Integrating a brief mental health intervention into primary care services for patients with HIV and diabetes in South Africa: study protocol for a trial-based economic evaluation

Introduction Depression and alcohol use disorders are international public health priorities for which there is a substantial treatment gap. Brief mental health interventions delivered by lay health workers in primary care services may reduce this gap. There is limited economic evidence assessing the cost-effectiveness of such interventions in low-income and middle-income countries. This paper describes the proposed economic evaluation of a health systems intervention testing the effectiveness, cost-effectiveness and cost-utility of two task-sharing approaches to integrating services for common mental disorders with HIV and diabetes primary care services. Methods and analysis This evaluation will be conducted as part of a three-armed cluster randomised controlled trial of clinical effectiveness. Trial clinical outcome measures will include primary outcomes for risk of depression and alcohol use, and secondary outcomes for risk of chronic disease (HIV and diabetes) treatment failure. The cost-effectiveness analysis will evaluate cost per unit change in Alcohol Use Disorder Identification Test and Centre for Epidemiological Studies scale on Depression scores as well as cost per unit change in HIV RNA viral load and haemoglobin A1c, producing results of provider and patient cost per patient year for each study arm and chronic disease. The cost utility analyses will provide results of cost per quality-adjusted life year gained. Additional analyses relevant for implementation including budget impact analyses will be conducted to inform the development of a business case for scaling up the country’s investment in mental health services. Ethics and dissemination The Western Cape Department of Health (WCDoH) (WC2016_RP6_9), the South African Medical Research Council (EC 004-2/2015), the University of Cape Town (089/2015) and Oxford University (OxTREC 2–17) provided ethical approval for this study. Results dissemination will include policy briefs, social media, peer-reviewed papers, a policy dialogue workshop and press briefings. Trial registration number PACTR201610001825405.

While the main trial protocol is cited in this protocol, it would be useful to have a little more detail here. In particular, can the authors provide some more detail on 'standard care' (MIND_TAU) -it's not wholly clear if it includes mental health services, and how these are delivered. Is MIND_TAU representative of typical care services in SA for HIV and diabetes? Can the authors also provide planned trial dates (perhaps in Figure 1)? Are baseline and endline assessments cross-sectional in the clusters or longitudinal?
For the sensitivity analysis, it would be useful to say how they would be usede.g. to generate ICER credible intervals or produce a CEAC. Additionally, how will the authors incorporate uncertainty about treatment effect size? Can the authors specify 'standard ranges' for discount rates?
For statistical analysis of costs and effects, could the authors be a little more specific about the analyses. How will costs be compared between arms? For the linear regression, I'm assuming no covariates from the text, but how will baseline and endline be accounted for, if at all? (e.g. outcome is differences, or baseline is controlled for).

The Protocol
The last sentence on page 3 is defining the study objective. There is scope to rephrasing this sentence, bringing greater clarity to the study objective. For example, is the study objective of economic evaluation including assessment of cost or consequences or combination of both? A diagram to portray the different arms in the research design would help in communicating the critical points in the protocol. More information about the rationale to compare the interventions of MIND-DED, MIND-DES and base line scenario would have been useful. Reference to previous literature to justify the 1-year time frame of economic evaluation will make the method of the protocol more credible. A brief description of the community health workers and facility staff such as professional qualification, years of experience and gender would help to repeat the study. Further clarity is required in the reporting of applicable discount rate on page 6 (3%), page 7 (8%) and Table 2 (Planned sensitivity analysis). The authors could elaborate on the plan for data collection. For example, elaborate the ingredients and step-down method for collection of institutional costs. The sentence "This clinical trial is powered for clinical rather than economic outcomes [38]" is confusing and needs greater explanation. It would be good to provide more information about presenting ICER from societal perspective. What factors are considered in this perspective to make it different from health services perspective? The other issue to clarify whether ICER from health services perspective represent the scope of primary care services? That is, primary care services assessing feasibility of mental health intervention. Or should the ICER perspective match the scope of the "facility counsellor", as discussed in the section of discussion. Clarity of this point is important for coherence of the protocol. Regarding analysis of ICER, providing specific examples of South African contexts and the cost-effectiveness thresholds would be very meaningful. The protocol will benefit from a specific section for analysis of budget impact. In absence of such section, the topic of "budget impact of equitably implementing the service" in the section of discussion is not very meaningful. The referencing style of the protocol needs more work. About 10% of the references are contemporary, belonging to publications in year 2017 and 2018. The protocol should be able to use higher share of contemporary literature, given the recent progress in the literature of integration of mental and physical care.

VERSION 1 -AUTHOR RESPONSE
Reviewer(s)' Comments to Author: Reviewer: 1 Reviewer Name: Sam Watson Institution and Country: University of Warwick, UK Please state any competing interests or state 'None declared': None declared Please leave your comments for the authors below This protocol describes the method for the economic evaluation of an intervention to integrate mental health services into care services for HIV and diabetes. The protocol is clear, comprehensive, and presents a sound method. My comments are therefore minor and are mainly points of clarification: 3. Can the authors clarify which costs fall into health service and societal perspectives?
Agreed clarity is needed here, health services or health systems should read provider perspective. On page 5 we have replaced "health systems" with "provider" and detailed what these costs include: i.
Full economic provider costs: of the mental health intervention; any cost offsets attributable to the intervention related to changes to the costs of HIV or Diabetes care at the primary care level; and changes to the costs of referral care (including tuberculosis, emergency department and inpatient care); Societal includes both the provider and the patient perspective, for the patient costs we have also added more detail: ii. Patient costs: associated with the intervention which will include direct OOP payments to private health providers (consultations and medications etc.), travel and subsistence costs and productivity losses. Under perspective on page 5, we have also added more detail: In line with good practise recommendations, the analyses will also be presented from a societal perspective (including both provider and patient perspectives). While public sector primary health care is free at the point of use, patients incur time and travel costs when accessing care and may experience productivity losses; in addition, they may incur costs when using private sector health services. These costs will be collected to inform the patient perspective within the economic evaluation.
We have also edited Table 1 to reflect measurement and valuation of provider costs the heading of the table has also been edited by replacing "health services" with "provider" Under Estimating costs for the providers perspective on page 6, we have also made some further editions for clarity on costs : Estimating costs for the provider's perspective Within economic evaluation, the appropriate scope of provider costs includes all costs incurred within the intervention and any changes in broader health system costs that can be attributable to the intervention. We have categorized these costs as intervention costs, HIV and Diabetes service costs and other related provider costs. As shown in Table 1, our approach to estimating these costs entails the measurement of quantities of resources that are utilized, and multiplying these quantities by the value (or unit cost) of each resource. These separate steps of measurement and valuation are described below.
Measurement of resource use For the intervention costs an ingredients approach will be used to estimate resources. Routine data linked to intervention protocols will be collected and used to assess resources directly consumed in the provision of the intervention, including supplies, manuals, and patient education materials.
4.While the main trial protocol is cited in this protocol, it would be useful to have a little more detail here. In particular, can the authors provide some more detail on 'standard care' (MIND_TAU) -it's not wholly clear if it includes mental health services, and how these are delivered. Is MIND_TAU representative of typical care services in SA for HIV and diabetes?
Yes, more care was needed here to explain standard care, which refers to standard of care for CMDs, on page 4 under Trial Design we have made the following insertion for clarity: Treatment as usual (TAU) is representative of typical primary care level services for CMD in the South African public health service facilities, which is typically limited to referrals [25], [26].
Under Intervention and comparator on page 4 the following insertion was made: In them TAU arm, the standard package of care will be provided to patients who are suspected of having mental health problems. In general, patients using the HIV or diabetes service are asked by a nurse or doctor attending to their care about their mental well-being, life stressors, and use of alcohol or other substances. Patients are provided with advice to change their lifestyles. Where the health care provider deems it necessary, patients are referred to a mental health nurse for further assessment or screening. The patient may also be referred to a social worker who may refer them to NGOs who provide counselling and support services [24].
Further reference to the TAU costs was also made on page 6: The TMT will enable an analysis of differences in counsellor time usage in the different intervention modalities, indicating opportunities for scale and scope efficiencies. It is anticipated that the costs associated with the TAU option will be minimal and any referral costs will be captured through patient self-report of services they have used (see next paragraph).
And finally the potential impact on cost-effectiveness of the limited referrals in the TAU arm is acknowledged on page 11: Another possible limitation is that only a few patients may be referred for mental health care for CMD in the TAU option so we may not fully capture referral costs thus potentially underestimating costeffectiveness.
Specifying trial dates is indeed helpful as is specification of the longitudinal nature of assessments, these editions have been added to the text under Outcome measures on page 5.
All outcomes are measured longitudinally at baseline, and at 6 and 12 month follow-up assessments, recruitment on the trial started in 2017 and will end in February 2019, and final outcome assessments will end in 2020 (see Figure 1). Figure 1 Economic Evaluation: Analyses, Outcomes, Measurement and Assessment Timing 6. For the sensitivity analysis, it would be useful to say how they would be usede.g. to generate ICER credible intervals or produce a CEAC.
Indeed, the sensitivity analysis will be used to produce a CEAC as detailed in the following edition on page 9: Probabilistic sensitivity analysis (which allows application of confidence intervals to point estimates) will be used for sensitivity analyses of all relevant individual level variables. The sensitivity analyses will be used to produce a cost-effectiveness acceptability curve (CEAC). 7. Additionally, how will the authors incorporate uncertainty about treatment effect size? Can the authors specify 'standard ranges' for discount rates? For statistical analysis of costs and effects, could the authors be a little more specific about the analyses. How will costs be compared between arms? For the linear regression, I'm assuming no covariates from the text, but how will baseline and endline be accounted for, if at all? (e.g. outcome is differences, or baseline is controlled for).

Analysis
Assessment

Please leave your comments for the authors below Abstract
The abstract needs to be amended. 9. In the introduction, the authors need to make it more distinct whether the study is conducted from the perspective of the service providers or the health services.
Thank-you, the need for clarity here is appreciated, we have amended the introduction as follows: This work presents a protocol for a cost-effectiveness and cost-utility analysis from a societal (i.e. provider and patient perspective) of a task-sharing strategy for delivering a brief mental health counselling intervention to patients with comorbid HIV or Diabetes in South Africa.
10. Another suggestion is to briefly clarify the two task sharing approaches of integrated services in the method section.
Clarity is indeed required: the difference between the 2 alternatives is that in the MIND_DED the CHW will be added to the facility staff complement and will only deliver the new counselling service, whereas in the MIND_DES a CHW already working in the facility will be chosen to deliver the new counselling service on top of their existing duties. Italics have been used to emphasise the difference on page 4.
The two models will be the dedicated and designated models of care. In the dedicated approach (MIND_DED), community health workers (CHWs) will be hired and added to the facility staff complement and will dedicate their time to only delivering the new counselling service. In the designated approach (MIND_DES) CHWs already working in the facility will be designated to deliver the service in addition to their other chronic disease-related activities such as adherence counselling for HIV and health promotion.
11. The abstract mentions of "budget impact analysis", but such analysis is not adequately described in the protocol. For example, the protocol did not adequately discuss plans to estimate the uptake of the proposed intervention and assess feasibility of budget in near future?
Apologies for this omission, this has been addressed on page 9 by adding a Budget Impact Analysis section after the Sensitivity Analysis section as follows: Budget Impact Analysis Trial based estimates of intervention uptake and population based estimates of number in need will be used will be used in budget impact approximations which will be conducted using a mathematical programming approach [52], [53]. These approximations will inform discussions on the fiscal implications of investing in primary care mental health services [54].
The Protocol 12. The last sentence on page 3 is defining the study objective. There is scope to rephrasing this sentence, bringing greater clarity to the study objective. For example, is the study objective of economic evaluation including assessment of cost or consequences or combination of both?
Agreed clarity would be helpful here, we have addressed this by the following edition on page 3: This work presents a protocol for a cost-effectiveness and cost-utility analysis from a societal (i.e. provider and patient perspective) of a task-sharing strategy for delivering a brief mental health counselling intervention to patients with comorbid HIV or Diabetes in South Africa. 13. A diagram to portray the different arms in the research design would help in communicating the critical points in the protocol.
Thank-you for this suggestionfigure 2 has now been added, and reference to it in the text is found on page 8: Figure 2 summarises the intervention arms, costs, and outcomes that will be included in this study.
14. More information about the rationale to compare the interventions of MIND-DED, MIND-DES and base line scenario would have been useful. Reference to previous literature to justify the 1-year time frame of economic evaluation will make the method of the protocol more credible.

Quality of Life
Co-morbidity

OUTCOMES
Thank-you this comment is appreciated and on page 4 under Intervention and comparator reference is made to relevant literature informing the content of the underlying counselling intervention, its efficacy and time frame for assessment informing the evaluation time frame : The underlying theory, content and evidence for the efficacy of the counselling approach for reducing symptoms of depression and hazardous /harmful alcohol use over the 1 year time frame have been described previously [29]- [31] 15. A brief description of the community health workers and facility staff such as professional qualification, years of experience and gender would help to repeat the study.
The need for clarity is acknowledged and the reader is referred to the trial protocol and this edition is made on page 4 under trial design: Descriptions of the CHWs and registered psychological counsellors roles in the MIND-DED and MIND-DES models and their qualifications and skills levels are detailed in the trial protocol [24].
16.Further clarity is required in the reporting of applicable discount rate on page 6 (3%), page 7 (8%) and Table 2 (Planned sensitivity analysis).
Sorry for the confusion here. On page 5, we refer to the use of a discount rate to capture time preferences, and here we aim to use 3% in line with other published studies and recommendations from guidelines for economic evaluation.
On page 7 we are referring to the annuitization of capital costs. Here, the recommendation from guidelines for economic evaluation is to use an interest rate that reflects the long-term return on government bonds; in South Africa this interest rate is approximately 8%. We have edited the section on page 7 to improve clarity, as outlined below: For capital items, costs will be estimated using the replacement value method [37]. In essence, the current replacement value of each items is calculated and annuitized using data on the estimated useful life of the item and an interest rate representing the opportunity cost component for capital, in order to estimate an equivalent annual cost. An estimated useful life of between 5-20 years for furniture and buildings will be applied as while an interest rate of 8% will be used as the annuitization factor based on the rate of return on government bonds in South Africa [37].
17. The authors could elaborate on the plan for data collection. For example, elaborate the ingredients and step-down method for collection of institutional costs.
Apologies for this omission, we elaborate on the ingredients method on page 6 and step down methodology on page 7 under Measurement of resource use as follows: Page 6 Measurement of resource use For the intervention costs an ingredients approach will be used to estimate resources. Routine data linked to intervention protocols will be collected and used to assess resources directly consumed in the provision of the intervention, including supplies, manuals, and patient education materials.
Page 7 For the HIV and Diabetes services, a step-down methodology will be used to establish resource use. Staff time will be determined through review of facility organograms and interviews with senior managers. In addition, utilisation of HIV and Diabetes medicine and diagnostics will be estimated by applying protocol-based facility utilisation statistics.
18. The sentence "This clinical trial is powered for clinical rather than economic outcomes [38]" is confusing and needs greater explanation.
Apologies, we agree, and the editions have been made on page 5 under Sample Size and Patient Population: This clinical trial is powered to detect clinical outcomes, specifically reductions hazardous/ harmful alcohol use and risk of depression at 12-month follow-up rather than economic outcomes [24], [35].
19. It would be good to provide more information about presenting ICER from societal perspective. What factors are considered in this perspective to make it different from health services perspective?
The other issue to clarify whether ICER from health services perspective represent the scope of primary care services? That is, primary care services assessing feasibility of mental health intervention. Or should the ICER perspective match the scope of the "facility counsellor", as discussed in the section of discussion. Clarity of this point is important for coherence of the protocol.
The need for clarity here is appreciated, we have edited the manuscript to clearly show the costs for the provider perspective (page 6) and the patient perspective (page 7) and the ICER measurements(page 9) as follows: Page 6 Estimating costs for the provider's perspective Within economic evaluation, the appropriate scope of provider costs includes all costs incurred within the intervention and any changes in broader health system costs that can be attributable to the intervention. We have categorized these costs as intervention costs, HIV and Diabetes service costs and other related provider costs. As shown in Table 1, our approach to estimating these costs entails the measurement of quantities of resources that are utilized, and multiplying these quantities by the value (or unit cost) of each resource. These separate steps of measurement and valuation are described below.
Page 7 Estimating costs for the patient perspective Measurement of resource use For the patient perspective, we will use a questionnaire at baseline, 6 and 12 months to estimate resource use. This will capture any user fees associated with public, private or NGO-provided health services as well as traditional and faith-based therapies. In addition, we will ask patients to estimate their time spent accessing health care services, as well as out of pocket payments for transport, subsistence and accommodation. Travel, subsistence and accommodation costs will relate to both the patient and their caregiver, if the patient is accompanied by a caregiver to any of these services.

Page 9
As the trial is focused on health systems strengthening for the delivery of mental health services, the decision rule applies to maximising health within the health care budget constraint. Therefore ICERs will be calculated from the provider perspective and , the cost per QALY gained will be compared to the cost-effectiveness thresholds (CETs) for LMIC settings [47]. If our intervention's ICER is less than the chosen CET this will mean that diverting resources to the intervention will increase population health, and if the ICER is more than the CET the intervention is not cost effective. Patient costs will be collected and reported on separately. In addition, to satisfy the requirements of common guidelines [46] and to allow for some degree of comparability with other studies, the ICERs will also be presented for the societal perspective, i.e. including both provider and patient costs.
20. Regarding analysis of ICER, providing specific examples of South African contexts and the costeffectiveness thresholds would be very meaningful.
Thank-you for this point, there are no CET for South Africa, we will make use of thresholds for LMICs , as per this insertion on page 9 under the section Determining Cost Effectiveness: Therefore ICERs will be calculated from the provider perspective and , the cost per QALY gained will be compared to the cost-effectiveness thresholds (CETs) for LMIC settings [47]. If our intervention's ICER is less than the chosen CET this will mean that diverting resources to the intervention will increase population health, and if the ICER is more than the CET the intervention is not cost effective. Patient costs will be collected and reported on separately. In addition, to satisfy the requirements of common guidelines [46] and to allow for some degree of comparability with other studies, the ICERs will also be presented for the societal perspective, i.e. including both provider and patient costs.
21. The protocol will benefit from a specific section for analysis of budget impact. In absence of such section, the topic of "budget impact of equitably implementing the service" in the section of discussion is not very meaningful.
Sorry, we have added the following insertion on page 9 after the sensitivity analysis section: Budget Impact Analysis Trial based estimates of intervention uptake and population based estimates of number in need will be used will be used in budget impact approximations which will be conducted using a mathematical programming approach [52], [53]. These approximations will inform discussions on the fiscal implications of investing in primary care mental health services [54].
22. The referencing style of the protocol needs more work. About 10% of the references are contemporary, belonging to publications in year 2017 and 2018. The protocol should be able to use higher share of contemporary literature, given the recent progress in the literature of integration of mental and physical care.