RESPONSE TO REVIEWERS
Manuscript: Community health workers’ perceptions of depression and barriers and facilitators to mental health care in Sierra Leone: A qualitative study
Article type: Original research | Decision: Revision
We thank the Editor and both reviewers for a careful and constructive reading of our manuscript. The comments were specific and actionable, and the paper is materially stronger for them. We have addressed every point raised. Where the two reviewers gave advice pointing in different directions on the reporting of frequency, we set out below the reconciliation we adopted and our reasoning, so that both reviewers can see that neither comment has been quietly set aside.
Summary of the principal changes
● A single overarching research question has been added to the introduction, and the three objectives are now presented as hierarchically ordered subsidiary questions serving it.
● The scope of the Jorm framework is now explicitly bounded to Theme 1, and a complementary access framework has been introduced to interpret Themes 2 to 5.
● A new methods subsection, Sample size and information power, appraises sample adequacy against all five dimensions of the information power framework rather than claiming saturation.
● A new methods paragraph reports the steps taken to contextualise the vignette and states plainly what was not done.
● The coding process is now reported phase by phase against Braun and Clarke, with the number of initial codes, subthemes, and themes stated and the fate of candidate themes described.
● Two new tables have been added: Table 2, the analytical progression from initial codes to subthemes and final themes; and Table 3, a summary of the five themes, 13 subthemes, and a representative extract for each.
● The thematic map, previously referred to but absent, is now included as Figure 1.
● A defined convention for reporting frequency has been added to the methods and applied consistently throughout the results. Percentages have been removed from the narrative text and from Table 1.
● Analytic commentary has been added at seven points in the results so that quotations are tied to the interpretive argument they are intended to support.
● The discussion now engages with two specific evidence-based interventions from comparable settings and sets out the implementation constraints specific to Sierra Leone.
● The abstract now reports the proportion who failed to identify suicidal ideation and the proportion endorsing psychosocial-only aetiologies.
● The limitations section has been substantially expanded, and the epidemiological claim about suicide is now sourced to WHO.
Reviewer 1
|
General assessment Thank you for the opportunity to review this manuscript. This article addresses an important and underexplored topic, namely the mental health literacy of community health workers (CHWs) in Sierra Leone, a country marked by a long history of collective trauma. The empirical contribution is genuine and valuable, particularly the original finding regarding the gap in recognising suicidal ideation among CHWs, a finding with direct implications for training and suicide prevention in resource-limited health systems. That said, the sample size (n=10, two sites only) limits the transferability of the findings, and I would encourage the authors to exercise caution when drawing broader conclusions, ensuring that generalizations remain within the boundaries of what the data can reasonably support. The suggestions below are intended to strengthen the impact of this article without altering its fundamental structure. |
Response
We are grateful for this generous assessment and for the recognition of the suicide recognition finding, which we agree is the most consequential result in the paper. We accept the caution regarding transferability without reservation and have acted on it in three places rather than one.
First, the new methods subsection on sample size and information power states explicitly that the dataset supports description of the range of accounts present among a specific and comparatively well-informed group of CHWs in two districts, and that it does not support estimation of how common any account is among CHWs nationally. We state that no such inference is drawn anywhere in the paper. Second, the limitations section now notes that Kono is atypical among rural districts in being diamond-producing, with associated patterns of migration and livelihood that may shape both distress and service use, so that the urban and rural contrast should not be read as a national one. Third, we have added a limitation noting that English-language ability was an eligibility criterion, which excluded CHWs working primarily in Krio, Temne, Mende, or Kono, and is likely to have selected for better-educated and better-trained workers. We now state directly that the recognition rates reported here probably sit at the upper end of what would be found across the cadre as a whole. We consider this last point an important correction to the reading of our own results.
Comment 1. Explicit formulation of the research question
|
Reviewer 1, comment 1 The article would gain considerably in clarity and impact if a central research question were explicitly stated in the introduction. As currently written, the manuscript presents three distinct aims without a clear hierarchy between them, which makes it difficult for the reader to identify the study’s overarching thread. I would encourage the authors to consider formulating a unifying research question that brings these three aims together and strengthens the overall coherence of the manuscript. |
Response
We agree, and this comment identified a real structural weakness rather than a presentational one. The three aims had been generated separately and were genuinely parallel in the original draft.
The introduction now states a single overarching research question: how do community health workers in Sierra Leone understand depression, and how do they account for the barriers to and facilitators of community mental health care, including under the conditions imposed by the COVID-19 pandemic? The three aims are now presented as subsidiary objectives, each addressing one component of that question.
We have also added a paragraph immediately after the objectives making the hierarchy explicit rather than leaving the reader to infer it. The first objective establishes what CHWs know and believe about depression; the second and third situate that knowledge within the service environment in which CHWs work, first under routine conditions and then under the conditions of an acute public health emergency. We state that the unifying thread is the relationship between what CHWs are able to recognise and what the system around them enables them to do about it. This thread is now carried through the discussion, which returns to it in interpreting the suicide recognition finding alongside the structural barriers.
The abstract objectives statement has been revised to open with the research question so that the framing is visible from the first page.
Changes made to the manuscript
● Introduction, objectives paragraph: overarching research question added; the three aims reframed as subsidiary objectives.
● Introduction: new paragraph added after the objectives setting out their hierarchical relationship to the central question.
● Abstract, Objectives: revised to lead with the research question.
Comment 2. Precision in quantification
|
Reviewer 1, comment 2 Throughout the results section, the manuscript relies on vague expressions such as "most participants" or "some participants" without specifying the corresponding figures. In a sample of ten individuals, this imprecision is analytically problematic: the difference between six and eight participants is substantive and can meaningfully affect the interpretation of findings. I would strongly encourage the authors to systematically replace these expressions with exact numbers or clear fractions across the entire results section. |
Response
We accept this comment and have acted on it. We note respectfully, and transparently, that it sits in tension with Reviewer 2’s comment 2 on the results, which asks us to reduce numeric precision and use language such as "most", "several", or "a minority", reserving numeric counts for clearly categorical findings. Rather than choose between the two reviewers, we have adopted a solution that we believe satisfies the substantive concern behind each.
A defined convention for reporting frequency has been added to the methods and is applied consistently throughout the results. Exact counts are given for the framework-coded findings under Theme 1, where every participant responded to the same standard prompt about the same vignette, so that a count is both meaningful and complete. For the inductively derived Themes 2 to 5, where accounts were volunteered rather than elicited by a fixed question and the absence of a statement cannot be read as disagreement, verbal quantifiers are used, but each now carries a stated numeric range: all (10 participants), most (7 to 9), several (4 to 6), a few (2 to 3), and one (1). No vague quantifier remains in the results that is not covered by this convention.
The effect is that a reader encountering "several participants" in Theme 3 now knows this means between four and six of the ten, which was precisely the ambiguity the reviewer identified. We chose ranges rather than point counts for the inductive themes because a point count would misrepresent the data: a CHW who did not raise poverty as a barrier did not thereby deny that poverty is a barrier, and reporting an exact denominator for a volunteered observation would imply a completeness the interview structure cannot support. For the vignette-based findings, where every participant did answer the same question, we have given exact counts throughout as the reviewer asks.
Specific changes include: seven of the ten identified depression, two described it as stress, and one as anxiety; none of the ten identified the suicidal ideation; all ten attributed the presentation to psychosocial circumstances, making psychosocial attribution unanimous at 10 of 10; all ten believed professional help was needed. The participant characteristics paragraph has been rewritten to give counts for every category, including those previously omitted.
Changes made to the manuscript
● Methods: new subsection Reporting of frequency, stating the convention and its rationale.
● Results, Theme 1: all findings now reported as exact counts out of ten.
● Results, Themes 2 to 5: all vague quantifiers replaced with convention terms carrying defined numeric ranges.
● Results, Participant characteristics: rewritten to report counts for every demographic category.
● Abstract, Results: revised to report exact counts.
Comment 3. Analytical progression table
|
Reviewer 1, comment 3 The transparency of the analytical process would be considerably strengthened by the addition of a simple table illustrating the progression from initial codes to subthemes and final themes. This addition would not require any further analytical work, since the analysis has already been completed. It would, however, allow readers to assess the rigour of the thematic process and would meaningfully enhance the credibility of the study. |
Response
We agree and have added this table. Table 2, Progression from initial codes to subthemes and final themes, now sets out the full analytical chain for all five themes and all 13 subthemes, with the initial codes collapsed into each subtheme listed and counted.
The table shows that 70 initial codes were generated at Phase 2 and collapsed into 13 subthemes and five final themes. Because Reviewer 2 raised a closely related point about the coding process, we have also added a narrative paragraph to the methods reporting the analysis phase by phase against Braun and Clarke, including the fate of candidate themes during Phase 4. Table 2 and that paragraph are intended to be read together, so that a reader can follow the analytical decisions rather than take them on trust.
Changes made to the manuscript
● New Table 2 added after the overview of themes, listing initial codes, subthemes, themes, and code counts.
● Methods, Data analysis: new paragraph reporting the analysis phase by phase, with code, subtheme, and theme counts and the fate of candidate themes.
Comment 4. Visual synthesis of the results
|
Reviewer 1, comment 4 Finally, I would suggest adding a summary table presenting the five themes, their respective subthemes, and one representative quotation for each. Such a visual synthesis would greatly improve the readability of the manuscript and strengthen its impact for readers who are less familiar with qualitative methods. The thematic map mentioned in the text but absent from the manuscript could also be included as a complementary figure, and I would encourage the authors to consider this addition. |
Response
Both additions have been made, and we apologise for the omission of the thematic map from the submitted version, which was an oversight in assembling the manuscript rather than a deliberate choice.
Table 3, Summary of themes, subthemes, and representative extracts, presents all five themes and all 13 subthemes with one representative extract for each, exactly as suggested.
Figure 1 now presents the thematic map. It shows equitable access to mental health care as the central need articulated by participants, shaped on one side by individual-level knowledge and disposition (Themes 1 and 2) and on the other by system, family, and community conditions (Themes 3 and 4), with Theme 5 represented as a cross-cutting condition in which an acute shock intensified the constraints described in the other four themes. The sentence in the results that previously referred to the map in the past tense has been rewritten to point to the figure.
We note that constructing the map prompted one substantive improvement to the analysis. Theme 5 had no named subthemes in the submitted version, which made it structurally inconsistent with the other four. We have now named three subthemes for Theme 5, namely fear and exposure, limited preparedness and disrupted delivery, and local adaptation, and added the corresponding subheadings to the results text. This is a presentational change only; no data have been recoded and no interpretation has changed.
Changes made to the manuscript
● New Table 3 added, summarising themes, subthemes, and one representative extract each.
● New Figure 1 added, presenting the thematic map, with an explanatory note.
● Results, overview of themes: sentence rewritten to reference Figure 1, Table 2, and Table 3.
● Results, Theme 5: three subthemes named and subheadings added for structural consistency with the other four themes.
Reviewer 2
|
Overall assessment Thank you for sharing such excellent body of work with the scholarly community. This manuscript represents a strong and timely public health contribution with generally solid methodological grounding. Overall Assessment: The paper addresses a genuine evidence gap on CHW mental health literacy in Sierra Leone and is situated within a well-articulated policy context. The use of COREQ, reflexivity statement, and member-checking via inter-rater coding are appropriate practices. However, ten participants across two sites is modest even for qualitative research. The manuscript does not discuss data saturation or information power (Malterud et al., 2016). The authors should explicitly address whether saturation was reached or explain why the sample size is sufficient given the study’s scope and aims. |
Response
We thank the reviewer for this assessment and for directing us to the information power framework, which we agree is the more defensible basis for appraising a sample of this kind.
We have added a new methods subsection, Sample size and information power. We state that saturation was not adopted as the criterion, and we give our reason: the concept originates within grounded theory, where it forms part of constant comparison, and its transfer to qualitative descriptive work is contested and inconsistently operationalised. We considered it more honest to appraise the sample against information power than to assert saturation retrospectively, which we could not have substantiated.
The subsection works through all five dimensions of Malterud et al. explicitly. The aim was narrow rather than exploratory, being restricted to CHWs’ understanding of a single presented condition and their accounts of a single service system. Sample specificity was high, since every participant was a serving CHW with prior exposure to mental health cases, recruited purposively against explicit criteria. Established theory supported the analysis, with the Jorm framework providing a deductive scaffold for the first objective. Dialogue quality was supported by a single experienced clinician interviewer, a standard vignette giving every participant an identical stimulus, and transcript checking against the recordings. The analysis strategy was cross-case rather than in-depth narrative. On these five grounds we judge ten interviews across two contrasting districts to hold sufficient information power for the stated aim.
We have been careful not to overclaim. The subsection closes by stating that this is a boundary condition on the findings rather than a claim of completeness, and the limitations section repeats the point in the terms the reviewer used, that ten participants across two sites is modest even for qualitative work.
Changes made to the manuscript
● Methods: new subsection Sample size and information power addressing all five dimensions of Malterud et al. (2016) and explaining why saturation was not claimed.
● Limitations: first limitation rewritten to restate sample adequacy in information power terms and to disclaim any prevalence inference.
● Abstract, Methods: sentence added noting that sample adequacy was appraised using information power rather than data saturation.
● References: Malterud et al. (2016) added.
Abstract
|
Reviewer 2, abstract The abstract does not report the proportion of CHWs who failed to identify suicidal ideation (100%), or the proportion who endorsed psychosocial-only aetiologies. These are among the most clinically significant findings and should appear in the results section of the abstract. |
Response
We agree entirely. Both findings now appear in the abstract results.
The abstract now states that none of the ten participants identified the suicidal ideation described in the vignette as a clinical risk or marker of severity, and that all ten attributed the presentation exclusively to psychosocial and life circumstances, with no participant offering a biological, spiritual, or supernatural explanation or mentioning medication or longer-term treatment planning.
Following Reviewer 1’s comment on quantification and the reviewer’s own comment on percentages, we have expressed these as counts out of ten rather than as percentages. We judged that "none of the ten participants" carries the finding more forcefully than "100%" while avoiding the implied precision the reviewer objects to elsewhere. We hope the reviewer agrees, but we will readily convert these to percentages if preferred.
The abstract conclusion has also been sharpened. It previously said that gaps remained in identifying suicide risk, which understated the finding. It now states that CHWs demonstrated partial recognition of depression but a complete absence of spontaneous suicide risk identification.
Changes made to the manuscript
● Abstract, Results: both findings added, expressed as counts out of ten.
● Abstract, Conclusions: revised to state the complete absence of spontaneous suicide risk identification.
Introduction and conceptual framework, comment 1
|
Reviewer 2, introduction comment 1 The introduction invokes Jorm et al.’s (1997) mental health literacy framework as the primary theoretical lens, yet the discussion extends substantially beyond it into service systems, structural barriers, and COVID-19 experiences. The manuscript should either more explicitly bound the scope of the Jorm framework to Theme 1 and use a complementary framework for Themes 2 to 5, or justify the framework’s applicability to the full study. |
Response
This is a fair criticism and we have taken the first of the two options offered, since we do not think the second is defensible. The Jorm framework was not designed to explain how services are organised, financed, or reached, and stretching it to cover Themes 2 to 5 would have been a post hoc rationalisation.
A new paragraph has been added to the introduction which states that the Jorm framework is bounded in this study to the first objective, the assessment of CHWs’ knowledge and beliefs about depression, and is not treated as a complete account of the treatment gap.
For Themes 2 to 5 we have introduced the patient-centred access framework of Levesque et al. (2013), which conceptualises access as a stepwise process spanning approachability, acceptability, availability and accommodation, affordability, and appropriateness, each paired with a corresponding ability on the population side. We chose this framework because it accommodates both the supply-side and demand-side barriers our participants described, and because it allows mental health literacy to be positioned as one determinant of access among several rather than as the sole explanation for low service use. That repositioning is, in our view, the substantive gain from the reviewer’s comment.
The framework is not merely announced and then abandoned. It is used analytically in the results, where the structural barriers subtheme is now read as describing simultaneous failures of availability, affordability, and appropriateness rather than a single point of breakdown, and it supplies the vocabulary for the corresponding discussion section. The paragraph explaining the objective hierarchy also makes clear which framework applies to which objective.
Changes made to the manuscript
● Introduction: new paragraph bounding the Jorm framework to Theme 1 and introducing Levesque et al. (2013) for Themes 2 to 5.
● Introduction: objective hierarchy paragraph specifies which framework applies to which objective.
● Results, Theme 3, structural barriers: new analytic paragraph applying the access framework.
● References: Levesque et al. (2013) added.
Introduction and conceptual framework, comment 2
|
Reviewer 2, introduction comment 2 The statement that suicide is "among the leading causes of death among young people aged 15–29" cites Khatami & Khodabakhshi-Koolaee (2021), but WHO global data are the standard reference for such claims. The authors should verify and cite the most current and authoritative epidemiological source. |
Response
The reviewer is correct and we have corrected this. The original citation was to a small qualitative study of unsuccessful suicide attempts in Iran, which is not an appropriate source for a global epidemiological claim.
We verified the current WHO position and have replaced both the claim and the citation. The sentence now reads that suicide was the third leading cause of death among people aged 15 to 29 years globally in 2021, that an estimated 727,000 people died by suicide in that year, and that 73% of these deaths occurred in low- and middle-income countries, cited to the current WHO suicide fact sheet. We note that this is an update on the position as well as a change of source, since suicide ranked fourth in this age group on 2019 data. The additional detail on the low- and middle-income country share strengthens the framing of the paper, so we have retained it rather than citing the rank alone.
The Khatami and Khodabakhshi-Koolaee reference is no longer cited anywhere in the manuscript and has been deleted from the reference list.
Changes made to the manuscript
● Introduction, paragraph 1: claim and citation replaced with the current WHO source, with updated figures.
● References: WHO (2025) suicide fact sheet added; Khatami and Khodabakhshi-Koolaee (2021) deleted.
Methodology
|
Reviewer 2, methodology The vignette was originally developed and validated in high-income, Western contexts. The authors acknowledge this as a limitation but do not describe any steps taken to culturally adapt or cognitively test the vignette for Sierra Leonean CHWs prior to use. This is a substantial methodological gap given that aetiological attributions and symptom vocabularies differ markedly across settings. |
Response
We accept this comment and have responded to it in full, including by stating plainly what we did not do. We considered it important not to present the limited contextualisation we undertook as though it constituted formal adaptation.
A new methods paragraph now reports that the protagonist was given a locally common Sierra Leonean forename and that the descriptive content was reviewed by the first author, a Sierra Leonean psychiatrist and Mental Health Programme Manager, for consistency with the symptom vocabularies and social circumstances CHWs encounter in routine practice. We also explain that the clinical structure of the vignette was deliberately retained so that findings would remain comparable with prior studies using the same instrument in Scotland and Kenya.
The same paragraph then states without qualification that no formal cognitive interviewing or structured cultural adaptation procedure was undertaken with CHWs before fieldwork, and that local idioms of distress were not systematically elicited or incorporated. We describe this as a genuine methodological limitation rather than an omission in reporting, and we note, citing Faregh et al. (2019) and Mendenhall et al. (2019), that a vignette not cognitively tested locally may under-represent culturally salient framings of distress. The instrument is now framed as a standardised elicitation device supporting cross-study comparison rather than as a culturally validated measure.
The reviewer’s comment also prompted us to re-examine one of our own findings, and we think this is the most valuable consequence of the point. The complete absence of spiritual and supernatural attributions in response to the vignette sits in direct tension with our participants’ own reports, under Theme 3, that such attributions are widespread in the communities they serve. The expanded limitations section now sets out three candidate explanations: a genuine professional distinction between CHWs’ explanatory models and those of their communities; social desirability in an interview conducted by a senior mental health professional; or a vignette whose framing cued a biomedical register. We state that these cannot be distinguished with the present data and that the finding should be treated as provisional until replicated with a locally adapted and cognitively tested instrument. We would not have surfaced this tension without the reviewer’s comment.
Changes made to the manuscript
● Methods, Interview guide and vignette: new paragraph reporting contextualisation steps taken and stating explicitly what was not done.
● Limitations: second limitation substantially expanded, including the three candidate explanations for the absent spiritual attributions.
● References: Faregh et al. (2019) added; Mendenhall et al. (2019), previously listed but uncited, is now cited in text.
Results, comment 1
|
Reviewer 2, results comment 1 The results present five themes but do not describe how many initial codes were generated, how codes were collapsed into themes, or whether any themes were discarded. A codebook or supplementary coding tree would substantially improve transparency. Braun and Clarke’s (2006) phases are referenced in the methods but the results section does not map back to them. |
Response
We agree and have addressed each element of this comment.
A new methods paragraph now reports the analysis phase by phase. It states that coding was conducted at the semantic level across all ten transcripts; that Phase 2 generated 70 initial codes; that in Phase 3 these were sorted into 13 candidate subthemes and five candidate themes; that in Phase 4 the candidate structure was checked against both the coded extracts and the full dataset, with refinement involving relabelling of subthemes and reallocation of codes between them; that no candidate theme was discarded or collapsed into another; that Phase 5 fixed the final definitions and names; and that Phase 6 produced the written account.
The coding tree the reviewer asks for is now provided as Table 2, which sets out the full progression from initial codes through subthemes to final themes, with the codes collapsed into each subtheme listed and counted. This also answers Reviewer 1’s third comment, and we have placed it in the main text rather than in supplementary material because both reviewers identified analytical transparency as central to the credibility of the study.
On mapping the results back to the phases, we have kept the phase-by-phase account in the methods, where reporting conventions place it, but have made the results section traceable to it in two ways: the overview of themes now directs the reader to Table 2 and Table 3 for the analytical progression and the theme summary respectively, and Theme 5 has been given named subthemes so that all five themes now display the same structure that Table 2 describes. If the reviewer would prefer the phase mapping restated within the results section itself, we will move it.
Changes made to the manuscript
● Methods, Data analysis: new paragraph reporting code, subtheme, and theme counts, the collapsing process, and the fate of candidate themes across all six phases.
● New Table 2 providing the full coding tree.
● Results, overview of themes: cross-references to Figure 1, Table 2, and Table 3 added.
● Results, Theme 5: subthemes named and subheadings added.
Results, comment 2
|
Reviewer 2, results comment 2 Percentages are used extensively (e.g., ‘70% correctly identified depression’, ‘20% described it as stress’). While descriptive counts are acceptable in qualitative descriptive research, the repeated use of percentages derived from a sample of 10 risks misleading precision. The authors should use language such as ‘most’, ‘several’, or ‘a minority’ where appropriate, reserving numeric counts for clearly categorical findings. |
Response
We accept this comment and have implemented it. As noted under Reviewer 1’s second comment, the two reviewers gave advice pointing in different directions here, and we have sought a resolution that honours the substance of both rather than favouring one.
Percentages have been removed from the narrative text of the results entirely. The methods now state the reason: with a denominator of ten, a percentage conveys no more information than the count while implying a precision the design cannot support, and it invites inappropriate comparison with survey estimates derived from far larger samples.
We have followed the reviewer’s guidance on reserving numeric counts for clearly categorical findings. Counts are now used only for the Theme 1 vignette responses, which are categorical in exactly the sense the reviewer intends: every participant answered the same standard prompt, and each response was assigned to one mutually exclusive category. For the inductively derived Themes 2 to 5 we use the verbal quantifiers the reviewer suggests, with each term given a defined numeric range in the methods so that Reviewer 1’s concern about ambiguity is also met.
We have gone further than the comment strictly required in two places. First, the Percent column has been deleted from Table 1, since with ten participants it simply restated the frequency column; a note explaining this has been added beneath the table. Second, the discussion previously compared our 70% recognition rate with percentages reported in larger studies. That comparison is now expressed as seven of ten and is explicitly qualified: we note that the comparator studies had substantially larger samples in which a proportion carries statistical meaning that it cannot carry here. We have also added the observation that our deductive coding frame, which treated any explicit use of the term depression as correct recognition, applies a lower threshold than instruments requiring elaboration of symptoms or severity, and that on a stricter criterion including recognition of the suicidal ideation, correct recognition in this sample would have been zero.
Changes made to the manuscript
● Methods: new Reporting of frequency subsection.
● Results: all narrative percentages removed.
● Table 1: Percent column deleted; explanatory note added.
● Discussion, mental health literacy section: cross-study comparison reworded and qualified; new sentences added on the recognition threshold.
Results, comment 3
|
Reviewer 2, results comment 3 Several quotations are presented in sequence without explicit analytic commentary linking them to the theme or subtheme they are intended to support. The narrative analysis would benefit from tighter integration between quotation and interpretive argument throughout. |
Response
We accept this and have revised the results substantially in response. Blocks of quotations previously stood without interpretation, and we agree that this left the analytic work implicit.
Analytic commentary has been added or substantially strengthened at seven points, following each block of quotations rather than at the end of each theme, so that the interpretive argument sits alongside the evidence for it. In each case the commentary names the participants whose extracts are being interpreted, so the link between quotation and argument is explicit rather than implied.
The additions are analytic rather than descriptive. Under Theme 1, we now observe that P004 named the condition correctly but described it as a state rather than as a condition carrying risk, while P002 and P005 substituted non-diagnostic labels that would not trigger referral in routine practice, and that across all three the element carrying the greatest clinical urgency went uncommented on. Following the course of action extracts we note that P006 and P007 could both articulate suicide as a downstream consequence of untreated illness yet neither had identified the suicidal ideation already present in the vignette, so that suicide was available to participants as an abstract prognostic category but not as a present-state clinical observation. This locates the training gap in risk detection rather than in awareness of risk as a concept, which we regard as a genuine sharpening of the paper’s central contribution and one we owe to this comment.
Comparable commentary has been added under Theme 2, distinguishing the unpredictability of the CHW role from the attitudes of individual community members; under Theme 3, distinguishing stigma as a sanction following disclosure from an explanatory barrier operating before distress is construed as a health matter, and noting that participants attributed responsibility to successively higher tiers of the system without describing any functioning line of accountability; under Theme 4, noting what participants did not propose and why that is consistent with the narrow treatment repertoire seen in Theme 1; and under Theme 5, noting that adaptation was described in the first person plural of a local team rather than as an instruction received from any level of the system.
Changes made to the manuscript
● Results, Theme 1: analytic commentary added after the conceptualisation extracts, the aetiology extracts, and the course of action extracts.
● Results, Theme 2: new analytic paragraph after the family and religious leader extracts; closing paragraph of the second subtheme rewritten.
● Results, Theme 3: new analytic paragraphs after the awareness extracts and the structural barrier extracts.
● Results, Theme 4: new analytic paragraph following the policy extracts.
● Results, Theme 5: analytic commentary added to the closing paragraph.
Discussion
|
Reviewer 2, discussion Recommendations for training, sensitisation, and policy reform are appropriate but generic. The discussion would be strengthened by reference to specific, evidence-based interventions in comparable LMIC settings (e.g., Friendship Bench in Zimbabwe, mhGAP adaptations in Nigeria) and by acknowledging implementation challenges specific to Sierra Leone’s health system context. |
Response
We agree that the recommendations were generic and have rewritten this part of the discussion around the two examples the reviewer suggests, together with a section on Sierra Leone specific constraints. Three new paragraphs have been added.
The first discusses the Friendship Bench (Chibanda et al., 2016) and identifies three features of that model that are directly transferable to the cadre we studied. The intervention was manualised, so delivery did not depend on the individual provider exercising clinical judgement; it was supervised, through regular structured contact between lay providers and clinical supervisors; and it was sited, so that the lay provider held a defined physical and institutional place within the service rather than an ambiguous community role. We then observe that precisely these three features are absent from our participants’ accounts, and in particular that no participant described any supervision structure for mental health work, which indicates where investment would have most effect.
The second discusses the Nigerian STEPCARE trial (Gureje et al., 2019), which found a stepped-care package delivered by lay health workers similarly effective to usual care enhanced with the mhGAP Intervention Guide. We draw the implication for Sierra Leone directly: since some CHWs have had exposure to mhGAP but no standardised national training pathway exists for this cadre, the evidence argues for consolidating and systematising mhGAP-based support with adequate supervision rather than for designing a novel intervention, and suggests that the marginal return on a more elaborate psychological package is likely to be small where the basic enhancement has not been reliably delivered.
The third paragraph sets out the implementation constraints specific to Sierra Leone rather than assuming them away. The specialist base available for supervision is extremely small, with two psychiatrists and 27 mental health nurses for approximately 7.5 million people, so any supervision model must be cascaded, nurse-led, and partly remote rather than psychiatrist-led. The decentralised nurse-led mental health units are the most plausible supervisory anchor but their caseloads and coverage are uneven, so coverage would need mapping before any national rollout. Psychotropic medication supply is intermittent, which constrains what a referral can deliver and, if unaddressed, risks training CHWs to detect conditions the system cannot then treat, with predictable effects on both patient trust and CHW morale. Finally, mhGAP materials require adaptation to local explanatory models rather than direct transfer, and reviews of mhGAP implementation have repeatedly identified inattention to culture, context, and community as a principal cause of weak uptake. We conclude with a sequenced recommendation: mhGAP-informed CHW training incorporating an explicit suicide risk module, anchored to nurse-led units for supervision, and phased to match medication availability rather than introduced ahead of it.
The implications for policy and practice section has been revised to match, replacing the general call for training with a specific one.
Changes made to the manuscript
● Discussion, facilitators section: three new paragraphs on Friendship Bench, STEPCARE and mhGAP, and Sierra Leone specific implementation constraints.
● Implications for policy and practice: revised to specify mhGAP-based training with a suicide risk module and cascaded nurse-led supervision.
● References: Chibanda et al. (2016), Gureje et al. (2019), and Faregh et al. (2019) added.
Closing comment
|
Reviewer 2, closing The manuscript makes a valuable contribution to a neglected area and demonstrates appropriate qualitative methods in principle. However, it requires stronger justification of sample adequacy, transparent reporting of the coding process, caution around quantification of qualitative data, and supporting the quotation |
Response
We have addressed all four requirements. Sample adequacy is now justified against the five dimensions of information power in a dedicated methods subsection. The coding process is reported phase by phase with code, subtheme, and theme counts, and the full coding tree is given in Table 2. Quantification is now governed by a stated convention, with percentages removed from the narrative text and from Table 1. Quotations are supported by analytic commentary at seven points in the results, each naming the participants whose extracts are being interpreted.
We thank both reviewers again. The revisions have improved the paper in ways we would not have reached unaided, particularly the repositioning of mental health literacy as one determinant of access rather than the explanation for the treatment gap, and the recognition that the absence of spiritual attributions in vignette responses stands in tension with our participants’ own accounts of their communities. We hope the revised manuscript is now suitable for publication and would welcome any further guidance.