Building upon five decades of CHW programs in Tanzania, the Integrated MNCH CHW Program sought to establish a volunteer cadre of MNCH CHWs that provide a range of socially oriented, preventive, and promotive services including village mapping, pregnancy surveillance, counseling through home visits, and health promotion meetings. This model is in contrast to other current CHW initiatives like the iCCM model, which expands on preventive and promotive activities to include curative services, requiring clinical oversight, training, and support inclusive of more sophisticated supply chain mechanisms. We examine how time allocation, service delivery, satisfaction, and motivation are effected by the following critical design elements (1) CHW profile and MNCH knowledge, (2) CHW to population ratio, (3) program monitoring and supervision, and (4) incentives.

CHW characteristics, including age, gender, education, and martial status, may influence performance [14,15]. At inception, the Integrated MNCH CHW program sought to train an equal proportion of male/female CHWs, with secondary school or higher education, who resided in and were selected by the communities where they would eventually work. In practice, MNCH CHWs were nearly evenly split in gender (55% male, 45% female), under 35 years of age (63%), nearly one third was unmarried, and only half met the MoHSW requirement of Form 4, secondary education or higher. Elsewhere globally, the sex of CHWs has been show to influence the reported frequency of counseling [16] and uptake of services, particularly for reproductive health [17] and child nutrition [18], as well as record keeping [16]. Education has been listed as an influencing factor in five prior CHW studies, and while higher education may lead to better performance, it may also correspond to higher rates of attrition [14]. Studies exploring the influence of age on performance have found evidence of poorer performance among younger and older CHWs [14]. In Kenya, the optimal CHW age range was 30–40 years [16]. While we explore the implications of these social characteristics elsewhere (Intersectionality implications of scaling up MNCH CHVs in Tanzania: examining how gender, age and educational determinants combine to influence CHV experience, to be submitted.), further analyses found no significant differences in the mean number of households visited monthly (service delivery) and in the mean composite scores for overall knowledge by CHW education, gender, or age, although qualitative data indicated that CHW education, gender, and age did influence CHW communication and visits with community members [19].

Overall mean knowledge scores for MNCH CHWs were observed to be poor at less than 50% for 8 of 10 MNCH domains assessed. However, set standards for interpreting knowledge scores for CHWs are not available within the literature to enable comparison between studies. Job aids may temper the effect of these gaps in CHW knowledge during counseling, and supportive supervision and fresher training help to overcome them. However, additional efforts are needed to prioritize key messages based on evidence and to assess whether changes in content or duration of training might lead to improvements in knowledge, ultimately translating to better quality of counseling. A 2010 review by WHO and the Global Health Workforce Alliance identified 19 studies which assessed CHW knowledge, attitudes, and practices [20], but only one included data on frequency of CHW recall of critical MNCH content [21]. However, details on the content of knowledge domains assessed are not described, rendering comparison with our study difficult.

In the absence of global standards for CHW knowledge and a wider array of examples in the literature [20], we used the Integrated MNCH CHW Program guidelines and job aids as a reference point and compared MNCH CHW knowledge against health center providers and CHWs interviewed prior to the program’s start in 2011. Among these three different populations, CHWs interviewed in 2011 had a mean overall knowledge score of 64% as compared to 50% for MNCH CHWs and 48% for RCH health center providers. While these single-point estimates do not consider the time lapse between RCH provider training and interview, the finding of lower knowledge scores among RCH health center providers compared to MNCH CHWs is surprising given the longer pre-service training of RCH providers. The performance of CHWs interviewed in 2011—a cadre of providers older in age by a median of 9 years and for whom only 17% had secondary school education or higher—suggests the potential for CHWs existing within the community and/or trained as part of prior vertical programs independently of education levels to be utilized as CHW candidates if they are able to meet certain competency requirements.

Beyond the contextualization of knowledge scores against those observed for other providers, we note that our data do not allow us to link knowledge scores to outcome and impact level indicators. Further, findings from Morogoro Evaluation Project (MEP) facility assessment activities in 2012 suggest that provider knowledge may not translate to improvements in the content of services provided. Rather, they suggest greater complexities may influence service delivery extending above and beyond what one “knows,” including provider and client perceptions, client characteristics, and the availability of provider time for service delivery given the high patient volume and competing time demands, among other factors (Quality of postnatal counseling in primary health care centers in Morogoro, Tanzania: effects of additional training and supervision, submitted for publication 2015) [13]. The poor quality of ANC and PPC in health center services also raises concerns about the implications of CHW efforts to generate increased demand for care seeking in health facilities that are often understaffed, overburdened, and ill-equipped (Content and duration of antenatal counseling and associated factors in selected health centers in Morogoro Region, Tanzania, to be submitted). This highlights the need for CHW programs to consider facility-based improvements parallel to the training, establishment, and ongoing support to community-based cadres. The importance of this has been echoed elsewhere in the literature as part of broader calls to recognize the health systems within which CHW programs are embedded [14,22] and evidence which suggests that relationships between CHWs and providers may strongly effect performance.

Studies elsewhere suggest that CHW performance is higher when the CHW to population coverage ratio is lower [14]. Wide variations in MNCH CHW to population were observed because the program sought to train a fixed number of CHWs per village. In the villages where they were established, MNCH CHWs reported providing services to a mean of 186 households, a figure comparable to the 150 recommended by the Millennium Development Villages [23] and an improvement from the 1 to 3438 households covered by CHWs in the same geographic area in 2011. This population to household coverage ratio roughly corresponds to an estimated 1 CHW per 1000 population. When translated into program activities, CHWs would need to make an estimated minimum of 471 home visits annually or 39 visits monthly above and beyond routine pregnancy and delivery surveillance activities. While steady declines were observed in the mean number of monthly CHW home visits from 21 in May to 15 in September 2013 (median 14 to 12, range of 0–207), results suggest that CHWs may be exceeding visits targets for pregnant women and children 1–59 months but falling short during the postnatal period (0–28 days). While the declines in mean household visits per month could be in part attributed to sharing of workload marked by the successive addition of new CHWs (some of whom went to new villages, others to villages where MNCH CHW had already been deployed), it may too reflect a continuing downward trend in outputs often characteristic of program implementation over time. Moving forward, efforts need to be made to use HMIS and population coverage data to improve its quality and use for CHW performance monitoring, reducing variability in data and service delivery outputs across CHWs and over time. Data on the timing of home visits, and in particular their proximity to the date of delivery for postpartum care, also needs to be measured along with reported uptake of facility-based MNCH services.

At the national level, as dialogue continues on how many CHWs to train per village or population, the implications on individual workload need to be considered given the wide variations in village sizes across Tanzania. While we did not observe a significant association between mean households served and mean monthly home visits reported (Pearson’s correlation coefficient of 0.0128, P < 0.8), MNCH CHWs provided services for a mean of 2 days per week—effectively part time. For alternative, more intensive CHW models, setting the number of CHWs to a population ratio versus village number may serve to reduce variability in coverage, content, and quality of care.

Assuming a CHW to population ratio of 1 per 1000, the national scale of this CHW model would require 43 625 CHWs to reach all mainland Tanzanians. The feasibility of identifying and recruiting such a high volume of new providers will need to be determined, particularly if required to have secondary school or higher levels of education. The Integrated MNCH CHW Program’s strategy of training an initial group of CHWs in each facility catchment area and later returning to train more may allow for a stepwise approach to implementation which eases the strain on facility providers and allows communities/MoHSW to identify candidates that meet eligibility criteria over time. This also has implications for CHWs already working in the villages, requiring that they re-adjust their catchment area according to the total number of CHW working in the village.

During the time motion data collection, we sought to better understand the competing demands upon MNCH CHWs’ time, which may in turn have implications for service delivery. While few studies have reported on CHW time spent on service delivery [14], findings from an assessment of community health volunteers in Madagascar suggest a correlation between CHW performance and time spent on the job [24]. In our study, beyond exploring the association between time and home visits, we sought to understand the linkage between financial incentives and total time spent per week on MNCH CHW activities. CHWs were observed to work for a mean of 5 h per week, of which less than 2 h was spent on home visits. When considered in context with the financial incentives received, the overall time spent working on MNCH CHW activities compares favorably to the estimated hourly wage earned through time spent on alternative income-generating activities. This may suggest that CHWs’ programmatic inputs directly correspond to the financial compensation they receive through incentives for initial training and attendance of supervisory meetings. Moving forward, if MNCH CHWs are asked to serve in a full-time capacity, the financial incentive structure would need to be adjusted to ensure comparability with their current earning potential in other sectors.

The availability, frequency, and location of supervision and its linkages with CHW motivation and quality of work have been discussed with limited rigor in the literature [14]. Following the identification, training, and deployment of MNCH CHW activities, the Integrated MNCH CHW Program provided ongoing supportive supervision through support to (a) two facility-based providers per facility, who were encouraged to conduct monthly meetings, and (b) quarterly regional/district MoHSW and Jhpiego supervisory visits. The latter were found to occur with less regularity, in part because Jhpiego/MoHSW may have been engaged in the training of subsequent batches of CHWs and unable to simultaneously initiate supportive supervision of those previously trained. Among facility providers, supervision occurred nearly universally every month and was complemented by visits by facility supervisors to CHWs in the community every other month. CHW attendance of supervisory visits was high, a factor which may be attributed to financial incentives disbursed, the amount of which corresponds to ~50% of the average the MNCH CHW’s household income.

Monthly and quarterly supervisory visits focused largely on HMIS registers. Despite this emphasis, inconsistencies were pervasive in CHW recordkeeping and nearly 25% of CHWs did not maintain MCH registers. This raises concerns about the quality of register data and suggests that a review of register format and content may be warranted to reduce complexity and ease routine documentation of home visits by CHWs. Given the added use of registers to facilitate CHW work flow planning, added attention should be paid to reviewing the accuracy of calculations for home visit scheduling and timely execution of these scheduled visits. To overcome some of these barriers, future CHW programs should consider the use of mobile platforms, which provide frontline health workers with simple tablets or mobile phone devices that facilitate client registration, tracking, and workflow planning and, ultimately, can be linked with reminder and alert systems which send messages to clients and can also offer refresher training information [25,26].