Abstract
Objectives:
The COVID-19 pandemic disrupted routine healthcare services, disproportionately affecting people living with chronic conditions such as type 2 diabetes (T2D). In response, the Western Cape Government Health implemented home delivery of medication (HDM) via community health workers (CHWs) to maintain continuity of care. This study aimed to evaluate the association between socioeconomic factors and access to HDM among T2D patients in Cape Town, South Africa, during the pandemic, with a focus on equity and health system responsiveness.
Methods:
A descriptive cross-sectional survey was conducted via telephone interviews with 267 patients receiving care at 4 public primary care facilities. Sociodemographic, economic, and treatment-related variables were collected. Fisher’s exact test and multivariable logistic regression were used to assess the associations between these variables and access to HDM.
Results:
Language, marital status, employment, access to piped water, distance from the clinic, and duration of diabetes were significantly associated with access to HDM. IsiXhosa-speaking and unmarried participants were less likely to receive HDM, while unemployed individuals and those with longer diabetes duration were more likely to benefit. Geographic and infrastructural barriers further limited access, suggesting that HDM implementation may have inadvertently excluded vulnerable groups.
Conclusion:
While HDM was a valuable innovation during the pandemic, its uneven reach highlights the persistence of health inequities. Language, social support, and geographic location emerged as key barriers. These findings underscore the need for inclusive, community-informed service design and the critical role of CHWs in delivering equitable, person-centred care. Future interventions should prioritise co-design with communities and address structural barriers to ensure equitable access to healthcare during crises and beyond.
Keywords
Background
Diabetes mellitus (DM) is a growing global health challenge, with prevalence projected to rise from 11.3% in 2030 to 12.2% by 2045, disproportionately affecting low- and middle-income countries (LMICs). 1 In South Africa, the burden of type 2 diabetes (T2D) mirrors this trend, driven by rapid urbanisation, lifestyle changes, and rising obesity rates, particularly among women.2,3 The Western Cape Province, including Cape Town, carries a significant share of this burden, with high rates of diabetes-related morbidity and mortality. 4
South Africa’s public health system has long grappled with the dual burden of communicable and non-communicable diseases (NCDs), compounded by deep-rooted socioeconomic inequities stemming from apartheid-era spatial and economic policies. 5 These inequities continue to shape access to healthcare, particularly in historically marginalised communities such as those in the Cape Flats. Before the COVID-19 pandemic, efforts to address NCDs were being conceptualised within a community-oriented primary care (COPC) model that shifted perspective from the clinic to the community. 6 The model included community-based services (CBS) delivered by community health worker (CHW) teams and nurse coordinators employed by local non-government organisations. This approach integrated primary care with public health and social support systems. 7
The COVID-19 pandemic disrupted routine healthcare delivery, particularly for people living with chronic conditions. 8 Lockdowns, mobility restrictions, and the reallocation of health system resources towards pandemic response disproportionately affected access to NCD services.9,10 In response, Western Cape Government Health (WCGH) leveraged its existing COPC model to implement home delivery of chronic medication (HDM) by CHWs, aiming to decongest facilities, protect vulnerable patients, and maintain facility continuity and medication adherence.10-12 This intervention was part of a broader strategy to ensure health system responsiveness and resilience during the crisis. 13
However, implementing HDM services occurred within a complex web of social determinants.13,14 Language, housing, employment, and geographic location influenced who received care and who was left behind. 15 These factors raised critical questions about equity of access to pandemic-era innovations and the extent to which health system adaptations were inclusive of the most vulnerable.
This study aimed to evaluate whether socioeconomic differences between patients determined the reach of this service innovation. In doing so, we consider the implementation of HDM not only as an adaptive response to COVID-19, but also as a lens through which to evaluate the health system’s capacity to deliver equitable, person-centred care under crisis conditions.
Methods
Study Design
This study employed a descriptive cross-sectional design.
Study Setting
Four public sector primary care facilities were purposively selected, 1 from each of the 4 health substructures in Cape Town. These public sector facilities provided services to the indigent and uninsured majority of the population. Patients dependent on accessing care at these facilities were from low socioeconomic groups, ranging from informal settlements to formal low-cost housing. Care protocols for people with type 2 diabetes (T2D) were differentiated into stable and unstable groups based on their HbA1c levels (unstable was typically defined as greater than 8%) and the clinician’s judgment informed by evidence-based guidelines. Stable patients were seen every 6 months and received monthly pre-packaged medication via the pharmacy or alternative pick-up points. Unstable patients were seen every 1 to 2 months, and treatment was modified to try to achieve glycaemic control.
Study Population
This paper forms part of a larger project that evaluated pandemic-specific health system interventions to reduce morbidity and mortality related to the COVID-19 pandemic among people living with diabetes in vulnerable communities in Cape Town, South Africa. For this larger project, a minimum sample of 200 was calculated to measure a 1% change in mean HbA1c over 1 year. The study population consisted of adult patients with T2D who had received care at 1 of the 4 selected primary care facilities for at least 6 months. Clinicians at these participating facilities assisted in generating lists of eligible patients returning to these facilities following the pandemic restrictions. The research team systematically selected every second patient from the clinician-generated lists for possible recruitment. The research team contacted these systematically selected patients telephonically and invited them to participate in the interview via a formal informed consent process.
Study Instrument
The research team designed a structured, interviewer-administered questionnaire to collect data on HDM access, social demographics, economics, and treatment-related variables. Operational definitions included “formal housing” as a brick house or flat; “strong social support” as self-reported using the Oslo Social Support Scale (OSSS-3) 16 ; and “access to piped water” referred to the availability of water inside the home versus in the yard or nearby.
Data Collection
Three independent (not involved in service delivery at these facilities) and trained research assistants collected data between June and September 2021. The research assistants were fluent in all the local languages, particularly Afrikaans and isiXhosa. The research team designed and piloted an interviewer-administered questionnaire. The research assistants administered the questionnaire via telephone interview, taking 20 to 30 min per participant. The information provided was captured in the REDCap (Research Electronic Data Capture) database by the trained research assistants during the interviews. The REDCap data entry forms included built-in field validation rules (e.g. range checks and required fields) to ensure data completeness and consistency.
Data Analysis
Data from REDCap were imported into the Statistical Package for Social Sciences (SPSS) version 28.0 (IBM Corporation, Armonk, New York, USA) and R version 4.3.0 (R Foundation, Vienna, Austria) for analysis by a statistician. Fisher’s exact test and multivariable logistic regression were used to assess the associations between sociodemographic, economic, and treatment-related variables and access to HDM. Participants who chose not to report their income were kept in the analysis as a separate group. This method was selected to maintain sample size and to prevent bias from excluding or imputing data. Variable categories (e.g. age, household size, distance to clinic) were selected based on the distribution of the data and considerations of statistical power. More detailed groupings were not possible due to small subgroup sizes.
Ethical Considerations
The University of Cape Town Health Research Ethics Committee (HREC) approved the study (reference: 480/2020), and the Western Cape Government approved access to the 4 study sites (reference: WC_202009_013).
Results
The study analysed the responses of 267 people living with T2D. The sociodemographic characteristics of the study participants are shown in Table 1. Most participants were female (64.8%), older than 50 years (71.9%), married (56.9%), and lived in households with more than 4 people (96.3%). Most (79.1%) spoke either Afrikaans or isiXhosa as their home language. There was a significant association (P < .001) between receiving HDM and being an older adult, as well as being married. Patients who spoke isiXhosa were significantly less likely (P < .001) to receive HDM. There was no significant association between gender, household size, or primary care facilities.
Summary of Sociodemographic Variables Stratified According to Home Medication Delivery.
p is statistically significant (<.05).
Table 2 shows the socioeconomic characteristics of the patients with T2D. Most participants were unemployed (70.4%), had not completed high school (95.5%), earned less than $4,304 per year (70.4%), and relied on public transport (55.1%). A substantial minority lived in informal housing (20.2%) and did not have piped water in the home (17.2%). Most lived within 5 km of the primary care facility (57.7%). Over half of the patients (57.7%) reported strong social support and had access to a smartphone (79.4%).
Summary of Socioeconomic Variables Stratified According to Home Delivery of Medication.
p is statistically significant (<.05).
Overall, 76.6% received HDM. Receiving HDM was significantly associated with being unemployed, having a lower annual income, having private transport, living in a formal structure with running water, residing closer to the facility, and having a smartphone (Table 2). Education, access to electricity, and social support were not significantly associated with HDM.
Treatment-Related Factors
Table 3 presents diabetes and treatment-related factors associated with HDM. Most participants had lived with T2D for more than 5 years (71.5%), had access to a glucometer (58.4%), and nearly half were on both oral medication and insulin (46.8%). Having diabetes for longer and having access to a home glucometer were significantly associated with HDM.
Summary of Treatment-Related Variables Stratified According to Home Delivery of Medication.
p is statistically significant (<.05).
Table 4 presents the results of the multivariable logistic regression analysis, which examines the association between predictors identified in the bivariable analysis (Tables 1-3) and the odds of having HDM. The odds of receiving HDM were reduced if you spoke isiXhosa (OR 0.10, 95% CI: 0.04-0.27), were not married (OR 0.27, 95% CI: 0.12-0.57), had piped water outside your home (OR 0.30, 95% CI: 0.10-0.84), and lived further away from the facility (OR: 0.42, 95% CI: 0.20-0.84). Conversely, the odds of receiving HDM were increased if you were unemployed (OR 2.50-fold, 95% CI: 1.14-5.58) or had T2D for longer (OR 2.23, 95% CI: 1.01-5.02). Age, transportation, smartphone, and glucometer access were not significantly associated with HDM.
Results of Multivariable Logistic Regression Models Exploring the Association Between Predictors and Odds of Home Medication Delivery.
p is statistically significant (<.05).
Discussion
Summary of Key Findings
Our findings suggest that access to HDM was not equitable, raising questions about the service’s design and implementation. Speaking isiXhosa, being unmarried, being employed, not having piped water in the home, living further away from the facility, and having T2D for less than 5 years were associated with not receiving HDM. These results highlight how social determinants determined the reach of a pandemic-era service innovation intended to ensure adherence to medication and some continuity of care.
Discussion of Key Findings
IsiXhosa-speaking participants were significantly less likely to receive HDM, despite no site-level differences in language distribution. This suggests that language may be a proxy for more profound systemic inequities, including socioeconomic status, housing, and access to smartphones or digital devices. Smartphone access was included in the model to adjust the association between language and HDM for smartphone access. The odds of smartphone access were reduced by a factor of 0.08 (92%, P < .001) for isiXhosa participants compared to Afrikaans participants. Although the type of housing is not in the model, the effect of language is essentially adjusted for it through piped water access (housing type is closely related to piped water access). Thus, the observed association between language and HDM accounts for factors potentially linked to language. The isiXhosa-speaking communities in Cape Town tend to occupy the lowest socioeconomic bands, with the highest level of informal dwellings and services. Language has been a barrier to accessing healthcare services.17,18
Similarly, being unmarried was associated with lower odds of receiving HDM. While marital status may reflect social support networks, it also intersects with gender norms, household structures, and caregiving roles. However, the level of social support did not show a significant association with HDM, suggesting that marital status may capture broader relational and structural dynamics. Prior studies have demonstrated that social support is crucial for effective diabetes self-management and emotional well-being.19,20
Unemployed participants were more likely to receive HDM, which is likely a reflection of logistical realities. Unemployed individuals are more likely to be home during delivery hours and are more reliant on public sector services. However, this finding also highlights the importance of considering how employment status intersects with access to care, particularly in contexts where informal work and precarious livelihoods are the norm. Other studies have suggested that access to primary care is challenging for employed individuals due to limited opening hours and a “no work, no pay” policy in many businesses.21,22
Geographic and infrastructural barriers also played a role. Participants living more than 5 km from the clinic or those without piped water inside their homes were significantly less likely to receive HDM. These findings underscore the structural determinants of health and the limitations of service delivery models that fail to account for spatial and infrastructural inequities fully. Community health workers may have faced logistical challenges in reaching certain areas, particularly informal settlements, where addresses were not well organised or where people lived in another catchment area. 13 There may also have been safety issues for CHWs in the so-called “red zones” with high levels of crime and violence.23,24
Patients with a longer duration of diabetes were more likely to receive HDM, possibly reflecting their established relationships with the health facility and inclusion in chronic disease registries. This aligns with differentiated models of care, where stable patients are prioritised for decentralised services.23,25 However, it also raises concerns about excluding newly diagnosed or less-engaged patients from such innovations, especially if patient data is lacking or inaccurate in health information systems.13,26
These findings suggest that while HDM was a valuable intervention during the pandemic, its implementation may have inadvertently reinforced the impact of existing inequities.27,28 As noted in our reflections, receiving HDM serves as a proxy for evaluating the responsiveness and resilience of the health system. It also illustrates the tension between coordinating and integrating services, where attempts to ensure logistical efficiency may come at the cost of inclusivity. 11
The role of CHWs in this intervention deserves particular attention. Their ability to deliver care under challenging conditions speaks to the flexibility of community-based primary care models. Yet, their effectiveness depends on adequate support, training, and integration into broader health system planning. 29 The psychosocial toll on CHWs during the pandemic, including personal risk and community expectations, must also be acknowledged. 30
This study contributes to the growing literature on pandemic-era innovations and their implications for equity in primary health care.31-33 It underscores the importance of designing interventions that are not only efficient but also inclusive and responsive to the lived realities of diverse patient populations. Nevertheless, the health system’s ability to pivot at scale and in a very short period to offer HDM suggests a high degree of resilience, agility, and innovation.
Strengths and Limitations
It is essential to acknowledge that the study’s cross-sectional nature limits the ability to infer causality, and the reliance on self-reported data may introduce recall bias. Conducting telephone interviews with patients systematically identified by clinical staff returning to chronic care could introduce bias into the data collection and selection process, potentially excluding the most marginalised patients and those who were not contactable by phone. Research teams were unable to access facilities to oversee recruitment due to social distancing policies and research restrictions imposed during the COVID-19 pandemic. We were unable to verify self-reported access to HDM, which may also be subject to recall bias. The study, limited to 4 primary care facilities in Cape Town, may not represent the broader provincial or national settings. Furthermore, our findings may not be transferable to settings without established CHW networks or with different infrastructure. While the HDM model was feasible in Cape Town due to the existing COPC framework, its implementation in other settings would require adaptation to local health system structures and community dynamics. Finally, the study was conducted over several months during which HDM implementation may have evolved. This temporal variation could have influenced participant experiences and is noted as a limitation.
Despite the inherent limitations of a cross-sectional survey design, this study has several strengths. It provides valuable insights into the impact of socioeconomic factors on healthcare access during a critical period. Firstly, it gives a snapshot of the socioeconomic determinants influencing access to HDM during the COVID-19 pandemic among patients with T2D in Cape Town. Using a structured interviewer-administered questionnaire ensured consistency in data collection, and the telephonic method allowed for data gathering during a period when face-to-face interactions were limited. The study’s focus on a diverse sample from multiple primary care facilities across different Cape Town substructures enhances the relevance of the findings to similar urban settings.
Recommendations
Many public health experts believe that it is a question of “when” rather than “if” we will face similar pandemics in the future, and this study points to ways in which we can be better prepared. 32 Primary care facilities need to ensure that they have strong connections to the more marginalised and informal parts of the community in their catchment area. The COPC model of care, which links all households to CHW teams, can help address this issue, and ongoing implementation and scale-up are essential. Greater community engagement, through both formal and informal structures, can also increase trust, understanding, and connectivity.
Functional registration and empanelment systems with accurate addresses and contact details may enhance affiliation with the primary care facility. The value of this was highlighted during the COVID-19 pandemic, and such a system is planned for the rollout of national health insurance in South Africa. A specific primary care provider or facility would then know all patients.
Hiring CHWs who speak the local isiXhosa language, translating delivery instructions into other languages, and providing interpretation support to CHWs could help ensure that all patients fully comprehend and feel at ease with the HDM service. Future interventions should also accommodate the needs of employed workers by utilising technology such as e-lockers, which enable individuals to access medication at times and places convenient to them while maintaining physical distance during pandemics. These systems are being introduced in Cape Town as alternative pick-up points.
Future interventions must be co-designed with communities, informed by local data, and responsive to patients’ lived realities to promote equitable healthcare access. This study contributes to the broader discourse on health equity, offering a lens through which to evaluate the responsiveness of health systems during times of crisis and beyond.
Conclusions
This study highlights how social determinants influenced HDM for people living with T2D in Cape Town during the COVID-19 pandemic. Key determinants such as language (used as a proxy for socioeconomic status), marital status, employment, geographic location, and duration of diabetes shaped who benefited from this service innovation. While HDM was a critical intervention to maintain medication adherence, its uneven reach underscores the persistent inequities embedded within the health system. The findings suggest that health systems can be better prepared to respond to crises such as a pandemic so that crisis adaptations are more equitably utilised. Fully implementing COPC, with a robust data management system tailored to the African context, and ensuring proper registration and empanelment of populations, while considering the challenges of access faced by employed workers, can contribute to this preparation.
Footnotes
Acknowledgements
The authors express their sincere gratitude for the invaluable contributions of the research assistants who collected data through telephonic interviews: Mr Deon September, Ms LiTsoanelo Zwane, and Ms Naledi Makhafula. A special acknowledgement is extended to Dr Beverley Schweitzer for her support in facilitating gatekeeper approval for one of the facilities. The authors also wish to thank the clinical teams and management of the facilities for their support in generating lists of eligible patients and for assisting the fieldwork team during a period of immense pressure as routine care was reinstated following the pandemic restrictions. Additionally, the authors express special appreciation to all individuals living with diabetes who participated in the telephonic interviews and shared their experiences, contributing to this dataset; we trust that the insights gained will enhance access to equitable primary care.
ORCID iDs
Ethical Considerations
The University of Cape Town Health Research Ethics Committee (HREC) approved the study (reference: 480/2020), and the Western Cape Government approved access to the 4 study sites (reference: WC_202009_013).
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We wish to acknowledge the Foundation for Advancing Family Medicine (FAFM) and the Besrour Centre for Global Family Medicine (Besrour Centre) at the College of Family Physicians of Canada (CFPC) for funding the 2020 Besrour Centre Global Co-RIG Grants Program, which supported this research project. The views expressed in all published works and communications are those of the recipient and do not necessarily reflect those of the CFPC or FAFM.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.