Family planning

Two studies were identified whose primary aim was to examine family planning in conflict-affected settings in northern Uganda and Pakistan.20 ,21 Both indicated that when family planning services are available and accessible, women's use of modern family planning increased.20 ,21 In northern Uganda, a case-series study evaluating a mobile health outreach and public health centre strengthening programme recorded an increase in current use of modern family planning between 2007 and 2010 from 7.1% to 22.6% (adjusted OR (AOR) 3.34; 95% CI 2.27 to 4.92).20 Current use of long-acting or permanent methods of family planning also increased from 1.2% to 9.8% over the same period (OR 9.45; 95% CI 3.98 to 22.42). In the second study, female Afghan refugees in Pakistan receiving subsidised healthcare reported more than double the use of modern family planning methods compared to those who did not receive subsidised healthcare (54% vs 25%; p=0.000) and they were six times more likely to discuss the number of children they wanted with their partners (p=0.000).21 Despite these increases, the cross-sectional study design precludes attributing healthcare subsidies to increases in the uptake of family planning services.

Three other publications also reported family planning as secondary health outcomes of interest.19 ,22 ,23 Viswanathan,23 using a case–control study and Mullany,22 using a case-series study observed an increased use in modern contraception following the implementation of community-based health worker interventions in Afghanistan (OR 1.61; 95% CI 1.12 to 2.15) and among internally displaces populations in eastern Burma (23.9% to 45.0%; prevalence rate ratio (PRR) 1.88; 95% CI 1.63 to 2.17). McGinn reported that following a literacy intervention with Sierra Leonean and Liberian camp-based refugees in Guinea, 40% of women started using modern family planning methods after being exposed to an educational programme with a focus on safe motherhood, family planning, STIs, HIV/AIDS and gender-based violence.19

Prevention, treatment and care for STIs, including HIV/AIDS

Four studies were identified that aimed to improve prevention, treatment, and care for STIs. Two were education based24 ,25 and two focused on clinical management.26 ,27

Woodward et al25 assessed the impact of refugee-led peer education on refugees living in Guinea through a cross-sectional evaluation. Refugees exposed to peer education had over double the odds of having heard about HIV (OR 2.19; 95% CI 1.58 to 3.05) compared to unexposed individuals and were more likely to identify both correct and erroneous HIV transmission pathways, including only having one partner (OR 3.24; 95% CI 1.62 to 6.44 (adjusted for gender)), using condoms (OR 1.91; unadjusted 95% CI 1.15 to 3.16), and sharing food with HIV-positive people (OR 1.52; unadjusted 95% CI 1.10 to 2.10). Participants also had 2.5 times greater odds than non-participants of reporting increased HIV avoidant behaviours (72% vs 58%; OR 2.49, 95% CI 1.52 to 4.08 (adjusted for gender and age of sexual debut age)).25

A separate educational intervention targeting military and female commercial sex worker (CSW) communities in Port Loko, Sierra Leone evaluated by a case-series design, found promising trends in improved knowledge, attitude and behavioural changes (all reported p values <0.01).24 The study, however, did not include detailed information on the intervention contents and process. The CSW's knowledge of three or more ways of avoiding HIV rose from 5.0% to 70.3% between baseline and endline over a 2-year period. Knowledge of two or more STI symptoms also rose from 46.8% to 88.1% (p<0.01). Behavioural indicators, such as reported condom use at last sexual intercourse, also rose between baseline and endline surveys from 38% to 68%, although no significance statistics for this indicator were presented.24

Working in Tanzania with Rwandan refugees, Mayaud et al26 documented an early humanitarian response in a high HIV prevalence setting. The intervention used health promoters and mass-distributed education materials, peer educators among bar and brothel workers, condom distribution, an STI intervention including training and supervising health workers, a syndromic approach for STI case management, medicine provision, and antenatal syphilis screening. During the 18-month intervention period, 120 information and education campaigns were organised and monitoring data estimated 230 000 sexually active people were reached and 1.5 million condoms were distributed throughout the camp. More than 11 000 syndromically identified patients were treated in the first year, growing to 18 000 at the end of programme, showing an increase from the 20 cases/week counted at baseline. The study authors noted that sexual behaviours were not necessarily safer as there was an increase in paid and transaction sex, and that during the intervention period, levels of sexual violence increased. The evaluation design did not allow for a causal association to be made as the findings were primarily descriptive and no significance testing was reported.

A fourth study by Reid et al27 provided intervention details and a descriptive comparison of a Médecins Sans Frontières (MSF) programme implemented in 2008 in response to election-related violence in Kenya. During the political unrest, initiation of highly active antiretroviral therapy (HAART) was temporarily halted. An intervention to ensure continuity of care was implemented where patients who required HIV treatment were given personalised clinical and treatment summaries with the names and dosage of medications. These summaries could then be used to ensure continuity of care in other clinics if MSF clinics were unable to open due to insecurity. A comparison of service record data from January to March in 2008 to the same time period in 2007, found more than twice the number of treatment interruptions during the conflict than 1 year before (13.1% compared to 6.4%, no significance testing). However, by February 2008, the difference had shrunk to 0.5% and by March there were an equal percentage of delayed appointments. Lost to follow-up was similarly controlled by March 2008. Among those enduring treatment interruptions, slightly more (36%) did not miss any pills than did (31%). However, no significance tests were reported.

Maternal, newborn and child health, including obstetric care

Nine studies evaluated interventions focused on improving maternal, newborn and child health outcomes. One double-blind randomised control trial evaluated the efficacy of permethrin-treated bed nets compared with untreated nets or no bed net.28 The incidence of malaria and anaemia in pregnant women was assessed. Researchers found that the geometric mean parasite density was significantly lower in women assigned to the treated bed nets (517; 95% CI 286 to 932/mm³) than those who received no or non-treated nets (1096; 95% CI 726 to 1655/mm³; p=0.049, adjusted for gravidity).28 Treated bed nets halved the risk of subsequent anaemia requiring treatment compared to those randomised to receive no net (95% CI 1.2 to 2.4) and had a 1.6-fold risk reduction compared to those using a non-treated net (95% CI 1.1 to 2.4).

Three studies focused on increased community education, involvement, and birth preparedness and planning.19 ,29 ,30 The Reproductive Health Literacy Project in Guinea was carried out with Sierra Leonean and Liberian women living in refugee camps, evaluating the impact of participatory adult literacy education on improving health knowledge and health-seeking behaviour in the camp.19 A retrospective cross-sectional evaluation found that 40% of women had not used a modern family planning method prior to the health literacy course and 24% reported using a condom at their last sexual encounter. In total 92% of women who became pregnant since exposure to the intervention attended at least three of the four recommended antenatal visits. The study also examined changes in ‘boldness’ (understood as empowerment). Before the health literacy course, 32% of women considered themselves ‘more bold’ than other women but after the course 81% did (p=<0.001). Women who identified themselves as ‘more bold’ were more likely to use family planning (51% vs 36%, p<0.01) and a condom at last sex (26% vs 14%, p<0.05) than women who did not consider themselves ‘more bold’.19 However, the study is considerably weakened by the fact that it did not use baseline data. Participants were instead asked after the intervention to describe their knowledge and behaviours before and after the literacy course, which along with other weaknesses in the study's execution makes the conclusion susceptible to numerous biases and limits its ability to attribute a direct impact on changing associated SRH-related outcomes.

Purdin et al used an ecological study design to evaluate a programme with Afghan refugees in Pakistan which aimed to reduce maternal and neonatal mortality by increasing the use of emergency obstetric care through improving community knowledge about skilled delivery and obstetric danger signs. They observed that the maternal mortality ratio dropped from 291 to 102 per 100 000 live births from the first to the fifth year of the programme (95% CI 181 to 400). Neonatal mortality also decreased from 25 to 20.7 per 1000 births from the first to the seventh year, however no CIs were reported.30 However, the study uses routine observational data and so the authors’ attribution of the change in outcomes to the intervention should be treated with considerable caution given the study design cannot demonstrate causation and there is high probability of confounding.

McPherson et al evaluated a birth-preparedness package in Nepal using a case-series design. The intervention was a multilevel community education and healthcare response, seeking to reduce maternal and neonatal mortality and increase use of health services. Comparing endline to baseline surveys among mothers of infants, health behaviours improved—including not putting anything on the umbilical cord, wrapping the baby immediately after birth and delaying bathing (p=0.000). Attending at least one antenatal visit, making financial and transport preparations also improved (p=0.000). Breastfeeding within an hour of birth was not statistically improved (p=0.06). Of the seven chosen indicators, only the use of a skilled attendant at birth remained unchanged (p=0.55).29

Five studies attempted to improve maternal and neonatal health through system strengthening and capacity building.22 ,23 ,31–33 One study using a cohort design from Bo, Sierra Leone, examined if a transportation and referral system could be implemented to reduce delays in accessing emergency obstetric care from a referral hospital.33 In addition to improving the physical structures of the hospital, providing drugs, staff trainings and leading a community sensitisation campaign aimed at early recognition and care-seeking for obstetric emergencies, a transportation and radio communication referral system to bring women from remote areas into the hospital was established. During the intervention period of 16 months, the case fatality rate (CFR) (maternal deaths/obstetric complications) dropped from 20% to 10%, regardless of whether the patient was brought to hospital using the referral and transportation intervention or not. Utilisation of emergency obstetric care doubled, some of which may be attributable to improved service and community recognition of maternal danger. The decreased CFR could also be caused by shortened delays in seeking care. However, because the CFR decreased for all women regardless of how they came to the hospital, caution should be used before assuming causation of the referral system—particularly given the study design.

An initiative in Makeni, Sierra Leone, assessed the impact of numerous hospital-based improvements including hiring a new medical officer with training in obstetrics and a second medical officer who would receive obstetric training.31 Midwives and nurses were also trained and underwent refresher courses in recognising and responding to obstetric complications. An unused operating theatre was renovated, an electricity generator installed, and drugs and supplies were acquired and paid for using a revolving fund. Policies were also implemented which required treatment before payment, and community outreach workers were trained to refer women with complications to the hospital. Maternal CFR dropped from 32% in 1990 to 4% in 1994, while the number of patients increased from 31 in 1990 to 89 in 1994, although no CIs or other statistics were provided. In addition, the case-series design did not allow for the identification of the pathways leading to the drop in CFR.

Orach et al32 compared the cost and coverage of reproductive health services for refugees and host populations in three rural districts in Uganda using a cross-sectional study with an economic evaluation component. They found that per capita health expenditure was 2.7 times higher for refugees (US$13.12 compared to US$4.85) compared to the host population. Interventions for major obstetric emergencies were also more common among refugees than host populations (1.02%; 95% CI 0.79 to 1.25 vs 0.85; 95% CI 0.80 to 0.90; p<0.05). Cost differences were partly explained by the higher qualifications of staff serving refugee populations. Additionally, medicines, health supplies and laboratory testing and equipment were all more readily available and used more in healthcare settings serving refugees than the general population.

Two studies were identified that evaluated the impact of CHWs on reproductive and maternal health services.22 ,23 Using data from the 2006 Afghanistan Health Survey, a population-based survey, Viswanathan et al23 examined if the presence of a CHW increased the use of modern methods of family planning, attendance at antenatal care and skilled attendance at birth, all of which form part of the basic package of health services intended to deliver cost-effective primary healthcare in rural areas. When adjusting for individual-level and community-level factors, the presence of a female CHW was significantly associated with increased use of modern family planning (OR 1.61; 95% CI 1.21 to 2.15), antenatal care (OR 2.71; 95% CI 1.87 to 3.92), and skilled birth attendance (OR 1.75; 95% CI 1.18 to 2.58). The authors found no significant difference between having no CHW in the community and having a male CHW.

Mullany et al22 also found positive health outcomes associated with the implementation of a three-tiered health worker system which included trained birth attendants (TBA) undertaking antenatal care (ANC) and uncomplicated deliveries and health workers who, in addition to the responsibilities of the TBAs, distributed family planning and could administer misoprostol and antibiotics. The third tier consisted of maternal health workers who oversaw complicated deliveries and provided comprehensive emergency obstetric services, including blood transfusions. Compared to baseline, women in the study area were 1.83 times more likely (PRR) to receive ANC (95% CI 1.64 to 2.04) and 1.88 times more likely to use modern family planning (95% CI 1.63 to 2.17). Birth attendance with someone trained in emergency obstetric care increased from 5.1% to 48.7% (PRR 9.55; 95% CI 7.21 to 12.64).

Quality of reporting in studies

Of the 15 studies, 14 were assessed for quality of reporting using the STROBE checklist and the only randomised controlled trial (RCT)15 was assessed using CONSORT. Of the observational studies, three were found to be of low quality,26 ,30 ,31 seven moderate,19 ,20 ,24 ,27 ,29 ,32 ,33 and the remaining four high.21–23 ,25 The RCT was found to be of high quality.28

The STROBE assessment allowed specific analysis of common areas of low-quality reporting. For these studies, three broad themes emerged: confounding, bias and appropriately reporting findings. Confounding factors were, on the whole, not clearly described or discussed. Only two studies attempted a control,21 ,23 which is largely unsurprising in the context of observational studies, but other quality aspects addressed by STROBE were also consistently lacking. Particularly, only nine of the selected papers gave a clear account of outcomes, exposures, effect modifiers and confounders under assessment.21–23 ,25 ,27–29 ,31 ,33 This persisted in the low reporting of statistical methods to address confounding, which was only adequate in seven studies.20 ,21 ,23–25 ,28 ,29 For the specific STROBE criteria, only three papers adequately reported efforts used to control bias.20 ,22 ,24 The majority of studies did not justify sample size, with only 3 of the 15 describing sample size calculations.21 ,22 ,25 An important theme was identified throughout the studies in the way missing data were handled. Two studies reported how missing data were addressed statistically,23 ,27 with three reporting missing data for variables of interest.22 ,23 ,33 Four studies19 ,22 ,27 ,33 adequately reported non-participation at each stage, although none used a flow diagram to represent numbers of participants. While most of the studies19–25 ,27 ,31 ,32 discussed limitations appropriately, one of those that did not also was deemed to overstate the causative attribution of its findings.30 Two others31 ,22 were also deemed to overstate their findings. The RCT28 was scored 28 of 35 applicable points under the CONSORT quality reporting criteria: the main methodological issue was a lack of any detailed reporting of the randomisation process, although this study was considered high quality as compared proportionally to the STROBE results. Overall, there was a pattern of claiming effectiveness and causation where the study design precludes any such inference.