Introduction

In low- and middle-income countries (LMICs), Community Health Workers (CHWs) are a cadre of healthcare workers that can help alleviate shortages of healthcare staff and achieve the goals of the primary healthcare system [1]. CHWs can be found in many regions of the world, from Ethiopia to Indonesia [2]. However, CHW programmes differ in their design, including the type of worker, level of training, scope of work, nature of supervision and the extent to which basic equipment is provided [2]. In South Africa (SA), Community Caregiver (CCG) outreach teams, also previously known as Ward-Based Outreach Teams (WBOTs), were established in 2011 as part of the Re-engineering Primary Health Care (RPHC) reform [3]. The goal of the CCG initiative was to support the delivery of primary healthcare (PHC) services in South Africa. The CCG teams bridged the gap between households and health care facilities, strengthening preventive healthcare within communities [3]. The CCG initiative brought services to patient homes without requiring a clinic visit, reducing the time and financial burdens on households to receive routine care. CCGs were conceptualised as an extension of existing primary healthcare (PHC) facilities, with facility managers providing the oversight, support and supervision of teams [4].

CCG outreach teams are normally comprised of about five CCGs and a team leader, typically a professional nurse [5]. Each CCG team is assigned a municipal ward where they provide health services to the community such as HIV, tuberculosis (TB), maternal and child health, and chronic condition care through household visits, typically working in pairs of 2–3 individual CCGs [3]. However, there are challenges in operating community-oriented care programs such as CCGs, both in South Africa and elsewhere in Sub-Saharan Africa [6, 7]. In South Africa, CCG coverage is unevenly distributed, with complete coverage in only half of the country’s 4,277 wards [8]. In March 2017, there were 3,275 CCGs teams—representing only 42% of the estimated 7,800 teams required to provide comprehensive national care [4]. And, while CCGs do ease access to care in communities, they have limited understanding of certain interventions, can overemphasize care for certain diseases and strategies, have high workloads, and can experience pushback from communities based on stigma and perception [7, 9–11].

Community-based care interventions like CCGs are only implementable with an appropriate understanding of resource requirements [7, 9–11]. Ideally, these programs aim to maximize the number of successful visits on a per day basis, since the number of households covered by each CHW is based on distance and travel time between households, demographic structure, and burden of disease [4]. Yet, lack of understanding by management may affect the quality of support and supervision of the CCG teams [4, 9, 12]. The impact of these challenges to the healthcare sector include inefficient use of labor and financial resources for health, such that public health service delivery does not result in optimal healthcare outcomes [13, 14].

There is growing evidence demonstrating the value and contribution of CCG teams in improving gaps of community health programs in South Africa. However, few economic analyses have evaluated both the operationalization and costs of CCG outreach programs to inform evidence-based resource allocation. Uncertainty in human resource and financial requirements can lead to CCG programs that do not make the best use of available resources. The primary objective of this manuscript was to evaluate the human resource and financial requirements necessary for implementation of CCGs in diverse settings. In this study, we assessed workloads and operational costs of CCG teams operating from two public health clinics in Ekurhuleni district. We report differences in CCG operations, human resource needs in delivering CCG-based care, and activity-based costs of services provided by the CCG teams.

Materials and methods

Results

Table 1 provides data on the characteristics of each clinic. Between March and October 2018, 159 self-reported daily TAM forms were included in the analysis out of 198 submitted from 11 CCG pairs (seven at clinic 1, four at clinic 2), totalling 39,667 minutes of data. CCG pairs spent a median time of 236 minutes (interquartile range [IQR]: 215, 261) per day conducting field activities at clinic 1, and a median 235 minutes (IQR: 201, 272) in the field at clinic 2. The shortest and longest days in the field lasted 105 and 345 minutes at clinic 1 and 114 and 325 minutes at clinic 2, respectively. CCG pairs at clinic 1 spent 49.5% (range: 40.3% - 53.4%) of their time directly interacting with patients at their households, compared to 35.0% (range: 33.0% - 38.1%) at clinic 2. It took a median of 8 minutes (IQR: 5, 15) for clinic 1 CCG pairs to travel between households and the clinic, compared to 14 minutes (IQR: 8, 20) for clinic 2 pairs (difference: 6 minutes, 95% Confidence Interval (CI): 5.1–6.9).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Clinic characteristics.

https://doi.org/10.1371/journal.pone.0282425.t001

A total of 1,552 household visits and 2,083 household activities were recorded. CCG pairs at clinic 1 visited an average of 17.4 patients across 9.5 households per day (1.8 patients per household on average), spending a median of 11 minutes (IQR: 7, 15) at each house (6.7 minutes with each patient) (Table 1). At clinic 2, each CCG pair visited an average of 8.8 patients across 6.7 households per day (1.2 patients per household), spending a median of 13 (IQR: 9, 18) minutes at each household (12.1 minutes with each patient). CCG pairs at clinic 1 therefore visited an estimated 8.7 patients (95% CI: 8.4–9.1) and 2.8 households (95% CI: 2.2–3.4) more per day than CCG pairs in Clinic 2. Among successful household visits, 3.9% (range: 0, 21.1) at clinic 1 and 3.2% (range: 0, 6.5) at clinic 2 were exclusively for TB investigations, and 52.2% (range: 0, 89.7) of visits at clinic 1 and 2.7% (range: 0, 6.5) at clinic 2 involved some TB activities, but were not exclusive to TB care.

For both clinics, the most common activity performed at household visits were non-tuberculosis client encounters for screening/counselling/monitoring, taking 22.1% of total field time for clinic 1-based pairs and 25.5% for clinic 2-based pairs (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Frequency, median duration, and total duration of activity type by clinic.

https://doi.org/10.1371/journal.pone.0282425.t002

Using a bottom-up costing approach, the total annual cost of CCG operations was estimated as $71,780 at clinic 1 and $49,097 at clinic 2. Per CCG pair, activities at clinic 1 incurred an average cost of $10,254, or $33.69 per CCG pair-day and $3.58 per successful household visit (Table 3, Tables S3.2. and S3.3 in S1 Appendix). Corresponding estimates in clinic 2 were $12,274 per CCG pair, $41.52 per CCG pair-day, and $5.85 per household successfully visited.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Bottom-up cost estimates.

https://doi.org/10.1371/journal.pone.0282425.t003

Using a top-down approach, the average cost per CCG pair-day was $41.02 at clinic 1 and $49.10 at clinic 2. The average cost per household successfully visited was $4.37 (range: $4.15-$4.66) at clinic 1 and $6.88 (range: $5.93-$8.66) at clinic 2 (Tables S4.1 and S4.2 in S1 Appendix). The largest determinant of unit cost was the number of households visited, with the average cost per household visit decreasing as the number of household visits per day rose (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Average cost per household based on the number of households visited per day.

Cost estimates are derived from the top-down cost estimates. As the number of households visited per days increases, the average cost per household decreases. Clinic Two has a more expensive cost per household at all values, but the difference in the cost per household between the clinics converges as the number of households visited per day increases.

https://doi.org/10.1371/journal.pone.0282425.g001

Over the full observation period, there were 26 unsuccessful visits and 947 successful visits at clinic 1, a 97.3% success rate (Tables 1 and 4). Clinic 2 had 165 unsuccessful visits and 414 successful visits, for a success rate of 71.5%. The most common reasons for an unsuccessful household visit were the target patient being unavailable or having the wrong address.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Frequency of unsuccessful visits.

https://doi.org/10.1371/journal.pone.0282425.t004

Discussion

Our study included over 1500 household visits made by 11 CCG pairs in Ekurhuleni District of Gauteng Province, South Africa. We report that CCG pairs perform a substantial amount of activities–thereby reducing burden to patients in seeking healthcare–at a modest cost to the healthcare system. Specifically, CCG pairs at both clinics were able to successfully complete seven to ten household visits per day, at a cost of $3-$6 per household successfully visited ($10,000-$12,500 per CCG pair per year). Our study also revealed variability in the number of successful household visits per day and the cost per household visit, driven in part by the higher frequency of unsuccessful visits in a more densely crowded, informally settled location. CCG teams were able to synergize TB contact investigations with other household health activities, with over one-third of all household visits involving both TB and non-TB evaluations.

At both sites, CCG pairs spent an average of about four hours per day in the field for household visits—with a large proportion of this time (greater than 50% at both sites) spent traveling between households, a finding consistent with an earlier study in Sedibeng and Umzinyathi districts in South Africa [21]. Despite spending a similar amount of time conducting household visits, the CCG pairs at clinic 1 (peri-urban) visited more households per day, spent less time at each household, and saw on average twice as many patients. There are several possible reasons for inter-clinic variability in the proportion of successful household visits, including both clinic-specific and external factors. First, the overall wealth and socioeconomic status of each catchment area may affect the likelihood of success, depending on the accessibility of household members for communications (e.g. by cell phone), ability to locate houses, and engagement of household members in out-of-household work. For example, in clinic 2’s catchment area, although households are generally close together, there are no formal street addresses. This leads to challenges in locating households and increases the travel time required between households, even though households are much more closely spaced than in clinic 1’s catchment area (a more formally settled, affluent area). The CCG pairs for clinic 2 therefore had more unsuccessful visits and greater travel time, despite having fewer pairs and making fewer attempted household visits during the observation period. These navigation challenges can also help explain why median travel time was shorter at clinic 1 (8 minutes) compared to clinic 2 (14 minutes). CCGs in similar settings could optimize their efficiency by planning travel routes to avoid unnecessary transit times between homes and instituting processes to reduce the frequency of unsuccessful household visits (e.g. closer pre-visit communication with households and more flexible scheduling options).

It is important for program and clinic managers to understand these challenges and the variability in success for CCGs across settings. If management is not aware of these issues, they could be prone to selecting strategies and approaches that are not cost-effective or do not align with the available resources (human and financial), resulting in inefficient or ineffective systems. These considerations, however, must be balanced against the possibility of greater need for accessible healthcare in regions like that served by clinic 2. Future research should combine estimates of cost and efficiency (as done here) with estimates of effectiveness, in terms of health outcomes.

The daily average cost per CCG pair was lower at clinic 1 than at clinic 2, despite clinic 1 having a larger total operating cost for its CCG program. This can be explained by economies of scale: clinic 1 had more CCG pairs, so fixed costs were distributed over a larger number of individuals, leading to a decreased marginal cost per additional CCG pair. The largest costs at both clinics were staff salaries and overhead expenses, and the biggest driver of unit cost was the number of household visits that each CCG pair was able to make in a day. The difference in costs between clinics demonstrates an important operational consideration: as more households are visited, the average cost per household decreases. And, as the number of household visits per day increased, the difference in cost per household between the two clinics decreased (i.e. there are diminishing marginal returns in savings). This is captured in the difference in operating costs between the two clinics evaluated in this study: a clinic that performs 9.5 household visits per day would spend $3.58 per household visit, whereas just making a few less visits each day (6.7 visits per day) would increase the cost to $5.85 per household visit.

CCG programs are not meant to replace clinic operations. Households and individual household members may choose to attend clinics directly, and all individuals can access public clinics and receive free treatment for diseases such as TB. But, there are still costs to patients for this care-seeking, such as travel time and lost wages. And South Africa does not have a national health insurance scheme at present, so for most individuals, these costs and any other costs of care are borne by individuals directly. Thus, CCGs could be, in terms of total cost, a more efficient alternative for patients.

While there are limited studies looking at the costs of community outreach interventions, one similar study investigated operations and costs of CCGs in deep-rural and peri-urban communities in South Africa [21]. While cost estimates are not directly comparable to those in this study (due to differences in methods and units of analysis), both studies highlight the importance of considering differentiated resource allocation that is contextualized to differences in supervision/management structure, geographical and demographic conditions, and healthcare needs in the communities where CCG teams operate. Another study by Lebina and colleagues (2020) looked at the cost of CCG (formerly WBOT) services as part of a larger integration package [22]. They also found that the cost of operating the program varied greatly depending on the size of the clinic, ranging from $145,228 for a clinic with 50 CHWs to $11,618 for a clinic with 4 CHWs—a cost comparable to that observed in our similarly sized clinics (15 and 9 CHWs for Clinic 1 and Clinic 2, respectively). Both studies also highlight the need to adequately fund and support CCG teams, considering the important role they serve in PHC. For example, CCG teams provide important infrastructure which can be utilized for effective and cost-effective contact tracing operations in low-resource settings. Contact tracing serves an important role in identification, monitoring, and treatment for many diseases, such as tuberculosis and COVID-19. But synergies from new service integration cannot be expected when CCG teams already face considerable workloads and resource constraints.

As with any study, our research has certain limitations. First, our study was only conducted at two clinics and we can draw only limited statistically significant conclusions with our clinic sample size. This limitation was the direct result of our voluntary participation study design (in that sites were provided the option to opt into, or out of, the costing analysis). However, the two clinics where data collection did occur represented distinct populations and social contexts, allowing us to highlight and contrast the operational challenges and operating costs between settings and set upper and lower bounds on the likely costs of CCG teams in other South African clinics. The use of two clinics also limited our ability to vary team composition within and across sites. Individuals composing each CCG pairing did rotate within each clinic, so that workers with differing levels of experience and work styles would spend time in the field together at various points during data collection. However, we could not randomly allocate CCGs across the two sites. We have made our data openly available on Github, should others wish to utilize our data in an expanded analysis of costs and operational requirements for CCGs in South Africa.

A second limitation was that time and motion data were self-reported, with the potential for bias and variability in the quantity and quality of data submissions over the study period. To enhance data quality, we provided routine feedback to CCG pairs on form submissions and developed a rating system to evaluate the quality of submitted forms, excluding those of the poorest quality. After review, more than 80% of submitted forms were considered of sufficient quality for inclusion in the analysis. Future studies can consider alternate ways of assessing time use for different commitments, such as time stamping data collection forms, automating logging systems, or direct observation of CCG activities by researchers [17]. Third, while the difference in per-visit costs between the sites was substantial, we do not have the ability—in the context of a study of two clinical trials sites—to disentangle the specific contextual factors that contributed to this difference, beyond the items (e.g. distance between clinics, frequency of unsuccessful visits) described above. Fourth, our data were limited to the evaluation of TB versus non-TB illness. Future research from a broader health system perspective could benefit from more detailed data on the specific non-TB illnesses evaluated.

Conclusion

Our study provides insight into the human resources required to maintain CCG services for general health promotion, a comparison of feasibility and efficiency in different settings, and cost estimates for CCG care. The CCG framework offers important infrastructure to expand contact tracing efforts for several diseases, including TB and Covid-19. This study can be helpful to future managers as they consider implementation of CCGs in different settings. One of the biggest challenges with implementation of CCGs across South Africa has been confusion about the duties of CCGs, and the financial needs necessary to operate such a program. The findings from this study can help managers appropriately plan resource allocation for CCGs, understand the likely drivers of human resource needs for CCG implementation, and raise awareness about the need to carefully consider local contextual and operational factors (i.e. infrastructure of surrounding locale, workloads of potential CCG team members, and the scale of implementation) to ensure efficient and optimized implementation and operation of these programs.

Supporting information

Acknowledgments

The authors would like to thank the CCG teams for their participation in this study. We would also like to thank the supervisory teams (both on and off site) for sharing the necessary information, the Ekurhuleni district and the Ekurhuleni Health District Research Committee (EHDRC) for allowing us to conduct the study in their districts.