Effectiveness of a health systems strengthening intervention to improve detection, treatment, and control of Multiple Long-Term Conditions in primary health care facilities in South Africa: study protocol for the ENHANCE Type 2 hybrid cluster randomised control trial

Patient and public involvement

Since the pre-grant proposal phase of the study, there has been active involvement and engagement with patient and public members, beginning with their participation in formative interviews and consultation during the pre-proposal stage. These early interactions were vital in ensuring the research questions were grounded in the lived experiences of healthcare fragmentation and the practical burdens of polypharmacy and poor treatment literacy. Following the study’s inception, these members alongside a named community co-applicant who served as a full member of the Project Executive Board formed the ENHANCE Advocacy Academy.²⁹ Throughout the study lifecycle, the Advocacy Academy met six monthly for a four-year tailored skills- enhancement programme, and ad-hoc to provide input into intervention and research design. Their involvement was critical in the design and conduct of the study through participation in co-development of the intervention tools, including the patient-health diary and treatment literacy guides, to ensure they addressed the “informational gap” identified during the formative phase. Furthermore they advocated for a focus on patient acceptability and exploration of condition control barriers, informing the outcome measures and mixed methods approach. They also provided guidance on recruitment of the study through reviewing screening and enrolment procedures and documentation and advising on feasibility of recruiting patients. The aim of this active partnership ensured that patient and public views informed the intervention and trial design to ground the study in the real-world context of MLTC care.³⁰ The Academy remains centrally involved in exploring findings and dissemination, assisting in the interpretation of findings and preparation of culturally appropriate research information to be shared with participants and linked communities.

Study design

ENHANCE is a randomised type 2 hybrid trial testing the effect of the intervention on improving clinical outcomes, while giving equal focus to investigating implementation processes and outcomes using the RE-AIM (reach, effectiveness, adoption, implementation, maintenance) framework.³¹ The unit of randomisation was PHC facilities, with outcomes measured on individual patient participants. Stratification by location (4 strata in KwaZulu-Natal and 3 in the Western Cape) was used to account for differences in prevalence of, and levels of chronic conditions control at baseline. Prevalence of conditions varies between South African provinces, with KwaZulu-Natal having a higher prevalence of HIV than the Western Cape, which has a higher burden of NCDs.³² Provinces also differ in rates of control of conditions including HIV, hypertension and diabetes documented.^33–35

Setting

The trial was conducted in 32 public sector PHC facilities located within the Cape Metro, Western Cape (n = 16) and uMgungundlovu district, KwaZulu-Natal (n = 16) provinces of South Africa. These provinces represent diverse health system contexts, with varying resources, patient volumes, and disease burdens. uMgungundlovu district covers an area of 9189 square kilometres that comprises of seven urban, peri-urban and rural municipalities with a population of one million. This district has 46 clinics, 17 mobile clinics, 4 district hospitals and one state aided clinic.³⁶ The Cape Metro health district has a population of over 4 million and is situated on the South Peninsula of the Western Cape Province. It incorporates eight sub-districts that are peri-urban or urban and has approximately 152 PHC facilities and 8 district hospitals.³⁷ Routine PHC services in these settings include the management of NCDs such as hypertension, diabetes, and chronic respiratory conditions, as well as communicable diseases including HIV. The study setting was chosen to ensure representation of typical South African PHC environments where multimorbidity and polypharmacy are prevalent.

Study population

The study population comprised 1): the facility managers and health care workers employed at fixed PHC clinics with substantive patient loads (headcounts >4000 per year) and surrounding areas, (including nurses, doctors and CHWs), 2) adult patients attending the clinics for care of at least two chronic conditions.

Inclusion criteria for PHC facilities

Eligible clinics were public sector PHC facilities that provide routine chronic disease management services including for NCDs, HIV and mental health. Clinics were considered for inclusion based on patient load, staffing capacity, and their ability to implement the intervention. Because a key element of the co-designed implementation strategy included supplementation of clinic-based care with CHW-provided treatment and condition literacy in the home, facilities had to have a designated CHW team. In the Western Cape, such CHWs are employed and supervised through non-profit organisations and are entirely community-based, whereas in KwaZulu-Natal, CHWs have been assimilated into the provincial health department staff as part of the Ward-Based Primary Health outreach team (WBPHCOT) and provide services at clinic, school and community level. Other provincial differences included a higher proportion of facility-based doctors in the Western Cape, a metropole area, whereas care in KwaZulu-Natal was largely nurse-led, with outreach doctors from district hospitals and local private sector doctors providing services at primary healthcare (PHC) facilities on a referral basis, typically for a maximum of four hours per day.

Inclusion criteria for patients

Patients were eligible for inclusion if they were aged 40 years or older, were receiving care for at least two of the following conditions, with at least one of the first five listed conditions being uncontrolled: HIV, hypertension, diabetes, asthma, depression, previous myocardial infarction or previous stroke. Patients were excluded if they were unable to participate in an hour-long interview, were planning to relocate during the study period, were unwilling to receive a CHW visit in their home, were pregnant or on current tuberculosis treatment.

Interventions

Standard of Care

Participants in control facilities continued to receive standard care. In South Africa, PHC for long-term conditions is delivered free-at-point-of-care and includes the management of HIV, NCDs and mental health according to South Africa’s Ideal Clinic and Integrated Clinical Services Management model.³⁸ This integrated approach in theory allows for simultaneous management of multiple conditions, enabled by the Essential Medicines and Standard Treatment Guidelines for Primary Healthcare³⁹ and the PACK/APC clinical decision support tool. Patients typically attend routine monitoring every 3–6 months, while chronic medication is collected monthly through fast-track queues or decentralized dispensing systems, including community venues, wellness clubs, trailers, retail pharmacies (in KwaZulu-Natal) or e-lockers.

During the COVID-19 pandemic patients were discouraged from collecting their medication from group settings (clinics, halls, clubs) and CHWs delivered a large proportion of refill prescriptions directly to patients’ homes. Home deliveries were slowing at the time of the trial, and are being reserved for patients with mobility issues, but dispensing outside of the clinic has continued to grow, posing a challenge for linkage with monitoring and support programmes. MLTC patients, like usual patients, intermittently attend general acute care services within the same clinic. Many who receive care for a long-term chronic condition from public sector facilities also seek care from doctors in private general practice, typically for acute problems.

Intervention

Participants enrolled in intervention clinics continued with usual care as described for control clinics, and in addition, received the ENHANCE health systems intervention. Intervention components are shown in Figure 1 and included:

Figure 1. ENHANCE intervention resources.

Following recruitment, participants in intervention clinics were seen by an ENHANCE trained clinician who was asked to complete a clinical review using the ENHANCE clinical decision support tool and link the participant to CHW services for follow-up in the home. In this way participants attending intervention clinics received at least one clinical review of their MLTC care and were expected to have two household support visits during the follow-up period. Additional visits were determined based on need (additional diagnoses, level of control, requirement for home-based support).

Implementation strategies comprised delivering of online and in-person onsite facility team training using case-based curricula to equip nurses and doctors to use the ENHANCE clinical decision support tool to review care of patients with MLTCs and refer them for follow-on support provided by CHWs at home. CHWs were trained in a simple approach to treatment literacy using a brief lay person formulary, and condition literacy using posters and the patient diary. Facility support visits focused on completion of training, the embedding of the intervention, clinical queries and review of referral pathways from clinicians to CHWs using a continuous quality improvement approach. Representatives from district and sub-district management, intervention facilities and CHW teams met quarterly in various forums to share learning around implementation.

Trial outcomes

Primary outcome

The primary outcome, measured at individual participant level, is a binary composite outcome variable for patient participants with any combination of the eligible index conditions which is aligned with the health system’s mandate to detect, treat and control chronic conditions to prevent adverse health outcomes. It will be considered present if a patient participant meets at least one of the following criteria at 12-month follow-up:

Certain of these criteria align closely with clinical effectiveness, such as improved control in a condition not optimally controlled at baseline, while others align better with an assessment of implementation, such as intensification or change or treatment. Diagnosis and initiation of treatment of another chronic condition could arguably be considered both a clinical and implementation outcome.

We gave equal weight to the three components of chronic condition care - diagnosis, treatment and control - which in our view represent the core mandate of PHC facilities in relation to people living with these conditions.

Secondary clinical effectiveness outcomes

We will report a diverse set of secondary effectiveness outcomes⁴³ including the primary outcome disaggregated by component; physical and social functioning assessed using WHODAS-2.0 (World Health Organisation Disability Assessment Schedule)⁴⁴; health-related quality of life assessed using the EuroQol 5D-5L⁴⁵; patient experience of care assessed using the PACIC (Patient Assessment of Chronic Illness Care)⁴⁶; depressive symptoms assessed using the PHQ-8 (patient health questionnaire-8)⁴²; patient experience with treatment and self-management assessed using a shortened version of the PETS tool⁴⁷ and health care utilisation and costs. Table 1 lists the outcomes and time points for their collection.

Secondary implementation outcomes

Implementation outcomes were collected at clinic level from intervention implementation data, semi-structured interviews and the trial patient participant questionnaires to assess reach, adoption and implementation and maintenance of the ENHANCE intervention.⁴⁸ Measures of reach and adoption will capture exposure of PHC staff and patients to elements of the intervention and implementation strategy (e.g. proportion of intervention group participants visited by a CHW) and integration of intervention components into care (e.g. patient diary use). Planned subgroup analyses across intervention and control group patients will assess appropriateness of new diagnoses and treatment (e.g. in individuals with elevated blood pressure at baseline). Semi-structured interviews will explore patient, clinician and CHW perceptions of the acceptability of the ENHANCE intervention and associated barriers and facilitators to its implementation.

Safety and monitoring

We immediately referred patient participants who had severely elevated blood pressure (≥180/110) to the clinicians at the clinic, irrespective of intervention allocation. At the conclusion of data collection, we also referred patients whose control remained poor to the clinic staff for review. We systematically monitored hospitalisations and deaths, which we anticipated would be common among individuals with MLTCs, and reported these to the independent Trial Steering Committee (TSC) at six-month intervals. This committee played both a supervisory and advisory role and considered the study from the perspective of the participants, the public affected by the issues being researched, the research team and the funder. The TSC reviewed cumulative study data to evaluate safety, study conduct, scientific validity and data integrity. Specifically, it assessed the progress of the trial, monitored adherence to the protocol, the safety data, considered new information as it becomes available, the critical effectiveness endpoints and provided recommendations to the Principal Investigators.

Members of the TSC consisted of both independent members and members of the research team. The former served in an individual capacity and provided their expertise and recommendations. A charter outlined the roles and responsibilities and served as the Standard Operating Procedure (SOP) for the TSC.

Sample size and statistical power

Our sample size calculations were informed by outcome prevalence data and intra-cluster correlation coefficients (ICC) from our PC101,⁴⁹ COBALT⁵⁰ and PACK Brazil trials⁵¹ in similar settings, and similar global trials⁵² ( Table 2).

We estimated that a sample size of 1888 participants from 32 clinics (59 patients per clinic, 944 per group) would provide 85% power to detect a 14% (50% vs 64%) difference in the binary composite primary outcome, allowing for 15% loss to follow-up, with an intra-cluster correlation coefficient (ICC) = 0.05 and 5% significance.

Procedures

Randomisation

Clusters were defined at the clinic level to minimise contamination between clinicians within the same clinic. Clinics were randomised to intervention or control groups in a 1:1 ratio, stratified first by province and by substrata. In the Western Cape facilities all in urban or peri-urban areas and were stratified by Metro Health District defined substructures. The four substructure strata are Northern and Tygerberg (NTSS) with three clinics, Mitchells Plain and Klipfontein (KMPSS) with six clinics, Khayelitsha and Eastern (KESS) with four clinics and Southern and Western (SWSS) with three clinics. The NTSS and SWSS strata were pooled to achieve an even number of clinics. In Kwa-Zulu Natal, two stratifications were also implemented to ensure balance between urban and rural areas and between clinics receiving substantial support from a local NGO. There are two urban strata, Msunduzi (urban) with nine clinics and other (less urban) with seven clinics. The latter stratum consisted of clinics from Richmond, uMngeni, Impendle, Mkhambathini and uMshwathi. Within the urban stratum six of the nine clinics were supported by the NGO and for the other stratum four of the seven clinics were supported by the NGO. The total number of strata within the KwaZulu-Natal province is therefore four.

The research team sought approval from the provincial health departments for clinics that met the eligibility criteria to be included in the trial. In both provinces the number of clinics approved for research exceeded the planned number of clusters. Additional clinics were earmarked for pilot work or to function as replacement clinics should a clinic be found not to be suitable after initial randomisation but before data collection started. PHC facilities were randomised by the trial statistician using nQuery advisor using province and substrata. Randomisation occurred before enrolment of participants or delivery of the intervention. Two clinics had to be replaced. The Western Cape intervention clinic was deemed unsafe for access by research staff due to community violence, and a KwaZulu-Natal intervention clinic could no longer start training due to severe staff shortages.

Additional clinics were randomised in each province and the first clinic allocated to the intervention arm was selected as the replacement clinic. After commencement of data collection, no clinics were replaced. The replacement strategy was presented to and approved by the TSC.

Allocation concealment and blinding

It was not possible to blind health workers, participants and fieldworkers administering questionnaires about the allocation to the intervention or control arms, however, the trial statistician was blinded up to the completion of the analysis. Outcomes for viral load and blood glucose levels were measured blindly by the National Health Laboratory Services (NHLS).

Recruitment procedures

People living with MLTCs

Eligible participants were identified by facility clinicians and, if interested in participating, were referred to a fieldworker who administered a brief screening questionnaire to determine eligibility. All potential participants received full information about the study in their preferred language (English, Afrikaans, isiXhosa, isiZulu) as described in the information sheet. In addition, they were given an opportunity to ask questions and to decide whether they chose to take part in the research. They were reassured that they could chose to not to take part or to withdraw from the study at any time and that they would continue to receive care and would not be penalised in anyway. If a participant chose to participate in the study, they were asked to sign a consent form.

Health care providers

Healthcare providers were invited to participate in the process evaluation. Eligible individuals were approached by research staff, provided with an information sheet, and given the opportunity to ask questions. Participation was voluntary, with the assurance that declining or withdrawing would not affect their employment or performance reviews. Those agreeing to take part provided signed consent.

Data collection

Trained fieldworkers administered standardised data collection forms using a REDCap (Research Electronic Data Capture) application embedded on handheld devices. Data were collected at enrolment, after consent was provided and before the participants was seen by a clinician.

Follow-up interviews took place 6 and 12 months later. The interview schedule is summarised in Table 1 and includes socioeconomic details, several generic and condition-specific instruments most of which had already been translated, validated or at least used in South Africa ( Table 1), questions on healthcare utilisation and costs, and careful documentation of participants’ medication including name of each medicine, dose and frequency with photographs of original scripts captured alongside. Fieldworkers were also trained to complete physical measurements including weight (using digital scales), height, waist circumference, and blood pressure (BP). BP was measured on three occasions at each interview using an OMRON BP monitor with readings at least two minutes apart in accordance with American Heart Association guidelines.⁵³ The first reading was discarded, and the mean of the second and third reading was included in the analysis.

At enrolment and at 12 months, participants living with HIV and/or diabetes were asked to provide a blood sample of 5 ml (one teaspoon) for measurement of their viral load and/or HbA1c. Blood samples were drawn by the clinics’ health care workers who routinely drew bloods and we compensated clinics with additional consumables.

Participants’ medication data, collected at the time of interviews, were cross-checked with electronically collected data [Chronic Dispensing Unit (CDU) in the Western Cape, Central Chronic Medicines Dispensing and Distribution (CCMDD) Programme in KwaZulu-Natal] and any differences interrogated and resolved.

Data management

Interview data was uploaded as soon as questionnaires were completed and stored in a secure Structured Query Language (SQL) server at the trial co-ordinating centre based at the Knowledge Translation Unit, University of Cape Town. Questionnaire data were enriched with laboratory results and electronic prescription data from patient clinic folders and maintained as part of provincial records. Participant level consent was obtained to access routine health information, and access was limited to a small number of supervisory staff following agreements with the respective health departments. All personal information handling complied with laws around personal data including POPIA (Protection of Personal Information Act) in South Africa and General Data Protection Regulation (GDPR) in the United Kingdom, and staff received relevant training.

Quality control measures included in-person and telephonic supervision of fieldworkers, refresher trainings, range checks and automated skips.

Data analysis

Descriptive statistics such as means, standard deviations, frequencies, percentages, counts, medians will be compiled and reported by study arm (intervention or control). We will test and estimate the effects of the intervention on primary effectiveness and implementation outcomes and secondary effectiveness outcomes, by comparing outcomes in intervention and control groups. Analysis will be by intention to treat. We will use multilevel mixed effects regression, with clinic and participants as the random effects, and with province and strata as covariates.

The regression models, adjusted for stratification and clustering, and depending on outcome format, will be:

Estimation of the regression models will be done using maximum likelihood to utilise the full maximum likelihood approach as out imputation model for missing values post enrolment. For longitudinal outcomes the enrolment value will be part of the longitudinal modelling and hence the test for an interaction between intervention and time will be the formal test for a significant intervention effect. Sensitivity analysis will be conducted adjusting for unbalanced baseline confounders in the models.

Subgroup analyses

By condition: our prior hypothesis was that some conditions would be less amenable to change in medication or control than others.

By level of control: our prior hypothesis was that patients with poorly controlled conditions would be more amenable to changes in medication or functioning than patients with well-controlled conditions.

By province: our prior hypothesis was that effects of the intervention may differ between Kwa-Zulu Natal and better-resourced Western Cape facilities, as some elements of care are arranged differently by province.

Economic evaluation

We will conduct a within-trial cost effectiveness of treatment of MLTC according to South Africa’s Ideal Clinic and Integrated Clinical Services Management model versus treatment with the ENHANCE intervention. The economic evaluation will be carried out based on the full enrolled trial sample, for the duration on the trial and adopt a societal perspective. The analysis will be reported according to the Consolidated Health Economic Evaluation Reporting Standards.⁵⁴ Resource use will be measured through questionnaires and medical records, and unit costs will be sourced from health department data. Incremental costs of developing ENHANCE will be calculated, converted to mean cost per consultation, and incorporated into trial arms. Incremental Cost- Effectiveness Ratos (ICERs) will be the arithmetic mean difference in cost between ENHANCE and usual care divided by the arithmetic mean difference in effect based on trial outcomes.⁵⁵

The primary trial outcomes will inform the ICERs expressed as an incremental cost per additional participant with control of at least one condition that was not controlled at baseline and incremental cost per additional participant diagnosed or initiated onto treatment. To enable comparability with other studies, quality-adjusted life years (QALYs) will be calculated based on participant responses to the self-reported EQ-5D-5L questionnaire and used to estimate an ICER expressed as an incremental cost per QALY gained. EQ-5D-5L responses will be converted into utility scores using the newly developed South African tariff.⁵⁶ Quality adjusted life years (QALYs) will be calculated using the utilities and the area under the curve method.⁵⁷

Univariate and two-way sensitivity analyses will investigate the impact on results of varying outcome effect sizes within confidence bounds and key cost drivers, such as staff or medicines for provider costs and travel or medical costs for participants. A key consideration for policymakers and implementers of new interventions is the affordability and feasibility of available options. A budget impact analysis over five years will model financial implications for the health system, applying inflation adjustments and scenario testing.⁵⁸

Should there be no statistically or clinically meaningful difference in the primary clinical outcome between trial arms, we will undertake a cost-consequence analysis. Cost-consequence analysis transparently displays all costs (e.g., intervention, healthcare utilization) against the full spectrum of consequences in a disaggregated manner to support decision-making.⁵⁹

Process evaluation

A parallel mixed-methods ethnographic process evaluation was conducted to examine the delivery of the ENHANCE intervention within the PHC context. It assessed fidelity, acceptability, and implementation barriers and facilitators for health facilities, staff and patients, identify sources of contamination in usual care, and explain trial outcomes to inform recommendations for scale-up. Eligible participants included policy makers, facility managers, doctors, nurses, and CHWs. We observed and audio-recorded 40 (5 per facility, 20 per arm) clinical consultations, supported by observational field notes. We conducted semi-structured interviews at 12 months after randomisation with the key actors involved in delivering the intervention to obtain diverse perspectives of acceptability, barriers and facilitators to implementation and impact of the intervention on access and quality of healthcare.

Quantitative components included facility profile questionnaires and analysis of routine aggregate data, while qualitative components involved semi-structured interviews and direct observations of care delivery. To assess intervention fidelity, we draw on Medical Research Council United Kingdom (MRC UK) process evaluation guidance that understands adaptation to the local context is more appropriate than a strict assessment of fidelity.⁶⁰ Interviews are transcribed verbatim, translated, and are being thematically analysed with the aid of NVivo software.

Ethical considerations

Ethical approval for the trial has been obtained from the Universities of Cape Town [HREC 686/2022] and KwaZulu-Natal [BREC/00005033/2022], and King’s College London [LRM-23/24–34161]. Permission has also been obtained from the KwaZulu-Natal and Western Cape Departments of Health. Written informed consent was obtained from all participants (patients and providers) before any research data collection was undertaken. The protocol complies with the Declaration of Helsinki.

Dissemination

We shall disseminate the study findings using various modalities:

Trial status

Patient level recruitment started in September 2023 and ended in June 2025. The data were cleaned blinded from allocation and the dataset locked. Data analysis has not started.