001). Lower patient activation was associated with worse physical health (β = 0.19, p<0.001), depression (β = −0.27, p<0.001) and anxiety (β-0.24, p<0.001).
The most common measures of health literacy and patient activation are weakly correlated with each other, but also independently correlated with health outcomes. This suggests health literacy represents a distinct skill-based construct, supporting the Institute of Medicine’s definition. Deficits in either construct could be useful targets for behavioral intervention.
The field of health literacy has expanded over the last two decades
This rapid growth has led to new definitions and interpretations of the term itself
People who are motivated and confident in their ability to use their knowledge and skills are more likely to be active participants in maintaining and improving health. The term ‘patient and consumer activation’ has come to represent this, and is specifically defined as those who ‘…
In a non-clinical national sample, the impact that patient activation could have on population health was demonstrated, with fewer than half (41%) of the population reaching the highest level of patient activation
Despite interest in expanding the meaning of health literacy to include factors such as patient activation
To date, few studies have investigated the relationship of both patient activation and health literacy with health outcomes
The present study reports a secondary analysis of baseline cross-sectional data from the LitCog cohort. Details of the main outcomes from this data and detailed procedures are available elsewhere
The baseline phase of LitCog recruited participants aged 55–74 from one primary care clinic and three federally qualified health centers in Chicago, Illinois. Recruitment ran from August 2008 through October 2010. A sample of 1768 eligible patients were reached by research staff and invited to participate in the study. Initial screening deemed 192 subjects as ineligible due to severe cognitive or hearing impairment, limited English proficiency, or not being connected to a clinic physician (defined as <2 visits in two years). In addition, 738 refused, 14 were deceased, and 20 were eligible but had scheduling conflicts. The final sample included 804 participants, giving a cooperation rate of 56% based on American Association for Public Opinion Research guidelines. A sub-sample of participants (n = 719) were also asked to complete a measure of patient activation. Data from these participants will be reported here. There were no missing data for gender or age. Participants were excluded from analyses if they had missing data for race, comorbidities, health literacy or patient activation (n = 22); giving a final sample for analyses of 697 patients.
Participation took place across two days, however all measures reported here were ascertained on the first day. Participants completed socio-demographic items, a health literacy measure, a measure of patient activation, and a series of health status measures. Participants were compensated $100 for their time. The Northwestern University Institutional Review Board approved the study procedures and all participants gave informed consent.
The study was approved by the Northwestern University’s Institutional Review Board. Participants provided written informed consent to participate in the study, which included permission to use the data for research.
Health literacy was assessed using the Test of Functional Health Literacy in Adults (TOFHLA). The TOFHLA is an objective measure of health literacy which uses materials similar to those that patients encounter in healthcare situations
To assess activation, the shortened version of the PAM was used
We assessed physical health using the SF-36 physical health summary subscale
Anxiety and depression were measured using the Patient Reported Outcomes Measurement Information Service (PROMIS) short form subscales
Participant characteristics were recorded. These included age, gender, marital status (married, unmarried) income (<$10,000, $10–24,999, $25–49,999, >$50,000) ethnicity (black, white, other) and comorbidities (0, 1–2, 3+).
One-way Analysis of Variance (ANOVA) was used to compare mean performance on the TOFHLA and PAM by participant characteristics. Associations between patient activation, health literacy, anxiety, depression and physical health were assessed using Pearson correlations. A series of multivariable linear regression models were conducted to examine the independent associations between health literacy, patient activation, and each of the physical and mental health status measures controlling for age, gender, race, and comorbidity. Standardized regression coefficients are reported throughout. Models were run first adding health literacy (Model 1) or patient activation (Model 2) alone in order to isolate the contributions of each, and then together (Model 3) to examine their combined effects. Outcomes are reported in the order of: (A) physical health, (B) anxiety and (C) depression. For example model A1 reports the association between health literacy and physical health, controlling for participant characteristics. F-tests were used to determine whether the variance explained by each of the models (R2) significantly changed with the addition of the other variable (i.e. Model 1 vs. Model 3 and Model 2 vs. Model 3). The Durbin-Watson statistic was used to investigate the assumption of independence. Normal probability (P-P) plots were used to investigate the normality of error terms and homoscedasticity was tested by observing the scatter plot of the residuals and the predicted value. These checks identified no violations of multiple regression assumptions. All statistical tests were one-tailed and a significance level of p<0.05 was set for all analyses. SPSS version 18.0 was used throughout.
The sample is described in
Characteristic | N | % |
Male | 226 | 32.4 |
Female | 471 | 67.6 |
55–59 | 221 | 31.7 |
60–64 | 213 | 30.6 |
65+ | 263 | 37.7 |
Married | 308 | 44.4 |
Unmarried | 386 | 55.6 |
Missing | 40 | 5.7 |
<$10,000 | 85 | 12.2 |
$10–24,999 | 132 | 18.9 |
$25–49,999 | 98 | 14.1 |
>$50,000 | 342 | 49.1 |
Black | 309 | 44.3 |
Other | 52 | 7.5 |
White | 336 | 48.2 |
0 | 94 | 13.5 |
1–2 | 390 | 56 |
3+ | 213 | 30.6 |
Inadequate | 94 | 13.5 |
Marginal | 124 | 17.8 |
Adequate | 479 | 68.7 |
Level 1 | 20 | 2.9 |
Level 2 | 27 | 3.9 |
Level 3 | 68 | 9.8 |
Level 4 | 582 | 83.5 |
(* = missing data).
Higher levels of health literacy were found among females (p = 0.01,
TOFHLA | Patient Activation Measure | |||
Mean (SD) | p-value | Mean (SD) | p-value | |
.011 | .828 | |||
Male | 74.18 (19.59) | 78.78 (13.25) | ||
Female | 77.54 (14.43) | 79.03 (14.61) | ||
.054 | .019 | |||
55–59 | 77.62 (16.15) | 76.97 (15.16) | ||
60–64 | 77.61 (15.63) | 78.92 (13.99) | ||
65+ | 74.53 (16.94) | 80.62 (13.28) | ||
.000 | .024 | |||
Black | 67.86 (18.01) | 77.46 (15.38) | ||
Other | 73.19 (9.29) | 78.05 (13.84) | ||
White | 84.85 (14.10) | 80.45 (12.90) | ||
.000 | .000 | |||
0 | 82.19 (13.96) | 81.44 (14.36) | ||
1–2 | 78.01 (14.57) | 80.23 (13.63) | ||
3+ | 71.06 (18.81) | 75.49 (14.51) |
Both lower health literacy and patient activation were associated with worse physical health in univariate analyses (health literacy: r = 0.30, p<0.001; patient activation: r = 0.29, p<0.001). In multivariable models controlling for age, race and comorbidities, lower health literacy was related to worse physical health (Model A1: β = 0.15, p<0.001,
Model A1 | Model A2 | Model A3 | |
β | β | β | |
−.07* | −.06 | −.07* | |
55–59 | – | – | – |
60–64 | −.04 | −.05 | −.05 |
65+ | .07 | .02 | .04 |
Black | −.12** | −.19*** | −.12** |
Other | −.01 | −.03 | −.01 |
White | – | – | |
0 | – | – | – |
1–2 | −.17*** | −.16*** | −.16*** |
3+ | −.59*** | −.57*** | −.56*** |
.15*** | – | .13*** | |
– | .20*** | .19*** |
Note: * = p<.05; ** = p<.01; *** = p<.001.
Model A1 – Gender, age, race, comorbidities, health literacy (F(8, 688) = 47.29***, R2adj = .347).
Model A2 – Gender, age, race, comorbidities, PAM (F(8, 688) = 52.04***, R2adj = .370).
Model A3 – Gender, age, race, comorbidities, health literacy and PAM (F(9, 687) = 48.74***, R2adj = .382).
For each outcome, in order to test whether including both health literacy and patient activation significantly improved the explanatory power of Models A1 and A2, the R2 change statistic was observed. There were significant differences between models A1 and A3 (R2 change = 0.04; F(1,687) change = 39.28, p<0.001) and between models A2 and A3 (R2 change = 0.01; F(1, 687) change = 14.25, p<0.001), indicating that including both health literacy and patient activation significantly improved explanatory power in the physical health outcome compared to either one alone. Interactions were tested but found to be non-significant.
Lower health literacy and lower patient activation were both significantly associated with greater anxiety in univariate analyses (health literacy: r = −0.11, p = 0.005; patient activation: r = −0.29, p<0.001). In multivariable analyses controlling for age, race and comorbidities, lower health literacy was independently associated with greater anxiety (Model B1: β = −0.09, p = 0.035;
PROMIS Anxiety | PROMIS Depression | |||||
Model B1 | Model B2 | Model B3 | Model C1 | Model C2 | Model C3 | |
β | β | β | β | β | β | |
.06 | .06 | .07 | .01 | −.01 | .01 | |
55–59 | – | – | – | – | – | – |
60–64 | −.07 | −.05 | −.05 | −.04 | −.02 | −.02 |
65+ | −.19*** | −.15*** | −.16*** | −.20*** | −.14*** | −.16*** |
Black | −.07 | −.04 | −.07 | −.02 | .05 | −.03 |
Other | .07 | .07* | .06 | .08* | .10** | .08* |
White | – | – | – | – | – | – |
0 | – | – | – | – | – | – |
1–2 | .13* | .12* | .12* | .11* | .11* | .10* |
3+ | .36*** | .33*** | .32*** | .39*** | .36*** | .34*** |
−.09* | – | −.07 | −.17*** | – | −.16*** | |
– | −.24*** | −.24*** | – | −.27*** | −.27*** |
Note: * = p<.05; ** = p<.01; *** = p<.001.
PROMIS Anxiety:
Model B1 – Gender, age, race, comorbidities, health literacy (F(8, 688) = 11.59***, R2adj = .109).
Model B2 – Gender, age, race, comorbidities, PAM (F(8, 688) = 17.33***, R2adj = .158).
Model B3 – Gender, age, race, comorbidities, health literacy and PAM (F(9, 687) = 15.80***, R2adj = .161).
PROMIS Depression:
Model C1 – Gender, age, race, comorbidities, health literacy (F(8, 688) = 19.05***, R2adj = .172).
Model C2 – Gender, age, race, comorbidities, PAM (F(8, 688) = 25.92***, R2adj = .223).
Model C3 – Gender, age, race, comorbidities, health literacy and PAM (F(9, 687) = 25.23***, R2adj = .239).
Adding patient activation to anxiety Model B1, which included health literacy alone, significantly improved its explanatory power (R2 change = 0.05; F(1, 687) change = 43.73, p<0.001). However, there was no significant difference between Models B2 and B3 (R2 change = 0.004; F(1, 687) change = 3.15, p = 0.077), indicating that health literacy did not explain a significant amount of additional variance in anxiety after patient activation had been entered in the model. We also tested for an interaction between health literacy and patient activation, and this was not significant.
Similar to anxiety, lower health literacy and patient activation were both significantly related to more depressive symptoms in univariate analyses (health literacy: r = −0.22, p<0.001; patient activation: r = −0.34, p<0.001). In multivariable analyses controlling for age, race and comorbidities, lower health literacy was independently associated with worse depression (Model C1: β = −0.17, p<0.001,
There was a significant difference between models C1 and C3 (R2 change = 0.07; F(1, 687) change = 61.28, p<0.001), and also between models C2 and C3 (R2 change = 0.02; F(1, 687) change = 15.35, p<0.001), indicating that the explanatory power of either health literacy or patient activation was significantly improved when the other was included. No interactions were found in the depression models.
In this sample of older American adults, health literacy and patient activation were independently associated with depression and physical health when included in the same statistical model. Health literacy was not significantly associated with anxiety, and patient activation was the stronger predictor of the two measures for all health outcomes. These findings are in line with previous general population studies
Collectively, these findings suggest health literacy, as it is currently measured by the most widely used assessment tool
Our findings have implications both for the individual treatment of patients, and for large-scale health interventions that affect the public more widely. For example clinicians attending to the health literacy needs of their patients by simplifying treatment regimens and clarifying instructions may be inadvertently missing opportunities to activate their patients. The assumption that an individual with the ‘skill set’ for how to act will automatically adhere to instructions, ignores the ‘mind set’ factors that underpin behavior change. Similarly, a focus on patient activation may fail to acknowledge the difficulties faced by those lacking the adequate ‘skill set’, despite being activated to self-manage their health. From a public health perspective, patient-centered interventions to improve health outcomes may be best served by incorporating elements of both health literacy and activation into their design and evaluation. The strong associations of health literacy
Mechanisms have been suggested through which health literacy and patient activation could be associated with mental and physical health. The strongest evidence suggests individuals with low health literacy find accessing and understanding health information more difficult
A strength of this study was the use of a large, socioeconomically diverse general population sample recruited from multiple sites, including academic and community-based services. Gold standard versions of the most commonly used measures of each of the dependent and independent variables were also used. Furthermore, previous models have typically only included either health literacy or patient activation. This study is also among the first to demonstrate the individual and combined effects of these constructs on important health outcomes, testing a hypothesis that was generated by the discrepancy between key health literacy definitions.
This study had limitations. The cross-sectional nature of the study prohibits causal inferences. Furthermore, estimates of associations were drawn from a sample that had higher levels of both activation and health literacy than normative estimates
The next step for research in this field would be to investigate whether similar effects are apparent in different health domains, such as complex health tasks, self-management and healthcare utilization. This will permit researchers to determine whether the relative importance of each construct varies in different circumstances, allowing specific policy recommendations to be made for each situation. As discussed previously, these findings strongly suggest there may be scope for behavioral scientists to develop a comprehensive measure that assesses both basic skills and activation within a single brief tool.
In conclusion, health literacy and patient activation are weakly correlated with each other, and also make independent contributions to health. Deficits in either domain could be useful targets for behavioral intervention. New measurement strategies are needed to evaluate both constructs and a combined approach may be attractive not only to researchers but also to clinicians who wish to identify patients who need further support. In the meantime, we recommend that health literacy and patient activation be treated as distinct and important constructs warranting assessment in public health and behavioral science research.
We would like to thank Elizabeth Bojarski, Rachel O’Conor, Emily Ross, and Rina Sobel for their determination and effort in recruiting and collecting data for the LitCog Study.