A comparative impact evaluation of two human resource models for community-based active tuberculosis case finding in Ho Chi Minh City, Viet Nam

Globally, 1.6 million people die of tuberculosis (TB) annually and an estimated 10 million people develop TB disease each year [1]. Despite concerted efforts, incidence is currently declining at less than 2% per annum [2]. If this trend continues, TB elimination will not be achieved until the year 2182 [3]. As for all infectious diseases, early detection and cure of incident cases is crucial to decrease transmission and halt the epidemic. In endemic countries, National TB Control Programs (NTP) are responsible for providing TB care and prevention to the population. However, in routine NTP operations case finding entails receiving health-seeking patients with little effort on actively searching for new TB cases. As a consequence, there are 3 million annual incident TB cases globally whom these NTPs fail to reach. This “detection gap” sustains transmission and mortality [4]. One strategy to close the detection gap is active case finding (ACF) [5]. This has led to a global surge in ACF initiatives beyond routine household contact tracing [6, 7]. These initiatives range from facility-based systematic screening to community-based interventions among vulnerable populations [8, 9]. These initiatives have also shown that ACF can find more people with TB at an earlier stage of disease progression [10, 11]. However, better evidence on impact and cost-effectiveness is needed if they are to be sustained through national public health policies and budgets [12].

A prerequisite for ending TB entails the optimal engagement of communities. Community engagement has long been recognized as a priority in health care [13, 14] and the fight against TB [15, 16]. Previously defined as “paraprofessionals or lay individuals with an in-depth understanding of the community culture and language” that “have received standardized job-related training of a shorter duration than health professionals” with the primary goal of providing “culturally appropriate health services to the community,” [17] community health workers (CHW) comprise a critical component of community engagement [18]. When effectively engaged, CHWs can raise health system capacity through decentralization and task-shifting [19, 20]. CHW’s often also reduce access barriers for vulnerable populations, so that their services are considered more patient-centric than institutionally provided care [21]. As such, CHW’s comprise an ideal group to implement ACF for TB in their communities. However, there remains substantial heterogeneity in the CHW models applied and gaps in understanding regarding their optimal means of engagement [17, 22].

Under most applied models, CHW receive either a salary or an incentive-driven compensation. Oftentimes performance-based incentives (PBI) are additionally offered to reward high performers [23]. While salaried CHWs have been associated with increased motivation and reduced attrition [23], volunteer CHW models are often more attractive to authorities and implementers due to their lower fixed costs. Irrespective of payment model, a consistent recommendation is that financial compensation should be commensurate with occupational demands [18]. It has further been noted that CHWs are generally poor and dependent on the remuneration for their livelihood. For this reason, the World Health Organization (WHO) recommends to limit the dominance of PBIs as part of the overall compensation package [24]. However, it remains unclear whether a higher fixed payment is associated with improved performance and greater impact. The comparative effectiveness between salaried and volunteer CHWs was therefore identified as a key knowledge gap in need of further studies [25].

Between 2014 and 2016, a community-based ACF project named PROPER CARE was piloted in Go Vap district, Ho Chi Minh City (HCMC), Viet Nam. This project employed CHWs on a full-time basis with a commensurate salary for ACF and case management. Despite positive results [26], aspirations of other district health authorities to scale up this ACF initiative were stifled by the resource demands of these salaried CHWs. In response, we devised the IMPACT-TB study (Implementing proven community-based active TB case finding interventions). Its aim was to scale up the ACF activities piloted under the PROPER CARE project, to measure changes in TB case notifications resulting from these ACF activities, and to compare the relative changes in TB case notifications between ACF implemented by incentivized volunteers and salaried employees.

Our study demonstrated that community-based ACF can generate a substantial yield of previously undetected TB, even in a setting with a well-functioning TB program. Our ITS analyses substantiated the additionality calculations of the intervention by adjusting for changes in population size and seasonality. Although the trend differences in case notification rates were modest, they were statistically significant, supporting the case for a positive intervention impact. The post-hoc comparison between the study’s control and selected non-study districts further substantiated this finding by showing that the notification declines recorded in the study’s control areas were experienced across the majority of all non-study districts in HCMC.

Our results are concordant with those of community-based ACF studies conducted in other settings [31‐36]. A key success factor of these studies was expanding screening coverage to large portions of vulnerable populations and enabling access to the more sensitive CXR-Xpert diagnostic algorithm. Another commonality was leveraging existing healthcare structures. Successful community engagement projects tended to be complementary to facility-based case finding [37]. The engagement of established networks has been shown to be effective in other health areas as well [38]. A final similarity entailed the strong community linkages. Our CHWs coordinated with neighborhood leaders and commune TB officers to conduct household contact investigations. This raised community confidence in the study and lowered access barriers [39].

In evaluating the two human resource models, we found both to be successful in improving TB detection and notification. Our analysis further showed that these yields did not displace routine activities in either model. This suggests that there is a general benefit to engaging CHWs for community-based TB service coverage expansion. We further found that despite a lower case detection yield, the increases in All Forms TB notifications in the salaried employee districts was significantly higher than the gains in the volunteer districts. This finding was substantiated by the ITS analysis.

One explanation for the dichotomy in yield and additionality was the difference in screening activities. It is well documented that to find more cases a greater number of people need to be screened [40]. The higher number of screening encounters by the employees can be a proxy for time spent on outreach. The ability of full-time CHWs to achieve higher population coverage compared to volunteers has been noted in other settings [41]. Greater population coverage and community outreach have been cited as a catalyst for increased notifications [42].

Reports from the field also suggest that volunteers placed greater reliance on public health staff and neighborhood leaders to refer persons with suspected TB. This is intuitive given the lower remuneration and time commitment to ACF activities. Consequently, a greater proportion of individuals initiated on the TB care pathway by volunteers would have potentially been notified without this ACF intervention. This displacement of passive case finding has been documented among household contacts in Viet Nam [43].

However, as evinced by the PROPER CARE pilot [26], there is limited utility in an effective community engagement model that government stakeholders deem untenable for scale-up. As such, another positive finding of our study is that the volunteer human resource model still resulted in a significant increase in TB notifications for a lower cost. However, if resource-constrained programs choose to implement a volunteer-based human resource model, it will be necessary to consider other inherent risks besides lower additionality that may offset anticipated cost savings. Volunteers tend to require greater supervisory efforts to improve performance and ensure accurate reporting [25, 44]. Past studies have also found volunteerism to be associated with lower value perception and job satisfaction [45]. Concordantly, the attrition in volunteer districts was higher than in salaried employee districts. This was also the case among district supervisors. The attrition required continuous recruitment and capacity building, and possibly impaired quality of care. These downsides of volunteerism have been noted elsewhere [46, 47]. A health economic analysis that incorporates the cost implications of the attrition and other operational challenges encountered in our study will be provided in a separate manuscript.

The Viet Nam NTP has already successfully replicated the volunteer model in cities with a lower TB burden and cost of living compared to HCMC, the economic center of Viet Nam [48]. In these cities the model generated comparable notification increases to the employee model in HCMC (manuscript in preparation). This suggests that the volunteer model could be appropriate in settings, where the workload and the opportunity cost of volunteering are lower. Appropriately powering incentives for the socioeconomic context and employing a blend of monetary and non-monetary incentives are further means to optimize volunteer models [49]. Lastly, future research may seek better ways to draw on the altruistic capital of volunteer CHWs to overcome operational challenges [50].

Our study had limitations. Conducting research in a programmatic setting exposed our study to supply chain interruptions. These interruptions affected the integration of study activities into routine program operations. However, this also reflects the reality of field scale-up of an intervention. While ITS analyses aim to increase validity and interpretability of the results by compensating for secular trends, we may have missed nuanced confounders due to the non-randomized study design. A strength of our study was the length and scale of our outreach as well as integration within the existing NTP activities. As such, we believe in the transferability of our findings that effective engagement of CHWs can produce a positive notification impact and that full-time employment of CHWs can generate superior outcomes to other high TB burden, resource-limited settings to inform local strategies to end TB.

Leveraging community networks to expand TB service coverage is both feasible and effective in diagnosing and treating additional persons with TB. Engaging full-time, salaried employee CHWs in TB ACF schemes can lead to greater impact, and this human resource model should be prioritized for scale up where resources permit. While further studies are needed to optimize this community engagement model, it can be a powerful and readily available tool for advancing the global End TB Strategy targets.