Article Text
PDF
Abstract
Background Health services were severely affected during the many years of instability and conflict in Afghanistan. In recent years, substantial increases in the coverage of reproductive health services have been achieved, yet absolute levels of coverage remain very low, especially in rural areas. One strategy for increasing use of reproductive health services is deploying community health workers (CHWs) to promote the use of services within the community and at health facilities.
Methods Using a multilevel model employing data from a cross-sectional survey of 8320 households in 29 provinces of Afghanistan conducted in 2006, this study determines whether presence of a CHW in the community leads to an increase in use of modern contraceptives, skilled antenatal care and skilled birth attendance. This study further examines whether the effect varies by the sex of the CHW.
Results Results show that presence of a female CHW in the community is associated with higher use of modern contraception, antenatal care services and skilled birth attendants but presence of a male CHW is not. Community-level random effects were also significant.
Conclusions This study provides evidence that indicates that CHWs can contribute to increased use of reproductive health services and that context and CHW sex are important factors that need to be addressed in programme design.
- Community health worker
- multilevel analysis
- Afghanistan
- reproductive health
- health impact assessment
- measurement
- international health
- access to healthcare
- environmental health
- epidemiology
- primary healthcare
- public health
Statistics from Altmetric.com
- Community health worker
- multilevel analysis
- Afghanistan
- reproductive health
- health impact assessment
- measurement
- international health
- access to healthcare
- environmental health
- epidemiology
- primary healthcare
- public health
Introduction
Few countries are on target to meet the Millennium Development Goals, especially the reduction of maternal mortality.1 In April 2009, the World Health Assembly passed a resolution on primary healthcare and issued a statement calling for a return to community-based approaches to improving population health.2 A substantial body of evidence supports the importance of community-based healthcare in improving child health outcomes, but there is little evidence indicating whether such approaches have similar potential to improve reproductive health outcomes.
In 2002, Afghanistan had dismal levels of coverage of reproductive health services and one of the highest maternal mortality ratios in the world.3 In recent years, substantial increases in coverage of reproductive health services have been achieved, yet absolute levels of coverage remain very low, especially in rural areas. A Basic Package of Health Services (BPHS) was developed in 2003, comprising a set of cost-effective primary healthcare services with emphasis on the needs of rural areas and women and children. Community-based healthcare is a foundation of the BPHS, aiming to link the community to the formal healthcare system through community health workers (CHWs). Prior to 2004, very few CHWs were in place, and most of the CHWs were men. Beginning in late 2004, CHW recruitment and training programmes were scaled up rapidly. By 2006, 14 136 CHWs had been deployed across Afghanistan, with slightly more than half being women.4 CHWs are recruited from the community to deliver a set of basic preventive, promotive and curative health services, including provision of home visits, distribution of micronutrient supplementation and anti-malarials to pregnant women, distribution of contraceptives, diagnosis and treatment of malaria, diarrhoea and acute respiratory infection, health education activities, support for immunisation activities, providing referrals to health facilities, encouraging the use of skilled birth attendants at delivery, teaching family members to recognise danger signs and complications of pregnancy and childbirth, assisting in preparing for an emergency referral, and strengthening linkages between communities and the public health sector.5 CHWs undergo a training programme that includes three 3-week theoretical and clinical training sessions that focus on common infectious diseases, maternal health and common childhood illnesses. CHWs report to the health facility and are supervised and supported by a community health supervisor.4
Studies from diverse settings have demonstrated that CHWs can contribute to improvements in child health, including uptake of immunisations and reduction in child morbidity and mortality.6 ,7 Much less is known about the role of CHWs in increasing the use of reproductive health services. There is experience showing that CHWs can contribute to increases in use of contraception, but evidence concerning the use of antenatal care (ANC) and skilled birth attendance (SBA) is scant. An evaluation in Nigeria found that village health workers (VHW) contributed towards improved child health outcomes but had little effect on maternal health.8 A recent study in Bangladesh showed reduction in neonatal mortality when CHWs identified pregnant women, made antenatal and postnatal home visits, promoted birth preparedness and referred or treated sick neonates.9 Lady village workers (LVW) in Bangladesh were able to contribute to a reduction in fertility when participating in a contraceptive distribution programme.10 ,11 In Pakistan, lady health workers were associated with increased use of modern contraception12 but not ANC.13 A VHW programme in Tanzania was associated with observed increases in women's knowledge of danger signs during pregnancy but only a modest increase in use of SBA.14
The objective of this study is to determine if presence of a CHW in the community is associated with increased use of modern contraception, ANC and SBA, using data from Afghanistan, where severe constraints have resulted in some of the poorest reproductive health conditions in the world.3 ,15
Methods
Sampling and data collection
Data for this study were derived from the Afghanistan Health Survey (AHS) 2006, a multistage cluster survey that is representative of rural Afghanistan. Data for the survey were collected between September and December 2006. Five of 34 provinces (Helmand, Kandahar, Nuristan, Uruzgan and Zabul) that could not be surveyed due to insecurity were excluded from the sampling frame, in addition to the six major cities of Kabul, Mazar-e-Sharif, Herat, Jalalabad, Kunduz and Kandahar. The sampling frame was obtained from the Central Statistics Office in July 2006 and included over 45 000 enumeration units, which are villages and subvillages in rural areas and urban blocks in urban areas.
Since the primary interest of the MOPH was to assess the impact of BPHS implementation, which is largely targeted at rural areas, for the purposes of sampling, all districts in the country outside of the six largest cities were classified into one of the five strata: (1) the three ‘contracting out’ strata consisting of districts where financing from the European Commission, the United States Agency for International Development and the World Bank is used to contract non-governmental organisations to deliver services, (2) the ‘contracting in’ strata where the government uses World Bank financing to strengthen government delivery of services and (3) the ‘other’ strata, which is a residual category including all other districts in the country not falling into one of the previous strata.
Two-stage cluster sampling was used to select households. Within each stratum, enumeration units (clusters) were selected from a list of all enumeration units by probability proportional to size sampling. In the second stage, a compact segment method was used to sample households within a cluster.16 The cluster was partitioned into segments. Each segment contained an equal and predetermined number of compounds. One segment was randomly chosen. As the number of households within each compound was not known (separate household listing exercise was not conducted due to time and financial constraints), all compounds in a segment were listed and a fixed number of compounds were selected based on simple random sampling. If a compound contained multiple households, all of them were interviewed to ensure that all households in a segment had the same probability of being selected.17 The AHS sample included a total of 425 enumeration units (clusters), of which 397 were completed. Twenty-eight clusters were not completed because the survey teams were unable to move safely in those areas. Out of the total 8320 households sampled, 42 households were either absent or refused to participate in the survey, making the household response rate 99% and the final sample size 8278 households. In each sampled household, all ever-married women between 10 and 49 years of age were interviewed. There were a total of 8659 eligible women in the sample, of which 8281 were interviewed, making the individual response rate over 95%.
Ethical approval was obtained from both the Ethical Review Board at the Afghanistan Ministry of Public Health and the Committee for Human Subjects Research at the Johns Hopkins Bloomberg School of Public Health.
Dependent variables
Three reproductive health outcomes were modelled: current use of modern contraceptives, ANC and SBA. Modern contraceptive use was measured among currently married non-pregnant women, and the latter two outcomes among all ever-married women aged 10–49 years who had delivered a live birth in the 2 years prior to the survey. The definition of modern contraception includes female sterilisation, intra-uterine device, contraceptive pill, contraceptive injection and condom. CHWs dispense condoms and contraceptive pills to women in their communities. The definition for ANC was the receipt of at least one ANC visit with a doctor, nurse, midwife or CHW during the woman's most recent delivery. The definition for use of SBA includes having one's last delivery assisted by a doctor, nurse or midwife. CHWs and traditional birth attendants are excluded from the definition of SBA.
Independent variables
Individual, household and community levels are used in the models. The individual variables include woman's education (whether she had any education), parity and age (both continuous variables categorised into four and five groups, respectively). Household-level variables include wealth status and presence of other married women in the household (not including the respondent). Wealth status is calculated using the asset-based measure developed by Filmer and Pritchett.18 Households were categorised into three groups: the poorest 40%, middle 40% and wealthiest 20%. The presence of other married women in the household is categorised into three groups: no other married woman, one to two other married women and three or more other married women. Community-level variables include the presence of at least one CHW in the community and the average walking time from the village to the nearest health facility. CHW presence was measured by asking all interviewed households whether they were aware of any CHWs within the community. If no respondents in the cluster reported that a CHW was present, the community was coded as not having a CHW. If one or more respondents in the cluster reported CHW presence, the community was coded as having a CHW. The sex of all CHWs reported to be present was recorded. The average walking time from the village to the nearest health facility is calculated from only those respondents in the sampled cluster who walked to the health facility.
Analytical methods
A robust two-level logistic regression model was used to estimate the independent effects of the variables of interest on the three outcomes. SEs for community-level effects cannot be properly estimated in marginal models.19 The presence of a CHW in a community is common to all women residing in the community and should not be treated as an individual characteristic. Individuals and households represent the level 1 unit and clusters represent the level 2 (community) unit.
V.10.1 of STATA20 and the gllamm command21 were used for the multilevel analyses.
Results
Table 1 presents a description of the sample and table 2 shows the bivariate relationships between the independent variables and each of the three dependent variables. From a total of 3294 reported deliveries, missing data reduced the effective ANC and SBA sample size to 3098 and 3118, respectively.
Description of sample
Prevalence of community-, household- and individual-level variables among respondents
Community-level effects
Table 3 presents the results of the multilevel analysis. The presence of a male CHW is no different than having no CHW present, in terms of the three outcome variables. In contrast, presence of a female CHW in the community is associated with increased use of contraceptives (OR=1.61, 95% CI 1.21 to 2.15), ANC (OR=2.71, 95% CI 1.87 to 3.92) and SBA (OR=1.75, 95% CI 1.18 to 2.58), holding individual- and household-level factors constant.
Results from two-level multivariate regression models
Model 1 in table 3 includes only the fixed effects of the community-level variables and the community-level random intercepts, whereas model 2 also includes individual- and household-level variables. Comparing models 1 and 2, the association between CHW presence and the outcome variables does not appear to be mediated by a respondent's individual-level factors.
Respondents living two or more hours' walking distance from the nearest health facility had lower odds of using contraception, ANC and SBA use compared with respondents living closer to health facilities. However, the odds of using contraception, ANC and SBA improved with the addition of individual-level factors for all categories of the distance variable compared with the reference group of respondents living <2 h from a health facility. This indicates that women living farther away are also more likely to have individual characteristics associated with non-use of reproductive health services, including lower wealth status and lack of education.
Random effects
The significance of the random intercepts is tested when the differences between the log likelihood for the ordinary logistic regressions (without the random intercept) are compared with the log likelihood of the model with the multilevel structure (model 2 in table 3). The χ2 distribution is used to test this difference with the degrees of freedom equal to the difference in the total number of parameters in both models. In this case, the random intercept is the only extra parameter in the multilevel model, so the degrees of freedom is 1 for all outcomes. For each of these outcomes, the χ2 statistic with one degree of freedom yields a p value <0.0001. The highly significant p values for all three outcome variables indicate that a significant degree of the variance in each of the models is explained by living in different communities. In other words, unmeasured community heterogeneity accounts for a significant degree of the difference in use of reproductive health services in rural Afghanistan.
An intraclass correlation (ICC) coefficient was estimated to examine the proportion of variance attributable to individual and community factors. The ICC is the ratio of cluster level/community variance to total variance and shows how clustered the outcomes are at the community level. An empty model without any explanatory variables was run to estimate how clustered the outcomes are at the community level. The empty model indicated clustering at the community level with ICC of 0.28 for modern contraceptive use, 0.42 for ANC and 0.41 for SBA. When explanatory variables were added to the model, the ICC reduced to 0.23, 0.33 and 0.30 for contraceptive use, ANC and SBA, respectively (shown in model 2 ICC). Whereas the explanatory variables included in the model reduced the ICC, there was still considerable variation between clusters with respect to the outcome variables.
Household and individual effects
Household wealth status and receipt of formal education were consistently associated with increased use of all three services. Older age groups tended to be associated with higher use of contraception and lower use of ANC compared with younger age groups, with no difference observed for SBA use. Higher parity was associated with higher use of contraception and lower use of SBA, with no difference observed for ANC. Having more than two other married women in the house was associated with lower use of ANC and SBA, with no difference observed for contraception.
Discussion
This study's finding that community presence of female CHWs is associated with higher utilisation of modern contraception, ANC and SBA, indicates that female CHWs can contribute to increases in coverage of reproductive health services. The magnitude of the effect is large. The ORs for female CHW presence on all three outcome variables are similar in magnitude to those for household wealth status and mother's education—two variables consistently found to be strongly associated with use of reproductive health services in Afghanistan15 and other settings.22–24
This study builds upon the earlier study from Afghanistan as well as studies from other settings through a multilevel analysis that more accurately estimates the effect of community-level variables. None of the other studies identified in the literature that examined the effect of CHW presence in the community on use of reproductive health services used a multilevel analysis. The findings from this study contrast with results from the earlier study from Afghanistan, which found no association between presence of female CHWs and use of SBAs and a negative association between presence of male CHWs and use of SBAs.15 This study improves upon the earlier study by employing multilevel analysis and demonstrates the importance of having sufficient time to assess the effects of a CHW programme on use of reproductive health services, in this case 3 years rather than 1 year.
Few studies in the literature have systematically examined the differential effect of CHW presence on use of reproductive health services by CHW sex. The study of VHWs in Nigeria concluded that one likely contributory factor to the low impact of VHWs on maternal health outcomes was the disproportionate number of male VHWs.8 The contraception distribution studies in Bangladesh only used lady village workers, with no male village workers involved.10 ,11
Previous studies have hypothesised that male CHWs in Afghanistan may contribute to increased use of reproductive health services by influencing male decision makers in the community.15 This is not supported by evidence from this study. Many decision makers in Afghanistan—male or female—are likely to prefer that women receive reproductive health services from a female rather than male provider. This is unlikely, however, to fully explain the finding since this finding applies to services for which CHWs are direct providers (ie, contraception and ANC) and to SBA for which CHWs are not direct providers. The importance of CHW sex may lie in a combination of preference for female providers and the ability of female CHWs to persuade other women in the community to use modern skilled reproductive health services.
We believe that the results are relevant to other countries even if the specific estimates may not be generalisable to settings outside Afghanistan. Provider sex may be a more important factor in Afghanistan relative to some other settings since women in much of Afghanistan rarely interact with men outside the family. The importance of CHW sex is likely to be greater in environments with strong normative pressure regulating the movement of women and their interactions with men. Other settings may need to pay attention to the sex of the health worker to establish their local relevance. There is also need for systematic investigation of household decision making related to use of reproductive health services to more fully establish the mechanisms through which female CHWs contribute to increases in use of reproductive health services and to identify opportunities for intervention. Another potentially generalisable finding is the significant effect of community-level heterogeneity, such as geographical variation, differences in security, perceptions about quality of care and local beliefs about appropriate care during delivery, suggesting that these factors should be considered more systematically when explaining differences in reproductive health services use.
Results from this study show that only half of rural communities in Afghanistan had an active CHW and fewer than 40% had a female CHW in 2006. Although the CHW programme in Afghanistan was still being scaled up, this raises questions regarding how active CHWs in Afghanistan are. According to government statistics, the drop out rate for CHWs is low, typically under 10%.4 These results may indicate that many CHWs remain nominally active but at activity levels that are so low that many in the community do not report that a CHW is present in the community. CHWs are not salaried. They may receive in-kind or monetary incentives based on their workload. There is need to understand the types of support needed by CHWs and the factors that influence CHW retention and activity levels, especially for women.
The associations between the other independent variables and the outcomes of interest are largely similar to other studies. Similar to the earlier study from Afghanistan,15 distance from the nearest health facility, household wealth status and receipt of formal education are associated with increased use of SBA. These factors are also associated with use of modern contraception and ANC, which is consistent with results from other settings.22–24 The earlier study from Afghanistan found that even small distance increments within a facility catchment area are associated with differences in use of SBAs; the current study confirms that larger distances are associated with larger decreases in use of skilled ANC and SBA. Women living >4 h from the nearest health facility were four times less likely to use SBA and three times less likely to use ANC compared with women living <2 h from the nearest facility. Future studies should examine the interaction between distance from the nearest health facility and CHW presence. CHW presence may be especially important in remote communities that are distant from the nearest health facility.
This study has a number of limitations. Since CHWs are not recruited, deployed or retained on a random basis, the association observed between CHW presence and use of reproductive health services may be due to potential selection bias. Communities that have female CHWs may be different in other ways from communities with only male CHWs or no CHWs. Such communities may have fewer normative restrictions on the movement of women outside the household and their engagement with modern heath services. This study has controlled for distance to the nearest health facility but not for other community-level effects. Nonetheless, even after accounting for the heterogeneity of communities through the random effects model, the influence of female CHWs on use of reproductive services was strong.
Another limitation is the lack of independent verification of CHW presence in the community and levels of activity. The true effects may be larger than observed in this study.
The validity of household reports of CHW presence in the community was not assessed. Some trained and deployed CHWs may not be working at all, while others may be working but covering a smaller population than the target population of 1000–1500 households. Other CHWs may be trying to cover the target population, yet barriers may prevent them from engaging with households in such a way that awareness and utilisation of their services is widespread. As a consequence, the true effect of CHWs may be larger than we estimated. Furthermore, as AHS covered only 72% of the rural population, the effect of CHWs on the outcomes in the excluded provinces is not known. Insecurity may modify the effect of determinants of the use of reproductive health services. Despite its limitations, this study contributes to the evidence on the effectiveness of interventions to improve reproductive health outcomes. Female CHWs, in settings like rural Afghanistan, can contribute substantially to improvements in the coverage of modern contraception, ANC and SBA.
What is already known on this subject
Evidence demonstrates that community health workers can contribute to improved child health outcomes, but there is little evidence regarding whether community health workers can contribute to increased use of reproductive health services.
What this study adds
This study demonstrates that community health workers can contribute to increased coverage of reproductive health services.
The sex of the community health worker, in some settings at least, is an important determinant of community health workers' ability to contribute to increases in service coverage.
Other community-level factors that are unobserved in this study determine levels of service utilisation.
Context and community health worker sex are important factors that need to be taken into account in programme design.
Acknowledgments
We would like to acknowledge the many contributions to this study made by members of the Monitoring and Evaluation Technical support team from the Johns Hopkins Bloomberg School of Public Health and Indian Institute of Health Management Research and colleagues from the Ministry of Public Health of the Islamic Republic of Afghanistan. We would also like to thank the 250 surveyors who collected data in difficult circumstances.
References
Footnotes
Funding This work was supported by the Third Party Evaluation Contract (MoPH/AFG/GCMU/19/04) between the government of Afghanistan and the Johns Hopkins Bloomberg School of Public Health with the Indian Institute of Health Management Research. The Ministry of Public Health of Afghanistan contracted Johns Hopkins Bloomberg School of Public Health with the Indian Institute of Health Management Research as independent third party evaluators of the Afghanistan health sector.
Competing interests None.
Patient consent This paper does not contain any personal medical information about an identifiable living or dead individual.
Ethics approval Ethical approval was obtained from both the Ethical Review Board at the Afghanistan Ministry of Public Health and the Committee for Human Subjects Research at the Johns Hopkins Bloomberg School of Public Health.
Provenance and peer review Not commissioned; externally peer reviewed.