Avalilação PREreview de Perceived Health Numeracy Difficulty and Health Self-Efficacy Among U.S. Adults with Cardiovascular Risk Factors: The Role of Patient–Provider Communication

This study examines the relationship between perceived health numeracy difficulty and health self-efficacy among U.S. adults with cardiovascular disease (CVD) risk factors, using data from the 2022 Health Information National Trends Survey (HINTS 6). The authors conducted a cross-sectional analysis involving 3,961 participants and employed survey-weighted ordered logistic regression to assess the association, including the potential influence of patient-provider communication.

The findings demonstrate a substantial association between perceived numeracy difficulty and lower health self-efficacy. Participants who reported difficulty understanding medical statistics had significantly lower odds of being in a higher health self-efficacy category (aOR 0.29; 95% CI 0.23–0.36). Following adjustment for patient-provider communication, the association remained statistically significant, although attenuated (aOR 0.37; 95% CI 0.29–0.46).

Notably, the proportion reporting complete confidence in managing their health was 28.9% among participants who found medical statistics easy to understand, compared with 12.9% among those reporting difficulty (p<0.001). Patient-provider communication was also independently associated with higher self-efficacy (aOR 2.31; 95% CI 1.96–2.74).

The study contributes to existing literature by examining perceived health numeracy, patient-provider communication, and health self-efficacy together in a population with cardiovascular risk factors. Its findings highlight the potential importance of communication quality in addressing barriers to patient confidence in managing health. However, the contribution should be interpreted as evidence of associations rather than evidence that improved communication causally mitigates numeracy-related barriers. The study also does not directly establish whether patients understand cardiovascular risk estimates or make better treatment decisions as a consequence.

Major issues

• 1. Construct validity and measurement of health numeracy

  The primary exposure, perceived health numeracy difficulty, is measured using a single self-reported question: how easy or difficult respondents find it to understand medical statistics. This captures perceived difficulty rather than objectively measured numerical comprehension.

  The distinction is important because a patient's confidence in understanding medical statistics may not necessarily correspond to their actual ability to interpret numerical health information. Furthermore, the study does not measure general health literacy, which could independently explain some of the observed association.

  The authors should acknowledge more explicitly that their findings concern perceived difficulty with medical statistics rather than objectively established health numeracy. They should also consider discussing how the use of a single-item measure and the dichotomization of responses into easy/very easy versus hard/very hard may affect measurement sensitivity and potentially obscure differences across the four original response categories.

  2. Potential selection bias arising from missing data

  The participant selection process requires greater scrutiny. Of the initial 6,252 HINTS respondents, 5,041 had at least one CVD risk factor. However, the final analytical sample comprised only 3,961 participants.

  This represents the exclusion of 1,080 otherwise eligible participants (approximately 21.4%) because of missing information. Of particular concern, 641 participants were excluded because of missing patient-provider communication scores, while a further 406 were excluded because of missing demographic covariates.

  The exclusion of a substantial proportion of eligible respondents raises concerns about potential selection bias. Participants with complete communication data may differ systematically from those without such information, particularly in healthcare access and healthcare utilization.

  The authors should clarify the reasons for missing communication data, establish whether respondents without a healthcare encounter were included among those with missing scores, and compare the characteristics of included and excluded participants. They should also explain whether the survey weighting adequately addresses the additional exclusions arising from the complete-case analytical approach. A sensitivity analysis using an appropriate missing-data approach would strengthen confidence in the findings.

  3. Interpretation of patient-provider communication and potential mediation

  The authors report that the association between numeracy difficulty and health self-efficacy attenuated following adjustment for patient-provider communication. Specifically, the estimated marginal difference in the probability of complete health confidence decreased from -18.1 percentage points (95% CI -21.2 to -15.0) to -14.8 percentage points (95% CI -17.9 to -11.7).

  Although this represents a 3.3-percentage-point reduction in the estimated difference, approximately 18% of the original marginal difference, it does not establish that communication mediates the relationship.

  Given the cross-sectional design, the temporal ordering of numeracy difficulty, communication experiences, and health self-efficacy cannot be established. It is equally possible that individuals with lower health self-efficacy perceive communication less favorably or that an unmeasured factor influences all three variables.

  The authors appropriately acknowledge that the design precludes causal mediation conclusions. Nevertheless, the discussion and conclusion should consistently maintain this distinction. A more explicit explanation of the statistical attenuation, without implying a demonstrated mechanism, would improve the interpretation.

  4. Statistical assumptions underlying the ordered logistic regression

  The study employs survey-weighted ordered logistic regression to model health self-efficacy as a five-level outcome. However, the manuscript does not report whether the proportional odds assumption was assessed.

  This is a fundamental assumption of the conventional ordered logistic regression model. The adjusted odds ratios presented in Table 2 assume that the estimated association is consistent across the cumulative thresholds of the ordinal outcome.

  The authors should report the diagnostic procedure used to assess this assumption and provide the results. If the assumption is violated, an alternative specification, such as a partial proportional odds or generalized ordered logistic model, should be considered.

  Additionally, the authors should clarify whether the patient-provider communication score was assessed for linearity on the log-odds scale before being entered as a continuous predictor.

  5. Conceptual alignment between the communication measure and the conclusions

  The patient-provider communication composite comprises six general communication domains, including opportunities to ask questions, attention to patients' emotions, clarity of explanations, adequate consultation time, understanding of self-care, and support in managing uncertainty.

  Although these domains are relevant to patient-centered care, they do not specifically measure how healthcare professionals communicate numerical cardiovascular risk information.

  Consequently, the manuscript's conclusions regarding clear and personalized communication of numerical CVD risk information extend beyond what was directly measured. Similarly, shared decision-making is discussed as an important implication, but the study does not directly assess patients' participation in treatment decisions.

  The authors should distinguish between general patient-provider communication and communication specifically addressing numerical risk information. Their conclusions should be framed accordingly, unless additional supporting analyses or measures are available.

  6. Potential residual confounding

  Although the models adjust for age, sex, race and ethnicity, education, household income, and CVD risk-factor burden, other potentially important factors are not adequately addressed.

  For example, general health status, mental health, cognitive functioning, healthcare access, health insurance status, and healthcare utilization could influence perceived numeracy difficulty, communication experiences, and self-efficacy simultaneously.

  The authors should provide a stronger justification for the selected adjustment variables and discuss the possibility of residual confounding. A clear conceptual framework identifying the presumed relationships between the exposure, outcome, communication variable, and confounders would improve the methodological rigor of the analysis.

Minor issues

• Clarification of adjusted odds ratios: The authors should ensure that the interpretation of the aOR of 0.29 is precise. It represents 71% lower odds of being in a higher health self-efficacy category, subject to the proportional odds assumption, rather than a 71% reduction in the probability of being completely confident.

• Interpretation of the communication effect: The reported aOR of 2.31 represents the association with a one-point increase in the communication score on its original 1–4 scale. The authors should explain the clinical meaning of a full-point increase and consider reporting an estimate for a more practically interpretable increment.

• Risk-factor burden: The study categorizes participants into one, two, or three or more CVD risk factors. Although the analysis adjusts for this burden, the authors should further discuss the implications of assigning equal weight to five distinct risk factors when constructing the cumulative measure.

• Subgroup analyses: The authors report no significant interactions across individual CVD risk factors, educational attainment, and race/ethnicity. However, the interaction p-values for individual risk factors range from 0.056 to 0.482. The borderline result of p=0.056 deserves appropriate interpretation, and the authors should avoid equating a nonsignificant interaction with evidence that the associations are identical across subgroups.

• Presentation of descriptive findings: The communication score was lower among participants reporting numeracy difficulty (68.6 versus 79.9 on the 0–100 scale; p<0.001). Reporting this difference more prominently would help readers appreciate the magnitude of the observed communication disparity.

• Discussion of practical implications: The authors could strengthen the discussion by specifying potential strategies, such as visual risk presentations, absolute rather than relative risk explanations, plain-language communication, and teach-back methods. These should be presented as potential interventions requiring further evaluation, rather than established effects of the current study.

Competing interests

The author declares that they have no competing interests.

Use of Artificial Intelligence (AI)

The author declares that they did not use generative AI to come up with new ideas for their review.