학술

Civic and community engagement among health profession students and professionals compared to the general population: A systematic review and meta-analysis protocol

PLOS ONE
CC BY
이 매체는 공공·자유 라이선스로 본문을 직접 표시합니다.

Figures
Abstract
Background
Physicians, nurses, and other health professionals encounter poverty, inequality, and the structural roots of ill health as a routine part of clinical work — an exposure that differs from that of most other occupations. The World Health Organization, the Global Consensus for Social Accountability of Medical Schools, and many professional bodies describe civic and community engagement as a professional responsibility. Whether health profession students and practitioners show higher, lower, or comparable levels of civic engagement relative to the broader public in non-crisis conditions has not been systematically reviewed. The existing literature is fragmented, limited to single professions, focused largely on pandemic volunteerism, and more concerned with attitudes than with behaviour.
Methods
We present a protocol for a systematic review and meta-analysis of observational studies reporting behavioural civic and community engagement among health profession students and practising professionals across all disciplines, with or without a general-population or non-health comparator. Seven databases (PubMed/MEDLINE, Scopus, Web of Science, CINAHL, ERIC, PsycINFO, and Embase) and grey-literature sources will be searched from inception without date or language restrictions. Two reviewers will independently screen records, extract data, and assess risk of bias using JBI checklists matched to prevalence and comparative analyses (and the Newcastle-Ottawa Scale for cohort studies), together with an appraisal of measurement-instrument quality. Within pre-specified, conceptually comparable analysis groups, prevalence will be pooled with random-effects generalized linear mixed models, and odds ratios and standardized mean differences with restricted maximum-likelihood random-effects models and Hartung-Knapp-Sidik-Jonkman confidence intervals; prediction intervals will be reported. Certainty of evidence will be graded with GRADE, separately for prevalence and between-group comparisons. Reporting will follow PRISMA 2020.
Expected outputs
No results are yet available; the review is at the protocol stage. The completed review will report pooled prevalence of each engagement domain, unadjusted and adjusted comparisons with general-population, education-matched, and non-health student comparators, and the certainty of evidence for each estimate.
Citation: Ghasedi A, Vahidi R (2026) Civic and community engagement among health profession students and professionals compared to the general population: A systematic review and meta-analysis protocol. PLoS One 21(11): e0363170. https://doi.org/10.1371/journal.pone.0363170
Editor: Javier Fagundo-Rivera, University of Huelva: Universidad de Huelva, SPAIN
Received: July 20, 2026; Accepted: November 4, 2026; Published: November 20, 2026
Copyright: © 2026 Ghasedi, Vahidi. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: No datasets were generated or analysed during the current study, which is a protocol for a systematic review and meta-analysis. All relevant data from this study will be made available upon study completion: all data extracted from included studies, the risk-of-bias and instrument-quality judgements, the GRADE evidence profiles, the comparator estimates derived from the World Values Survey, and the R code used for all analyses will be deposited in a public Open Science Framework repository with a persistent DOI under a CC BY 4.0 licence, and cited in the Data Availability statement of the completed review. The full search strategy is provided as Supporting information (S1 File).
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: CI, Confidence interval; CINAHL, Cumulative Index to Nursing and Allied Health Literature; ERIC, Education Resources Information Center; GDP, Gross domestic product; GLMM, Generalized linear mixed model; GRADE, Grading of Recommendations, Assessment, Development and Evaluation; HKSJ, Hartung-Knapp-Sidik-Jonkman; I2, I-squared statistic (measure of statistical heterogeneity); JBI, Joanna Briggs Institute; MeSH, Medical Subject Headings; NGO, Non-governmental organization; NOS, Newcastle-Ottawa Scale; OR, Odds ratio; OSF, Open Science Framework; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; PRISMA-P, PRISMA for Protocols; PROSPERO, International Prospective Register of Systematic Reviews; REML, Restricted maximum likelihood; SD, Standard deviation; SMD, Standardized mean difference; SWiM, Synthesis Without Meta-Analysis; WHO, World Health Organization; WVS, World Values Survey
Introduction
Through clinical training and day-to-day practice, health professionals encounter poverty, structural exclusion, and the social conditions that drive ill health in ways that most working people do not. This proximity has long been used to argue that the health professions carry a particular civic obligation — that a doctor, nurse, or pharmacist who witnesses the downstream consequences of inequality bears some responsibility to engage with its upstream causes [1,2]. Whether this argument is reflected in behaviour is an empirical question that has not been tested systematically.
This expectation is also institutional. Medical, nursing, dental, and allied health curricula worldwide now include social accountability and community engagement among their expected graduate outcomes [3]. The World Health Organization, the Global Consensus for Social Accountability of Medical Schools, and a wide range of national licensing bodies treat civic participation as part of what it means to be a health professional [4,5]. For the purposes of this review, civic and community engagement is understood as voluntary, behavioural participation in activities oriented toward the common good, undertaken in a personal (citizen) capacity: formal and informal volunteering, voting and other political acts, community activism, charitable giving, and involvement in civil society or non-governmental organizations [6,7]. Activities organised, required, or incentivised by an employer, training institution, or professional body, and advocacy undertaken as a designated representative of a professional organization, are conceptually distinct and are classified separately (see Eligibility criteria).
Despite this institutional framing, empirical evidence on how the civic engagement of health professionals compares with that of the general public is limited. Four gaps stand out. First, the studies that exist tend to focus on a single profession — usually medicine — making cross-disciplinary comparisons impossible [8,9]. Second, much of the available evidence deals with willingness to volunteer during the COVID-19 pandemic, which reveals crisis behaviour rather than ordinary civic conduct [10]. Third, a great deal of attention has gone to what health professionals think about social responsibility, while the question of what they actually do — whether they vote, join organizations, donate, or show up for their communities — has been largely sidestepped [11]. Fourth, even the comparative studies that exist are confined to single professions without any synthesis across medicine, nursing, dentistry, pharmacy, and allied health; the civic engagement of most health disciplines against a public benchmark remains largely unexplored.
This review is designed to address these four gaps. By bringing together evidence from medicine, dentistry, nursing, pharmacy, physiotherapy, and allied health, and by measuring behavioural outcomes against general-population and education-matched benchmarks, it will provide a systematic answer to a simple question: does the civic participation of health professionals, as a group, differ from that of the general population — and if so, in which direction and in which domains?
The answer matters for at least three reasons. If no differences are observed, educators and accreditors have reason to re-examine how social accountability is taught and assessed. If differences exist, in either direction, understanding where they arise — by profession, career stage, country, or type of engagement — provides an evidence base for curriculum and policy. And at a broader level, establishing what health professionals actually do in civic life, rather than what institutions expect them to do, is a precondition for any evidence-based evaluation of the profession’s social contract.
Objectives
Primary objective. To determine the prevalence and level of civic and community engagement among health profession students and practising professionals across all disciplines, and to compare these with corresponding data from the general population or non-health comparator groups.
Secondary objectives.
To estimate the pooled global prevalence of civic and community engagement — including volunteering, political participation, activism, charitable giving, and social responsibility behaviours — among health profession students and professionals.
To compare civic engagement levels between health profession populations and comparator populations (the general population, education-matched adults, and non-health university students) using odds ratios and standardized mean differences, with and without adjustment for confounders.
To examine whether engagement levels differ across health disciplines: medicine, dentistry, nursing, pharmacy, physiotherapy, and allied health.
To compare students and practising professionals, as an indication of whether civic engagement differs by career stage.
To explore variation by sex, geographical region, country income level, and type of engagement activity.
To map the instruments used to measure civic engagement in health profession populations, appraise their measurement properties, and assess their comparability.
To identify gaps in the evidence base and offer recommendations for future research on social accountability in health professions education and practice.
To explore, as an exploratory aim, the association between empathy and civic engagement within health profession samples.
Materials and methods
Study design
We present the protocol for a new systematic review and meta-analysis; it is not an update of a previous review. Conduct and reporting will follow the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines [12]. The protocol was developed in accordance with the PRISMA-P 2015 checklist [13], which is provided as Supporting Information (S1 Checklist). The protocol was registered with PROSPERO (CRD420261348391); amendments made since registration are listed under Protocol amendments.
Eligibility criteria
Population.
Inclusion. Health profession students — medical, dental, nursing, pharmacy, physiotherapy, and other allied health disciplines — and practising healthcare professionals of any age, sex, nationality, or career stage. No restrictions are placed on country or healthcare system type.
Exclusion. (i) Non-health university students, except as comparators; (ii) administrative or non-clinical healthcare staff without patient-care responsibilities; (iii) samples in which health profession participants cannot be separated from a mixed population; (iv) samples whose only measured engagement is mandatory, curriculum-based community service.
Exposure.
Inclusion. Voluntary civic or community engagement behaviour occurring outside mandatory clinical, educational, or professional duties: formal or informal volunteering; political participation (voting, signing petitions, attending civic meetings, contacting elected officials); community activism and advocacy; charitable giving; and membership or active involvement in NGOs or civil-society organizations. Engagement may be measured as a binary (yes or no) or frequency outcome (e.g., hours per month) using validated scales, standardized survey items, or structured self-report questions.
Exclusion. (i) Mandatory, credit-bearing, or remunerated service, including curriculum-based service reported only as an educational output; (ii) activities performed as part of clinical or professional duties; (iii) measures of attitudes or intentions only, with no behavioural measure; (iv) crisis-specific volunteering only (e.g., COVID-19 or disaster response), with no measure of everyday engagement.
Capacity in which engagement occurs. Each eligible behaviour will be classified at extraction as (a) personal-capacity engagement, undertaken as a private citizen; (b) institution-organised or institution-incentivised engagement, i.e., optional activities organised, sponsored, or rewarded by an employer, training institution, or professional body (e.g., employer-supported volunteering or school-run outreach); (c) professional-role advocacy, i.e., political or policy advocacy undertaken as a designated representative of a professional organization; or (d) unclassifiable, where the instrument does not distinguish capacity. Because general-population surveys generally do not distinguish capacity, primary analyses will include categories (a), (b), and (d). Category (c) will not be combined with individual political participation and will be summarised separately. A sensitivity analysis will restrict comparisons to category (a) where data allow.
Instruments combining attitudes and behaviour. When an instrument contains both attitudinal and behavioural items, only behavioural items or subscales will be extracted (e.g., the behaviour subscale of the Civic Engagement Scale [14]). Where only a composite score mixing attitudes and behaviour is reported, authors will be asked for subscale data; if these are unavailable, the study will be classified as using a mixed instrument, analysed in a separate analysis group, and excluded from the primary behavioural analyses.
Crisis periods. Studies will be classified by their actual data-collection period and by the recall period of the engagement measure, not by publication date. The COVID-19 pandemic period is defined as 11 March 2020 (WHO pandemic declaration) to 5 May 2023 (end of the public health emergency of international concern). Studies conducted during this period that measure everyday, non-pandemic-specific engagement are eligible and will be flagged; pandemic-specific items (e.g., volunteering in the COVID-19 response) will not be extracted. A study is classified as pandemic-overlapping if its data-collection or recall period overlaps this window in whole or in part. Where dates are not reported, authors will be contacted; studies that remain unclear and were published in 2020 or later will be treated as pandemic-overlapping in the corresponding sensitivity analysis.
Comparator.
Three comparator types are pre-specified and will be analysed separately rather than combined: (i) the general adult population or community-based samples; (ii) education-matched adults, i.e., general-population respondents with tertiary education; and (iii) non-health university students or graduates. Because educational attainment is a strong determinant of civic engagement [7,15], comparisons with education-matched comparators (types ii and iii) are treated as the principal test of whether any difference is specific to the health professions rather than to higher education. Students will be compared preferentially with non-health students, and practising professionals with education-matched adults or the general population. Studies without a comparator will contribute to pooled prevalence estimates only.
Comparator source. Comparator data may come from the same study (within-study comparator) or from an external population dataset such as the World Values Survey (WVS) [16] or a national survey (external benchmark). Within-study comparisons form the primary comparative analysis. An external benchmark will be used only when both reviewers agree that it matches the index study on country, data-collection period (within three years), item wording or construct, and recall period. For the WVS, comparator estimates will be derived from individual-level microdata restricted to the index study’s country and survey wave and, where possible, to its age range and to tertiary-educated respondents. External-benchmark comparisons will be analysed separately from within-study comparisons, and comparator source will be both a pre-specified subgroup analysis and a sensitivity analysis.
Outcomes.
Primary outcomes. Volunteering rate: prevalence of any formal or informal volunteering (binary: yes or no); effect measures — pooled prevalence with 95% CI and odds ratio (OR) versus comparator. Civic engagement score: continuous score on a behavioural civic engagement scale or behavioural subscale (e.g., the behaviour subscale of the Civic Engagement Scale [14]); effect measure — standardized mean difference (SMD, Hedges’ g) with 95% CI. Political participation rate: prevalence of voting, petition signing, or attendance at civic meetings; effect measures — pooled prevalence with 95% CI and OR versus comparator. Where a study reports several recall periods, the following hierarchy will be applied: past-12-month behaviour, then the recall period closest to 12 months, then lifetime behaviour; for voting, turnout at the most recent national election will be used.
Secondary outcomes. NGO or civil society participation rate (%); charitable giving rate (%) and mean amount donated (converted to international dollars at purchasing-power parity for the survey year); activism and advocacy participation rate (%); frequency and intensity of engagement (hours per month); and, as an exploratory outcome, the association between empathy and civic engagement within health profession samples (see Exploratory analysis of empathy).
Study designs.
Inclusion. Cross-sectional and survey-based studies, prospective and retrospective cohort studies, and mixed-methods studies with an extractable quantitative component.
Exclusion. Case reports, case series, qualitative-only studies, editorials, letters, and opinion pieces; studies without a quantitative measure of civic engagement.
Context and setting.
Studies from any country are eligible, with no geographical limits. Both high-income and low- and middle-income countries (World Bank classification) are included to allow subgroup analyses by national economic context. There are no restrictions on publication date, language, urban or rural setting, or healthcare system type.
Information sources and search strategy
Seven databases will be searched from inception to the date of search execution: PubMed/MEDLINE, Scopus, Web of Science (Core Collection), CINAHL (via EBSCOhost), ERIC (via EBSCOhost), PsycINFO (via APA PsycNet), and Embase. Additional searches will cover Google Scholar (first 200 results), the WHO Institutional Repository for Information Sharing, ProQuest Dissertations and Theses, and the preprint servers medRxiv and OSF Preprints (including SocArXiv and PsyArXiv). Reference lists of all included studies will be hand-searched, forward citation tracking will be carried out via Scopus for key papers, and corresponding authors of included studies will be asked about relevant unpublished or ongoing work. The World Values Survey database (worldvaluessurvey.org) will be consulted for general population comparator data [16]. The date of each search will be reported.
No language restrictions apply. For non-English articles, the methods and results sections, including tables, figures, and supplementary material, will be translated in full using machine translation (DeepL or Google Translate), which has shown acceptable accuracy for data abstraction in systematic reviews [17]. Eligibility decisions and the extraction of key items — population and comparator definitions, behavioural definitions, recall periods, adjustment variables, and numerical data — will then be verified by a reviewer or colleague fluent in the language or, where none is available, confirmed with the study authors. Studies whose key items cannot be verified will be flagged and examined in a sensitivity analysis. The complete search strings for all databases appear in Supporting Information (S1 File).
The strategy is organized into three concept blocks joined by AND. Block A defines the population (health profession students and professionals); Block B covers the exposure (all forms of civic and community engagement); Block C addresses the comparator and study design. Within each block, synonyms are linked by OR and truncation (*) captures morphological variants. The PubMed example is shown in Table 1.
Study selection
All records will be exported in RIS format and deduplicated in Rayyan (rayyan.ai). After a calibration exercise on a random sample of 100 records, both reviewers (AG and RV) will independently screen titles and abstracts in Rayyan blind mode, followed by independent full-text assessment of potentially eligible records using a pre-piloted inclusion checklist. Disagreements at either stage will first be discussed between the two reviewers; where discussion does not resolve the matter, a third party will be consulted. Inter-rater agreement will be quantified using Cohen’s kappa at both the title/abstract and full-text stages. Multiple reports of the same study population will be linked and treated as a single study. A PRISMA 2020 flow diagram will document all decisions.
Data extraction
Both reviewers will extract data independently using a structured form developed and piloted before the search is run. The following will be recorded for each included study: study identity (first author, year, country, journal); design, sampling approach, response rate, and use of survey weights; data-collection period and recall period of each engagement measure; participant profile (profession, career stage, training year, age, sex); how civic engagement was operationalized and measured, including the capacity category and whether the instrument combines attitudes and behaviour; the scale or instrument used and its validation status; comparator type and source; outcome data, with adjusted and unadjusted estimates recorded separately together with the variables adjusted for; and, for the exploratory analysis, empathy measures and their correlation with engagement. Any discrepancies will be resolved by discussion, with a third party consulted where needed. If a paper reports insufficient detail for meta-analysis, we will contact corresponding authors and allow four weeks for a response, with one reminder, before proceeding with available data.
Risk of bias assessment
Risk of bias will be assessed independently by both reviewers, using tools matched to the purpose of each analysis. Studies contributing prevalence estimates will be appraised with the JBI Critical Appraisal Checklist for Studies Reporting Prevalence Data (9 items) [18], which addresses the sampling frame, sampling method, sample size, coverage, response rate, and the validity and reliability of outcome measurement. Studies contributing between-group comparisons will be appraised with the JBI Critical Appraisal Checklist for Analytical Cross-Sectional Studies (8 items) [19], which addresses the comparability of groups, the identification of and adjustment for confounders, and the measurement of exposure and outcome. A study contributing to both types of analysis will be appraised with both tools. The Newcastle-Ottawa Scale (maximum 9 stars) will be used only for genuine cohort studies with longitudinal follow-up; cross-sectional analyses of cohort data will be appraised with the relevant JBI tool. Disagreements will be resolved through discussion or, where necessary, by consultation with a third reviewer.
Overall judgements will be derived from pre-specified critical items rather than summed scores. For prevalence, the critical items are an appropriate sampling frame, an adequate response rate (≥60%, or evidence of no material non-response bias), and valid and reliable measurement of engagement. For comparisons, the critical items are the selection of comparison groups from comparable sources and the valid, reliable, and identical measurement of engagement in both groups; confounding is addressed separately by analysing adjusted and unadjusted estimates in different groups and is considered in GRADE. A study will be rated at high risk of bias if any critical item is judged ‘no’, moderate if any is ‘unclear’ (and none ‘no’), and low if all are ‘yes’. Cohort studies with Newcastle-Ottawa ratings of 7–9, 4–6, and 0–3 stars will be rated low, moderate, and high risk of bias, respectively. Item-level judgements will be reported for every study.
Measurement-instrument quality. Because most included studies are expected to rely on self-reported engagement, which is susceptible to social-desirability bias (for example, over-reporting of voting [20]), each instrument will additionally be classified as (i) validated — published evidence of reliability (e.g., internal consistency or test–retest) and validity (e.g., structural or construct validity) in a comparable population, judged against the criteria of Terwee et al. [21]; (ii) a standardized survey item — taken unchanged from an established population survey (e.g., the WVS or a national election study); (iii) adapted — modified from a validated instrument without re-validation; or (iv) ad hoc — developed for the study without reported psychometric evaluation. This classification will inform the measurement items of the JBI tools, will be reported for each study, and will define a sensitivity analysis excluding studies that use ad hoc instruments.
Data synthesis
Meta-analysis.
Analysis groups and criteria for pooling. Before data extraction, analysis groups will be defined by conceptual and methodological comparability: engagement domain (volunteering, political participation, NGO or civil-society involvement, charitable giving, activism and advocacy); career stage (students, practising professionals); comparator type (general population, education-matched adults, non-health students); comparator source (within-study, external benchmark); and effect measure (prevalence, OR, SMD, correlation). Adjusted and unadjusted estimates will never be combined. Pooling within an analysis group requires at least three studies (an author-defined criterion) and agreement between both reviewers that populations, behavioural definitions, and comparators are substantively comparable. Statistical heterogeneity will not be used as a threshold for pooling; between-study variance (τ2), the direction and consistency of individual study effects, and 95% prediction intervals will instead be used to judge how informative a pooled estimate is.
Proportions. Prevalence will be pooled using a random-intercept logistic regression model (a binomial-normal generalized linear mixed model) fitted by maximum likelihood [22,23]. This model uses exact within-study binomial likelihoods and accommodates proportions of 0% and 100% without continuity corrections. The Freeman-Tukey double arcsine transformation will not be used because of its known back-transformation problems [24]. As a sensitivity analysis, logit-transformed proportions will be pooled with an inverse-variance random-effects model, applying a 0.5 continuity correction to studies with 0% or 100%. Where a study reports only a survey-weighted prevalence with a design-based standard error, the logit inverse-variance model will be used for that analysis group.
Odds ratios. ORs will be analysed on the log scale. Where 2 × 2 data are available, unadjusted ORs will be computed from counts; otherwise, reported ORs and 95% CIs will be used. Unadjusted and adjusted ORs will be analysed separately. Adjusted ORs will be pooled only when the adjustment set includes at least age and sex; where several adjusted models are reported, the most fully adjusted model will be used, and the variables adjusted for will be tabulated. Zero cells will be handled with a 0.5 continuity correction, with a binomial-normal generalized linear mixed model as a sensitivity analysis.
Continuous outcomes. Continuous outcomes will be expressed as Hedges’ g, which incorporates a small-sample correction [25]. Scale direction will be harmonized before analysis so that higher values always indicate greater engagement; reverse-scored outcomes will be multiplied by −1. Missing standard deviations (SDs) will be derived from standard errors, confidence intervals, t statistics, or exact P values; medians with interquartile ranges or ranges will be converted to means and SDs using the methods of Wan et al. [26]. If an SD still cannot be obtained, authors will be contacted and, failing that, the SD will be imputed from the pooled SD of other included studies using the same instrument. Studies with converted or imputed SDs will be excluded in a sensitivity analysis. Scores from different instruments will be combined only when they measure the same behavioural construct.
Estimation. For ORs, SMDs, and correlation coefficients, random-effects models will be fitted with restricted maximum-likelihood (REML) estimation of τ2 [27], and confidence intervals will be calculated with the Hartung-Knapp-Sidik-Jonkman (HKSJ) method [28]. Results using the DerSimonian-Laird estimator, which was specified in the registered protocol, will be reported as a sensitivity analysis. A 95% prediction interval will be reported for every pooled estimate [29].
Dependent effect sizes. Within any single meta-analysis, each independent sample will contribute one effect size. When a study reports several eligible measures for the same domain, the hierarchy described under Outcomes will be applied. When several health profession groups are compared with one shared comparator group, the profession groups will be combined for the main analysis using the formulae in the Cochrane Handbook [30]; in profession-specific subgroup analyses, the shared comparator’s sample size will be divided across the comparisons to avoid double-counting. Where an analysis includes several dependent effect sizes per study (for example, meta-regression across engagement domains), a correlated and hierarchical effects model with robust variance estimation and small-sample (CR2) corrections will be used [31,32], assuming a within-study correlation of 0.6, with values of 0.2 and 0.8 examined in sensitivity analyses.
Heterogeneity will be described using τ2, the I2 statistic, and Cochran’s Q test. I2 will be interpreted using the overlapping ranges suggested in the Cochrane Handbook [30] and alongside prediction intervals, not as a fixed threshold. Analyses will be run in R (version 4.2.0 or later) using the meta, metafor, and clubSandwich packages [33].
Reporting bias.
Small-study effects will be assessed separately for each effect measure when an analysis includes at least 10 studies, using methods suited to that measure. For ORs, funnel plots will be inspected and asymmetry tested with Peters’ regression test, which avoids the artefactual correlation between log ORs and their standard errors [34]. For SMDs, funnel plots and Egger’s regression test will use a precision term based only on sample sizes, √(1/n1 + 1/n2), because the standard error of an SMD depends on the effect itself [35]. For single-group prevalence estimates, conventional funnel plots and Egger’s test are unreliable [36] and will not be used; instead, the association between study size and prevalence will be examined by meta-regression on the inverse square root of sample size. Funnel-plot asymmetry will be interpreted as an indication of small-study effects requiring further investigation — which may arise from heterogeneity, chance, or the relationship between effect estimates and their standard errors, as well as from publication bias — rather than as proof of publication bias [37]. The trim-and-fill procedure will not be used to correct pooled estimates and will be reported, if at all, only as a sensitivity analysis for ORs and SMDs. Our search design — which includes grey literature and preprints and imposes no language restrictions — is intended to reduce selective retrieval from the outset.
Certainty of evidence.
The GRADE framework will be applied to each primary outcome via GRADEpro GDT software (gradepro.org), separately for the two types of question addressed by this review. For participation prevalence, following GRADE guidance for estimates of event rates in broad populations [38], evidence from observational studies will start at high certainty and may be rated down for risk of bias (mainly sampling and measurement concerns), inconsistency (judged from prediction intervals and the spread of study estimates rather than I2 alone), indirectness (e.g., unrepresentative samples, pandemic-period data, or mixed or institution-organised measures), imprecision (a 95% CI wider than 20 percentage points), and publication bias. For between-group comparisons, evidence from observational studies will start at low certainty [39] and may be rated down for the same five domains, with indirectness including the use of external benchmarks or comparators not matched for education, and imprecision judged against the null effect (OR of 1 or SMD of 0) and the optimal information size. Rating up will be considered for a large magnitude of association (OR above 2 or below 0.5, one level; above 5 or below 0.2, two levels), a dose-response gradient, or when all plausible residual confounding, notably by education, would reduce an observed difference [40]. Each judgement and its rationale will be documented in Summary of Findings tables accompanying the final publication.
Narrative synthesis.
If meta-analysis is not appropriate for an analysis group — because fewer than three studies are available, or because the reviewers judge populations, behavioural definitions, or comparators insufficiently comparable — we will conduct a structured narrative synthesis following the SWiM (Synthesis Without Meta-Analysis) reporting guideline [41], grouping studies by the pre-specified analysis groups, tabulating effect estimates with their direction, and using vote counting based on the direction of effect where appropriate.
Subgroup analyses.
The following subgroup analyses are pre-specified and will be conducted wherever at least three studies per subgroup are available (Table 2). Differences between subgroups will be tested formally with a random-effects meta-regression that includes the subgroup as a categorical moderator (omnibus test with the Knapp-Hartung adjustment); a statistically significant estimate in one subgroup and a non-significant estimate in another will not, by themselves, be interpreted as a subgroup difference [42]. Subgroup findings will be treated as observational and hypothesis-generating.
Sensitivity analyses.
The following sensitivity analyses are pre-specified: (i) excluding studies rated at high risk of bias; (ii) a leave-one-out analysis, removing each study in turn, to detect any single study that exerts undue influence; (iii) excluding studies whose data-collection or recall period overlaps the COVID-19 pandemic period (11 March 2020–5 May 2023), classified as described under Eligibility criteria; (iv) excluding studies using ad hoc instruments; (v) restricting comparative analyses to within-study comparators; (vi) restricting analyses to personal-capacity engagement; (vii) excluding studies with converted or imputed SDs; (viii) excluding non-English studies whose key items could not be verified; and (ix) alternative estimation methods (the DerSimonian-Laird estimator; logit inverse-variance pooling of proportions; and within-study correlations of 0.2 and 0.8 in robust variance models).
Meta-regression.
Where at least 10 studies contribute to an analysis, univariable random-effects meta-regression will be used to explore heterogeneity, one covariate at a time; multivariable models will not be fitted, to limit overfitting. The pre-specified covariates are year of data collection, country GDP per capita (World Bank, purchasing-power parity), proportion of female participants, and mean participant age. To limit false-positive findings across covariates, P values will be obtained from permutation tests [43]. Because these covariates are study-level aggregates, the associations are ecological and will not be interpreted as individual-level or causal effects; for example, an association with the proportion of female participants would not show that women engage more than men [44].
Exploratory analysis of empathy.
Empathy is retained as an exploratory correlate for three reasons: empathic concern is central to theoretical accounts of altruistic and prosocial behaviour [45]; it is associated with volunteering in general-population studies [46,47]; and it is both an explicit target of health professions education and a frequently assumed link between clinical exposure and civic behaviour. Studies of health profession samples reporting both a validated empathy measure (e.g., the Jefferson Scale of Empathy [48] or the Interpersonal Reactivity Index [49]) and an eligible civic engagement outcome will contribute to this analysis. The effect measure will be the correlation coefficient between empathy and engagement. Pearson’s r will be extracted directly; Spearman’s ρ will also be included, with a sensitivity analysis restricted to Pearson’s r. Coefficients will be transformed to Fisher’s z, pooled with REML random-effects models and HKSJ confidence intervals, and back-transformed. Regression coefficients will not be pooled, because adjustment sets and scaling differ, and will be summarised narratively; between-group differences in empathy are outside the scope of this review. Few eligible studies are expected; if fewer than three are available, findings will be summarised narratively. This analysis is exploratory, will not inform GRADE ratings for the primary outcomes, and will not be used to support causal claims.
Protocol amendments
This version of the protocol (version 1.2) incorporates amendments made in response to peer review, before any data extraction or synthesis. Relative to the registered protocol (PROSPERO CRD420261348391), the amendments are: (1) classification of the capacity in which engagement occurs, and rules for instruments combining attitudes and behaviour; (2) three pre-specified comparator types, with the education-matched comparison moved from an exploratory to a formal subgroup analysis, and rules for external comparators; (3) classification of studies by data-collection period, with the pandemic period redefined as 11 March 2020–5 May 2023; (4) risk-of-bias tools matched to the purpose of each analysis, with appraisal of measurement-instrument quality; (5) replacement of the DerSimonian-Laird method with generalized linear mixed models for proportions and REML with HKSJ confidence intervals for other effect measures, with explicit rules for dependent effect sizes; (6) removal of the I2 > 75% threshold for pooling; (7) assessment of small-study effects tailored to each effect measure; (8) separate GRADE approaches for prevalence and between-group comparisons; (9) restriction of meta-regression to univariable models, with year of data collection replacing year of publication; (10) a specified analysis plan for the exploratory empathy outcome, broadened from the Jefferson Scale of Empathy to any validated empathy measure; and (11) additional grey-literature sources. The PROSPERO record has been updated to reflect these amendments. Any further amendments, including changes to the anticipated timeline, will be dated and recorded in PROSPERO, and all deviations from this protocol will be reported, with reasons, in the completed review.
Ethics and data management
Ethics. Ethical approval was not required for this study. This is a protocol for a systematic review and meta-analysis of previously published, aggregate data; no primary data will be collected from human participants and no individual-level or identifiable data will be accessed.
Data management and availability. Records will be managed in Rayyan for screening and in a piloted structured extraction form for data collection. No primary human-subjects data will be generated. On publication of the completed review, all data extracted from included studies, the risk-of-bias and instrument-quality judgements, the GRADE evidence profiles, the comparator estimates derived from the WVS, and the R code used for all analyses will be deposited in a public Open Science Framework repository with a persistent DOI under a CC BY 4.0 licence, and cited in the Data Availability statement of the completed review. The full search strategy is provided as Supporting Information (S1 File).
Study status and timeline. At the time of submission the review is at the protocol stage and database searches had not yet been executed. We anticipate completing the searches and title/abstract and full-text screening by the fourth quarter of 2026, data extraction and risk-of-bias assessment by the first quarter of 2027, and quantitative synthesis and GRADE assessment by the second quarter of 2027, with the completed review expected to be submitted for publication by mid-2027. If the timeline changes, the PROSPERO record will be updated accordingly.
Discussion
Unlike previous work, which has been confined to single professions or crisis settings, this review takes a cross-disciplinary approach and centres on everyday behavioural outcomes — voting, NGO participation, charitable giving, activism — measured against general-population and education-matched benchmarks. In doing so, it will assess whether the civic expectations embedded in health professions education are reflected in what health professionals actually do.
The practical stakes are considerable. Curriculum planners in health professions education will be able to ground social accountability teaching in comparative behavioural evidence rather than normative assumption alone. Professional bodies and licensing councils that invoke civic engagement as a component of good professional standing will, for the first time, have a systematic basis for that claim — or for revising it. Health ministries and international agencies that depend on health worker volunteerism and community engagement for public health programmes need reliable baseline data; this review is designed to provide them.
We acknowledge several methodological challenges. Substantial heterogeneity is expected given the diversity of countries, professions, and measurement tools; pooling will therefore be restricted to conceptually comparable analysis groups, and pooled estimates will be interpreted alongside prediction intervals, with narrative synthesis where pooling is not appropriate. Most evidence is likely to be cross-sectional and self-reported, so social-desirability bias and residual confounding — particularly by education and socioeconomic status — cannot be excluded; the education-matched comparisons, adjusted analyses, and appraisal of instrument quality are intended to mitigate, not eliminate, these concerns. Comparisons with external population datasets may be affected by differences in item wording, recall periods, and sampling frames and are therefore analysed separately. Although the search includes grey literature, preprints, and dissertations and imposes no language restrictions, unpublished data — particularly from settings without strong thesis or preprint repositories — may still be missed, and non-reporting bias cannot be ruled out. Finally, the search was not designed to identify studies of empathy specifically, so the exploratory empathy analysis may be incomplete.
Supporting information
S1 Checklist. PRISMA-P 2015 checklist.
Completed checklist indicating where each item is addressed in this protocol, with page and section references.
https://doi.org/10.1371/journal.pone.0363170.s001
(DOCX)
S1 File. Full search strategy.
Verbatim search strings for all seven databases and grey-literature sources.
https://doi.org/10.1371/journal.pone.0363170.s002
(DOCX)
Acknowledgments
The authors used Claude (Anthropic) for translation support and for editing the grammar and fluency of the English text. The authors reviewed and edited all output and take full responsibility for the content of this manuscript.
References
- 1. Gruen RL, Campbell EG, Blumenthal D. Public roles of US physicians: Community participation, political involvement, and collective advocacy. JAMA. 2006;296(20):2467–75. pmid:17119143
- 2. Wynia MK, Latham SR, Kao AC, Berg JW, Emanuel LL. Medical professionalism in society. N Engl J Med. 1999;341(21):1612–6. pmid:10577119
- 3. Reeve C, Woolley T, Ross SJ, Mohammadi L, Halili SB Jr, Cristobal F, et al. The impact of socially-accountable health professional education: A systematic review of the literature. Med Teach. 2017;39(1):67–73. pmid:27797293
- 4.
Global Consensus for Social Accountability of Medical Schools. 2010. Available from: https://healthsocialaccountability.sites.olt.ubc.ca/files/2011/06/11-06-07-GCSA-English-pdf-style.pdf
- 5. Boelen C, Woollard B. Social accountability and accreditation: A new frontier for educational institutions. Med Educ. 2009;43(9):887–94.
- 6. Adler RP, Goggin J. What do we mean by civic engagement? J Transform Educ. 2005;3(3):236–53.
- 7.
Verba S, Schlozman KL, Brady HE. Voice and Equality: Civic Voluntarism in American Politics. Cambridge: Harvard University Press; 1995.
- 8. Dharamsi S, Ho A, Spadafora SM, Woollard R. The physician as health advocate: Translating the quest for social responsibility into medical education and practice. Acad Med. 2011;86(9):1108–13. pmid:21785306
- 9. Grande D, Asch DA, Armstrong K. Do doctors vote? J Gen Intern Med. 2007;22(5):585–9.
- 10. Umar TP, Samudra MG, Nashor KMN, Agustini D, Syakurah RA. Health professional student’s volunteering activities during the COVID-19 pandemic: A systematic literature review. Front Med (Lausanne). 2022;9:797153. pmid:35928294
- 11. Hafferty FW, Franks R. The hidden curriculum, ethics teaching, and the structure of medical education. Acad Med. 1994;69(11):861–71. pmid:7945681
- 12. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. pmid:33782057
- 13. Shamseer L, Moher D, Clarke M, Ghersi D, Liberati A, Petticrew M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: Elaboration and explanation. BMJ. 2015;350:g7647. pmid:25555855
- 14. Doolittle A, Faul AC. Civic engagement scale: A validation study. SAGE Open. 2013;3(3).
- 15. Dee TS. Are there civic returns to education? J Public Econ. 2004;88(9–10):1697–720.
- 16.
World Values Survey Association. World Values Survey Wave 7 (2017-2022). [cited 2026 Mar 24] Available from: worldvaluessurvey.org
- 17. Jackson JL, Kuriyama A, Anton A, Choi A, Fournier J-P, Geier A-K, et al. The accuracy of google translate for abstracting data from non-english-language trials for systematic reviews. Ann Intern Med. 2019;171(9):677–9. pmid:31357212
- 18. Munn Z, Moola S, Lisy K, Riitano D, Tufanaru C. Methodological guidance for systematic reviews of observational epidemiological studies reporting prevalence and cumulative incidence data. Int J Evid Based Healthc. 2015;13(3):147–53. pmid:26317388
- 19.
Moola S, Munn Z, Tufanaru C, Aromataris E, Sears K, Sfetcu R. Chapter 7: Systematic reviews of etiology and risk. In: Aromataris E, Munn Z, editors. JBI Manual for Evidence Synthesis. JBI; 2020.
- 20. Bernstein R, Chadha A, Montjoy R. Overreporting voting: Why it happens and why it matters. Public Opin Q. 2001;65(1):22–44. pmid:11264053
- 21. Terwee CB, Bot SDM, de Boer MR, van der Windt DAWM, Knol DL, Dekker J, et al. Quality criteria were proposed for measurement properties of health status questionnaires. J Clin Epidemiol. 2007;60(1):34–42. pmid:17161752
- 22. Stijnen T, Hamza TH, Ozdemir P. Random effects meta-analysis of event outcome in the framework of the generalized linear mixed model with applications in sparse data. Stat Med. 2010;29(29):3046–67. pmid:20827667
- 23. Lin L, Chu H. Meta-analysis of proportions using generalized linear mixed models. Epidemiology. 2020;31(5):713–7. pmid:32657954
- 24. Schwarzer G, Chemaitelly H, Abu-Raddad LJ, Rücker G. Seriously misleading results using inverse of Freeman-Tukey double arcsine transformation in meta-analysis of single proportions. Res Synth Methods. 2019;10(3):476–83. pmid:30945438
- 25. Hedges LV. Distribution theory for Glass’s estimator of effect size and related estimators. J Educ Stat. 1981;6(2):107–28.
- 26. Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol. 2014;14:135. pmid:25524443
- 27. Veroniki AA, Jackson D, Viechtbauer W, Bender R, Bowden J, Knapp G, et al. Methods to estimate the between-study variance and its uncertainty in meta-analysis. Res Synth Methods. 2016;7(1):55–79. pmid:26332144
- 28. IntHout J, Ioannidis JPA, Borm GF. The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method. BMC Med Res Methodol. 2014;14:25. pmid:24548571
- 29. IntHout J, Ioannidis JPA, Rovers MM, Goeman JJ. Plea for routinely presenting prediction intervals in meta-analysis. BMJ Open. 2016;6(7):e010247. pmid:27406637
- 30.
Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ. Cochrane Handbook for Systematic Reviews of Interventions. 6.5 ed. Cochrane; 2024.
- 31. Hedges LV, Tipton E, Johnson MC. Robust variance estimation in meta-regression with dependent effect size estimates. Res Synth Methods. 2010;1(1):39–65. pmid:26056092
- 32. Pustejovsky JE, Tipton E. Meta-analysis with robust variance estimation: Expanding the range of working models. Prev Sci. 2022;23(3):425–38. pmid:33961175
- 33. Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat Softw. 2010;36(3):1–48.
- 34. Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Comparison of two methods to detect publication bias in meta-analysis. JAMA. 2006;295(6):676–80. pmid:16467236
- 35. Pustejovsky JE, Rodgers MA. Testing for funnel plot asymmetry of standardized mean differences. Res Synth Methods. 2019;10(1):57–71. pmid:30506832
- 36. Hunter JP, Saratzis A, Sutton AJ, Boucher RH, Sayers RD, Bown MJ. In meta-analyses of proportion studies, funnel plots were found to be an inaccurate method of assessing publication bias. J Clin Epidemiol. 2014;67(8):897–903. pmid:24794697
- 37. Sterne JAC, Sutton AJ, Ioannidis JPA, Terrin N, Jones DR, Lau J, et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ. 2011;343:d4002. pmid:21784880
- 38. Iorio A, Spencer FA, Falavigna M, Alba C, Lang E, Burnand B, et al. Use of GRADE for assessment of evidence about prognosis: Rating confidence in estimates of event rates in broad categories of patients. BMJ. 2015;350:h870. pmid:25775931
- 39. Balshem H, Helfand M, Schünemann HJ, Oxman AD, Kunz R, Brozek J, et al. GRADE guidelines: 3. Rating the quality of evidence. J Clin Epidemiol. 2011;64(4):401–6. pmid:21208779
- 40. Guyatt GH, Oxman AD, Sultan S, Glasziou P, Akl EA, Alonso-Coello P, et al. GRADE guidelines: 9. Rating up the quality of evidence. J Clin Epidemiol. 2011;64(12):1311–6. pmid:21802902
- 41. Campbell M, McKenzie JE, Sowden A, Katikireddi SV, Brennan SE, Ellis S, et al. Synthesis without meta-analysis (SWiM) in systematic reviews: Reporting guideline. BMJ. 2020;368:l6890. pmid:31948937
- 42. Borenstein M, Higgins JPT. Meta-analysis and subgroups. Prev Sci. 2013;14(2):134–43. pmid:23479191
- 43. Higgins JPT, Thompson SG. Controlling the risk of spurious findings from meta-regression. Stat Med. 2004;23(11):1663–82. pmid:15160401
- 44. Thompson SG, Higgins JPT. How should meta-regression analyses be undertaken and interpreted? Stat Med. 2002;21(11):1559–73. pmid:12111920
- 45.
Batson CD. Altruism in Humans. New York: Oxford University Press; 2011.
- 46. Wilson J. Volunteering. Annu Rev Sociol. 2000;26:215–40.
- 47. Bekkers R. Participation in voluntary associations: Relations with resources, personality, and political values. Polit Psychol. 2005;26(3):439–54.
- 48. Hojat M, Mangione S, Nasca TJ, Cohen MJM, Gonnella JS, Erdmann JB. The Jefferson Scale of physician empathy: Development and preliminary psychometric data. Educ Psychol Meas. 2001;61(2):349–65.
- 49. Davis MH. Measuring individual differences in empathy: Evidence for a multidimensional approach. J Pers Soc Psychol. 1983;44(1):113–26.

관련 뉴스