Search strategy

The databases searched were Medline (via PubMed), COCHRANE library, PsycINFO, SCOPUS, Web of Science, Scientific Electronic Library Online (SciELO), Literatura Latino-Americana e do Caribe em Ciências da Saúde (LILACS) and the Banco de Dados em Enfermagem (BDENF). The search strategy entailed the use of Health Descriptors (Descritores da Saúde - DeCS) and Medical Subject Headings (MeSH), in their possible permutations, using the Boolean operators AND/OR: Health Literacy and health personnel. Although the problem (P) was related to older adults with low health literacy, it was decided not to include the descriptor (MeSH) “aged” (entry terms Elderly) in the search because it acted as a limiter. The references of the articles included in the review, and of other relevant reviews, were hand-searched (^{Clement et al, 2009}; ^{Lowe et al, 2013}). The selection process commenced in December 2017 and was finalized in April 2020, where 2000 was defined as the lower limit for publication date. The Medline search strategy (via PubMed) applied is outlined in Table 1.

Study selection

After reading article titles and abstracts, studies were screened according to the eligibility criteria for inclusion: a) original study articles in which health professionals planned, applied, or evaluated interventions for older adults with low health literacy; b) involving populations aged ≥ 50 years and/or with mean age ≥ 60 years. The age bracket for older adults was broadened to reflect the importance of health promotion and disease prevention in individuals aged under 60 years (^{WHO, 2015}); c) randomized clinically-controlled trials; d) published in Portuguese and English; and for exclusion: e) Studies which involved the planning, application or evaluation of health education interventions but did not draw on the health literacy concept, despite including older participants; f) non-clinically-controlled trials, discussion articles, editorials, summaries, notes, books, book chapters, abstracts presented at annals of congresses, dissertations, theses, qualitative studies, bibliographic studies, documental (desk)-based studies, case studies, and surveys.

Study selection was performed by searching the scientific databases, identifying potential studies based on title, abstract and key words. The articles retrieved were then screened by reading the full texts. Selection differences were discussed and resolved by consensus. When no consensus could be reached, a third researcher was consulted to deliberate on the issue.

Quality assessment

The Cochrane Handbook for Systematic Reviews of Interventions was used to assess the quality of the articles (^{Higgins, 2011}) and the following study characteristics were rated: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, description of losses and exclusions, and intention-to-treat analysis. Studies lacking a clear description of these characteristics were classified as unclear or not reporting these items.

Data analysis

The meta-analysis was performed using the random effects model and measures of effect were based on post-intervention values. For the meta-analysis, a comparison of health literacy strategies versus control group or usual care in the assessment of glycemia and hypertension outcomes was carried out. An alpha value = 0.05 was adopted as statistically significant. The heterogeneity statistic of treatment effect among studies was assessed using Cochran´s Q-test and the inconsistency test (I2). All analyses were performed using the software Review Manager 5.1 (Cochrane Collaboration).

Description of studies

A synthesis of caractheristics of the studies, including information on setting and methodology, is given in Table 2.

Of the studies that carried out an assessment of health literacy, 2 (10.53%) used the results to select the study sample, comprising older adults with low health literacy (^{DeWalt et al., 2006}; ^{Lam et al., 2017}). Low or limited health literacy levels were found in most of the studies (^{Bosworth et al., 2008}; ^{DeWalt et al., 2012}; ^{Graumlich et al., 2016}; ^{Kiser et al., 2012}; ^{Kripalani et al., 2012}; ^{Miller et al, 2018}; ^{Protheroe, 2016}; ^{Seligman et al., 2005}). A longitudinal study found major disparities in scores among individuals with different age, skin color and gender (^{Verney et al., 2019}).

Figure 1 Flowchart of Study Selection Process 

Intervention strategies

The interventions employed different technical resources and methodologies (Table 3). Approaches included cognitive training sessions (^{Verney et al., 2019}), an individual self-management educational intervention (^{Kiser et al., 2012}), use of the teach-back method (^{Liu et al., 2018}) and FamLit (^{Wu et al., 2019}), delivery of written material containing practical accessible information about the health/disease status of the target subjects of the intervention, complemented by consultations with professionals (^{DeWalt et al., 2006}; ^{Rudd et al., 2009}), use of graphically-enhanced interventions (^{Kripalani et al., 2012}), videos accompanying booklets (^{Di Palo et al., 2017}; ^{Ferreira et al., 2005}; ^{Muir et al., 2012}; ^{Wilson et al., 2010}), and use of telemedicine or telephone for educational intervention or support/follow-up (^{Bosworth et al., 2008}; ^{Bove et al., 2013}; ^{DeWalt et al., 2012}; ^{Muir et al., 2012}; ^{Protheroe, 2016}). Some interventions employed tailored interventions via an iPad app, mPATH-CRC, as a tool for improving Colorectal Cancer Screening rates (^{Miller et al., 2018}) and the digital Medtable program (^{Graumlich et al., 2016}) and Talking Pill Bottles (^{Lam et al, 2017}), both for pharmacy care. One study used the group method to promote interactive exchanges, scenario simulations and other activities (^{Zhang et al., 2019}).

The period for which strategies were applied ranged from short-term (1-3 months), medium-term (6-12 months) to long-term (24-36 months and 10 years).

Regarding intervention follow-up, most studies assessed results before and immediately after interventions, where follow-up periods for 8 (36.36%) studies were 1-, 6- or 12-months post-intervention, whereas 2 (9.09%) interventions involved ongoing assessments of results every 30 days. One longitudinal study involved assessments shortly after intervention and again at one, two, three, five- and ten-years post-intervention.

Concerning the aims of interventions, 11 (50%) studies focused on older adults diagnosed with heart disease, 3 (13.64%), 3 involved interventions for older patients with diabetes mellitus and 3 (13.64%) focused on cancer. In addition, there were interventions aimed at patients with glaucoma, rheumatoid arthritis, chronic obstructive pulmonary disease and at improving level of health literacy. The studies which applied interventions in older adults with heart disease sought to improve health outcomes by optimizing treatment, behavior and control of blood pressure, self-management and reducing rates of treatment or readmission. The strategies planned for older adults with diabetes mellitus sought to improve self-management in care and in controlling blood sugar levels, encouraging behavioral changes through healthy life-style choices and self-management of the condition (^{Protheroe, 2016}), promoting better use of prescribed medications (^{Graumlich et al., 2016}) or exploring physician-patient communication (^{Seligman et al., 2005}). The studies on cancer employed strategies for increasing screening and self-care behaviors. However, primary, and secondary outcomes differed methodologically in terms of the means of assessment adopted, where authors applied published scales, specific instruments for the condition investigated, or used instruments they had devised and published. Adherence to medication and treatment was the most assessed outcome class (50.0%). Other outcomes assessed included health knowledge held (22.73%), self-efficacy (22.73%), quality of life (18.18%), physical and mental health (9.09%) and health behaviors (27.27%). Patient-professional communication, lifestyle and social support were measured (4.55%), as well as perception of the disease and self-care (9.09%), cognitive domains, memory, and reasoning (4.55%). Satisfaction was rated by 5 studies (22.73%), one of which measured patient satisfaction with pharmacy services and total direct costs. Concerning clinical outcomes, 3 studies assessed heart failure measures (13.64%), 3 blood pressure values (13.64%), 3 diabetes control (13.64%), while 2 assessed mortality and all-cause readmission for heart failure (9.09%).

Effects of Interventions

With regard to results, of the 22 studies reviewed, 3 (13.64%) reported statistically significant differences in primary outcome measures favoring the intervention groups (^{Ferreira et al., 2005}; ^{Liu et al., 2018}; ^{Wilson et al., 2010}) and 5 (22.73%) found significant results for 1, 2 or 3 secondary outcomes (^{Bosworth etal., 2005}; ^{DeWalt et al., 2006}; ^{Graumlich et al., 2016}; ^{Protheroe, 2016}; ^{Zhang et al., 2019}). Thirteen studies (59.09%) reported no statistically significant differences between groups. One study (4.55%) (^{Murray et al., 2004}) observed improvement in primary outcomes at the time of the intervention, but loss of these gains during the post-intervention period.

Although not all outcomes differed statistically between intervention and control groups in the studies, most interventions had a positive effect on intragroup scores, as evidenced by comparisons of baseline versus post-intervention values.

Regarding significant results of interventions reviewed, a study investigating outcome adherence to colorectal cancer screening promoted a significant increase in disease screening rates (^{Ferreira et al., 2005}). In the intervention group, 41.3% of patients underwent fecal occult blood tests, flexible sigmoidoscopy, or colonoscopy versus 32.4% of control group subjects (P<0.003). According to the authors, this result confirmed that a health care provider-directed intervention based on training workshops and individualized feedback on screening rates significantly increased adherence to colorectal cancer screening among older adults treated at a general clinic of a large urban area, supporting wider implementation of this type of intervention. A study addressing the outcome self-care behaviors in managing radiation side effects in prostate cancer found that self-care behaviors significantly increased in the intervention group compared to the control group from baseline to 6 months (P=0.05) (^{Wilson et al., 2010}). Another study, assessing patients´ general health status, self-management skills and disease-management abilities identified an interaction between time and group (P<0.001) and statistically significant differences between the 2 groups for some variables in controlling behavioral risk factors such as drug compliance, physical activity, and diet (P<0.05) (^{Zhang et al., 2019}).

Notable significant results for intervention groups include some actions centered on patients with heart failure, such as a study which applied the American College of Cardiology Patient Navigator Program (^{Di Palo et al., 2017}). Specific education for HF was associated with statistical difference (P=0.0002), and documented education increased by 59% in the intervention group. There was a statistically significant difference in HF specific education (P=0.0002) and documented education increased by 59.0% among the intervention group. In addition, there was a statistically significant increase in scheduled 14-day follow-up prior to discharge in the intervention group (P=0.0044). The results showed that, of the scheduled appointments, patients included in the Navigator program were more likely to follow-up with a cardiologist (56.8%) than the control group (18.6%). The use of this program in the discharge process resulted in a 53.2% decrease in the number of patients discharged without a scheduled post-discharge follow-up visit. In another study, comparing the efficacy of a heart failure self-management program (^{DeWalt et al., 2006}) results in the intervention group relative to the control, although only statistically significant for self-care behavior in terms of weight measurement, showed intragroup improvements for knowledge (mean difference in score improvement of 12 percentage points) (P<0.001) and self-efficacy (mean difference in score improvement of 2 points) (P=0.0026). A difference was also found for mortality and hospitalization rates, which were lower among intervention group patients.

The intervention group also had improved results in a study assessing knowledge and blood-pressure reading (^{Lam et al., 2017}) showing significantly higher scores at day 90 post-intervention compared to baseline (P<0.001). Regarding self-efficacy of intervention participants, the study which tested the effect of a product launched 12 years ago in the USA called “Talking Pill Bottles” (^{Lam et al., 2017}), consisting of a device with a base sized to accommodate most common prescription and a 60-second recording capacity, found that 101 of the 134 participants reported the highest level of confidence when faced with a situation of not having someone there to remind them to take their medication (M = 2.73, maximum score 3.0).

Self-efficacy, a secondary outcome of a study which applied an intervention for improving blood pressure control (^{Bosworth et al., 2005}) showed a significant increase after treatment in the intervention group from baseline to 6-month follow-up, whereas the usual care group shown a decrease (P=0.007). Another study, aimed at enhancing health outcomes in arthritis patients, assessed the self-efficacy variable both at baseline and 6-months post-intervention (^{Rudd et al., 2009}). Results showed improvement in the intervention group from baseline to 12 months compared to the standard care group. Differences between the groups were statistically significant at 6 and 12 months (P=0.05). By contrast, a study evaluating post-visit self-efficacy scores of diabetes patients found no significant difference between intervention and control groups (P=0.60) (^{Seligman et al., 2005}). According to the authors, this lack of a difference in self-efficacy scores between the intervention and control group patients suggests that the management strategies physicians employed need to be reinforced over several patient visits.

Regarding health literacy level, one of the studies aimed at improving literacy found a significant result for the total health literacy score on 4 dimensions (health knowledge, health beliefs, health behaviors and health skills) in the intervention group relative to the control group (P<0.005) (^{Liu et al., 2018}). Although detecting differences in health literacy in subgroups was not necessarily a goal of the other studies, some included analysis of subgroups according to literacy. For example, the study on self-management for heart failure demonstrated that, among other outcomes, the lower rate of hospitalization or death in the intervention group was greater for patients with low literacy (^{DeWalt el al., 2012}). In a study aimed at reducing heart failure readmission rates, the intervention was tailored to patient´s health literacy and social needs (DiPalo et al., 2017). Results showed a strong correlation between the education intervention and readmission rate, with a lower rate of readmission in the group receiving the intervention which almost reached statistical significance (P=0.15).

In the study of ^{Muir et al. (Muir et l., 2012}) aimed at improving glaucoma medication adherence, subjects with inadequate or marginal health literacy skills were more likely to report the presence of a physical disability which made proper drop instillation more difficult (P=0.020). However, the number of days without medication in the 6 months following enrollment was similar for the control and control groups. For each literacy level, the number of days without medicine was fewer in the intervention than in the control group, and the magnitude of the difference increased as literacy decreased.

^{Ferreira et al. (2005}), investigating colorectal cancer screening, found that among patients with health literacy skills less than ninth grade, screening was completed by 55.7% of patients in the intervention group versus 30% of controls (P< .01/ p=0.002). However, the study by ^{Graumlich et al. (2016}) found that patients who received the intervention had greater knowledge about indications for medications, irrespective of literacy status; the study by ^{Protheroe et al. (2016}) found evidence that the health literacy intervention can have a positive impact on the mental health of patients, while Kiser et al. (2002) found, in the literacy subgroup, greater improvement among COPD patients with low literacy compared to those with higher literacy. The results of the study by ^{Lam et al. (2017}) suggested that the intervention applied was well accepted by patients with low health literacy.

Risk of Bias

Regarding risk of bias (Figure 2), 86.36% of studies reported performing random sequence generation, under half of studies (42.1%) reported blinding of participants and 31.57% carried out blinding of outcome assessment. Description of losses and exclusions were included in 54.54% of the studies, whereas intention-to-treat statistical analysis was present in only 5 studies (27.27%).

Meta-analysis

The studies reviewed were highly heterogeneous for intervention strategies and outcomes analyzed. Consequently, the meta-analysis included only 3 studies comparing strategies for health literacy strategies versus control groups or usual care for the glycemia outcome (^{Graumlich et al., 2016}; ^{Protheroe, 2016}; ^{Seligman et al., 2005}) and 2 studies comparing health literacy strategies versus control group or usual care for the hypertension outcome (^{Bove et al., 2013}; ^{Lam et al., 2017}).

Review of the studies investigating glycemia outcomes revealed no statistically significant effect of the intervention for improving glycemia in this population [(0.19; 95%CI: -7.60, 7.98)] (Figure 3).

Figure 2 Risk of Bias 

Figure 3 Strategies for Health Literacy vs. Control Group or Usual Care in the Evaluation of Blood Glucose Outcome 

Similarly, the results of the meta-analysis of studies comparing health literacy strategies versus control group or usual care for the hypertension outcome (^{Bove et al., 2013}; ^{Lam et al., 2017}) showed no significant effect of the interventions for absolute change in systolic and diastolic blood pressure in the population investigated - SBP [(-2.07;95%CI:-6.04,1.09)] (Figure 3); DBP [(-1.35;95%CI:-3.77,1.08)] (Figures 4 and 5).

Figure 4 Strategies for Health Literacy vs. Control Group or Usual Care in the Evaluation of Diastolic Blood Pressure Outcome 

Figure 5 Strategies for Health Literacy vs. Control Group or Usual Care in the Evaluation of Systolic Blood Pressure Outcome 

Limitations

This study has several methodological strengths, such as formulation of a specific research question, conducting of a systematic, comprehensive, sensitive literature search by 2 reviewers independently; where study selection, data extraction and analysis of methodological quality of the articles was also performed by 2 independent reviewers.

The RCTs reviewed were methodologically limited, whereby only 3 out of the 22 studies selected fully reported all the items rated in the bias assessment. The interventions reported in the studies reviewed varied greatly and targeted different outcomes, severely hindering the process of meta-analysis.

Comparison with other reviews

The present review differs to that of ^{Clement et al. (2009}) in scope, sample age and publication period, given the present study restricted its focus to intervention goals (health concerns) and effects, on older adults and more recent RCTs, having included 3 of the 16 studies covered in the cited review. Notwithstanding these differences, some results of the present review were similar to those found by Clement et al. (2009), namely, the methodological disparity among the studies reviewed in terms of length of the interventions, besides differences in complexity and in the measures employed, hampering the determination of evidence of effect.

Additionally, the samples involved resembled those of most of the studies in the systematic review by ^{Lee et al. (2012}), whose objective was to identify effective intervention strategies to improve health outcomes for heart disease patients with low health literacy skills.

These differences aside, the conclusions of the studies, including those of the present review, were largely in agreement, confirming that no reliable conclusive evidence is currently available on the most effective intervention strategy for improving the health of individuals with limited literacy (^{Clement et al., 2009}; ^{Lee et al., 2012}).

The results of the present systematic review and meta-analysis suggest that the available evidence on the effect of interventions used by health professionals for older adults with low health literacy failed to promote statistically significant improvements in glycemia and hypertension outcomes. Adherence to medication and treatment, followed by self-efficacy and satisfaction, were the outcome classes assessed, but not included in the meta-analysis owing to the wide variety of different technical and methodological procedures employed in the studies reviewed. However, the positive effects observed for each type of intervention applied, yielding significant results for only some outcomes and improvements in intragroup scores, demonstrated the interventions applied had good acceptability by older adults with limited health literacy. It was not possible, however, to conclude which intervention strategies had a significant positive effect on improving health outcomes in these patients. Further randomized clinical trials of high quality and with greater methodological rigor in assessing results are warranted.