PROTOCOL: Family and Community Interventions under Integrated Management of Childhood Illness Strategy for Reduction of Neonatal and Under‐five Mortality among Children in Low‐And‐Middle‐Income Countries: A Systematic Review

Abstract

Millennium Development Goal 4, as defined by the United Nations in 2000, aims to reduce the 1990 under-five mortality rate (U5MR) by two thirds in the year 2015. A reduction U5MR will reflect the impact of child survival interventions and global policies aimed at international development (Mosley 2003). Nearly 98% of the 7.6 million child deaths globally each year occur in Low-and-Middle-Income countries (LMICs) and 64% of them are from preventable infectious causes namely sepsis, pneumonia, diarrhea and Malaria (Liu 2012). About 40% of the child deaths occur among neonates (0-28 days), from preventable causes such as preterm birth and intrapartum complications and sepsis (Liu 2012 and Lawn, 2012). Though U5MR has decreased from 9.6 million in 2000 (Black 2003) to 7.6 million in 2010, reduction of U5MR could still improve in many developing countries where U5MR remains high (Liu 2012). A slower decline of U5MR has been attributed to stagnant neonatal mortality rate (NMR) (Save the Children 2013) and prevailing high childhood morbidity rates (Murray 2007). Annual decline of NMR during 2000-2010 was slower than annual mortality decline among children 1-59 months when compared to 1990-2000 (Lawn 2012). Hence, child survival strategies should focus on infectious and neonatal causes of mortality (Liu 2012) and direct the resources towards vulnerable socio-economic groups for improved neonatal survival (Lawn 2012). Available evidence suggests that interventions such as exclusive breastfeeding, clean delivery and skilled attendance at delivery, tetanus toxoid immunization to pregnant mothers, newborn resuscitation, and appropriate management of infections can prevent most neonatal deaths (Darmstadt 2005). In addition, prompt and appropriate management of acute diarrheal diseases (ADD), acute respiratory infections (ARI), and malaria and childhood under-nutrition can also prevent child mortality (Jones 2003). Research has shown that comprehensive training of health care providers (HCPs) on management of childhood morbidities at health facilities and in outreach services, can prevent a significant proportion of childhood mortality (Gove 1997). Studies have reported that available child health services are under-utilized, care-seeking behavior during acute illnesses is neither prompt (timely) nor appropriate (Sreeramareddy, 2006a; Sreeramareddy, 2012) which could be a reason for a slower progress towards achieving MDG-4 in many LMICs. Stagnation in neonatal mortality rate has been attributed to unsafe delivery and newborn care practices particularly in underserved rural communities and urban slums (Osrin, 2002; Sreeramareddy, 2006b; and Seward, 2012). Counseling the parents/care takers about symptoms and danger signs of child illness may improve the care-seeking behavior during the episodes of childhood illness. Therefore, behavior change communication aimed at improving utilization of child health services, seeking prompt and appropriate treatment during illness, improving perinatal and child care practices assume importance in LMICs where U5MR remains high. Previously child survival interventions were usually implemented as separate disease specific programs for common causes of child mortality. These interventions and others such as exclusive breastfeeding, vaccination and oral rehydration therapy (ORT), are effective, children attending primary care settings are often known to have multiple and overlapping morbidities which need an integrated approach to treat them adequately (Bryce 2005b). The World Health Organization (WHO) and United Nations Children Fund (UNICEF) jointly with other technical partners developed the Integrated Management of Childhood Illness (IMCI) strategy to reduce U5MR in 1997 (Tulloch 1999). IMCI strategy aims to address the limitations encountered in disease specific child health programs by integrating treatment of common childhood morbidities that cause child mortality. Since its conception and after piloting the strategy in a few countries, more than 70 countries have fully implemented IMCI strategy (Lambrechts 1999). Three components of the IMCI strategy are:1) improving case management skills of healthcare providers (HCPs) particularly in the outpatient facilities (health worker interventions), 2) strengthening health systems (health system interventions) and 3) improving family and community health practices (WHO 2003; WHO 2005) (community interventions). Training component of Integrated Management of Childhood Illness (IMCI) is expected to improve case management skills of healthcare providers, thus reducing mortality from acute respiratory infections, acute diarrheal diseases and malaria. The health system interventions would improve the infrastructure of health facilities i.e. health staff, job aids, and treatment facilities (drugs) for treatment of childhood illnesses. While the health worker training and health system interventions together would improve the quality of care provided at health facilities; the family and community interventions may improve family's care seeking behavior and community childcare practices. Improved quality of care as a result of training and health system interventions together with community interventions may lead to increase in utilization of child health services. All the three IMCI interventions are expected to act synergistically towards reducing child mortality (Ahmed et al. 2010; Lulseged 2002). A causal chain (figure 1) shows various family and community interventions of IMCI strategy and ensuing programmatic pathways of intermediate and final health outcomes. Family and community interventions are expected to improve care seeking behaviors of family members/caregivers for common childhood illnesses (choosing appropriate care and seeking prompt care (within 24 hours), community practices regarding perinatal and newborn care, caregiver's compliance to HCP's treatment and advice. Family interventions are also expected increase uptake of preventive services, improve child feeding practices also. In addition to these expected change, social mobilization through involvement of women's groups, community leaders will raise the general awareness and demand for improving household and community practices with regards to child health. All these improvements will increase the utilization of preventive and curative services and overall improvements in health and nutritional status of under-five children, thus decreasing child mortality (Bryce, Victora, Habicht, Black, & Scherpbier 2005b). Evidence about impact of Integrated Management of Childhood Illness (IMCI) is available from community trials, WHO multi-country evaluation report and a few systematic reviews. Primary studies which have assessed the impact of IMCI strategy on health care providers' performance have demonstrated that IMCI strategy can improve the quality of care at the health facilities (Amaral et al. 2004; Amaral et al. 2005; Amaral et al. 2008; Arifeen et al. 2005; Bryce et al. 2005a; Huicho et al. 2005). Primary studies which have assessed the impact of health care provider training alone as an intervention to strengthen IMCI program have shown a variable effect on outcomes of childhood illness (Ahmed, Mitchell, & Hedt 2010; Goga & Muhe 2011; Lulseged 2002). However, evidence from primary studies about the impact of one or more of IMCI interventions on child mortality is inconclusive (Amaral, Leite, Cunha, &Victora 2005; Arifeen et al. 2009; Bhandari et al. 2012; Huicho, Davila, Gonzales, Drasbek, Bryce, &Victora 2005). A report about Multi-Country Evaluation on Effectiveness of Integrated Management of Childhood Illness (IMCI), Cost and Impact (MCE-IMCI) suggested that IMCI delivery systems should be expanded to include other potential channels such as private health sector and adopt stronger community-based approaches rather than being just health facility-based (Bryce, Victora, Habicht, Black, & Scherpbier 2005b). A mixed methods multi-country survey has highlighted that financial shortages, logistic difficulties in follow-up after training in IMCI case management training and human resource shortages (in terms of number and/or skills) as challenges to IMCI scale-up (Goga et al. 2009; Goga and Muhe 2011). To date, we found three systematic reviews on this topic by searching in Index Medicus, Scopus and Cochrane library. A narrative review that focused on parents/caregiver's role in IMCI strategy has underscored that ICMI trained workers lacked communication skills about counseling the parents/caregivers about danger signs and homecare of sick child (Paranhos et al. 2011). A systematic review comparing the effect of standard 11-days training versus shortened 5-10 days training on quality of care has concluded that standard (11-days) IMCI training was marginally better and underscored the need for implementing other IMCI interventions irrespective of the training duration. This review also found that for a third of sick children, the HCPs did not adhere to IMCI guidelines (Rowe et al. 2012). A Cochrane systematic review protocol aims to synthesize the evidence on effect of integration of individual IMCI components as a healthcare package on both HCPs and its beneficiaries (Gera et. al. 2012). However, a Cochrane review of both experimental and quasi-experimental studies has shown that there is very little evidence that integrating primary health care services at the point-of-delivery may improve the utilization and outputs of healthcare delivery (Dudley & Garner, 2011). However, among the published reviews, one review did not perform meta-analysis but only summarised the results on impact of IMCI interventions on mortality rates and utilization rates of child health care services (Rowe et al. 2012) while another review qualitatively synthesized the caregiver's (mothers and/or family members) child health care practices (Paranhos et al. 2011). A published Cochrane systematic review protocol aims to synthesize impact of various combinations of three IMCI interventions on mortality, quality of care and key IMCI indicators. Though family and community interventions of our review overlap with interventions to be included in a review by Gera et al. we aim to focus only on various components of family and community interventions (see section 3.1.2) and assess entire intermediate outcomes related to proximate determinants to childhood mortality. Moreover, none of the above reviews have focussed on the impact of IMCI interventions on community child health care practices as underscored by another review (Paranhos et al. 2011) or focussed the review on the impact of standalone community interventions on child mortality rates plus childcare practices of the family and community. Moreover, published reviews have not compared the impact of supply side interventions (health worker and health system interventions) with demand side interventions (community interventions). In addition, from the existing reviews evidence is lacking about the impact of family and community interventions on compliance to homecare of a sick child, care seeking behavior and preventive practices (Thompson and Harutyunyan 2009). Hence, synthesizing the evidence about impact of family and community interventions of IMCI strategy on its beneficiaries is timely and policy relevant. We aim to report the effect of household and community interventions inclusive of social mobilization on under-five mortality and other intermediate outcomes such as care seeking and community child care practices. We also aim to compare the outcomes demand side interventions (community) with outcome of supply side interventions (health worker plus health system). If the studies/trials were randomized at the level of villages/wards (i.e. cluster randomized trial) then all participants (as listed above) living in the clusters studied will be eligible. All the populations to be included will be residing in the low-and-middle-income countries (LMICs) as defined by World Bank (The World Bank Group, 2011). We will include all potential studies which have examined the effect of health worker training together with household and community interventions or family and community intervention only or health worker training only. We will also include the studies which have examined the effect of all the three IMCI interventions. Comparison groups in controlled study designs will be those individuals and clusters or communities where routine mother and child health services were ongoing without implementation of IMCI household and community interventions. Studies which have adopted the following designs and are addressing the household and community interventions as described in UNICEF's IMCI document will be included: We will include either individual or cluster RCTs where in unit of randomization is at individual or cluster (village or district) level and in each study comparison group will differ only in their exposure to the intervention i.e. baseline characteristics and potential confounding factors are adjusted for. qRCT or NRT the trials should have a concurrent comparison group, (for example no IMCI intervention) and groups adjusted for baseline characteristic and investigator allocates into groups by non-random methods (in qRCT allocation into groups is done by pseudo-random sequence). In CBA studies allocation into the different comparison groups are not made by the investigators and outcomes measurement is done in both intervention and control groups before the intervention is introduced and once again after the intervention has been introduced. RDD studies are pretest-posttest design which the investigators demonstrate interventions effects by assigning an intervention using a cut-off i.e. above or below a threshold In ITS studies, investigators obtain data on outcomes from the same population several times before and after intervention. We will include cRCT, qRCT, NRT and CBA studies, if these studies have studied at least two intervention sites and two control sites. We will include ITS studies, if the time point of intervention was clearly defined and data on outcomes is available from at least three time points before and after the intervention. Considering that risk of bias and measures of effect to be different according to types of studies we will synthesize the results separately for RCTs and non-randomized trials (qRCT, NRT, CBA, RDD and ITS studies) All mortality rates will be calculated as number of deaths per 1000 live births. (Any of the outcome measures listed below and/or any other outcomes of interest that fit into the categories listed below if identified during the search for studies and/or data extraction will be included and be given an operational definition) Newborn care practices such as (% of newborns) Child nutrition and feeding practices such as (% of children) Practices related to utilization of child health services and care during illness of child (ex: acute respiratory infections, diarrhea and Malaria) such as Outcomes measuring the efficacy of community mobilization program (Indicators that measure if IMCI strategy has been successful in educating community members/caregivers about childcare and healthcare seeking behavior) We will perform a comprehensive electronic search for primary studies in both ‘mainstream’ databases and ‘specialist databases'. To minimize the publication bias, we will search all available databases and the titles of the journals covering the fields of child health, public health, and global/international health. To avoid language bias and publication bias, we will perform the searches without any language and publication status restrictions. The following ‘mainstream’ electronic databases will be searched for primary studies. PubMed EMBASE (Athens) Popline CINAHL (Cumulative Index to Nursing and Allied Health Literature) Ovid Index Copernicus CAB-Direct (Global Health) LILACS (Latin American and Caribbean health sciences) Web of Science SCOPUS Science Citation Index Expanded Social Sciences Citation Index Sosiological Abstracts World Health Organization Library Information System (WHOLIS) African Healthline (bibliographic databases on African health issues) African Index Medicus IndMed Western Pacific Region Index Medicus Index Medicus for South-East Asia Region Australasian Medical Index International Bibliography in Social Sciences (Athens) The Campbell Library The Cochrane Central Register of Controlled Trials The database of Abstracts of reviews of Effectiveness FRANCIS - bibliographic database in social sciences BDSP (Banque de Donnees en Sante Publique) – French database on public health literature MEDCARIB - Caribbean health sciences literature JSTOR, Wiley Inter-science, Science Direct HINARI (Health InterNetwork Access to Research Initiative) We will develop the search strategy in MEDLINE as detailed in the annexure-5. The MEDLINE search strategy will be translated into other databases using appropriate vocabulary. Some suitable primary studies may have been published as monographs and reports or as research articles in journals that are not indexed in electronic databases listed above. So we will search the following ‘specialist’ electronic databases and non-governmental organizations' websites as further sources for primary studies. WHO/CAH reports - World Health Organization's, Department of Child And Adolescent health (www.who.int/child_adolescent_health/en/) HCPP (Health Care Provider Performance) database UNICEF (United Nations Children's Fund) JOLIS Library Catalog - World Bank & IMF Libraries of the World Bank British Library for Development Studies (BLDS) - a database about economic and social issues in developing countries at Institute for development studies ID21 – a database reporting the UK-based international development research Database on Prince Leopold Institute of Tropical Medicine, Antwerp, Belgium WHO Regional Index Medicus for Eastern Mediterranean Region (EMRO) PAHO - PAHO HQ Library Catalog WHO Global Health Library DFID (Department for International Development) ELDIS (Electronic Development and Environment Information System) OpenGrey Partnership for Maternal, Newborn and Child Health Global Strategy for Women's and Child Health Healthy Newborn Network Concern International Hellen Keller International World Vision International Concern Worldwide Save the Children We will also use ‘Google’ and ‘Google Scholar’ search engines with one more of the following search terms; ‘IMCI’, ‘care-givers' counseling’, ‘community mobilization’, ‘social mobilization’, ‘Millennium Development Goal 4’, ‘MDG-4’, ‘under-5 mortality’, ‘child mortality’, ‘infant mortality’, and ‘neonatal mortality’. In Google search we will use cited references feature to identify further studies. We will hand search the journal titles and conference proceedings that have not been hand searched on behalf of the Campbell collaboration. We will refer to Cochrane Collaboration's master list of journals and conference proceedings. The reference lists of retrieved studies will be searched to identify additional studies. Authors of the primary studies and investigators of IMCI evaluations will be contacted to identify any further published or unpublished studies. Experts in the field of health systems research or authors of any other relevant reviews will be contacted to know if they are aware of any relevant studies. For the ongoing trials, we will look into the registered trials at the following trial registries: a) International Clinical Trials Registry Platform (ICTRP), World Health Organization (WHO) http://www.who.int/ictrp/en/ and b) ClinicalTrials.gov, US National Institutes of Health (NIH) http://clinicaltrials.gov/ List of journals for hand searching Lancet Bulletin of World Health Organization British Medical Journal BMC Public Health, BMC Pediatrics, BMC International Health and Human Rights, BMC Health services Research PLoS Medicine PLoS One Global Health Action List of conference proceedings for hand searching International Conference of Pediatrics World Congress of Public Health World Congress of Epidemiology The search results from various sources i.e. electronic databases, hand searching and other sources will be merged into a reference manager file. Duplicate references will be deleted and a list of titles, abstracts and full text articles will be generated. Eligibility of the retrieved studies will be assessed by a two-stage process by two review authors (CTS and TNS). At first, two review authors, CTS and TNS will independently screen the lists of titles, abstracts and full text articles by applying pre-defined screening criteria (appendices 1 and 2). At this stage, if any titles/abstracts are deemed not eligible will be excluded. If uncertain in the first stage, we will reassess by reading the full text at the second stage. If any disagreements have to arise during study selection, they will be resolved either by discussion or consulting a third review author (HNHK). If insufficient or ambiguous information is present in the studies, we will contact primary authors for further information or seeking clarification. All these processes will be guided and entered into flow diagram provided in the Review Manager 5.1. At least two independent reviewers (among CTS, TNS, and HNHK) will extract the data using pre-designed data extraction form. Data extraction form will be piloted on a few studies and revised if necessary (see Appendix 1). Extracted data will be stored electronically in RevMan or Stata software. We will ensure inter coder-reliability by following methods: pilot testing the coding form, 2) creating a detailed codebook and 3) training about use of codebook and coding forms. The following information about details of the included studies will be recorded; 1) characteristics of the study, 2) participants, 3) comparison groups, 4) outcomes, 5) statistical analysis (measure of effect size reported) and 5) conclusions (for details see appendix 3). Data to be extracted for the possible meta-analysis includes the number of withdrawals, exclusions and loss to follow-up, and number included in the final analysis, baseline characteristics, types of statistical analysis used, specific measures of effect size reported i.e. both unadjusted and adjusted for overall sample and sub groups. We will also extract data about frequency of dichotomous outcomes, mean and standard deviation for the continuous outcomes. Data about relative risk as a measure of effect size for RCTs NRTs, and CBA studies. If the same data has been published as more than one publication, the study with the most complete results will be included into meta-analysis. Disagreements will be resolved by consulting a third reviewer or an independent reviewer having expertise in content and methods. Any disagreements that cannot be resolved would be addressed by contacting the study authors. Finally, if all these attempts were unsuccessful, the disagreement would be reported in the review. Two review authors (CTS and TNS) will independently assess the risk of bias for each included study. Disagreements will be resolved by a third reviewer who has methodological and statistical expertise (HNHK). Assessment of quality of the studies will be done by keeping Preferred Reporting Items for Systematic Reviews (Moher 2009) guidelines as reference and using the risk of bias model in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008) for methodological quality of RCTs. For non-randomized studies, we will pay particularly attention to selection bias, such as baseline differences between the groups, and the potential for selective outcome reporting (Higgins 2008). The risk of bias assessment will be based on five dimensions (described below). The questions about assessment of risk of bias will be piloted and modified (appendix 3) and risk of bias will be rated as low risk, high risk, and uncertain risk of bias. Assessment of risk of bias will inform the sensitivity analysis to be conducted (see 5.6.2 Sensitivity analysis). We will report the assessment of risk of bias for each included study. Selection bias is systematic baseline differences between the groups (i.e. observable factors that are not adequately accounted for) which may compromise comparability between the groups. Performance bias refers to systematic bias and confounding related to intervention fidelity and/or exposure to factors other than the interventions and comparisons of interest that may confound the outcome measured. Blinding of participants and intervention delivery is not applicable for IMCI strategy due to the nature of this intervention. Detection bias deals with systematic differences between groups in relation to how outcomes are determined, including blinding of outcome assessors. Participants who do not undergo IMCI family and community interventions before the end of the study should be censored from the outcome data and if not adequately accounted for may have the potential for introducing bias. Therefore, censoring of the participants is a potential source of detection and attrition bias (see below). Attrition bias deals about completeness of the sample and follow up data. This bias refers to systematic differences between the participants who did not (drops outs, migrated etc) complete the follow-up and those who completed the follow-up. Reporting bias refers to both publication bias (see 5.5.3 assessment of publication bias) and selective reporting of outcomes data and results. We will examine for other potential sources of bias after extraction of data about study designs and the statistical analyses in the included studies. The focus will be to assess if the study authors have reported other potential sources of bias and if they have dealt with these biases adequately. Dichotomous outcomes will be analyzed using relative risk (RR) ratio and its 95% confidence intervals (CIs). For continuous data, we will report mean and standard deviation for the outcome measures. If continuous data for the outcomes were measured on a same scale between trials, we will use mean difference (MD) with 95%CIs otherwise we will calculate standardized mean difference (SMD) and its 95% CIs. If the means, standard deviations, and/or effect sizes are not available, we will use chi-squared values and correlation coefficients. For rare outcome binary or continuous data (ex: prelacteal feeds, not seeking any healthcare), we will use Peto OR and Mann Whitney test. Hedges' ‘g’ will be used to correct for small sample size. For non-randomized studies, adjusted effect sizes will be calculated to account for the influence of confounding factors by propensity score matching. Exact criteria for selection of effect sizes will be a developed ad hoc during the coding phase The included studies are cRCTs or CBA studies, we will use the reported cluster adjusted risk ratios and their 95% CIs. If unadjusted for clustering, we will use intracluster correlation coefficient (ICC), if available or else we will impute ICC from any other included study. If multiple interventions groups are included in the trial, community interventions paired with training and health system intervention will be compared with the control group. If the included studies report variable and multiple time points of outcomes assessments, we will only focus on comparison of baseline and final endpoint even though it is known that effects of interventions may diminish over time. We will attempt to contact the authors to obtain the missing data and any other information irrespective of whether, it can assumed to be ‘missing at random’. If it will not be possible to obtain the missing data, we will impute the missing values, and all the assumptions made for any imputation will be recorded. We will either use intention-to-treat analyses, if reported or contact the authors to carry out re-analyses. In the absence of intention-to-treat analyses and fair indications that data are not missing at random, we will consider missing data to constitute a risk of bias and report in risk of bias assessment. Imputation made will be for missing aggregate data such as missing SD of for change-from-baseline, from SD for the same outcome from another study, or impute SD using recommended methods (Abrams 2005). If necessary, we will impute the missing data using appropriate statistical methods (White, 2009). In addition to this, we will report attrition rates of >50% in any groups and explore its effect on the outcomes by doing a sensitivity analyses. We will measure heterogeneity using I2 statistic. I2 test measures the statistical heterogeneity across the studies. I2 statistic describes the percentage of total variation across studies due to heterogeneity rather than chance (Higgins, 2002). I2 can be readily calculated from basic results obtained from a typical meta-analysis. I2 is given by 100%×(Q - df)/Q, where Q is Cochran's heterogeneity statistic and df is the degrees of freedom (Higgins, 2002). Negative values of I2 are considered as zero so that I2 lies between 0% and 100%. (Higgins, 2002). We will also calculate Cochran's Q-test and τ2 statistic. Publication bias will be evaluated using a quasi-statistical method of drawing a Funnel Plot. Funnel plots will be drawn to investigate relationships between effect size and study precision using the ‘trim and fill method’. Meaningful Funnel plots will require an adequate number of studies with a diverse range of sample sizes (Hayashino 2005). Hence, a funnel plot analysis will be drawn if there are at least ten studies with appropriate data. We will also conduct formal statistical tests for funnel plot asymmetry, using Begg's and Egger's methods (Egger 1997). Analysis will be conducted in RevMan 5 and/or Stata as a