This investigation was a prospective, monocentric, quantitative, analytic cross-sectional observational study conducted in the ED of a German university hospital. The aim was to investigate the influence of HL on patients’ subjective assessment of emergency condition severity. The study took place between November 21, 2023, and April 9, 2024. Written informed consent was acquired from all participants. The analytical cross-sectional science approach of this study was in compliance with the statement of Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) (28). A pilot study was conducted prior to the main study to confirm feasibility (n = 10). The study was approved by the ethics committee of the Friedrich-Schiller-University Jena, Germany (Registration number: 2023-3106_1-Bef, date: 9^th October 2023) and preregistered on drks.de (DRKS00032962). All methods were carried out in accordance with relevant guidelines and regulations. All methods were carried out in compliance with the Declaration of Helsinki II.

All adult patients admitted to the ED of a German university hospital were eligible for inclusion. This ED handles approximately 37,000 visits per year. Sampling was performed daily during study hours, which alternated randomly between five time slots: 10–12 AM, 12 AM-2 PM, 2–4 PM, 4–6 PM, 6–8 PM. For further information regarding randomization see Supplementary 1. Patients were approached consecutively. 324 patients were included. The inclusion criteria required participants to be at least 18 years of age and classified under the Emergency Severity Index (ESI) triage category 2, 3, 4, or 5. Patients were excluded if they were admitted via emergency medical services or had a cognitive deficit that would impair their ability to complete the questionnaire. Of the 324 patients included, 257 patients had a complete data set. Only this data was used for analysis.

Upon admission and following the initial ESI triage process, patients were approached by the study team, including the principal investigator, to inform them about the study and invite their participation. After obtaining informed consent, patients were asked to complete the study questionnaire assessing their subjective perception of their emergency condition severity. This was done privately, in a quiet and separate room. Completion of the questionnaire took most patients around 5 min (range: 3–8 min). If a medical necessity arose during the data collection process, the process was terminated and not resumed. After the patient had completed their questionnaire, their ED nurse and physician were asked independently for their assessment of the emergency condition severity based on their first impression. The nurse and physician did not discuss this patient prior to their assessment. The composition of the medical team of physicians and nurses varied from shift to shift. If more than one interaction occurred between the medical team and the patient before the assessment, the data were excluded from the analysis to avoid mutual interference or influence by already collected test results. The final diagnosis of each patient and—if applicable—their treatment during their hospital stay was obtained through a retrospective examination of medical records 30 days post-admission performed by emergency medicine specialists (i.e., retrospective case assessment). See Figure 1 for a flowchart of the data collection process.

The minimum required sample size was calculated to be 248 patients (see Supplementary 4 for details).

All data were analyzed using descriptive methods. Normality was assessed using the Shapiro–Wilk test and Q-Q plots. Spearman rank-order correlations (ρ) quantified how closely assessments matched. Fisher’s z values indexed effect size. Discrepancy was contrasted across HL levels with Kruskal–Wallis tests; significant omnibus results were followed by Dunn post-hoc contrasts with Bonferroni correction, and effect size was expressed as eta-squared (η²). Multiple OLS regressions predicted each Discrepancy from HL score, age, education, and gender, reporting unstandardized Bs, SEs, and R². See Supplementary 5 for a detailed description. All statistical analyses were performed using SPSS Statistics version 29.0.0 build 241 (IBM, New York, United States) and in collaboration with a faculty biostatistician who remained blinded to the study conditions. Figures were created using Prism version 10.2.1 build 339 (GraphPad Software, LLC., Boston, United States). All results were evaluated at a two-sided significance level of α = 0.05, unless otherwise specified.

341 patient cases were screened upon admission. 17 screened patients declined participation. 324 patients were included. Data of 67 patients (21%) was later omitted due to incomplete HL questionnaires, which occurred due to organizational processes (e.g., call for X-ray). 14 of those excluded patients had other missing data, mainly the physician’s or nurse’s assessment. Data from 257 patient cases was analyzed, see Figure 1. Only complete data sets were analyzed.

The age of included patients followed the usual distribution for our ED. Mean age of participants was 53 years (Md 55, SD 22.05, range: 19–93). Most patients included reported to have completed high school (42%).

Patients were stratified into three groups based on their HL level as determined by HLS-EU-Q16: adequate (n = 95), problematic (n = 119), and inadequate (n = 43) HL. Problematic and inadequate HL were collectively referred to as the limited HL group (30). 63% of the patients included scored in the inadequate or problematic HL categories. There were no demographic differences between the groups. See Table 1 for a detailed group description.

Group characteristics are described in Table 1. The overall distribution of diagnoses in the included patients was representative of the common reasons for presentation to our ED, Supplementary 6, 7.

The duration of initial contact between patient and ED physician was 3–6 min, while it was 9–12 min for ED nurses.

Assessments were covariate-adjusted (Age, Education, and Gender) for analysis, see Supplementary 8. Only the regression analysis was calculated using unadjusted values. Mean severity assessment by the patients across the whole sample was 5.33 (Md 5.42, SD 1.74), while it was 4.27 (Md 4.54, SD 1.48) for the ED nurses, 3.95 (Md 3.70, SD 1.59) for the ED physicians, and 3.66 (Md 3.77, SD 1.93) for the emergency medicine specialists.

Assessments were compared between HL levels, Table 2. Only the severity rated by the patients differed statistically significant between the HL levels, indicating an association between self-assessment and HL.

Correlation analysis was performed for all assessments and between HL levels, see Table 3. Across the full sample, all severity assessments except patients’ and specialists’ were positively correlated (Spearman’s ρ = 0.18–0.34). Patient self-assessments showed a small association with ED nurses’, ρ = 0.20, p < 0.01, and ED physicians’ (ρ = 0.18, p = 0.01). Clinician assessments were interrelated: nurse–physician, ρ = 0.24, p < 0.001; physician–specialist, ρ = 0.44, p < 0.001; and nurse–specialist, ρ = 0.22, p < 0.01.

A binary logistic regression examined whether a Severe Outcome (specialists’ assessment > 5.59) could be predicted using patient-related data. The overall model was significant, R² = 0.09, N = 249. Controlling for age, education, and gender, higher HL test (HLS-EU-Q16) scores were associated with lower odds of a severe outcome, OR = 0.84, 95% CI [0.75, 0.93], p = 0.001. Conversely, higher patient self-assessments increased the odds, OR = 1.21, 95% CI [1.07, 1.37], p = 0.003. Greater educational attainment was also protective, OR = 0.83, 95% CI [0.70, 0.98], p = 0.025. Age and gender were non-significant (p ≥ 0.23). Thus, both HL and the patient’s own severity perception meaningfully contribute to predicting whether a specialist will judge a case as severe. The models explained 10% of the variance in the Severe Outcome variable, indicating relevant predictive value, see Supplementary 9.

Discrepancy between patients’ and clinicians’ assessments in all variations averaged between 1.5 and 2 in most cases, see Table 4.

The size of the discrepancy between patients’ self-assessments and clinicians’ assessments depended on HL level. For Discrepancy All, the difference across inadequate-, problematic-, and adequate-HL groups was significant, H (2) = 7.07, p = 0.029, ε² = 0.02, indicating a small effect. Post-hoc tests revealed that adequate-HL patients exhibited smaller gaps than both inadequate-HL (p = 0.011) and problematic-HL patients (p = 0.021), whereas the latter two groups did not differ (p = 0.83). The same pattern emerged for Discrepancy Medical Team, H (2) = 7.02, p = 0.030, ε² = 0.02; adequate-HL patients differed from inadequate (p = 0.014) and problematic (p = 0.027) patients, while inadequate and problematic groups were similar (p = 0.79). By contrast, the Discrepancy Specialists did not vary by HL level, H (2) = 4.23, p = 0.121, ε² = 0.01. Thus, lower health literacy is associated with modestly larger mismatches between patient and nurse/physician assessments, but not between patient and specialist assessments.

Using absolute values to index the magnitude of disagreement (ignoring direction) yielded very similar results. Discrepancy All still differed across health-literacy groups, H (2) = 7.76, p = 0.021, ε² = 0.02, indicating a small effect. Post-hoc contrasts confirmed that adequate-HL patients disagreed less with the full clinical team than did both inadequate-HL (p = 0.013) and problematic-HL patients (p = 0.025); the latter two groups did not differ (p = 0.77). An analogous pattern emerged for Discrepancy Medical Team, H (2) = 6.92, p = 0.031, ε² = 0.02, with the adequate-HL group again showing smaller mismatches than the other two (p < 0.05), see Figure 2. In contrast, Discrepancy Specialists did not vary by HL level, H (2) = 4.12, p = 0.127, ε² = 0.01. Thus, even when only the size of the gap is considered, lower HL remains associated with larger patient–clinician disagreements for nurses and physicians, whereas agreement with specialists is unaffected by HL level.

Multivariable regression analysis was performed to evaluate the relationship between HL, discrepancy, and the covariates age, education, and gender, Supplementary 10. The numerical value of the patient’s HLS-EU-Q16 score was used for HL.

Controlling for age, education, and gender, higher HL scores reliably predicted smaller patient–clinician mismatches. For Discrepancy All, HL was a significant negative predictor, b = −0.15, t(240) = −2.63, p = 0.009, explaining 4% of the variance (R² = 0.04). A comparable, marginally significant relation appeared for Discrepancy Medical Team (b = −0.11, t(229) = −1.84, p = 0.067, R² = 0.02), and the association was strongest for Discrepancy Specialists (b = −0.23, t(219) = −3.83, p < 0.001, R² = 0.07).

Age and education did not affect any discrepancy indices except Discrepancy Specialists, where older and less-educated patients showed greater results (age: p = 0.022; education: p = 0.008). Gender was unrelated to all three outcomes. In sum, higher HL scores consistently correspond to smaller patient–clinician disagreements, particularly with specialists, while demographic factors add little explanatory power. The models accounted for only 2–7% of the variance, implying that HL is only partly accounting for patient–clinician incongruence.

Concordance was most prevalent among patients with adequate HL for Discrepancy All (34%) and Medical Team (33%), see Table 5. For Discrepancy with the Specialists, the problematic HL group achieved a slightly higher prevalence of concordance (26 vs. 22%). Positive discrepancy was around 10% more frequent in inadequate and problematic HL patients than in adequate HL patients. Negative discrepancy was slightly more frequent in inadequate HL patients compared to the other groups. The combination of negative discrepancy and severe outcome was frequent.

Severe cases were as expected: 12–15%. The dangerous combination of Severe Outcome and negative discrepancy (indicative of patients underestimating their symptoms) was not rare (3–11%) and prevalent on all HL levels yet emphasized for inadequate HL.

In the multivariable logistic model, a Discrepancy Medical Team independently increased the likelihood of a Severe Outcome: each one-unit rise was associated with 30% higher odds, OR = 1.30, 95% CI [1.07, 1.59], p = 0.008, after controlling for age, gender, and education. Higher nurse (OR = 1.20, 95% CI [1.01, 1.43], p = 0.037) and physician assessments (OR = 1.22, 95% CI [1.01, 1.47], p = 0.040) provided comparable incremental risk, whereas a higher health-literacy score lowered the odds by 12% (OR = 0.88, 95% CI [0.78, 0.99], p = 0.028). Age showed a marginal protective trend, p = 0.063, and neither patient self-assessments nor education reached significance. The model accounted for 24% of the variance in the severe-outcome indicator (R² = 0.24).

This study is not devoid of limitations: Clinical outcomes were only partially included in our results as they are part of the retrospective case assessment by the emergency medicine specialists and in the form of the variable Severe Outcome. Regardless of this, no delays, diagnostic errors, or adverse events were measured. These aspects are significant but are not part of our research question. We compared self-reported severity in the context of different HL levels, subjective assessments, and retrospective chart-based evaluations. Therefore, our data cannot be used to draw conclusions about the direct impact of HL on emergency medical care. In general, subjective evaluations inherently carry the risk of bias. However, since the professional assessments were conducted by a specialized team, biases should have been minimized. As these biases (e.g., confirmation or anchoring bias) are part of subjective assessments and opinion formation, they are ultimately unavoidable and to a certain extent necessary. For the chart-based retrospective analysis the specialists worked independently and were blinded to potentially influential contextual factors (e.g., presentation time or triage). The study population was limited to patients seen during daytime hours to achieve broad coverage and feasibility of the procedure. However, this approach may have introduced some degree of inclusion bias, as typical nocturnal emergencies such as myocardial infarction or intoxication were inevitably less prevalent. However, the most prevalent emergency conditions are represented and the cases included cover a broad time period. In addition, the groups are consistent with other surveys in terms of gender, age, HL, education, and time as well as day of presentation to the ED (5, 16). HL was assessed using the HLS-EU-16 self-assessment questionnaire. Self-assessments, especially those based on questionnaires, are susceptible to generalized over- or underestimation. Despite these limitations, this questionnaire was employed because it is a validated scientific instrument, and it is suitable for a fast-paced setting such as the ED. As nearly all tools for evaluating HL are based on self-assessment, no more suitable alternative was identified for this trial. Even though some questionnaire items have been debated in the literature (46), these were deemed irrelevant for the subject of this trial.

During data collection, a high number of participants (21%) did not complete their questionnaire in full and therefore had to be excluded, as there was no plan to complete the questionnaire later. In all cases, this was due to organizational procedures in the emergency department (e.g., call-up for X-ray examination). No case had to be excluded for medical reasons. Ideally, obtaining subjective assessments of condition severity from patients before triage could help lower the procedural bias and the potential for unconscious influence. However, this was not feasible within the constraints of routine ED operations.