Bulletin of the World Health Organization

Home visits by community health workers to prevent neonatal deaths in developing countries: a systematic review

Siddhartha Gogia a & Harshpal Singh Sachdev a

a. Sitaram Bhartia Institute of Science and Research, B-16 Qutab Institutional Area, New Delhi 110016, India.

Correspondence to Harshpal Singh Sachdev (e-mail: hpssachdev@gmail.com).

(Submitted: 02 July 2009 – Revised version received: 16 January 2010 – Accepted: 17 January 2010 – Published online: 10 May 2010.)

Bulletin of the World Health Organization 2010;88:658-666B. doi: 10.2471/BLT.09.069369


The last three decades have witnessed a significant fall in mortality rates among children under 5 years of age in developing countries, whereas neonatal mortality rates have decreased at a slower pace.1,2 Estimates published in 2001 suggest that about 38% of all under-5 mortality occurs in the neonatal period and accounts for 4 million deaths worldwide each year.3 Ninety-nine per cent of global neonatal mortality occurs in developing countries.4 It is widely recognized that lowering neonatal mortality is vital for achieving further reductions in infant and child mortality.1,58

Among neonatal deaths, three fourths occur during the first week of life, while 25−45% occur within the first 24 hours after birth. The majority occur at home.1,5,9,10 A strategy that promotes universal access to antenatal care, skilled birth attendance and early postnatal care has the potential to contribute to sustained reductions in neonatal mortality. To complement facility-based care, home-based strategies to promote optimal neonatal care practices have been proposed. Two related modalities for this purpose have been attempted in programmes and research trials in the last decade. The first involves home visits for the promotion of optimal neonatal care; the second includes home-based management of neonatal infections and other neonatal problems arising during birth, including neonatal resuscitation if required, plus the promotion of preventive interventions.

Information on the effectiveness of these complementary community-based approaches for reducing neonatal mortality is needed to frame policy for their inclusion in public health programmes. Further, the relative value of preventive or promotive and treatment interventions is unclear. We have therefore performed a systematic review for the purpose of determining whether home visits for neonatal care by community health workers can reduce infant and neonatal deaths and stillbirths in resource-limited settings with poor access to health facility-based care.


Inclusion criteria

We only looked for trials comparing groups that received different experimental interventions, including home visits for neonatal care by community health workers, with a control group that did not receive any home-based intervention by community health workers during the neonatal period. Trials having a random, quasi-random or non-random design, with individual or cluster allocation, were eligible for inclusion. However, trials evaluating interventions for the home-based follow up of infants born and initially cared for in a hospital were excluded, as were single-intervention trials.

The trial population had to be composed of neonates (i.e. infants ≤ 28 days old or in the first month of life if age not specified in days) born in resource-limited settings with poor access to health-facility-based care.

Trials were required to include home-based experimental interventions by community health workers in the neonatal period. However, they could also include additional home-based interventions by community health workers during pregnancy or delivery.

Interventions during the neonatal period could include one or more of the following: (i) the promotion of optimal neonatal care practices, such as exclusive breastfeeding, keeping the baby warm and clean umbilical cord care; (ii) caregiver education to improve caregiver recognition of life-threatening neonatal problems and appropriate health care seeking behaviour; (iii) the identification of signs of severe neonatal illness and referral to a health facility; or (iv) home-based management of neonatal conditions.

Interventions during pregnancy could comprise one or more of the following: (i) promotion of antenatal care; (ii) health education and/or counselling of the mother regarding desirable practices during pregnancy; (iii) promotion of delivery in a hospital or at home by a skilled birth attendant; and (iv) education about safe and/or clean delivery practices.

Interventions during delivery could include the implementation by community health workers of safe delivery practices at home and proper care of the neonate immediately after birth, such as keeping the baby warm, providing neonatal resuscitation (if required) and initiating breastfeeding early.

A community health worker was defined as any paid village health worker or unpaid volunteer, or any auxiliary health professional working in the community.

Outcome measures

The primary outcome was the all-cause neonatal mortality rate, defined as the number of deaths from any cause in infants up to the age of 28 completed days (or 1 month) divided by the number of live births in the study population.

Secondary outcomes included: (i) all-cause infant mortality rate, defined as the number of deaths from any cause during the first year of life divided by the number of live births in the study population; (ii) cause-specific neonatal mortality: deaths due to sepsis, tetanus, asphyxia or prematurity (as defined by authors, irrespective of single- or multiple-cause assignment); (iii) stillbirth rate; and (iv) care practices during pregnancy and delivery and in the postnatal period in trials providing data on neonatal mortality. Such practices included the following: > 1 antenatal care visit; 2 doses of maternal tetanus toxoid injection; money saving for childbirth; skilled care at birth; clean umbilical cord care; breastfeeding initiation within 1 hour of birth; bathing of the neonate no less than 24 hours after birth; and skin-to-skin care after birth.

Search methods

We searched PubMed, the Cochrane Controlled Trials Register in the Cochrane Library, Excerpta Medica Database (EMBASE), Health Services Technology, Administration, and Research (HealthSTAR), the ISI Web of Science, the Cumulative Index to Nursing and Allied Health Literature (CINAHL) and clinical trials web sites. Included were articles in any language published from the beginning of each database up to 5 October 2008. For all included articles, we performed a lateral search in PubMed by using the “related articles” link. We also hand searched for reviews and for conference proceedings/abstracts.

Since neonatal care practice indicators were not a primary outcome and were examined only as explanatory variables for any effect on mortality, we did not search for them independently. We did not employ any filter to limit the search to developing country (resource-limited) settings. However, we included only trials that had been conducted in countries with a low or middle level of human development.11

Quality assessment

The quality of the identified trials was assessed on the basis of the methods used for sampling and for allocation into intervention and control groups.12 Randomization was classified as: (a) adequate, (b) unclear, (c) inadequate and (d) not used; allocation concealment as: (a) adequate, (b) unclear, (c) inadequate and (d) not used.

Data abstraction

Both authors extracted data separately. The data were then compared and any differences were resolved through mutual agreement. When necessary, the original investigators were asked for additional data or clarifications. Data entry and initial analysis were performed on SPSS version 14.0 software (SPSS Inc., Chicago, United States of America).


We performed meta-analysis using Stata® software version 9.2 (StataCorp LP, College Station, USA). The presence of bias in the extracted data was evaluated quasi-statistically using the funnel plot13 and formally with the “metabias” command.14,15 To be able to appropriately combine individual and cluster randomized trials, we made pooled estimates (relative risk, RR, with 95% confidence intervals, CIs) and calculated the heterogeneity of the evaluated outcome measures by the generic inverse variance method using the “metan” command14,16,17. The effect size of the intervention (summary RR) was calculated by comparing mortality rates at the end of each intervention or observation period, since baseline and/or change data were not available for all included trials. For completeness, we analysed both random effects and fixed effects model estimates; however, a random effects model was preferred if substantial heterogeneity was present (I2 > 50%).

The following pre-specified subgroup analyses were performed for all-cause neonatal mortality as a hypothesis generating exercise: (i) random (individual or cluster) versus non-random or quasi-random allocation to examine the effect of trial quality on the RR of death; (ii) preventive interventions versus preventive and curative interventions (e.g. injectable antibiotics for neonatal sepsis) to examine the potential effect of adding curative treatment; (iii) high (> 45 deaths per 1000 live births) versus low (≤ 45 deaths per 1000 live births) baseline neonatal mortality to examine the possibility of a greater benefit in populations with higher baseline mortality; and (iv) proportion of neonates receiving a postnatal visit (< 50% versus ≥ 50%) to assess the effect of intervention coverage.


Trial flow

We identified 60 potentially eligible references, 47 of which were excluded (Fig. 1) for reasons detailed in Table 1 (available at: http://www.who.int/bulletin/volumes/88/9/09-069369). The remaining 13 references, which pertained to 5 trials, were included in the review.1830

Fig. 1. Study selection in systematic review of randomized controlled trials (RCTs) of home-based interventions to reduce neonatal and infant deaths and stillbirths
Fig. 1. <b>Study selection in systematic review of randomized controlled trials (RCTs) of home-based interventions to reduce neonatal and infant deaths and stillbirths</b>
RCTs, randomized controlled trials.

Trial characteristics

Table 2 summarizes the characteristics of included trials, all of which were conducted in southern Asian countries with high baseline neonatal mortality rates (> 45 deaths per 1000 live births). Sylhet18 and Shivgarh20 trials were cluster-randomized and provided cluster-adjusted mortality data. The other three trials, from Hala,19 Gadchiroli21 and Barabanki,30 were non-randomized or quasi-randomized and had a concurrent control group. End-line evaluation provided data on 17 675 and 14 251 live births, and on 746 and 779 neonatal deaths in the intervention and control arms, respectively.

Intervention Package

Table 2 describes the training received by the health-care workers who delivered each intervention package. Table 3 summarizes the intervention packages used in the trials.

Quantitative data synthesis

Five trials provided neonatal mortality data1821,30 and three provided data on stillbirths.1921 One trial provided infant mortality data and cause-specific mortality data.21

Neonatal mortality

All five trials provided neonatal mortality data.1821,30 The funnel plot appeared symmetrical, which suggests the absence of publication bias. This was confirmed using Egger’s method (P = 0. 974). There was evidence of a reduced risk of death during the neonatal period in association with home-based neonatal care; the pooled relative risk was 0.62 (95% CI: 0.44–0.87; I2 = 86.4%; P = 0.000) in the random effects model (Fig. 2).

Fig. 2. Forest plot (random effects model) for relative risk of neonatal death in trials of home-based interventions to reduce neonatal and infant deaths and stillbirths, as identified through systematic review
Fig. 2. <b>Forest plot (random effects model) for relative risk of neonatal death in trials of home-based interventions to reduce neonatal and infant deaths and stillbirths, as identified through systematic review</b>
CI, confidence interval.

On performing pre-specified subgroup analyses we found evidence of significant heterogeneity among subgroups with respect to randomization and coverage (Table 4). Subgroup analyses for baseline neonatal mortality were not feasible because all trials were classified as having high mortality. Trials with adequate randomization (RR: 0.54; 95% CI: 0.39–0.75), showed a greater reduction in neonatal mortality than non-randomized or quasi-randomized trials (RR: 0.67; 95% CI: 0.40–1.13; heterogeneity P = 0.006). A statistically non-significant trend towards a greater effect on mortality was observed with both curative (injectable antibiotics) and preventive interventions (RR: 0.51; 95% CI: 0.30–0.85), as compared to only preventive intervention (RR: 0.70; 95% CI: 0.44–1.12; heterogeneity P = 0.088). Higher (≥ 50%) coverage with home-based neonatal care was associated with better survival (RR: 0.54; 95% CI: 0.42–0.70) than lower (< 50%) coverage (RR: 1.06; 95% CI: 0.81–1.38; heterogeneity P < 0.001).

On performing univariate meta-regression analyses, none of these variables emerged as a significant predictor of heterogeneity (results not shown).

Infant mortality

Data on infant mortality were available from only one trial,21 and it showed a significant decline (RR: 0.41; 95% CI: 0.30–0.57).

Cause-specific mortality

Only one trial21 presented cause-specific mortality data for neonates. The reported reduction in neonatal cause-specific mortality due to sepsis, asphyxia, prematurity and hypothermia was 89.8% (95% CI: 78.6–101.0), 53.3% (23.8–82.8), 38% (4.3–71.6) and 100% (one-sided 95% CI not stated), respectively.

Stillbirth rate

Data was pooled from 3 trials.1921 There was evidence of a reduced risk of stillbirth; the pooled RR was 0.76 (95% CI: 0.65–0.89; I2 = 0%; P = 0.766) in random and fixed effects models.

Care practice indicators

Antenatal and neonatal practice indicators improved significantly (> 1 antenatal checkup, 2 maternal doses of tetanus toxoid, clean umbilical cord care, early breastfeeding and delayed bathing) (Table 5).


This systematic review of controlled trials, of which 5 satisfied the inclusion criteria, indicates that home visits for neonatal care by community health workers are associated with reduced neonatal mortality in resource-limited settings with poorly accessible health-facility-based care when conducted along with community mobilization activities. Data from three trials showed a reduction in the stillbirth rate. Only one trial showed evidence of reduced infant mortality and neonatal cause-specific mortality (from sepsis, asphyxia, prematurity and hypothermia). While on meta-regression no variable emerged as a significant predictor of an effect on neonatal mortality; subgroup analyses suggested that the survival benefit is higher as intervention coverage increases and possibly when curative care (injectable antibiotics for neonatal sepsis) is provided in addition to preventive or promotive interventions.

Strengths and limitations

In this up-to-date systematic review that incorporated relevant subgroup and meta-regression analyses, no evidence of publication bias was found. With the sole exception of the Gadchiroli trial,2129 in which the intervention and control groups had only one cluster each, all cluster- and individual-randomized trials were appropriately combined by correcting for a design effect on mortality outcomes. Both random and fixed effects models were used for pooling the data, and the results were invariably synchronous.

The review also had several limitations. First, data on stillbirths were limited to three trials, while only one trial had investigated infant mortality and cause-specific mortality. Second, all trials were conducted in parts of southern Asia with high baseline neonatal mortality rates (> 45 deaths per 1000 live births),31 which impedes generalization to other regions, particularly to sub-Saharan Africa or to areas with lower neonatal mortality. Finally, the subgroup and meta-regression analyses showed discordance, perhaps because some subgroup results could have been falsely positive or because the number of trials may have been too small. Any significant predictor identified should therefore only be considered as exploratory.

We excluded trials that exclusively evaluated the effect of home-based follow-up of infants born in and recruited from hospitals because they were not central to framing policy on home-based neonatal care in settings with poor access to health facilities. Nevertheless, the conclusion regarding reduced mortality remained stable even after we included two such trials32,33 from developing countries (Zambia32 and the Syrian Arab Republic33). Upon assuming that all deaths in these two trials occurred in the neonatal period, the pooled RR of neonatal death in 7 trials was 0.64 (95% CI: 0.46–0.90; I2 = 81.8%; P < 0.001) in a random effects model.

We depicted both random-effects and fixed-effects model estimates for completeness; however, we preferred a random-effects model because substantial heterogeneity (I2 > 50%) was observed for neonatal mortality. Nevertheless, inferences regarding neonatal mortality and stillbirths remained stable irrespective of the model chosen, and this finding in better quality trials is reassuring. However, it may also indicate that effects in programme rather than research settings may be smaller. Subgroup analyses also suggested a greater neonatal survival benefit with higher (≥ 50%) intervention coverage levels, as expected. In the only trial (Barabanki30) with low postnatal intervention coverage (39%), intention to treat analysis did not reveal any reduction in neonatal mortality (RR: 1.06; 95% CI: 0.81 to 1.38). However, neonates who received a postnatal home visit within 28 days of birth had 34% lower neonatal mortality (design effect, unadjusted: 35.7 deaths per 1000 live births; 95% CI: 29.2–42.1) than those who received no postnatal visit (53.8 deaths per 1000 live births; 95% CI: 48.9–58.8).30 From a programmatic perspective it would have been useful to get some insight into the optimal number and timing of neonatal visits, but unfortunately this was not possible from the available data.

In the 5 trials under review, the intervention was delivered as a package comprising three components: home visits during pregnancy (all trials), home visits for neonatal care (all trials) and community mobilization efforts (4 trials). Thus, we were unable to differentiate the independent effects of the three intervention components on neonatal mortality. Other trials from similar settings, some of which are listed in Table 1, suggest that community mobilization alone, without home-based neonatal care, improves neonatal health outcomes, including survival.3440 However, in the only direct comparison of the two approaches,18 neonatal mortality was reduced in the home-based care arm (RR: 0.66; 95% CI: 0.47–0.93) but not in the community-mobilization arm (RR: 0.95; 95% CI: 0.69–1.31). It was also impossible to differentiate the independent effects of antenatal and postnatal home visits. However, programmatically this is not crucial because in practice antenatal visits are required to establish contact with pregnant women before postnatal visits and health workers can also provide community mobilization services.

The effects on mortality observed in these trials is supported by significant improvements in antenatal and neonatal care practices whose association with reduced mortality has been demonstrated in previous reviews.7

Implications for policy

Home visits for neonatal care by community health workers, when accompanied by community mobilization efforts, are associated with reduced neonatal deaths and stillbirths in settings with high neonatal mortality rates (> 45 deaths per 1000 live births) and poor access to health-facility-based care. This provides evidence in support of adopting a policy of home-based neonatal care provided by community health workers in such settings. High intervention coverage (≥ 50%) is essential for achieving meaningful reductions in neonatal mortality. No concrete recommendations can be formulated from the available evidence regarding the optimal timing of home visits and specific responsibilities of community health workers. It would be prudent to remember that all the evidence pertains to southern Asia; however, there are no obvious reasons to suspect different results in other regions with similar neonatal mortality rates and access to health care.

Implications for future research

The following gaps in the evidence base need to be urgently addressed to guide policy: (i) the effectiveness of the intervention package in high-mortality settings in other regions, particularly sub-Saharan Africa; (ii) the effectiveness of the intervention package in settings with lower neonatal mortality rates (15–29 and 30–45 deaths per 1000 live births31); (iii) the benefit of adding a curative component (especially the treatment of neonatal sepsis) to preventive or promotive neonatal care; (iv) the relative efficacy of home visits of a certain number and timing (e.g. 1 versus 2–3 in the first week of life); and (v) ways to achieve high coverage and an intervention of high quality in programme settings.


We are grateful to Clive Osmond, MRC Epidemiology Resource Centre, Southampton, the United Kingdom, for helping with the statistical analysis in relation to the calculation of cluster-adjusted relative risks.


External: Department of Child and Adolescent Health and Development, World Health Organization, Geneva. Internal: Sitaram Bhartia Institute of Science and Research, New Delhi, India. The funding sources had no involvement in the study or the decision to publish the manuscript. There was no agreement with the funders that could have limited our ability to complete the research as planned, and we had full control of all primary data.

Competing interests:

None declared.