Description of the condition
HIV attacks the immune system and weakens immunological response against infections and disease. It can be passed on through blood, semen, pre-seminal fluid, rectal fluids, vaginal fluids, and breast milk (CDC 2013). The most common ways HIV is passed on are: unprotected (without a condom) vaginal or anal sex with someone living with HIV, sharing infected needles, syringes or other injecting drug equipment, and from an HIV-positive mother (to her child) during pregnancy, childbirth or breastfeeding (CDC 2013).
HIV epidemic still carries a huge burden of morbidity and mortality in a large part of the world, and according to the estimates of the Joint United Nations Programme on HIV/AIDS (UNAIDS), 34 million (31.4 – 35.9 million) people worldwide were living with HIV at the end of 2011 (UNAIDS 2012). In the same year 2.5 million (2.2 – 2.8 million) people were newly infected with HIV and 1.7 million people (1.5 – 1.9 million) died from acquired immunodeficiency syndrome (AIDS)-related causes (UNAIDS 2012). Although, those numbers are still significant, in comparison to 2001 it is a 20% decline in the number of newly infected and a 43% decline in AIDS related deaths (UNAIDS 2002; UNAIDS 2012).
However, not all parts of the world have recorded the same declining trends in HIV epidemics. In 2009, 91% of the world’s total HIV infected population lived in the low- and middle-income countries, also sometimes referred to as developing countries (Shao 2012; UNAIDS 2010; World Bank 2012). The region most affected by HIV epidemic is still sub-Saharan Africa where 4.9% of the adult population is HIV-infected, and the region itself accounts for 69% of people living with HIV globally (UNAIDS 2012). Other predominantly low- to middle-income regions that are most seriously affected by HIV are South and Southeast Asia, Eastern Europe and Central Asia, Latin America and the Caribbean, together accounting for 20.6% of people living with HIV globally (UNAIDS 2012). The highest increase in the number of people newly infected with HIV in the period from 2001 to 2011 was recorded in the Middle East and North Africa (37%), East Asia (18%) and Eastern Europe and Central Asia (7%), all predominantly low- to middle-income regions (UNAIDS 2012).
Even though reasons for variations in epidemic trends within different world regions are complex and vary a lot depending on the context, there are certain common structural factors that significantly influence HIV transmission. Different studies have suggested that HIV epidemic can be strongly influenced by poor socioeconomic conditions, economic and political displacement of communities, certain cultural practices, gender inequalities, high level of intimate partner violence and lack of access to health care, all of which are frequently found in low- and middle-income country settings (Abramsky 2012; Shao 2012; UNAIDS 2012).
Description of the intervention
Key approaches for HIV prevention include: (i) male and female condom use, (ii) testing and counselling for HIV and sexually transmitted infections (STIs), (iii) voluntary medical male circumcision, (iv) anti-retroviral treatment-based prevention, (v) harm reduction for injecting drug users, (vi) elimination of mother-to-child transmission of HIV (eMTCT) (WHO 2013). Different strategies for promoting HIV prevention have been designed to deliver interventions to individuals, groups or communities. Additionaly, structural community interventions have wider scope and aim to increase individual and community control over their well-being to improve quality of life and social justice (Blankenship 2000; Campbell 2014; Charania 2010; King 1999 ; Wallerstein 1992).
There are different types of community-based interventions and in a review of behavioural change interventions in HIV prevention King splits them into six categories that are most often used in public health: (i) social influence and social networks, (ii) outreach programmes, (iii) school-based programmes, (iv) community mobilization and empowerment, (v) social marketing, and (vi) policy level interventions (King 1999). What makes community mobilization and empowerment different from other interventions in the list is the aim to detect a source of health problems in the social, economic and political environments and empower their members to detect health issues of importance to the whole community, and generate a wider social change in order to change a context in which health is shaped (Blankenship 2000; Wariki 2012). Additional value of this approach is that by changing harmful social and cultural norms, addressing policy and economic issues and empowering communities to conduct changes by themselves it endorses sustainability of changes over time (Adimora 2010; King 1999).
On the other hand, individual-level approach in HIV prevention is more diseases-based and focuses mainly on improving an individual’s ability and self-efficacy in protecting oneself from acquiring HIV infection (such as individual counselling and testing service) (Bandura 1986; Kerrigan 2013). Group-level interventions are delivered to more individuals at once, with the aim to influence every member of the particular group (such as computer based HIV education for children of certain age in a defined geographic area) (Klein 2011; Teti 2010). However, both individual-level and group-level interventions do not consider broader social action and group involvement in the development and delivery of the interventions, and that is what differs them significantly from community mobilization and empowerment interventions.
In the review of the use of community mobilization for HIV prevention in the African context Lippman et al. have defined community mobilization using the following six main components: (i) defining shared concern (target of the change) about a health problem; (ii) rising community sensitization and interest in the issue; (iii) defining organizational structure connected to groups or communities that are targeted with the intervention; (iv) promoting leadership inside of a community; (v) organising collective or shared activities for addressing a defined health issue; (vi) and promoting social cohesion among individuals and different groups in communities (Lippman 2013). Similarily, Tedrow et al. used qualitative analysis of ACCEPT project (randomised controlled trial designed to test the efficacy of community mobilization in three African countries and Thailand) to describe main strategies for community mobilization utilization. According to their analysis, the first step should always be gaining stakeholders support and forming community coalitions within local communities and that should lead to increasing community engagement and participation of members of targeted community in rising community awareness of a certain health problem (Tedrow 2012). Finally, the effective strategy for community mobilization should promote participation of community leaders in decision making and should aim to create partnerships with community organizations (Tedrow 2012).
In health care settings these interventions promote collaboration among health workers and communities with the aim to empower community members in social participation around health issues important for that community (Campbell 2014). It is important to mention that community mobilization is often used in different domains of health and there is still no clear consensus about its uniform definition and strategies for its utilization (Tedrow 2012). Other community based behavioural interventions (i.e. outreach programmes, school based education, health promotion, peer education) aim to change risky behaviours at the community or individual level by increasing the knowledge about HIV in population sub-groups such as school children or by using peer education for promoting safe sexual behaviour among commercial sex workers (King 1999). However, these interventions do not advocate behaviour change from within the community throughout the active involvement of community members who design, execute, and evaluate their projects, like community mobilisation and empowerment does (Campbell 2014; Wariki 2012). Therefore, community mobilization and empowerment represents one of the approaches for achieving a desired behavioural change in the community (i.e. condom use or compliance to antiretroviral treatment), but differently from alternative approaches, it is primarily guided from within that same community and with minimal involvement of outsiders (King 1999; Wariki 2012).
Reporting time for the results in community mobilization and empowerment interventions varies a lot depending on the settings in which the study is conducted. For example, in the systematic review of community mobilization and empowerment interventions among female sex workers Kerrigan et al. reported follow-up time in different studies to range from 2.5 to 6 years (Kerrigan 2013). Additionally, in two big studies (IMAGE study and Stepping Stones Project) on the effectiveness of community mobilization authors assessed the impact of the intervention two years after the start of intervention (Abramsky 2012; Kerrigan 2013), and Gutierrez et al. took a 3-year follow-up period in their study (Gutierrez 2010). Due to the complexity of delivering and implementation of community mobilization interventions it is more likely their realisation will take longer than individual or group level interventions. It is also expected that most of the benefits (primarily decrease in HIV and STI incidence or prevalence) will be seen later than in the regular two to three years evaluation periods that are commonly used for similar interventions (Campbell 2014). We plan to include all eligible studies regardless of the length of the follow-up period, but if we find big heterogeneity in reporting we will consider conducting a subgroup analysis (please see Subgroup analysis and investigation of heterogeneity section).
The most important component of every community-based intervention is a community itself. This is even more significant for community mobilization where the whole process of social change and empowerment has to come from inside of the community and it should be led by its members. Freire sees potential in every marginalized group that is united by common sense of exclusion and solidarity to serve as an agent of change, capable of transforming society through their collective action (Campbell 2014; Freire 1972). In practice, in the community mobilization programmes communities are predominantly defined by geographical area, assuming that people living in the same area are affected by the same health problems, and therefore have the motivation to tackle them together as a community (Campbell 2014). Whereas, in reality geographical communities are very diverse and a feeling of belonging to one can be influenced by gender, age, ethnicity, education level, membership in a certain political party and many other factors strongly dependant on the local context (Lippman 2013). As a result different communities in the same area can have different needs and expectations from the same intervention, and therefore different motivations to participate may even undermine community mobilization programmes (Gruber 2005).
For that reason, rather than using a universal definition of community that could be applied to every context, it is more realistic to adjust the definition specifically to the local conditions and context in which the intervention is being delivered (Campbell 2014; Lippman 2013). For the purpose of this review we will broadly define community to be a group of people that are living in the same setting, share common social values, have certain attitudes, responsibilities and interests in common and therefore consider themselves to be part of a collective (Aarts 2014).
Taking into consideration all available definitions, for the purpose of this review we decided to define community mobilization and empowerment as interventions consisting of:
(i) Community preparedness and involvement activities with the aim to gain community acceptance;
(ii) Provision of support to the community to facilitate and promote change in perceptions and knowledge of HIV and HIV-related attitudes and behaviours, and increase awareness of HIV prevention interventions;
(iii) Implementation of local collective activities such as public events, self-help groups, meetings and discussions that empower community members to adopt safer HIV-related behaviours and increase the access to and uptake of HIV prevention services.
How the intervention might work
HIV epidemic is highly dynamic in terms of disease progression, treatment options, strategies of prevention and surveillance systems. It also depends strongly on people’s behaviours that are diverse and depends on individual desires, social and cultural relationships and environmental and economic processes (Fishbein 2000; King 1999).
Behaviours that as a consequence have increased likelihood of acquiring HIV are usually in the domain of intimate (sexual behaviour) or illegal (drug use) behaviours, or like in the example of commercial sex work have both illegal and intimate components (Blankenship 2000). In addition, individuals are sometimes willing to undergo risky behaviours aiming to obtain economic compensation (i.e. more money that sex workers can get for sex without a condom), or have limited control over their behaviours (i.e. women’s inability to negotiate condom use with sex partner due to violence in a relationship or disempowerment of women in society) (Blankenship 2000). Because of these complexities in risky behaviours and due to the strong influence that environment and social norms can have on HIV transmission, only direct regulation (on the individual level) of the risky behaviours is likely to be insufficient for effective long term HIV prevention strategy (Blankenship 2000).
In the Cochrane review on effects of behavioural interventions on reduction of transmission of HIV among sex workers and their clients in low- and middle-income countries Wariki et. al conducted a meta-analysis of studies that were comparing a community empowerment intervention with standard prevention approaches (Wariki 2012). The analysis showed that community empowerment has a significant effect on increase in condom use with regular partners, but no effect in reducing prevalence of STIs (herpes simplex virus type-2 (HSV-2) or syphilis) or on condoms use with clients (Wariki 2012). Among the studies included in this meta-analysis neither one measured HIV prevalence or HIV incidence to assess effectives of community empowerment (Wariki 2012).
Projects such as IMAGE, Stepping Stones and Aastha have used community mobilization and empowerment interventions in order to reduce inter-personal violence, increase self-confidence of women and sex workers in negotiating condom use, improve financial autonomy of women and reduce risky sexual behaviour; all with the goal to reduce the number of new HIV infections among the most vulnerable populations (Abramsky 2012; Gaikwad 2012; Kim 2009; Jewkes 2008). Similarly, researchers in China used community mobilization to increase coverage of voluntary HIV counselling and testing (VCT) service among rural migrants in Shanghai (Zhang 2013).
In Uganda a community mobilization intervention in Networks model aimed to reduce HIV transmission from mother to child (vertical transmission). That was done by increasing coverage and access to HIV services (i.e. pre-test counselling, home based care, prevention-of-mother-to-child-transmission (PMTCT) services for women and men), supporting linkages of women with healthcare facilities such as child health services, early infant diagnosis and prophylaxis, immunization and infant feeding support) and addressing structural barriers by strengthening sexual health and social protection (i.e. improving PMTCT literacy and reducing stigma) (Mburu 2012).
Different trials have shown the efficacy of early antiretroviral treatment (ART) for HIV positive individuals in preventing HIV transmission to uninfected partners as well as providing ART to high risk HIV negative individuals to prevent acquisition of the virus (Cohen 2011; Grant 2010). Many agree that in order for these treatment-as-prevention approaches to be successful and feasible, broad support of all involved parties will be needed. One of the proposed approaches to attain the success found in these studies in much more demanding real-life conditions is to develop effective community mobilization strategies and empowerment of those living with HIV or having an elevated risk for acquiring HIV, their providers, families and community networks (Lippman 2013). Likewise, in Iringa Region in Tanzania researchers successfully used community mobilization to generate a demand and to increase coverage with voluntary medical male circumcision (Mahler 2011). They used community workers for bringing out pre-procedure tasks like counselling, testing and scheduling for procedures from hospitals into the community. They also formed local, district and regional-level committees composed of public officials, health facility staff, and international and community-based organizations with the task to recruit clients and sensitize community political and administrative leaders (Mahler 2011).
Therefore, we believe that community mobilization and empowerment could have a significant role in the long term HIV prevention strategy because it allows individuals (particularly members of marginalized and underpowered groups) to modify norms of their peer networks by themselves. This may influence changes in behaviours of the whole community and eventually lead to decreased HIV transmission in the population (King 1999; Lewycka 2010).
Why it is important to do this review
Significant differences in HIV incidence or prevalence can be found among different social groups and across different communities. These differences suggest that despite the substantial reduction in individual-level risky behaviours, for successful and sustainable HIV prevention (that is to be achieved at a large scale and for a longer time), it is necessary to apply structural community-level programmes (Blankenship 2000; Blankenship 2010; King 1999; UNAIDS 2012).
Although individual studies have shown significant positive impact of community mobilization and empowerment interventions (such as reduction of incidence of HSV-2 (Jewkes 2008); reduction of risky sexual behaviours (Gaikwad 2012; Zhang 2013); improvement in knowledge about HIV and increased uptake of VCT service (Zhang 2013); increased autonomy and financial stability of women and reduction of inter-personal violence (Abramsky 2012)), a rigorous systematic review with narrative and quantitative synthesis to evaluate effectiveness of such interventions on reduction of HIV incidence and prevalence has not been done. Furthermore, satisfactory progress in applying structural community-level interventions in developing countries is limited because of a lack of rigorous evidence on their effectiveness in preventing HIV (Gupta 2008).
In order to enable programme planners and policy makers to make evidence-based decisions in development of their HIV prevention programmes, it is essential for this systematic review to assess all available community mobilization and empowerment interventions, and wherever possible offer quantitative estimate of their effectiveness in preventing HIV. It is also important to include and assess only high quality studies like Randomised Controlled Trials (RCT) or cluster Randomised Controlled Trials (cRCT) in order to provide clear and unambiguous guidance for policy makers and future researchers in this field.
Criteria for considering studies for this review
Types of studies
We will include RCTs and cRCTs that compare either two or more alternative community mobilization and empowerment interventions, or community mobilization and empowerment interventions with one or more alternative interventions (please see Types of interventions section for details), or with no intervention.
Types of participants
We will include studies done in communities in low and middle income countries, as defined by the World Bank (World Bank 2012), with any of the following types of participants, regardless of their age and gender:
People living with HIV (PLWH) and their partners
People infected with sexually transmitted infections (STI) or with history of STI
Drug users, including injection drug users (IDU) as well as other drug-using populations
Sex workers (SW)
Clients of commercial sex workers
Transgender or transsexual population
Men who have sex with men (MSM)
The Cochrane review by Wariki et al. assessed the effectiveness of community empowerment interventions in low- and middle-income countries targeted exclusively to FSWs and their clients (Wariki 2012). We will assess effectiveness of the community mobilization among female sex workers (FSW) and their clients as well, but we will include additional secondary outcomes that were not covered by Wariki et al. (such as uptake of harm reduction programmes, ART and PMTCT programmes, retention in care 12 month after the intervention, cost of interventions and potential harms due to interventions). The studies included in the meta-analysis done by Wariki et al. that used the concordant outcomes measures to the ones defined in our review will be excluded from our meta-analysis. We have identified studies by Basu 2004, Gutierrez 2010 and Swendeman 2009 to match this criteria as they have used STIs prevalence, condom use and high risk sexual behaviour as their outcome measures, all of which are defined as outcomes in both Wariki et al. and our review (please see Types of outcome measures section). However, we plan to present and discuss results of these studies in conclusion of our review.
Types of interventions
We will include studies that clearly explained the three components (listed in the section Description of the intervention) of community mobilization and empowerment interventions that we defined to be essential.
Alternative intervention is any intervention offered in comparison arm of a study, to either individuals (such as counselling and testing service), groups or communities (such as peer education and needle and syringes distribution for injection drug users) that did not meet inclusion criteria for community mobilization and empowerment interventions listed above.
In cases of multi-armed studies, with one community mobilization and empowerment intervention arm, we will consider all control arms relevant, group them and include pairwise comparison of the community mobilization and empowerment arm with the grouped control arm. In cases of multi-arm studies with more than one community mobilization and empowerment intervention arm, we will make pairwise comparisons between different community mobilization and empowerment intervention arms. When appropriate, we will group different community mobilization and empowerment intervention arms and make pairwise comparisons with single control intervention arms, or grouped control intervention arms, in case there are more than one.
Types of outcome measures
Primary outcomes will not be part of eligibility criteria. We will consider studies addressing one or more secondary outcomes.
1. HIV incidence
2. HIV prevalence
3. STI incidence
4. STI prevalence
1. Condom use (male/female; commercial/regular partners)
2. Uptake of VCT service
3. Uptake of harm reduction programs for drug users (including IDUs as well as other drug-using populations)
4. Uptake of ART
5. Uptake of medical male circumcision
6. Uptake of PMTCT programmes
7. Involvement in high risk sexual behaviour (such as early sexual debut; more than one sexual partner in last 12 months; unprotected sex with commercial sex worker, anal sex without condom)
8. Retention in HIV care measured at least 12 months after the intervention ends
9. Comparative costs of interventions
10. Possible harms due to interventions, if any
We will include primary and secondary outcomes in the Summary of Findings tables.
Timing of outcome assessment
For the purpose of this review we will take into consideration all reported results regardless of the time period between beginning of the intervention implementation and measurement of outcomes. If appropriate, we will use subgroup analysis to compare results from the studies that have used the same follow-up period for measurement of outcomes.
Studies which were not RCTs or cluster RCTs
Studies which were not conducted in low and middle-income countries
Studies of individual level interventions
Studies of community-level interventions other than mobilisation and empowerment such as:
(i) social influence and social networks
(ii) outreach programmes
(iii) school-based programmes
(iv) social marketing
(v) policy level interventions (King 1999)
Search methods for identification of studies
We will aim to identify all relevant RCTs or cRCTs, regardless of language or publication status (published, unpublished, in press or in progress). We will consult Cochrane HIV/AIDS Review Group’s Trial Search Coordinator for developing the search strategy. We will include both peer reviewed studies and grey literature.
We will search the following databases for the period 1 January 1980 to the search date:
• The Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library)
• MEDLINE (Ovid);
• PubMed Central (PMC)
• EMBASE (Ovid);
• ISI Web of Science which includes: Social Sciences Citation Index (SSCI), Conference Proceedings Citation Index Science (CPCI-S), Science Citation Index Expanded (SCI-EXPANDED), Arts & Humanities Citation Index (A&HCI).
Searching other resources
Clinical Trials Registers
We will search the following registers:
We will look through the reference lists of published studies and reviews for references to relevant studies.
We will attempt to find unpublished studies through searching grey literature sources such as Open Grey and Google Scholar.
We will search the following databases:
• Conference on Retroviruses and Opportunistic Infections (CROI)
• International AIDS Society (IAS)
• AEGIS (AIDS Education Global Information System)
• International AIDS Conference on HIV Pathogenesis, Treatment and Prevention
In order to identify unpublished studies relevant for the review we will also consult trial authors of included and excluded trials, other experts in the field, non governmental organisations (NGOs), as well as international agencies such as UNAIDS, World Health Organization (WHO), United Nation Population Fund (UNFPA) World Bank etc.
Data collection and analysis
Selection of studies
Two authors (SH and JB) will independently screen titles and abstracts according to pre-defined inclusion criteria to evaluate eligibility of the studies. In cases of uncertain eligibility based on title and abstract alone, full texts of studies will be retrieved and assessed. Disagreements will be resolved by discussion and consulting a third review author if necessary. We will report kappa scores for each phase of the selection process. We will consider values of kappa between 0.40 and 0.59 to reflect fair agreement, between 0.60 and 0.74 to reflect good agreement and 0.75 or more to reflect excellent agreement (Higgins 2011).
Data extraction and management
We will retrieve full texts of the included studies. Two authors (SH and JB) will independently extract data using a standardized data extraction form which will include the following domains:
Study Citation (including complete citation, study ID that will be used as a reference in the text of the review, corresponding author’s contact details).
Methods (including type of study design, when the study was conducted, whether ethical approval was obtained, data on potential conflict of interests and funding sources).
Setting and participants (including country and city where the study was conducted, categorization of country according to its GNI per capita, whether the study was conducted in urban or rural setting, definition of population from which participants were drawn, recruitment methods, exclusion and inclusion criteria, number of sampled participants, significant baseline imbalances, data on withdrawals and exclusions, stratification of population by age and gender).
Intervention (including detailed description of all components that we used to describe community mobilization and empowerment interventions and HIV prevention that interventions address; description of all control interventions).
Outcomes (including primary and secondary outcomes, timing and methods of outcome assessment, defining risk of biases and using quotes from the text as a proof if available)
Results (all primary and secondary outcomes of review interest at all time points as prespecified in the Methods’ section of our protocol)
We will write to the study authors for clarification or additional data when necessary. Differences in opinion will be discussed and resolved by contacting a third author. One author (SH) will add the data into RevMan and another author (JB) will recheck the data
Assessment of risk of bias in included studies
Two authors will use the Cochrane Collaboration's tool for assessing risk of bias, described in section 8.5 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) and Consolidated Standards of Reporting Trials (CONSORT) 2010 checklist of information when reporting cRCT (Campbell 2012) to independently assess the methodological quality of each included study. The following issues will be taken into account when determining 'low risk of bias', 'high risk of bias', or 'unclear risk of bias':
1. Sequence generation
2. Allocation concealment
Mechanism used to implement the random allocation sequence
Steps taken to conceal the sequence until interventions were assigned
Specification that allocation was based on clusters rather than individuals and whether allocation concealment (if any) was at the cluster level, the individual participant level or both
When appropriate, we will consider risk of bias separately for participants, personnel and outcome assessors. We will report on who was blinded, which methods were used and whether we judged them to be effective.
4. Incomplete outcome data
Completely defined pre-specified primary and secondary outcome measures, including how and when they were assessed
Any changes to trial outcomes after the trial commenced, with reasons
Whether outcome measures pertain to the cluster level, the individual participant level or both
Results at the individual or cluster level as applicable and the estimated effect size and its precision (such as 95% confidence interval) and a coefficient of intra cluster correlation (ICC) for each primary outcome
Completeness of outcome data, including exclusions from the analysis and attrition.
We will consider studies with outcome measures obtained for less than 80 % of randomised participants in any of the trial arms to be at high risk of bias.
We will report on reasons for missing data, when available, and whether appropriate intention-to-treat (ITT) analysis was performed.
5. Selective outcome reporting
6. Other potential biases
We will narratively describe risk of bias across studies. We will compare the assessments, discuss and resolve any disagreements in the grading between the reviewers. We will also write to the study authors for clarification or additional data when necessary.
Measures of treatment effect
For dichotomous outcomes we will present risk ratios (RR) with the 95% confidence interval (CI). In the case of continuous outcomes we will use the mean difference (MD) (with 95% CI) and standardised mean difference (SMD) (with 95% CI) if some of the outcomes are measured on the different scale. We will consider differences in compared outcomes to be significant if P-value calculated using t-test is less than 0.05.
If the trials were examining the same intervention and the same populations and methods were judged to be sufficiently similar we will use fixed-effect model for meta-analysis. Where heterogeneity between trials’ treatment effects will be significantly different we will conduct meta-analysis using random-effect model (Higgins 2011; Wariki 2012).
Unit of analysis issues
Studies with similar units of analysis will be grouped together for the purposes of analysis. We will check if studies with cRCT design accounted for clustering effect. We will preferentially use the intra-cluster correlation coefficient (ICC) provided in the included study. In the absence of an available ICC from the included study, we will use an ICC from an external estimate from similar studies to calculate the design effect. We will also perform a sensitivity analysis including only studies with original ICCs (or design effect) to evaluate the impact of the use of ICCs from external sources has on the summary estimates. In the absence of a reported design effect in an included study, the design effect will be calculated using the study’s ICC (whether provided or estimated from external sources). The design effect for study i will be calculated as follows: 1+ (Mi - 1)ICCi, where M is the average cluster size for studyi, and the ICC is the intra-cluster correlation for study i (Gazi 2011; Higgins 2011; Rao 1992). Lastly, in the meta-analysis, we will approximate the analyses of cRCTs by inflating standard errors to account for clustering. To do this, we will inflate the standard error for each included study by multiplying the standard error by the design effect (inflated sei = sei * design effecti).
Dealing with missing data
We will contact trial authors trying to obtain unreported or missing data. When appropriate we will calculate missing statistics using methods described in the Cochrane Handbook (Higgins 2011). For each included study, and for each outcome or class of outcome, we will assess completeness of the data including whether attrition and exclusions were noted, along with the numbers included in the analysis at each stage (compared with the total number of randomised participants). We will conduct all outcomes analyses on an intention-to-treat approach. We will check if reasons for attrition or exclusions where reported, and whether missing data were balanced across groups or were related to outcomes. Where sufficient information was reported, or was supplied by the trial authors, missing data will be included in the analyses (Higgins 2011; Wariki 2012). We will use sensitivity analysis to measure impact of including trials with high levels of missing data in the overall assessment of the treatment.
Assessment of heterogeneity
We expect high degree of heterogeneity among studies such as different approaches for utilisation of community mobilisation, difference in target population, preventive approach and variations in types of outcome measures (Wariki 2012). We also expect differences in socioeconomic status among communities involved in studies, variations in study settings (rural versus urban), sample size and cultural and health service environment (Gazi 2011).
We will examine whether the intervention, population and methods between different studies were sufficiently similar to conduct meta-analysis by using fixed-effect model. If we notice high level of heterogeneity among studies we will assess heterogeneity qualitatively (based on study population context) or by using box plots (displaying medians and ranges). We will also calculate I2 test according to the mentioned variables and if substantial heterogeneity (I2 >50%) will be identified we will conduct subgroup analysis. Where heterogeneity between trials according to all mentioned characteristics will be significant we will conduct meta-analysis using random-effect model (Higgins 2011, Wariki 2012). We will use stratifying analysis for exploring reasons of potential differences in final results of meta-analysis after using fixed and random-effect model.
Assessment of reporting biases
We will test publication bias by the use of a funnel plot and Egger's Regression test if appropriate and when adequate data are available (Egger 1997).
After grouping the studies as described in the Assessment of heterogeneity section, we will determine level of heterogeneity between different studies. We will examine whether the intervention, population and methods between different studies were sufficiently similar to conduct fixed-effect inverse variance meta-analysis for combining the data. Where heterogeneity between trials’ treatment effects will be significantly different we will use random-effect model for meta-analysis (Higgins 2011, Wariki 2012). If appropriate we will use stratifying analysis for exploring reasons of potential differences in results after using fixed and random-effect models. In cases of multi-arm studies, we will include pairwise comparisons as described in Types of interventions section.
We will use the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to create Summary of Findings (SoF) tables when appropriate. In case meta-analysis is not possible we will use qualitative methods to perform narrative synthesis of results.
Subgroup analysis and investigation of heterogeneity
If possible, stratified analysis will be undertaken among participant subgroups. Subgroupings will include characteristics of community mobilization, types of HIV prevention and types of target population. We will include only HIV and STI incidence and prevalence as an outcome measures in this analysis.
Due to expected differences in the time period from the start of the intervention to the measurement of the outcomes, for the purpose of this review we will take into consideration all reported results, but if appropriate we will use subgroup analysis to compare results from the studies that have used the same follow-up period for measuring effect of interventions.
Additionally, we will divide studies according to methodological quality and risk of bias (high versus moderate/low risk of bias).
If enough studies will be retrieved we will conduct subgroup analysis to distinguish the effect of community mobilization separately from empowerment interventions.
In order to reduce the amount of heterogeneity potentially introduced by the type and different definitions of the communities in different studies we will perform a subgroup analysis that stratifies the included studies by their definition of community (i.e. group 1= schools/learning institutes, group 2=brothels/sex venues, group 3=villages/cities, group 4=prisons/jails, group 5=health facilities/hospitals).
We intend to undertake sensitivity analyses to determine the effects of excluding the poorer quality trials, those with a moderate or high risk of bias as defined in the Cochrane Handbook of Systematic Reviews of Interventions (Higgins 2011). We also plan to perform a sensitivity analysis including only studies with original ICCs (or design effect) to evaluate the impact of the use of ICCs from external sources has on the summary estimates.
We thank Tara Horwath and HIV/AIDS Cochrane group for help in the development of search strategy for this protocol.
Contributions of authors
SH was the contact person with the editorial base, coordinated contributions from the co-authors, wrote the background section and final draft of the protocol. JB and IB drafted the objectives section and methods section. AA provided expertise and supervision in section of data collection and data analysis. IB provided content expertise and supervision.