Discussion
The prospective evaluation of the multisite, global HealthRise programme, a community-based programme seeking to improve hypertension and diabetes care among underserved populations, demonstrates the complexity of assessing community-based interventions across diverse settings and variable data environments. The programme’s implementation and evaluation faced many challenges, including some associated with establishing and maintaining monitoring efforts, particularly within pre-existing systems in underserved communities. Yet, process evaluation findings indicate that more than 56 000 and 59 000 individuals were screened for diabetes and hypertension, respectively, in Brazil, India and South Africa; however, a much smaller proportion of patients received new diagnoses and were referred to care, highlighting potential limitations of large-scale screening programmes. While Brazil and most US sites showed patient-level progress, with increasingly more hypertension and patients meeting disease control targets since programme enrolment and substantial declines in patients’ A1c and SBP decreased since baseline, improved disease management was not detected among patients in HealthRise implementation areas compared with those in comparison areas in Shimla and South Africa. A relatively short implementation period in some sites, alongside the inherent tension between demonstrating impact within time constraints and properly capturing the often slower or complex changes of health behaviours and systems, may underlie these findings. Limited data availability on intervention adherence and fidelity precludes further assessments of the programmes’ implementation quality and its impact on endline findings in each site.
Qualitative data highlighted some positive views of integrating home-based health workers in NCD care to bridge geographical, linguistic and cultural divides, and the importance of effective care coordination across provider types, data platforms and between facilities and communities. Nonetheless, long-standing gaps in system infrastructure likely contributed to continued challenges with care provision and social determinants of health continued to play roles in patient abilities’ to access treatment and manage their conditions. In combination, these findings demonstrate the potential for community based, and particularly, CHW led, interventions to improve NCD outcomes, but also underscore how their reach and effectiveness can be hindered by broader health system, infrastructure and policy constraints. Irrespective of their increasingly vital role for underserved populations, community-based programmes cannot fully remedy inadequate prioritisation or investments in strong, well-coordinated primary care and NCD services.
Successes and challenges for HealthRise and broader community-based NCD interventions
HealthRise programmes referred thousands of screened individuals to care, yet relatively few new diagnoses occurred. Low yields from population-based screening activities are not uncommon,43 44 and these findings support guidelines recommending more selective screening of high-risk groups to improve cost-effectiveness.45 A primary focus on screening may also have contributed minimal community-level effects in India and South Africa. Interventions in Brazil and the USA were more oriented towards improving access to care and medication adherence, which could more quickly affect health outcomes than screening activities, which only initiate the process of bringing patients into care.
In Brazil, both HealthRise sites exhibited notable progress; nevertheless, since comparison patient data were not collected for Brazil sites, we cannot ascribe these patient-level patterns to HealthRise participation. Despite these positive trends, qualitative data indicated poor adherence to medication for both Brazil sites, as well as patient-reported sociocultural tensions around adopting dietary changes and health system-level obstacles to accessing multidisciplinary care and reliably stocked medication at health facilities. In the USA, HealthRise participation was associated with reductions in A1c or SBP and increases in patients meeting treatment targets at some sites relative to comparison patients. Several factors may have contributed to observable impacts at US sites, which had among the longer programme implementation durations and the most robust evaluation, relative to other sites. Interventions in the USA were targeted to address specific barriers to keeping patients in care; the number of patients reached was small, so each person received substantial focus; and the US health system is better organised and equipped to deal with NCDs and therefore did not face the same health system challenges experienced in other HealthRise sites. However, as demonstrated by recent analyses of ‘superutiliser’ patients with medically and socially complex conditions,46 replicating intervention impact to larger populations may be difficult, especially if the resource intensity and patient attention provided through the US HealthRise programmes are not feasible or sustainable. This is particularly relevant given some of the reported difficulties in early-stage programme implementation in the USA, such as recruiting and retaining CHWs and ensuring all providers could access and update electronic medical records.
Additional data and contextual information are needed to better understand why programme impact was not found in India and South Africa. Unlike the Brazil sites, where HealthRise interventions were incorporated into longstanding community-based healthcare structures and CHW-led service provision, India and South Africa HealthRise grantees often had to build systems—both physical and administrative—from the ground-up to support NCD care coordination, medication logistics and community engagement. This challenge, in combination with relatively short implementation periods, and screening being higher programmatic priorities in these sites, may underlie the negligible community-level effects on outcomes. Further, we could not fully account for other local or national initiatives to expand NCD care in both HealthRise and comparison areas; for instance, India launched national guidelines for NCD screening activities through CHWs and community platforms in 2017,47 48 potentially spurring the scale-up of broader community-based NCD programming throughout India after HealthRise began in Shimla. Other studies, including an evaluation of a CHW-managed intervention for patients in India with high cardiovascular disease risk,49 indicate patient-level barriers to care like cost, transport and medication availability could affect intervention impact as well. Substantial drop-offs in care cascades also emphasise the need for locally relevant mechanisms for coordinated care.50 These findings correspond with larger-scale assessments of diabetes care cascades in India and South Africa,14 18 both of which stressed the importance of strengthening NCD case detection and management for more rural, underserved communities.
Additional challenges were highlighted in qualitative analyses. Despite some indication of heightened patient empowerment, as measured by self-reported knowledge and confidence in at least some sites, adherence to recommended dietary and physical activity behaviour changes emerged as a source of tension between patients and providers, highlighting the difficulty of enacting meaningful cognitive and behaviour change amid strong social and environmental influences. Additional challenges that could negatively affect the adoption and scale-up of community-based NCD interventions in resource-constrained settings included the availability and quality of technologies for care coordination; minimal experience managing or working with in-home providers as members of care teams; and challenges in securing long-term funding for community-based NCD programmes, as well as broader health system capacities for NCD care (eg, functional diagnostic equipment, reliable stocking of NCD pharmaceuticals, accessible primary care services). Without greater prioritisation of NCDs in health financing—from government sources to development partners alike—the potential impact of community-based NCD programmes could be hindered by the lack of underlying infrastructure and resources.4 9
Potential implications for community-based NCD interventions
Building off of previous work, the present study offers some programmatic considerations, including facility-level and community priorities, as well as for national agendas on NCD prevention and treatment. In Brazil, India and South Africa, pre-existing health system challenges, ranging from medication stock-outs to long travel times to reach health facilities, posed obstacles to patients and providers. To more effectively treat the rising burden of NCDs, it is critical to address deficiencies in facility infrastructure, transportation, staffing and supplies. In the USA, many providers and administrators had limited previous exposure to home-based providers, which made programme implementation challenging at times, especially during the early stages of intervention. Identifying processes and supportive technologies by which care teams may incorporate home-based care more seamlessly, particularly in terms of sharing patient data and informing facility staff about findings from home visits, is likely to be beneficial. Finally, especially for LMICs, there is an urgent need for development partners to dedicate more funding to NCD care and strengthening health systems more broadly.4 Without a greater emphasis on these health financing areas, many communities in LMICs will remain ill equipped to provide effective NCD care.9
Limitations
Our study’s findings should be interpreted in light of its limitations. First, while HealthRise sites were selected to represent a range of underserved populations worldwide, findings are not generalisable to all underserved communities seeking to improve NCD care. Continued work is needed to understand which community-based NCD interventions may work best given local contexts and needs. Second, despite being incorporated into the initial process evaluation framework, comprehensive information on intervention reach and fidelity (ie, the degree to which interventions were implemented per protocol) were not available across sites and thus could not be included in the present study. To better understand intervention impact, ongoing and future community-based programmes could greatly benefit from ensuring adequate funding, personnel and infrastructure to establish and maintain data collection for evaluation indicator monitoring. Third, in the USA, comparison groups were constructed retrospectively by each grantee, using available patient record information and were not selected by random assignment. While efforts were made to ensure that included comparison patients generally represented individuals who would have been eligible for HealthRise enrolment, they may have differed from individuals who enrolled. Fourth, in Shimla, India and South Africa, we were limited to cross-sectional patient data at endline, and thus could not directly assess potential differences in cascades of care from baseline to endline among HealthRise implementation and comparison areas. Not being able to explicitly account for pre-intervention differences in these areas and how they changed over time may contribute to some findings in Shimla (ie, patients presenting at facilities in comparison areas having somewhat higher levels of current diabetes treatment and meeting treatment targets for hypertension than patients in HealthRise implementation areas). Also, based on our sampling strategy, we cannot rule out cross-contamination in Shimla and South Africa (ie, patients presenting at facilities in comparison areas engaged in HealthRise activities and/or patients in HealthRise implementation areas were not exposed to HealthRise programming). Fifth, the global evaluation team could not verify monitoring data accuracy for sites in India and South Africa, as only aggregated data were provided by grantees due to government regulations governing data use outside the country. Organisations were assigned by the government in each country to check the validity of data before it was transmitted to IHME. Sixth, for Brazil and the USA, we only included patients who remained enrolled at endline in the endline analyses; by taking this ‘as treated’ analytical approach, which provides insights into programme effects closer to full adherence, these patients may not represent all potential target populations for HealthRise interventions and results may be positively biased. Seventh, while results varied by site, the relative lack of differences for several indicators between patient groups and over time could be related to factors beyond programme effectiveness. It is possible that, on average, HealthRise programme implementation and patient duration in the programme was not long enough to detect positive effects; this may be particularly relevant for sites where community-based care and CHW networks had not previously been longstanding models of service provision and therefore required substantial time to establish. Changes in clinical outcomes also can lag behind intervention exposure and thus improvements may not have been fully realised by endline. Eight, information on the existence of preintervention services or programmes were not available across sites; subsequently, it was not possible to ascertain the potential effects of this factor on endline evaluation results.