Review Reports - Technology-Enhanced Learning for Pressure Injury Prevention in Nursing: A Systematic Review and Meta-Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

The topic is important for nursing education and pressure injury prevention. However, in the current version, there are fundamental inconsistencies regarding which studies were included in the meta-analysis, how comparative effect sizes were calculated, and how the results in the figures relate to the text. These issues are not merely presentational; they directly impact the reported pooled effect and the conclusions drawn from it. Without re-verifying the study-level data and analyses, the pooled estimate cannot be relied upon.

The identity of the studies included in the meta-analysis does not match across the text, table, and forest plot. The authors state that seven studies were included in the qualitative synthesis and six controlled comparisons from five studies were included in the meta-analysis. However, while Figure 2 displays Morente, two separate Karimian comparisons, Çakar, Chao, and Erdem, Bredesen—which is presented as a comparative study in Table 1—is missing from the plot. More importantly, according to Table 1, Çakar and Karadağ is a single-group, pretest–posttest design study; nevertheless, it is presented in Figure 2 as though it were a controlled comparison. The text excludes the Fok study from the pool because it lacks a comparison group; why the same rationale was not applied to the Çakar study is not explained. This is the most critical issue concerning the validity of the primary analysis. The authors should provide a data table for each forest plot line displaying the study ID, intervention and control groups, group sample sizes, outcome measure used, time of measurement, effect size, and standard error. If the Çakar study lacks a concurrent control group, it must be removed from the comparative meta-analysis. Why Bredesen is not included in the plot must also be explained; if its eligibility is confirmed, it should be included in the analysis. Following these corrections, the abstract, main text, table, and all figures must reflect the identical set of studies.
The independence and origin of the two comparisons reported from the Karimian study are unclear. Table 1 defines the Karimian study as comprising 30 intervention and 30 control participants. Figure 2, however, presents two distinct effects labeled "Karimian 2020 A" and "Karimian 2020 B." It is not specified whether these represent different groups, different knowledge scales, different time points, or two outcomes derived from the same participants. Although the methods state that the first post-intervention measurement was prioritized, the decision rule used to select two effects is also not explained. Pooling effects derived from the same participants or a shared control group as if they were independent observations may distort the study's weight and confidence interval. The authors must clarify the source of the two effects and their degree of dependency; if they are not independent, they should select a single pre-specified effect for the primary outcome or implement a method that appropriately accounts for dependence. Subgroup and sensitivity analyses should also be repeated in accordance with this decision.
The necessary data required to reproduce the SMD 1.51 result are not reported. The methods state that Hedges’ g values were calculated from post-intervention group means and standard deviations; however, these values, along with the study-level calculated effect sizes and variances, are not provided to the reader. Details regarding the random-effects model are also lacking: how between-study variance was estimated, how confidence intervals were derived, and the specific analysis commands or code utilized should be detailed. The data availability statement also does not provide an accessible repository for the extracted data and analysis code. To allow verification of the primary analysis, study-level numerical data, effect calculation rules, handling of multi-arm studies and multiple outcomes, and reproducible code or adequate methodological detail should be provided in supplementary materials. Without this information, the reported pooled effect, heterogeneity, and sensitivity results cannot be independently evaluated.
Subgroup confidence intervals in Figure 4 contradict the values reported in the text. The student subgroup is reported in the text as SMD 1.39; 95% CI 0.71–2.07, whereas Figure 4 shows 0.75–2.03. For ICU nurses, the text reports SMD 1.89; 95% CI 1.15–2.63, whereas Figure 4 displays 0.57–3.21. The latter discrepancy in particular is substantial for interpretation and is not merely a rounding difference. It must be determined which analysis output the figure reflects, and both the figure and the text should be revised according to the correct results. Furthermore, the student subgroup relies on two studies, while the ICU nurse subgroup, according to the text, relies on two comparisons from a single study. In the latter case, two plotted comparisons should not be interpreted as evidence from two independent studies.
The scope of the review does not fully align with its title and concluding statements. While the title focuses on technology-enhanced learning, the eligibility criteria also encompass "structured educational programs." In Table 1, the Karimian intervention is defined as "lecture + materials"; in what respect this constitutes technology-enhanced learning is not clear. Combining different educational modalities is not inherently invalid, but the research question and interpretation of the effect estimate must correspond to this scope. The authors must define technology-enhanced interventions with pre-specified, operational criteria and indicate which interventions meet this definition. If the primary pool includes structured education that is not technology-enhanced, the outcome should not be portrayed exclusively as the effect of technology-enhanced learning. If sufficient data are available, a separate analysis for technology-enhanced interventions could be considered.
The reporting of the search and selection process must be corrected in terms of the PRISMA flow diagram. The arithmetic in the text is straightforward: 2,427 records − 1,440 duplicates = 987 screened records. In contrast, Figure 1 states "Records screened (n = 1440)." The figure also contains duplicated numbers, truncated text, and difficult-to-read boxes. The manuscript's assertion that selection numbers were "reconciled" is incompatible with the current figure. Additionally, it is stated that of the 244 reports for which "retrieval was sought," 207 were excluded during the initial full-text/topic screening; the number of reports actually retrieved and the stage at which these 207 exclusions occurred should be reported more clearly. While 10 reports were excluded during the 30 detailed exclusions on the grounds of "no comparison group for quantitative synthesis," why two studies lacking a comparison group were retained in the qualitative synthesis should be explained with a clear and consistent selection rule.
The search strategy and review protocol are not sufficiently verifiable. The text notes that searches were conducted up to December 2025, but the exact search date for each database was not retained; it also indicates that the search strings in Supplementary S1 are reconstructed versions rather than verbatim records of historical queries. This represents a significant transparency limitation. If possible, original search records should be retrieved to provide the exact date of the final search, full search strings, applied filters, and retrieved record counts for each database. If these records cannot be located, reconstructed strings should not be presented as the original search; an updated, reproducible search should be conducted, and the methods and flow diagram updated accordingly. Furthermore, the statement regarding institutional research registration under MSSADER via the University of Buraimi does not in itself demonstrate the existence or content of a pre-established systematic review protocol. The authors must clarify whether a protocol existed, provide registration details if available along with any protocol deviations, or explicitly state the absence of a protocol.
The risk-of-bias assessment is not presented in a verifiable manner. The methods describe the use of RoB 2 and ROBINS-I; however, Table 1 predominantly contains brief, general statements. For instance, one row reads "High; RCT, clear allocation, validated knowledge tool"; this description does not indicate which domain gave rise to the high risk of bias judgment. For each study, the appropriate tool for the study design, the specific effect estimate assessed, domain-level judgments, and the rationales for these judgments should be provided. Assessments of single-group studies should be segregated from assessments of intervention effects in controlled studies. The textual summary of "four high risk, three moderate or some concerns" can only be validated through this detailed evaluation.
The GRADE assessment should be repeated after the primary analysis is corrected. The "low certainty" conclusion is currently based on a contested set of studies and an as-yet unverified pooled effect. The foundation for certainty judgments remains unclear until the distinction between controlled and uncontrolled studies, potential dependent effects, and high heterogeneity are resolved. The authors should first finalize the appropriate comparative evidence base, and subsequently update the baseline certainty, rationales for each downgrading decision, imprecision, and publication bias assessments. That the confidence interval does not cross zero does not alone resolve all certainty concerns across a small and heterogeneous set of studies.
In the presence of high heterogeneity, the interpretation of the pooled effect should be more cautious. The reported I² is 89.5%. Interventions, comparators, participants, and knowledge measures vary considerably. That the effect remains positive in a leave-one-out sensitivity analysis does not demonstrate that the pool is methodologically sound or that a single common educational effect exists; moreover, which comparisons are valid must be resolved first. Following re-analysis, the authors should clearly delineate which set of interventions and comparators the pooled estimate represents, and display between-study variance and study-level differences more transparently. Inferences extending beyond measured knowledge scores to clinical practice or patient outcomes should be avoided.

The research question of this review is valuable; however, in the current manuscript, the inclusion of an uncontrolled study in a controlled meta-analysis, the absence of another seemingly eligible study from the plot, potentially dependent comparisons, and discrepancies between figures and text regarding confidence intervals undermine the reliability of the primary finding. The authors should be requested to return to the raw extraction table to confirm eligibility decisions and all effect sizes, reproduce the analysis from scratch, and resubmit the PRISMA flow diagram, risk-of-bias assessment, and GRADE findings in full alignment with the corrected analysis. Prior to these revisions, I do not recommend accepting the conclusions based on the SMD 1.51 estimate.

Author Response

1.1 The identity of studies in the meta-analysis does not match across the text, Table 1, and the forest plot. Çakar and Karadağ is a single-group study but appears in Figure 2. Bredesen is missing from the plot. The reviewer asks for a data table for each forest-plot line.

Response. We agree that this was the most serious problem in the previous version, and we have corrected it.

Çakar and Karadağ was a single-group study of patient relatives. It meets neither our population criterion nor the requirement for a concurrent comparator. It has been removed from the meta-analysis and from the review; Table S8 lists it among excluded studies with the reason.

Bredesen et al. is an eligible randomized trial and is included in the review. It measured classification accuracy (proportions of correct photograph responses) and Braden accuracy, not a knowledge test score, so it cannot be combined with knowledge outcomes and is synthesized narratively (Section 3.8). The rule is stated in Section 2.3 and applied to every study in Table 1 (“Synthesis” column).

Data for each forest-plot line. Supplementary Table S2 (summary) gives, for each line, the study, intervention and comparator, group sample sizes, outcome instrument, time point, means, SDs, and the page and table of the source report. Supplementary Table S6 gives the pooled SD, d, correction factor J, g, variance, SE, 95% CI, and weight.

Consistency. The abstract, Section 3, Table 1, Figure 2, and Tables 2–3 now all refer to the same three pooled studies.

Changes: Sections 2.3, 3.1, 3.2, 3.4, 3.8; Table 1; Figure 2; Supplementary Tables S2, S6 and S8.

1.2 The origin and independence of “Karimian 2020 A” and “B” are unclear, and Table 1 gives 30 vs 30 participants.

Response. The two effects were the knowledge outcome (Pieper Pressure Ulcer Knowledge Test) and the attitude outcome (APuP) of the same participants. Pooling them as independent observations was an error. Only the knowledge outcome now enters the meta-analysis, and the attitude result is reported narratively (Section 3.8). The rule of one prespecified knowledge outcome per study is stated in Section 2.6, so no study contributes more than one effect size, and the subgroup and sensitivity analyses were repeated accordingly.

On re-checking the source report (Table 2, p. 93), the analysed group sizes are 32 and 35, not 30 and 30. Table 1 and all analyses now use 32 and 35.

Changes: Sections 2.6, 3.4, 3.5, 3.8; Table 1; Figure 2; Supplementary Table S2.

1.3 Data needed to reproduce the pooled effect are not reported: study-level means and SDs, effect sizes and variances, the τ² estimator, the CI method, code, and handling of multi-arm studies and multiple outcomes.

Response. Section 2.8 now specifies all of these:

Effect measure: Hedges’ g from post-intervention means and pooled SDs, with the small-sample correction and the variance formula.

Model: DerSimonian–Laird estimator of τ², with Wald-type 95% CIs.

Sensitivity checks on the model: the Hartung–Knapp–Sidik–Jonkman adjustment, the Paule–Mandel estimator, and a common-effect model.

Special designs: a matched-pairs study, a study with a three-month assessment only, and the multi-arm trial used in the sensitivity analysis.

Study-level data and calculations are in Supplementary Tables S2 and S6. The analysis code is provided as Supplementary Code S1. The Excel file recalculates the pooled result with live formulas, and its results match the code and an independent implementation (statsmodels). The Data Availability Statement now states where these materials are.

Changes: Section 2.8; Data Availability Statement; Supplementary Tables S2, S6 and S7; Supplementary Code S1.

1.4 Subgroup confidence intervals in the figure contradict the text (students 0.71–2.07 vs 0.75–2.03; ICU nurses 1.15–2.63 vs 0.57–3.21), and the ICU subgroup relies on two comparisons from one study.

Response. We could not identify which analysis produced the earlier figure, so both the figure and the text were regenerated from the same new analysis output. The ICU subgroup was based on the two dependent Karimian outcomes, which no longer exist as separate comparisons.

The subgroups are now:

Nursing students: one study (Morente et al.; g = 1.75, 95% CI 1.20–2.30).

Registered nurses: two independent studies (Karimian et al. and Chao et al.; g = 1.17, 95% CI 0.87–1.47; I² = 0%).

Difference: Q = 3.33, p = 0.07.

Figure 2 and Section 3.5 report identical values. Because one subgroup contains a single study, the analysis is described as descriptive. The text states the number of studies explicitly and draws no conclusion about learner-group differences.

Changes: Section 3.5; Figure 2; Abstract.

1.5 The scope does not align with the title: “structured educational programs” and Karimian’s “lecture + materials” are not clearly technology-enhanced. TEL should be defined with prespecified, operational criteria.

Response. We have removed “structured educational programs” from the eligibility criteria and defined TEL operationally (Sections 1 and 2.2). Table 1 now states the digital component of each included intervention and the nature of the contrast with its comparator.

The previous description of Karimian et al. as “lecture + materials” was incomplete. The programme included two virtual video sessions in addition to face-to-face teaching, brochures, and pocket cards, so it meets the blended-TEL criterion. Because its comparator received no specified education, its effect may partly reflect education versus no education; we discuss this in Section 4.2.

The primary analysis is now restricted to interventions that meet the TEL definition, which is the separate TEL analysis the reviewer suggested. Omitting Karimian et al. gives g = 1.41 (95% CI 0.81–2.00) (Table 2). We describe the result as the effect of TEL interventions, which often included face-to-face components, relative to their comparators. We do not describe it as an effect of technology alone.

Changes: Title; Abstract; Sections 1, 2.2, 3.2, 3.4, 4.1, 4.2, 5; Tables 1 and 2.

1.6 PRISMA flow diagram: 2,427 − 1,440 = 987, but Figure 1 states 1,440 screened. There are duplicated numbers and truncated text, and the stage of the 207 exclusions is unclear. The selection rule for single-group studies needs to be explained.

Response. Figure 1 has been redrawn using the PRISMA 2020 template, from the documented search described under comment 1.7. It now has no duplicated or truncated content, and every stage is shown separately:

Records identified: 4,195. Duplicates removed: 2,146.

Records screened: 2,049. Excluded at title and abstract: 1,874.

Reports sought: 175. Not retrieved: 0. Assessed in full text: 175.

Excluded at full text: 170 (not pressure injury education 151; no educational intervention 11; not nurses or nursing students 7; not TEL 1).

Included: 5 studies (3 in the meta-analysis, 2 narrative only).

Section 3.1 reports exactly these numbers. The selection rule is stated in Sections 2.2 and 2.3:

Single-group pre–post studies are eligible for the review but contribute only to the narrative synthesis.

Lack of a comparator excludes a study from the meta-analysis, not from the review.

Fok (2024) is therefore in the narrative synthesis. Çakar and Karadağ is excluded because of its population, not its design.

Changes: Figure 1; Sections 2.2, 2.3, 2.5 and 3.1.

1.7 The search is not verifiable: exact dates were not retained and the strings were reconstructed. If the original records cannot be located, an updated reproducible search should be conducted. The MSSADER institutional registration does not demonstrate a protocol.

Response. We agree, and we followed the reviewer’s recommendation. Because the original search histories could not be recovered, we did not report the reconstructed strings as the original search. We conducted a new, fully documented search:

Databases and dates: Scopus and CINAHL on 10–24 January 2026; PubMed on 4 April 2026; all four databases updated in October 2026 (PubMed and CENTRAL on 7 October 2026). The update found no additional eligible studies.

Documentation: Supplementary Table S1 reproduces each strategy verbatim from the database search history (screenshots retained), with dates, limits, and record counts. The Methods, flow diagram, and results are based on this search.

Sources not re-run: Embase, Web of Science, and grey-literature sources reported in the earlier version were not part of this documented search. Section 2.4 says so, and Section 4.4 lists it as a limitation.

Protocol: Section 2.1 states that no protocol was prospectively registered or published. It also explains that the MSSADER entry at the University of Buraimi is an institutional research record, not a review protocol, and identifies the changes made during peer review.

Changes: Sections 2.1, 2.4, 4.4; Supplementary Table S1; Figure 1.

1.8 Risk of bias is not verifiable: Table 1 contains general statements. Domain-level judgments and rationale are needed, with single-group assessments kept separate.

Response. The risk-of-bias column has been removed from Table 1. Each study is now assessed for the specific outcome and time point used, with the tool appropriate to its design:

RoB 2 for randomized trials (Table S3, Figure S1): Morente et al., some concerns; Karimian et al., high risk; Bredesen et al., some concerns at the immediate assessment and high risk at three months. Every domain has a judgment and a written rationale.

ROBINS-I for the non-randomized matched comparison (Table S4, Figure S2): Chao et al., serious risk.

NHLBI before–after tool, reported separately from the controlled studies (Table S5): Fok, fair quality.

Section 3.3 summarizes the domains that drove each overall judgment.

Changes: Section 3.3; Table 1; Supplementary Tables S3–S5; Supplementary Figures S1–S2.

1.9 GRADE should be repeated after the primary analysis is corrected; a CI that does not cross zero does not resolve all certainty concerns.

Response. GRADE was repeated on the corrected three-study evidence base (Table 3; Sections 2.9 and 3.9). Evidence from randomized trials started at high certainty and was downgraded twice:

Risk of bias (−1): one trial at high risk, the non-randomized study at serious risk, and some concerns in the remaining trial.

Imprecision (−1): three small studies with 277 participants, below a conventional optimal information size, and a CI that widens to 0.55–2.12 with the HKSJ adjustment. As the reviewer notes, this judgment is not based on whether the CI crosses zero.

Inconsistency was not downgraded (I² = 41.5%; Q p = 0.18; all estimates favour TEL). Publication bias could not be assessed with three studies. Certainty of evidence is low.

Changes: Table 3; Sections 2.9 and 3.9.

1.10 With high heterogeneity the interpretation should be more cautious; a positive leave-one-out result does not show the pool is sound; inferences to clinical practice or patient outcomes should be avoided.

Response. Once the ineligible and dependent comparisons were removed, heterogeneity fell to I² = 41.5%. It is explained largely by the larger effect in Morente et al. (omitting it gives g = 1.17, I² = 0%).

The text now states which interventions and comparators the pooled estimate represents (Sections 3.4 and 4.1). It reports τ², the prediction interval (uninformative with three studies), and every study-level estimate (Figure 2; Table S6). It interprets the magnitude as an upper estimate for content-aligned, well-resourced programmes (Section 4.2). We state explicitly that the evidence concerns knowledge test scores and does not establish effects on clinical practice or patient outcomes (Sections 4.1, 4.3, 5).

Changes: Sections 3.4, 3.6, 4.1–4.4, 5; Abstract; Key Points

Reviewer 2 Report

Comments and Suggestions for Authors

Manuscript: Technology-Enhanced Learning for Pressure Injury Prevention 2 in Nursing: A Systematic Review and Meta-analysis

Reviewer Comments

Methods

Scope and eligibility criteria: The title suggests a focus on technology-enhanced learning in nursing; however, the review includes nurses, nursing students, and caregivers/patient relatives, as well as both technology-enhanced and broadly defined “structured educational programs.” These populations differ substantially in baseline knowledge, educational needs, and professional responsibilities. Please clarify the rationale for pooling these populations and interventions and ensure that the title, objectives, eligibility criteria, and conclusions accurately reflect the scope of the review.

Meta-analysis methodology: Please specify the estimator used for between-study variance (τ²) and the method used to calculate the confidence interval for the pooled random-effects estimate. In addition, multiple comparisons from the same study were analyzed separately. Please clarify whether these comparisons shared participants or a common comparator and, if so, how this statistical dependency was handled in the meta-analysis.

PRISMA flow diagram: The study-selection numbers appear inconsistent. The text indicates that 2,427 records were identified and 1,440 duplicates removed, leaving 987 records for screening, whereas Figure 1 reports 1,440 records screened. Please verify and reconcile the PRISMA flow counts.

Results

Inconsistency between the Results and Figure 3: The subgroup confidence intervals differ between the text and Figure 3. For undergraduate nursing students, the text reports 1.39 (95% CI 0.71–2.07) versus 1.39 (0.75–2.03) in Figure 3; for ICU nurses, the corresponding values are 1.89 (1.15–2.63) versus 1.89 (0.57–3.21). Please verify the original analysis and correct these discrepancies.

Inclusion of Çakar and Karadağ [8]: Çakar and Karadağ [8] is included in the controlled meta-analysis, yet the qualitative synthesis states that the study lacked a control group. In contrast, Fok [11] was excluded from the pooled analysis because it lacked a concurrent comparator. Please clarify this inconsistency and confirm whether Çakar and Karadağ [8] meets the eligibility criteria for the primary meta-analysis.

GRADE assessment: The decision not to downgrade for imprecision appears to be based mainly on the pooled CI not crossing the null. Given the small evidence base of six controlled comparisons, please provide a clearer justification for the assessment of imprecision.

Discussion
Although the limitations and substantial heterogeneity are appropriately acknowledged, the main pooled effect (SMD 1.51, 95% CI 0.95–2.07) is not sufficiently discussed. Please compare the magnitude of this effect with relevant previous systematic reviews/meta-analyses and discuss possible explanations for the large effect and heterogeneity, particularly differences in learner populations, interventions, comparators, outcome measures, and risk of bias.

Author Response

2.1 The review pools nurses, nursing students, and caregivers, and both TEL and structured educational programmes. Please justify this and align the title, objectives, eligibility criteria, and conclusions.

Response. Caregivers and patient relatives are no longer eligible, and non-technological structured programmes have been removed (Section 2.2). The review now concerns TEL, defined operationally, for nursing students and registered nurses. We pool these two groups because both are being prepared for, or hold, professional responsibility for pressure injury prevention, and learner population was a prespecified subgroup (Sections 2.2, 2.8, 3.5). The title, objectives, abstract, and conclusions reflect this scope. Please also see Reviewer 1, comments 1.5 and 1.1.

2.2 Specify the τ² estimator and the CI method, and how dependency between comparisons from the same study was handled.

Response. The analysis uses the DerSimonian–Laird τ² estimator with Wald-type 95% CIs. The HKSJ adjustment and the Paule–Mandel estimator are reported as sensitivity analyses (Section 2.8; Table 2).

Dependency is no longer an issue. Each study now contributes one prespecified knowledge outcome (Section 2.6). The two previous Karimian effects were knowledge and attitudes from the same participants, and only knowledge is now pooled. The three-arm physical-simulation trial is used only in a sensitivity analysis, where its arms are either analysed one at a time or combined, so its control group is never counted twice.

2.3 The PRISMA counts are inconsistent (987 vs 1,440 screened).

Response. Corrected. Please see Reviewer 1, comment 1.6. Figure 1 and Section 3.1 now report identical counts from the documented search.

2.4 Subgroup CIs differ between the text and the figure.

Response. Corrected. Please see Reviewer 1, comment 1.4. Both are now generated from the same analysis output.

2.5 Çakar and Karadağ is in the controlled meta-analysis although it lacked a control group, whereas Fok was excluded for the same reason.

Response. We agree that this was inconsistent. Çakar and Karadağ studied patient relatives in a single-group design. It has been excluded from the review (population criterion) and listed in Table S8. The same rule now applies to all studies: single-group studies are not pooled (Section 2.3).

2.6 Justify the imprecision judgment in GRADE given the small evidence base.

Response. We now downgrade by one level for imprecision. The reasons are three small studies, a total of 277 participants (below a conventional optimal information size), and a CI that widens to 0.55–2.12 with the HKSJ adjustment (Table 3; Section 3.9).

2.7 Discuss the magnitude of the pooled effect: compare it with previous reviews and explain the large effect and heterogeneity.

Response. Section 4.2 has been rewritten. It compares our estimate with three earlier meta-analyses:

Lahti et al.: no significant difference in knowledge between e-learning and traditional learning.

Voutilainen et al.: scores about 5 points higher on a 0–100 scale, with very large heterogeneity.

Jin and Kim: SMD 0.63 for cognitive outcomes in synchronous e-learning.

It then discusses four likely reasons for the larger effect here:

a comparator that received no specified education;

knowledge tests closely aligned with intervention content, administered mostly immediately after teaching;

additional contact time and components in one study;

small studies at risk of bias.

It also explains the heterogeneity (driven by Morente et al.) and concludes that the magnitude should be read as an upper estimate. The two new references have been added.

Changes: Sections 1 and 4.2; References.

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you to the authors for the opportunity to review this manuscript. The study addresses a relevant topic and is generally well structured. However, I consider that some methodological and presentation aspects need to be clarified or revised. I present my comments below according to the sections of the manuscript:

Title: The title is clear and appropriately identifies the study as a systematic review and meta-analysis. However, the scope suggested by the expression “Technology-Enhanced Learning” does not appear to be fully consistent with the eligibility criteria and the interventions included in the review. The eligibility criteria also include “structured educational programs” and at least one of the included interventions is described as “lecture + materials vs. usual practice”, without an evident technological component. I suggest clearly defining what is considered Technology-Enhanced Learning and ensuring consistency between the title, the review objectives, the eligibility criteria, and the interventions included in the synthesis.

Abstract: I suggest revising the abstract to clarify and ensure consistency in the scope of the review:

In the Background, the study is framed specifically as an assessment of the effectiveness of Technology-Enhanced Learning (TEL) compared with conventional education. However, in the Methods, eligible studies are described more broadly as those evaluating “pressure injury education”, without specifying that the interventions should include a technological component. This discrepancy becomes even more evident in the Conclusion, where the authors refer to “TEL and structured educational interventions”, suggesting that the synthesis included educational interventions that do not necessarily correspond to the concept of TEL. This issue is particularly relevant considering that the title presents the review specifically as an assessment of Technology-Enhanced Learning. I therefore suggest clearly defining, also in the abstract, the type of interventions covered by the review and ensuring consistency between the title, Background, Methods, and Conclusions.
Despite the designation “Background/Objectives”, the objective of the review is not explicitly presented; I therefore suggest including a clear statement of the objective.
In the Methods, eligible studies are described as those reporting quantitative knowledge outcomes, although the eligibility criteria presented in the manuscript include other outcomes; this information should therefore be harmonised.
I also suggest replacing “qualitative synthesis” with “narrative synthesis”, since the included studies are quantitative.
Finally, subgroup analyses should be presented with greater prudence: in particular, it should be made explicit when a subgroup estimate results from multiple comparisons derived from a single study, rather than merely stating that it is based on “few comparisons”.

Keywords: I suggest revising the keywords to ensure that they adequately represent the actual scope of the review and are consistent with the title, eligibility criteria, and included interventions. The use of “Educational Technology”, “E-Learning”, and “Simulation Training” reinforces a framework specifically focused on Technology-Enhanced Learning, whereas the review also appears to include structured educational interventions without a clearly identifiable technological component.

Introduction: The Introduction is generally well structured and adequately contextualises the relevance of the topic, although some conceptual and methodological aspects require further clarification:

I suggest providing a clear definition of Technology-Enhanced Learning (TEL) and specifying the characteristics that an intervention must have to be considered TEL. This clarification is particularly important because the Introduction includes, alongside e-learning, simulation, serious games, and immersive teaching, “structured educational packages”, which do not necessarily correspond to technologically mediated interventions.
I also suggest strengthening the specific contribution of this systematic review by clarifying what has already been synthesised in previous reviews and what specific gap the present review aims to address. The statement that the evidence “has not been synthesized consistently across learner populations and educational modalities” should be supported and the additional contribution of this review made explicit.
Regarding the Objectives, I recommend revising them to ensure consistency with the eligibility criteria and the analyses actually performed:
- The first objective refers only to nurses and nursing students, although the review also includes caregivers.
- The third objective proposes exploring effects according to educational modality, but this analysis is not presented in the results. I suggest presenting this analysis, if it was conducted, explaining why it was not possible to perform it, or reformulating the objective accordingly.

Materials and Methods: The section is generally well organised, but several methodological aspects require further clarification to ensure the transparency and reproducibility of the review, namely:

Study Design and Eligibility Criteria:
- What type of systematic review was conducted? Was it an effectiveness review?
- What methodology was used to conduct the review? The use of the PICO strategy and the PRISMA 2020 recommendations does not replace the identification of the methodology adopted to conduct the review.
- The criteria relating to the intervention should be clarified, particularly the inclusion of “structured educational programs”. It should also be made explicit which characteristics are required for an intervention to be considered Technology-Enhanced Learning and whether educational interventions without a technological component were also eligible.
- Why were caregivers included?
- Please justify the decision to restrict study eligibility to the period between 2000 and 2025, particularly the choice of the year 2000 as the lower time limit.
Information Sources and Search Strategy:
- On what date was the search actually conducted in each database? The expression “The searches were conducted from January 2000 to December 2025” appears to refer to the eligibility/publication period of the studies rather than to the date on which the searches were conducted.
- The authors state that grey literature was consulted. I recommend describing the grey literature search in greater detail, identifying the institutional repositories consulted, the search strategy used, and the types of documents considered eligible.
- Where are the search strategies used? Were they the same across all databases? The authors state that these are available in the supplementary material, but this corresponds to the PRISMA 2020 checklist. This issue limits the reproducibility of the systematic review. I suggest presenting the original search strategies used in each database, including the respective limits/filters and the dates on which the searches were conducted.
- Although the authors state that no language restrictions were applied in the search, they subsequently establish as a requirement the availability of an abstract in English and sufficient information for extraction. How was this criterion applied during the study selection process? I suggest clarifying this issue and presenting it consistently in the eligibility criteria.
Study Selection:
- I suggest limiting this subsection to the description of the methods used for study selection. According to PRISMA 2020, the numbers of records identified, assessed, excluded, and included correspond to the results of the selection process and should be presented in the Results section, ideally accompanied by the PRISMA flow diagram.
- What criteria were used to determine which studies would be included in the quantitative synthesis and which would be retained only for narrative synthesis? I suggest clarifying this decision and ensuring that the same criteria were consistently applied to all studies.
Data Extraction and Outcome Measure:
- How was the dependence between effect sizes derived from the same study handled statistically, particularly when comparisons shared participants or a comparator group? The authors only state that this dependence was considered in the interpretation, which does not make it possible to understand whether it was incorporated into the statistical analysis.
- I also suggest justifying the decision to select the first post-intervention assessment time point when several assessment time points were available.
Risk of Bias Assessment:
- I suggest presenting, in the manuscript or supplementary material, the risk-of-bias assessment by domain for each study, according to RoB 2 and ROBINS-I, allowing readers to understand how the overall risk-of-bias judgements were obtained.
Effect Measure and Statistical Analysis:
- What estimator was used for the between-study variance (τ²), and what method was used to calculate the confidence intervals?
- How was the dependence between multiple comparisons derived from the same study handled statistically?
- What criteria were used to conduct the subgroup analyses?
- How was the existence of multiple comparisons derived from the same study taken into account in the subgroup analyses?

Results: Overall, the results are presented clearly. However, some aspects require clarification and revision to ensure consistency between the text, tables, and figures:

Study Selection:
- Figure 1 should be presented in the Results section, in accordance with PRISMA 2020.
- There are inconsistencies between the text and the PRISMA flow diagram. The text states that, after the removal of 1,440 duplicates, 987 records were screened, whereas the figure appears to indicate “Records screened (n = 1440)”. In addition, the flow diagram contains duplicated numbers and textual elements, as well as information that is not fully visible. I suggest thoroughly revising Figure 1 and confirming the consistency of all values presented with those described in the text.
Study Characteristics:
- Table 1: there is an apparent inconsistency regarding the study by Çakar and Karadağ [8]. The study is described as single-group quasi-experimental and without a randomised control group. However, it subsequently appears in the forest plot of the main meta-analysis. I suggest checking the design of this study and clarifying whether or not an eligible comparator exists.
- The presentation of risk of bias in Table 1 should be revised. Expressions such as “High; RCT, clear allocation, validated knowledge tool” do not make it possible to understand which domains determined the overall classification as high risk of bias.
Effect of Educational Interventions on Knowledge:
- The text refers to six controlled comparisons derived from five studies. However, Bredesen et al. [7] does not appear in Figure 2, whereas Karimian et al. [19] appears twice (“A” and “B”). What do these two estimates correspond to? I suggest clarifying which studies and comparisons were included in the meta-analysis.
- Çakar and Karadağ [8] is described as a single-group study, without a comparator group. How was the effect size presented in Figure 2 obtained, considering that the meta-analysis is based on between-group comparisons? I suggest clarifying this issue, particularly because Fok [11] was excluded from the quantitative synthesis because it did not include a comparator group.
- I suggest presenting, in the supplementary material, the data used to calculate each effect size included in the meta-analysis, in order to increase the transparency and reproducibility of the analysis.
Subgroup Analyses by Population Type:
- In the ICU nurses subgroup, the two comparisons are derived from the same study. How was the dependence between these comparisons taken into account in the subgroup analysis? I suggest clearly distinguishing the number of studies from the number of comparisons.
- The confidence intervals presented in the text do not match the values presented in Figure 4. I suggest checking and correcting these estimates.
Qualitative Synthesis of Study Findings:
- I suggest replacing “Qualitative Synthesis” with “Narrative Synthesis”, since this section presents a narrative synthesis of the results of quantitative studies.
Risk of Bias and Study Quality:
- Where are the risk-of-bias judgements by domain for each study presented? I suggest presenting the complete results of the assessment conducted using RoB 2 and ROBINS-I, allowing readers to understand the overall classifications assigned to the studies.
- I suggest clearly distinguishing the assessment of risk of bias in the individual studies from the assessment of the certainty of the evidence conducted using GRADE.
Other aspects:
- The initial objective also proposed analyses according to educational modality, which are not presented in this section. I suggest clarifying this issue.
- Figure 3, Figure 4, and Table 2 are not referred to in the text of the manuscript.

Discussion: The Discussion is generally well structured and contextualises the main findings in light of the existing literature. However, some aspects of the interpretation of the results should be revised:

I suggest revising the interpretation of the magnitude of the effect, taking into account the high heterogeneity, the small number of studies, and the low certainty of the evidence.
The implications for practice should reflect that the available evidence focuses predominantly on knowledge, with evidence regarding clinical performance and patient outcomes still being limited.

Conclusions: The conclusion is generally aligned with the main findings and limitations identified in the review.

Supplementary Materials: The authors refer throughout the manuscript to Appendix S1 for the database-specific search strategies. However, this appendix is not included in the supplementary material provided, with only the PRISMA 2020 checklist being indicated in the Supplementary Materials section. I suggest checking and providing all supplementary material referred to in the manuscript.

References: I suggest a thorough review of the references. Some DOIs are presented twice (as in references 9 and 10), and some references have formatting issues or incomplete information.

Author Response

Title. Make the scope consistent with the TEL focus.

Response. TEL is now defined operationally, and only interventions meeting the definition are included. The title, objectives, eligibility criteria, and included interventions are therefore consistent (Sections 1 and 2.2; Table 1). Please also see Reviewer 1, comment 1.5.

Abstract. Align the Background, Methods, and Conclusions on TEL; state the objective; harmonise outcomes; use “narrative synthesis”; present subgroups with prudence.

Response. The abstract has been rewritten. It now:

defines TEL;

states the objective explicitly (“This review estimated the effect of TEL on pressure injury knowledge among nurses and nursing students”);

states that controlled studies with continuous knowledge scores were pooled and that other outcomes were synthesized narratively;

uses “narrative synthesis” throughout;

no longer refers to “structured educational interventions”.

The subgroup estimates are not presented in the abstract, because one subgroup contains a single study. In the main text, the number of studies in each subgroup is stated explicitly (Section 3.5).

Keywords. Make them consistent with the actual scope.

Response. “Simulation Training” has been removed, because physical simulation is no longer within scope. The keywords are now: Educational Technology; E-Learning; Meta-Analysis; Nursing Education; Pressure Ulcer; Systematic Review. These match the revised TEL scope.

Introduction. Define TEL; clarify what previous reviews have synthesized and what gap this review fills; align the objectives (caregivers; modality analysis not presented).

Response.

Definition: TEL is defined in the second paragraph of the Introduction, which also states which interventions do not qualify.

Gap: the third paragraph summarizes previous meta-analyses of e-learning in nursing education (Lahti et al.; Voutilainen et al.; Jin and Kim), which pooled many clinical topics and found small or uncertain effects on knowledge. It then states the gap: controlled evidence on TEL specifically for pressure injury education has not been synthesized separately.

Objectives: these now refer only to nurses and nursing students. The objective of analysing effects by educational modality has been removed, because three pooled studies with three different modalities do not allow a meaningful modality analysis. Modality is described in Table 1 and discussed in Section 4.2.

Methods: design and eligibility. Type of review and methodology; criteria for “structured educational programs” and TEL; why caregivers; why 2000.

Response.

Type and methodology: Section 2.1 states that this is a systematic review of effectiveness with meta-analysis, conducted following the Cochrane Handbook for Systematic Reviews of Interventions and reported according to PRISMA 2020.

TEL criteria: Section 2.2 gives the operational TEL criteria; structured programmes without a digital component are not eligible.

Caregivers: caregivers are no longer eligible, as explained under Reviewer 2, comment 2.1.

Start year: the lower limit of 2000 is justified in Section 2.2. Web-based and computer-mediated instruction became widely available in nursing education from that period, so earlier technologies have limited relevance to current practice.

Methods: search. Actual search dates for each database; grey literature; the strategies (Appendix S1 was not provided); language criterion.

Response.

Dates: Section 2.4 and Table S1 give the search dates for each database (please see Reviewer 1, comment 1.7).

Grey literature: this was not part of the documented, updated search, and the text and limitations now say so explicitly.

Strategies: the complete, verbatim strategy for each database, with limits and record counts, is in Supplementary Table S1, which is now provided.

Language: Section 2.2 states that no language restriction was applied and that reports needed an English-language abstract and sufficient information for extraction.

Methods: study selection. Limit this subsection to methods and move the numbers to the Results; clarify the criteria for quantitative versus narrative synthesis.

Response.

Numbers moved: Section 2.5 now describes only the selection methods. All counts are reported in Section 3.1 with Figure 1, which now sits in the Results.

Quantitative versus narrative criteria: these are stated once in Section 2.3 and applied to every study (Table 1, “Synthesis” column).

Methods: data extraction. How was dependence handled statistically? Justify the first post-intervention time point.

Response.

Dependence: each study contributes a single prespecified knowledge outcome, so there are no dependent effect sizes in the analysis (Section 2.6).

Time point: the first post-intervention assessment was chosen in advance because it was available in all studies, is the most comparable across them, and avoids selecting among time points according to their results (Section 2.6).

Methods: risk of bias, statistics, subgroups. Domain-level RoB 2 and ROBINS-I; τ² estimator and CI method; subgroup criteria; multiple comparisons in subgroups.

Response.

Risk of bias: domain-level judgments with rationale are in Tables S3–S5 and Figures S1–S2.

Estimator and CIs: the τ² estimator and CI method are specified in Section 2.8.

Subgroups: the only subgroup analysis is the prespecified comparison by learner population. It now involves independent studies only, and is reported descriptively because the student subgroup contains a single study (Sections 2.8 and 3.5).

Results. Place Figure 1 in the Results; resolve the PRISMA inconsistencies; check the design of Çakar and Karadağ; revise the risk-of-bias presentation in Table 1; explain Bredesen and Karimian A/B; provide the data used for each effect size; distinguish studies from comparisons in subgroups; correct the CIs; use “narrative synthesis”; present domain-level risk of bias; separate risk of bias from GRADE; address the modality analysis; cite Figures 3–4 and Table 2 in the text.

Response.

Figure 1: now appears in Section 3.1, with corrected counts (Reviewer 1, comment 1.6).

Çakar and Karadağ: removed (Reviewer 2, comment 2.5).

Table 1: no longer contains risk-of-bias statements (Reviewer 1, comment 1.8).

Bredesen and Karimian: explained in Reviewer 1, comments 1.1 and 1.2.

Effect-size data: in Supplementary Tables S2 and S6.

Subgroups: now report numbers of independent studies, with corrected CIs (Reviewer 1, comment 1.4).

Narrative synthesis: Section 3.8 is titled “Narrative Synthesis of Other Educational Outcomes”.

Risk of bias versus GRADE: risk of bias for individual studies (Section 3.3) is now reported separately from the GRADE certainty assessment (Section 3.9; Table 3).

Modality analysis: addressed under the Introduction above.

Citations of figures and tables: every figure and table is now cited in the text: Figure 1 (Section 3.1), Figure 2 (Sections 3.4–3.5), Figure 3 (Section 3.7), Table 1 (Section 3.2), Table 2 (Section 3.6), and Table 3 (Section 3.9).

Discussion. Revise the interpretation of the magnitude in light of heterogeneity, the small number of studies, and low certainty; implications for practice should reflect that the evidence concerns knowledge.

Response.

Magnitude: Section 4.2 now compares the effect with previous reviews, explains why it is likely to be inflated, and presents it as an upper estimate (Reviewer 2, comment 2.7). The principal findings (Section 4.1) state the low certainty and the small evidence base.

Implications: Section 4.3 states that the evidence concerns knowledge test scores and does not establish effects on bedside assessment, prevention practice, or patient outcomes.

Supplementary materials. Appendix S1 is cited but not provided.

Response. All supplementary material is now provided and cited consistently:

Supplementary Material document:

Table S1: search strategies;

Table S2: data summary;

Tables S3–S5: risk of bias;

Table S6: effect-size calculations;

Table S7: sensitivity analyses;

Table S8: excluded studies;

Figures S1–S3.

Supplementary Table S2 (Excel).

Supplementary Code S1.

PRISMA 2020 checklist.

The Supplementary Materials section of the manuscript lists each item.

References. Some DOIs appear twice; some references are incomplete or badly formatted.

Response. All references have been checked against PubMed or the publisher’s record and reformatted consistently. Duplicated DOI prefixes have been removed, missing authors, issues, pages, and DOIs have been completed, and the thesis references now give the degree and institution. References no longer cited after the revision have been removed, and the list has been renumbered in order of citation.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

I thank the authors for their detailed responses and the substantial revisions made to the main manuscript. The removal of the ineligible study, the separation of dependent knowledge and attitude outcomes in the Karimian et al. study, the operational definition of technology-enhanced learning (TEL), and the more cautious interpretation regarding clinical practice and patient outcomes are all commendable improvements. The main text, Table 1, and the forest plot are now consistent regarding the three studies pooled in the meta-analysis.

However, the supplementary file provided for review is not aligned with the revised manuscript and the response letter. The following issues must be resolved before this manuscript can be considered for publication:

Provide the current supplementary materials. The evaluated supplementary file contains only the PRISMA checklist and search strategies. The study-level data, effect size calculations, domain-level risk-of-bias assessments, list of excluded studies, and analysis code referenced in the authors' response letter are missing from this document. If these were prepared as separate files, they must be made accessible to the peer review process. In their absence, the reproducibility of the revised meta-analysis cannot be independently verified.
Harmonize the supplementary file with the revised manuscript. The PRISMA checklist in the supplementary file still reports the outdated figures: seven included studies, six controlled comparisons, a pooled SMD of 1.51, and an I2I^2 of 89.5%. These directly contradict the revised main text, which reports five included studies (three pooled) and an I2I^2 of 41.5%. Descriptions of the analysis software, sensitivity analyses, and data availability must also be brought into full alignment with the revised manuscript.
Supply documentation for the updated search. The main text and response letter describe fully documented searches conducted in 2026. In contrast, the supplementary file presents reconstructed search strings dated December 2025. The actual search strings used, the exact calendar search dates for each database, and the corresponding record counts should be provided. If an outdated supplementary file was uploaded inadvertently, please ensure the correct and finalized version is submitted.
Clarify outcome classification and GRADE justifications. The rationale for pooling Morente et al.’s image-based assessment as a knowledge outcome while excluding Bredesen et al.’s image classification outcome should be made clearer. Furthermore, the authors should elaborate on the GRADE approach applied to the mixed pool of randomized and non-randomized studies, specify the assumptions underlying the optimal information size, and justify why indirectness was not downgraded despite substantial clinical heterogeneity across comparators.

In conclusion, while the revisions in the main text are in the right direction, the submitted documentation is insufficient to verify the reported corrections. The manuscript may be reconsidered once the complete, updated supplementary materials are provided and these inconsistencies are resolved.

Author Response

Comment 1. Provide the current supplementary materials. The evaluated supplementary file contains only the PRISMA checklist and search strategies. The study-level data, effect size calculations, domain-level risk-of-bias assessments, list of excluded studies, and analysis code are missing.

Response. We apologize; the outdated supplementary file was made available for review in error. File S1 now contains every item referred to in the manuscript and the response letter:

Item	Content	Location
Study-level data	Group sizes, means, SDs, instrument, time point, and source page or table for each study	Table S2; File S2 (full extraction matrix)
Effect-size calculations	Pooled SD, Cohen’s d, correction factor J, Hedges’ g, variance, SE, 95% CI, and weight for each study, with formulas	Table S6; File S2 (live formulas)
Sensitivity analyses	Leave-one-out, HKSJ, Paule–Mandel, common-effect, and addition of the excluded trial	Table S7; Figure S3
Domain-level risk of bias	RoB 2 (randomized trials), ROBINS-I (non-randomized study), NHLBI (single-group study), each with a judgment and rationale for every domain	Tables S3–S5; Figures S1–S2
Excluded studies	Studies that might appear eligible, with reasons	Table S8
Analysis code	Complete Python code	Code S1 (printed in File S1); File S3

The code, the spreadsheet, and an independent implementation (statsmodels) give identical results: g = 1.33 (95% CI 0.98–1.69); I² = 41.5%; τ² = 0.041. These match the manuscript.

Changes: File S1; File S2; File S3; Supplementary Materials and Data Availability sections of the manuscript.

Comment 2. Harmonize the supplementary file with the revised manuscript. The PRISMA checklist still reports seven included studies, six controlled comparisons, SMD 1.51, and I² 89.5%. Descriptions of the analysis software, sensitivity analyses, and data availability must also be aligned.

Response. The PRISMA 2020 checklist has been completed again for the revised manuscript (Table S10). Every location now refers to the current sections, tables, and figures, and the outdated figures (seven studies, six comparisons, SMD 1.51, I² 89.5%) no longer appear anywhere in the submission. The checklist entries now match the manuscript as follows:

Software (item 13d): Python 3.13 with NumPy 2.5 and SciPy 1.18, cross-checked with statsmodels 0.15 and a spreadsheet. The manuscript (Section 2.8) now reports the same versions.
Sensitivity analyses (item 13f): the same six analyses as Section 2.8 and Table 2.
Data and code availability (item 27): Files S2 and S3 and Code S1, as in the revised Data Availability Statement.

Changes: Table S10; Section 2.8 (Software); Data Availability Statement.

Comment 3. Supply documentation for the updated search: the actual search strings, the exact calendar search dates for each database, and the record counts.

Response. Table S1 now gives, for each database, the search strategy reproduced verbatim from the search history, the limits applied, the record count, and the search dates:

Database	First search	Final search	Records (final search)
MEDLINE via PubMed	4 April 2026	7 October 2026	1,424
Scopus	January 2026 (10–24 January)	7 October 2026	2,120
CINAHL (EBSCOhost)	January 2026 (10–24 January)	7 October 2026	351
Cochrane CENTRAL (Issue 10 of 12, October 2026)	7 October 2026	7 October 2026	300 (311 before the year limit)

Screenshots of the search histories of the final searches are provided as new Figure S4 (panels a–d). The note to Table S1 states that these searches replace the reconstructed strategies reported in the original submission. Section 2.4 now states that the final searches of all four databases were run on 7 October 2026, and the abstract gives this date.

Changes: Table S1; Figure S4 (new); Section 2.4; Abstract.

Comment 4. (a) Clarify why Morente et al.’s image-based assessment was pooled as a knowledge outcome while Bredesen et al.’s image classification outcome was excluded. (b) Elaborate on the GRADE approach for the mixed pool of randomized and non-randomized studies, specify the assumptions underlying the optimal information size, and justify why indirectness was not downgraded despite clinical heterogeneity across comparators.

Response (a). Section 2.3 now states the rule used to classify outcomes and applies it to both studies:

The rule. A knowledge outcome is a participant-level score on a test of pressure injury content, reported as group means and SDs. Classification accuracy is agreement with an expert standard when assigning pressure injury categories to photographs.
Morente et al. Although the questionnaire used images, its 0–22 score combined six domains: wound classification, detection of infection, tissue type, peri-ulcer skin condition, dressing selection, and design of care interventions. Classification was therefore one domain within a broader score of evaluation and management knowledge, and the score was reported per participant as means and SDs.
Bredesen et al. The study assessed only category assignment (and Braden scoring) against expert ratings. Classification was reported mainly as proportions of correct photograph responses pooled across participants. A participant-level sum score was reported for one photograph set only, compared with the Mann–Whitney U test, and without means or SDs. The outcome therefore differs in construct and cannot be converted to a standardized mean difference without unsupported assumptions.

Because the reviewer may reasonably question the decision for Morente et al., the text points to the leave-one-out analysis. Omitting Morente et al. gives g = 1.17 (95% CI 0.87–1.47), so the conclusion does not depend on this decision.

Response (b). Section 2.9 now explains the approach, and a full evidence profile has been added as Table S9. Table 3 has been expanded accordingly.

Mixed body of evidence. The non-randomized study was assessed with ROBINS-I. Following GRADE guidance 18 (Schünemann et al., 2019), evidence assessed with ROBINS-I may start at high certainty, with confounding and selection bias captured in the risk-of-bias domain. The body of evidence therefore started at high certainty. Its risk of bias was rated serious because the two studies carrying 72.7% of the weight were at high (RoB 2) or serious (ROBINS-I) risk. As a check, we repeated the assessment for the two randomized trials alone (g = 1.49, 95% CI 0.98–1.99; 139 participants). The certainty was also low, for the same two reasons.
Optimal information size. Following GRADE guidance 6 (Guyatt et al., 2011), we considered the optimal information size (OIS) because a large effect from a small sample can appear precise. The OIS was calculated for a two-sided α of 0.05, 80% power, and a minimal important difference of 0.3 SD, giving about 175 participants per group (350 in total). The review included 131 and 146 per group, and the HKSJ-adjusted CI (0.55–2.12) was wide, so imprecision was rated serious. Table S9 shows that the OIS is not met for any minimal important difference below about 0.35 SD.
Following GRADE guidance 8 (Guyatt et al., 2011), indirectness was judged against the review question: TEL compared with any of the prespecified comparator conditions. All three studies enrolled the target populations, measured the primary outcome, and used comparators within that set. Indirectness was therefore not rated down for this question. The differences among comparators are reflected in how the estimate is interpreted (as the effect of TEL-based educational packages) and in the sensitivity analyses.

We agree with the reviewer that the comparator differences matter for a narrower question. If the question were whether the digital component itself improves knowledge, only Morente et al. provides a direct comparison, and the evidence would be rated down for indirectness to very low certainty. This is now stated in Section 3.9, Table 3, and Table S9.

The three GRADE guidance papers have been added to the reference list (references 30–32).

Changes: Sections 2.3, 2.9, and 3.9; Table 3; Table S9 (new); references 30–32.

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have made the necessary revisions. I recommend accepting the manuscript in its current form.

Author Response

Comment. The authors have made the necessary revisions. I recommend accepting the manuscript.

Response. We thank Reviewer 2 for the careful reviews and for this recommendation.

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you to the authors for the work undertaken in revising the manuscript.

The revised version presents significant improvements. However, I consider that some issues still need to be addressed:

Search Strategies and Supplementary Material: The manuscript reports searches updated in October 2026, conducted in MEDLINE/PubMed, Scopus, CINAHL and CENTRAL, and refers to a supplementary table containing the search strategies actually executed. However, the material provided contains reconstructed search strategies dated December 2025, covering different sources. Therefore, I recommend providing the search strategies actually used in each database, including the respective dates, limits applied and number of records retrieved.
Reproducibility of the Meta-analysis: The restructuring of the meta-analysis, using a single effect size estimate per study, represents an important improvement compared with the previous version. However, although the authors state that the extracted data, individual effect size calculations and statistical analysis code are available in the supplementary material (Supplementary Tables S2, S6 and S7 and Supplementary Code S1), these elements are not included in the files to which I had access for the present review. I therefore suggest ensuring their availability to allow independent verification and reproducibility of the quantitative analyses.
Risk of Bias Assessment: The authors now present the overall risk of bias judgements and the main reasons supporting them. However, the domain-level assessments, referred to as available in Supplementary Tables S3–S5 and Figures S1–S2, are not included in the files provided. I recommend including these elements.
PRISMA 2020 Checklist: The PRISMA checklist provided does not correspond to the current version of the manuscript. It continues to present results from the previous version. I suggest updating it in full to ensure consistency with the revised manuscript.
Supplementary Materials: I recommend revising this section to ensure that all the supplementary materials identified correspond to the files actually provided.
Author Contributions: This section is incomplete. I suggest completing it in accordance with the journal's guidelines.

I acknowledge the methodological improvements introduced in this revised version. The issues identified do not negate this progress, but their resolution is necessary to ensure the transparency, documentary consistency and reproducibility of the systematic review and meta-analysis.

Author Response

Comment 1. Search strategies and supplementary material: provide the search strategies actually used in each database, with dates, limits, and record counts.

Response. Please see Reviewer 1, Comment 3. Table S1 gives, for each database, the verbatim strategy, the limits applied, the record count, and the search dates (final searches on 7 October 2026). Figure S4 shows the search-history records. The reconstructed strategies dated December 2025 came from the original submission and have been replaced.

Comment 2. Reproducibility of the meta-analysis: the extracted data, individual effect-size calculations, and statistical analysis code (Tables S2, S6, S7 and Code S1) were not in the files available for review.

Response. We apologize; these files were not made available as intended. They are now included in File S1:

Table S2: study-level data;
Table S6: effect-size calculations and formulas;
Table S7: sensitivity analyses;
Code S1: the complete analysis code, printed in full.

The code is also provided as a runnable script (File S3), and File S2 recalculates the analysis with live formulas. Please also see Reviewer 1, Comment 1.

Comment 3. Risk of bias: the domain-level assessments (Tables S3–S5, Figures S1–S2) were not included in the files provided.

Response. These are now included in File S1:

Table S3: RoB 2, five domains plus an overall judgment, with rationale, for each randomized trial;
Table S4: ROBINS-I, seven domains plus an overall judgment, for the non-randomized study;
Table S5: NHLBI before–after tool, 12 criteria, for the single-group study;
Figures S1–S2: the corresponding traffic-light plots.

Comment 4. The PRISMA 2020 checklist does not correspond to the current version of the manuscript.

Response. The checklist has been completed again for the revised manuscript (Table S10). It no longer contains any result from the original version. Please see Reviewer 1, Comment 2.

Comment 5. Revise the Supplementary Materials section so that the items listed correspond to the files actually provided.

Response. The Supplementary Materials section now lists File S1 (each table, figure, and code item it contains), File S2 (Excel), and File S3 (Python script). These are exactly the files uploaded with this revision. The Data Availability Statement refers to the same files.

Comment 6. Author Contributions: this section is incomplete.

Response. The Author Contributions section has been completed using the CRediT taxonomy and has been agreed by all authors.

We thank the reviewers again for their careful and constructive reviews.

Confirmation of authorship. As requested by the Editorial Office, we confirm the following:

The author list and affiliations are unchanged.
Both corresponding authors (A.S.I. and I.G.J.) are marked with an asterisk.
The e-mail addresses in the manuscript now match those in the submission system.