Perinatal health care services for imprisoned pregnant women and associated outcomes: a systematic review

Background Women are an increasing minority of prisoners worldwide, and most are of childbearing age. Prisons offer unique opportunities for improving the pregnancy outcomes of these high-risk women, and no systematic review to date has looked at their care. This systematic review identified studies describing models of perinatal health care for imprisoned women which report maternal and child health and care outcomes. Methods We systematically searched for literature published between 1980 and April 2014. Studies were eligible if they included a group of imprisoned pregnant women, a description of perinatal health care and any maternal or infant health or care outcomes. Two authors independently extracted data. We described relevant outcomes in prisons (including jails) under models of care we termed PRISON, PRISON+ and PRISON++, depending on the care provided. Where outcomes were available on a comparison group of women, we calculated odds ratios with 95 % confidence intervals. Results Eighteen studies were reported, comprising 2001 imprisoned pregnant women. Fifteen were in the US, two in the UK and one in Germany. Nine contained a comparison group of women comprising 849 pregnant women. Study quality was variable and outcome reporting was inconsistent. There was some evidence that women in prisons receiving enhanced prison care, PRISON+, were less likely to have inadequate prenatal care (15.4 % vs 30.7 %, p < 0 · 001), preterm delivery (6.4 % vs 19.0 %, p = 0 · 001) or caesarean delivery (12.9 % vs 26.5 %, p = 0 · 005) compared to women in prisons receiving usual care (PRISON). Women participating in two PRISON++ interventions, that is, interventions which included not only enhanced care in prisons but also coordination of community care on release, demonstrated reductions in long term recidivism rates (summary OR 0 · 37, 95 % CI 0 · 19–0 · 70) compared to pregnant women in the same prisons who did not participate in the intervention. Conclusions Enhanced perinatal care can improve both short and long-term outcomes but there is a lack of data. Properly designed programmes with rigorous evaluation are needed to address the needs of this vulnerable population. The cost to mothers, children and to society of failing to address these important public health issues are likely to be substantial. Trial registration PROSPERO registration: CRD42012002384.


Background
Women are a small but increasing minority of the 10 · 2 million people imprisoned worldwide [1]. There are around 100 000 women in prison in Europe on any 1 day, representing 5 % of the total prison population [2]. In the United States (US) there are nearly 215 000 women in prisons and jails, representing 9 % of the incarcerated population and an absolute increase of 30 % since 2000 [3]. Despite growing numbers, women's minority status means that their specific health care needs and those of their children may be overlooked or remain unmet. A review from the United States found that 38 states had inadequate or no prenatal care in their prisons [4], and a 2008 report from US Department of Justice notes that 46 % of pregnant imprisoned women reported they received no pregnancy care [5]. The World Health Organisation's (WHO) 2003 Moscow declaration recognises prison health as an important public health issue [6], and a 2009 WHO declaration acknowledges that current arrangements for dealing with women offenders often fail to meet their basic and health needs and specifically mentions inadequacies in provision for imprisoned pregnant women [2]. The 2010 United Nations Bangkok rules [7] and the 2015 Standard Minimum rules for the Treatment of Prisoners [8] provide guidance on perinatal care in correctional settings and state that pregnant women should be provided with a healthy environment and the same standards of health care that are available in the community. The provision of adequate perinatal care in prison is also the law in the US under the Eighth Amendment which prohibits "cruel and unusual punishment." As the number of women in prison grows, so does the number of imprisoned pregnant women and mothers; most imprisoned women are of childbearing age and an estimated 6 % are pregnant, although there is no recent or accurate statistic establishing this proportion [9]. However, although imprisoned pregnant women are at high risk of poor perinatal outcomes due to factors such as ethnicity, low levels of education, access to antenatal care, smoking, drinking alcohol and illegal drug habits [10], a review of perinatal health care in prisons found that there was a lack of available data on perinatal health care worldwide [11]. In a synthesis of the limited evidence available in 2005 we found that imprisoned pregnant women had poorer outcomes of pregnancy than the general population but better outcomes than similarly disadvantaged groups of women not imprisoned [12]. The former finding highlights this group as a vulnerable population worthy of further investigation. The latter demonstrates that prisons offer unique opportunities for improving the health care and pregnancy outcomes of a group of high-risk women when they need to be imprisoned, contributing to the health of both mother and child in the short and longer term; particularly given the growing evidence that events during early development, including the foetal period, have a profound impact on one's risk for development of future adult disease [13].
The health care provided to imprisoned pregnant women is of considerable public health importance, and no systematic review to date has looked at this care. This study aimed to identify effective models of care for these women. The specific review objectives were: to describe models of perinatal health care for imprisoned women which exist in the research literature and subsequent maternal and child health and care outcomes; and to examine, where possible, the effectiveness of models of perinatal health care for imprisoned women on subsequent maternal and child health and care outcomes.

Search strategy and selection criteria
We developed a protocol for the systematic review using PRISMA guidelines [14], which was prospectively registered in the PROSPERO database (International Prospective Register of Systematic Reviews), registry number CRD42012002384.
We searched Medline, Embase, PsycINFO, Global Health, CINAHL, The Cochrane Library database, Scopus, Web of Science, Applied Social Sciences Abstracts (ASSIA), Campbell Collaboration (C2-Spectr and C2-RIPE), CareDATA (Social Care Online), Health Management Information Consortium (HMIC), Intute (previously SOSIG) and the National Criminal Justice Reference Service Abstracts to identify relevant articles, searching from 1980 to April 2014. Search terms (Table 1) were identified from database thesauri, and included prisons and jails. The terms were combined within columns using the "or" operator, and between columns using the "and" operator.
Electronic database searches were supplemented with hand searches of the references of selected papers and relevant policy documents. We undertook extensive but targeted grey literature searching including contacting relevant prison health-related networks.
Throughout this paper, the term "imprisoned" is used for simplicity but includes women incarcerated in jails and prisons.

Data extraction
After removal of duplicates, we screened all abstracts and obtained full manuscripts of all possible eligible citations, irrespective of language. EB and EP independently assessed these manuscripts for inclusion using prespecified criteria (Table 2) and then independently extracted data from included studies using a proforma. MK mediated any disagreements related to eligibility, risk of bias, or data. Authors were contacted if further information was required. Data, extracted from papers using a pre-prepared proforma, included language, publication date, study design, setting, study duration and dates, details of participants, care received by participants, control selection, source of outcome measurement, outcomes, results of study, and funding source. The risk of bias was assessed for each paper as part of the data extraction process with a domain-based assessment adapted from Cochrane Database guidelines [15]. Risk of bias for various components of the studies was classified as high, low or unclear.

Data synthesis
We described all relevant outcomes across studies to present a picture of perinatal outcomes in prison under different models of perinatal health care. Using the descriptive data provided in the text, we classified the care in the intervention group of imprisoned women into three levels of care according to the services they received. PRISON described models of perinatal health care that represented usual care for that prison with no attempt having been made to improve perinatal health care or implement any intervention. PRISON+ described models of perinatal health care where some specific effort had been made to improve conditions or care for pregnant prisoners. Women receiving PRISON++ care are provided with alternative accommodation during pregnancy and co-residence with their children after birth, with strong links between these programmes and community services, recognising that the support for women must continue after release from prison. Where possible we examined differences in outcomes between PRISON, PRISON+ and PRISON++ groups.
Where outcome data were available on women or their babies in either of the three intervention groups described above and a comparison group of women receiving an alternative model of care, we calculated odds ratios or weighted mean differences with 95 % confidence intervals using fixed effects models (Mantel-Haenszel) or random effects models if there was evidence of significant heterogeneity between groups (evidenced by the I 2 statistic). Summary measures for odds ratios comparing outcomes in the intervention groups with comparison groups were calculated and presented where appropriate. We categorised comparison groups into disadvantaged controls: those experiencing similar social disadvantage to imprisoned women, through drug use, previous criminal conviction or imprisonment; and population controls: those selected from a general population in whom no such factors were identified. The data were analysed and presented with STATA (version 12). Forest plots were produced for the available perinatal outcomes, stratified by type of comparison group.

Models of perinatal health care in included studies
Seven studies described models of perinatal health care that represented usual care in that prison: PRISON [17, 24-26, 30, 32, 33]. Antenatal care generally involved health personnel intermittently visiting the prison and transfer to nearby hospitals for birth or if complications arose. Prenatal care appointments were provided on site in four studies [17,24,25,32] and it is unclear for the other three [26,27,33]. In one facility in Missouri women were admitted to the prison hospital from the eighth month of pregnancy until one month post-partum [30].
Six studies described models of perinatal health care where some specific effort had been made to improve conditions or care for pregnant prisoners: PRISON+ [20][21][22][27][28][29]. In three of these programmes, doulas/ birth companions supported pregnant prisoners before, during and after birth [22,27,29]. Two programmes provided enhanced prenatal care for all pregnant prisoners including increased nutrition relative to other inmates, vitamins and iron supplements, reduced physical duties, prenatal counselling and education, and transfer in the third trimester to separate accommodation [21,28]. One programme used outreach workers to identify women in jails across the state who were at risk of giving birth to an HIV-positive or substance-exposed infant and to link these women to prenatal care [20].
The five studies in the PRISON++ category refer to two programmes in which women are provided with alternative accommodation during pregnancy, co-residence with their child after birth and are linked to community services [16,18,19,23,31]. Two studies examined outcomes associated with a live-in nursery within a women's prison in Nebraska, US [18,19]. Women were transferred to the nursery 1 to 2 months before birth and their babies were able to stay with them after birth. The programme provided prenatal parenting, infant care and child development education, hands-on training, and coordinated community resources available for the mother during her prison stay and upon her release. The other three studies examined a programme in Michigan, US, named Women and Infants at Risk (WIAR) [16,23,31]. Imprisoned pregnant women with a history of substance abuse were transferred to a residential programme outside prison where Table 4 Risk of bias in eligible studies they were supported through pregnancy, birth and in the postpartum period with prenatal care, educational and therapeutic groups, employment enhancement services and substance abuse education. They stayed with their infants on-site until release into the community, which was facilitated by the programme through housing arrangements, coordination with social services and day care on release. This programme was not strictly in prison but perinatal outcomes in the WIAR group were compared to outcomes of women in prison before the programme existed, and to women in prison who were eligible for, but did not participate in WIAR. Therefore these studies were included because the comparison groups of pregnant women were in prison. When describing outcomes in prison under different models of care, we used only outcomes from women in the WIAR studies who were actually in prison (hence in the PRISON group). In the analysis comparing interventions to a comparison group, the intervention groups are the women in the WIAR programme (PRISON++ intervention) and the comparison groups are the women in the standard prison (PRISON).

Describing outcomes in prison
There was little consistency in the reporting of outcomes: the largest number of studies reporting any one outcome was six. Many outcomes were reported in only one study, and these are not all reported here. Fourteen outcomes were reported in more than one study, enabling us to describe outcomes across studies, and where possible, compare outcomes in PRISON+ to PRISON (Table 5). Sample sizes were often small, limiting our ability to detect statistical differences between groups, particularly for rare outcomes such as stillbirth, neonatal death, small for gestational age and APGAR score.
Five studies, all in the PRISON category, reported low birth weight, with rates ranging from 6 to 17 % [16,24,26,31,32]. Mean birth weight was reported in six studies, three in the PRISON group with values of 3100 g, 3165 g and 3299 g [16,24,25], and three in the PRISON+ group with values of 2495 g, 3153 g and 3299 g [21,28,29]. Only one study reported the associated standard deviation, making it impossible to summarise these figures [24]. Mean gestational age was reported in two studies: 38 · 8 weeks (standard deviation 2.2) in a PRISON study [16] and 39 weeks (no standard deviation reported) in a PRISON+ study [29].
There was some evidence that rates of caesarean delivery ( Fig. 2 and Table 5), inadequate prenatal care (Fig. 3 and Table 5) and preterm delivery ( Fig. 4 and Table 5) were lower in women in prisons receiving enhanced prison care (PRISON+) compared to women in prisons receiving usual care (PRISON). For all these outcomes, the p-value for heterogeneityusing the I 2 testwhen comparing PRISON and PRISON+ groups was less than 0 · 05. "Inadequate prenatal care" was defined slightly differently in each of the three studies. Cordero defines it as women receiving less than six prenatal visits [21], Bell defines it using an "Adequacy of Prenatal Care Utilisation Index" which takes multiple factors into account [17], and Mertens calls it "Low antepartum care" without defining it more clearly [26].
There was no significant difference in rates of stillbirth ( Fig. 5 and Table 5) or neonatal unit admission ( Fig. 6 and Table 5) between PRISON and PRISON+ groups. For low APGAR score and small for gestational age it was not possible to compare outcomes between PRISON and PRISON+ groups as there were not enough nonzero outcomes (Table 5). For breastfeeding rate, there was evidence of heterogeneity between the two studies in the PRISON group so we did not pool the data and could not compare outcomes in the PRISON and PRISON+ groups (Table 5).
Women's satisfaction with a particular prison intervention was measured in three studies. 5/5 women in one PRISON+ programme felt that "contact with the programme was extremely positive" [27], and 14/14 women in another were "very satisfied" [29]. In Carlson's PRISON++ study, 35/37 women said they would "go through the programme again" [18].

Outcomes in intervention groups compared to comparison groups
Of the nine studies with useable comparison groups, five were in the PRISON category [17,24,26,32,33] and four were in the PRISON++ category [16,19,23,31]. We present the analysis separately for the PRISON++ and PRISON studies as we did not deem the intervention groups to be similar enough to pool the results across those studies.
Of the five PRISON studies with comparison groups, three had population comparison groups [26,32,33], one had a disadvantaged comparison group [17], and one included both population and disadvantaged comparison groups [24]. Table 6 shows the outcomes of interest which were only reported in one study each, all of which had population comparison groups. There was no significant difference found between intervention and comparison groups for caesarean delivery, neonatal death, stillbirth, low APGAR score and small for gestational age. Women in the PRISON group were significantly less likely to breastfeed than population controls (OR 0 · 28, 95 % CI 0 · 14-0 · 56).
Mean birth weight was reported in one study with both population and disadvantaged comparison groups [24]. The mean birth weight of babies born to the women in prison was not significantly different to those born to women in the population control group (standardised mean difference −0 · 19, 95 % CI −0 · 61 to 0 · 23) or There is evidence of significant heterogeneity between PRISON and PRISON+ groups (p = 0 · 005 for caesarean, p < 0 · 001 for inadequate prenatal care, p = 0 · 001 for preterm delivery), therefore a random effects model has been used c No summary measure possible as only one non-zero result d No summary measure possible as significant heterogeneity in results within PRISON category the disadvantaged control group (standardised mean difference 0 · 39, 95 % CI −0 · 07 to 0 · 85). Low birth weight was found to be significantly more common in imprisoned women compared to population controls (OR 3 · 14, 95 % CI 1 · 50-6 · 58) but no difference was found comparing imprisoned women to disadvantaged controls (OR 0 · 40, 95 % CI 0 · 10-1 · 58) (Fig. 7). Inadequate prenatal care was reported in two PRISON studies [17,26] and was significantly more likely in intervention groups (summary OR 1 · 87, 95 % CI 1 · 25-2 · 81) than in comparison groups, which were one group each of disadvantaged and population controls. Preterm delivery rates were  significantly higher among imprisoned women in two PRISON studies compared to population controls (summary OR 2 · 06, 95 % CI 1 · 12-3 · 79) [32,33]. The four PRISON++ papers examining two residential interventions compared outcomes in the intervention groups to disadvantaged controls who were in fact themselves in prison [16,19,23,31]. The following outcomes were only reported in one paper each: caesarean delivery, neonatal death, stillbirth, low APGAR scores, NICU admission, small for gestational age, breastfeeding and  There was no significant difference in these outcomes comparing intervention women to comparison women (Table 7).
Low birth weight was reported in two PRISON++ studies [16,31] and was not significantly different between intervention and comparison groups (summary OR 1 · 22, 95 % CI 0 · 42-3 · 55). Recidivism rates were reported in two studies [19,23] each one examining a different interventionand there is evidence that the PRISON++ interventions reduced recidivism compared to women in prison who did not receive the intervention (summary OR 0 · 37, 95 % CI 0 · 19-0 · 70) (Fig. 8). Recidivism was defined in one study as "confined for any offence post-birth" and was measured using a large administrative database up to the year 2008 which was between eight and ten years after the birth of the women's babies [23]. In the other it was defined as "returned to the facility for violating parole or committing a new crime", measured from the facility records, and was sought up to the year 2007 which was between three and 13 years after the birth of the babies [19]. We considered these two measures of recidivism similar enough to pool the data across the two studies.

Discussion
This is the first systematic review to examine perinatal health care services for pregnant women. We reviewed comprehensively the research literature and examined the evidence for the effectiveness of models of care. We categorised the perinatal health care provided in prison into three distinct groups we termed PRISON, PRISON+ and PRISON++ depending on the level of care provided to the imprisoned women. There was some evidence that women in prisons with increased perinatal care provision had improved maternal and perinatal outcomes; women in prisons receiving enhanced prison care, PRISON+, appeared to be less likely to have inadequate prenatal care, a preterm delivery or a caesarean  delivery when compared to women in prisons receiving usual prison care. The two PRISON++ interventions, that is, interventions which included not only enhanced care in prisons and co-residence with children after birth, but also coordination of community care on release, demonstrated reductions in recidivism rates over the 10 years following release when compared to women in the same prisons who did not take part in the intervention. This finding suggests that a long-term outcome can be improved when interventions are designed in a way that supports women beyond just their time in prison. There is evidence that children of incarcerated parents are more likely to experience a range of negative outcomes than children of similar socioeconomic backgrounds who do not have an incarcerated parent [87]; thus reduced recidivism in the years following birth could impact positively on the lives of both mother and child.
One of the most striking findings of this review was the lack of data and in particular, a lack of high quality studies. Thirty five studies were excluded because although examining specific interventions, they did not quantify any outcomes. This suggests that that there is a paucity of meaningful data evaluating the quality and impact of programmes, even where the programmes exist. Of those studies that were included, the quality was variable ( Table 4). Risk of selection bias was assessed as generally high among the prison populations. In nine studies it was unclear how prisoners were selected [17, 20-22, 25, 27, 30-32], and in eight there was a high risk of selection bias [16,18,19,23,24,26,28,31], largely because of strict eligibility criteria for entering the intervention programmes which excluded a high proportion of imprisoned pregnant women. If or when critical justice reform happens in the US it will be vital that women with violent charges are not excluded from Mean gestation a Mean 38 · 9, SD 1 · 7 Mean 38 · 8, SD 2 · 2 0 · 05 (−0 · 41 to 0 · 52) a standardised mean difference instead of odds ratio presented Fig. 7 Low birthweight of babies born to women in PRISON intervention groups compared to controls, stratified by type of comparison group enhanced programmes, as otherwise very few women will be eligible to participate. Assessing performance bias, that is, whether intervention or comparison groups were exposed to care other than that described, was almost impossible as authors did not comment on this. There was also a high risk of selective outcome reporting, with eight studies reporting outcomes that were not described in their methods or omitting to report outcomes included in the aims [18,19,22,25,[30][31][32]. The overall risk of outcome measurement bias was low in nine studies [17, 20-22, 24, 25, 28, 29, 32], high in six studies [16,18,19,23,27,30], unclear in two [31,33] and mixed in one [26]. Risk of measurement bias was high for both measures of recidivism. There were no other outcomes that had consistently high risk of bias. In all studies with comparison groups except one [23]; it was felt that there were comparable methods measuring outcomes in intervention and comparison groups. All the studies were observational and thus there remains possibility of uncontrolled confounding accounting for some of the observed differences. We recognise the difficulties of conducting research in prisons. These are challenging environments where, for example, prison regimes make it difficult to access women, there are unique ethical considerations and research funding opportunities are limited. However, with careful planning and engagement of all stakeholders, it is possible for large scale and high quality research to be conducted.
Length of stay in prison has been demonstrated to have an effect on perinatal outcomes [88], and it is a limitation of our study that we were not able to adjust outcomes based on timing of entry into prison and length of imprisonment of pregnant women. This information was only available in one study [32], reflecting again the lack of high quality studies. Related to this is the difference between prisons (longer term inmates) and jails (shorter term inmates) in the US. The PRISON and PRISON+ studies from the US were based in a mixture of prisons and jails. All PRISON++ studies and their comparison groups were in US prisons.
Previous systematic reviews have shown that imprisoned women are a high risk obstetric group [10] and that imprisoned women may have improved pregnancy outcomes compared to similarly disadvantaged women outside prison [12]. One possible explanation for these improved outcomes -is that prison provides protection from the disarray of women's lives outside prison, which includes enabling them to access antenatal care. The findings of this latest review, although limited by the quality of the included studies, suggest that greater health and social care input leads to improved outcomes relating to adequate prenatal care, preterm delivery and caesarean delivery; and that programmes providing longer term support can reduce recidivism. These results do not endorse the imprisonment of pregnant women. They focus on a limited set of outcomes and do not examine the wider psychosocial or ethical aspects of imprisoning pregnant women. If women need to be incarcerated, they should be provided with excellent care in a correctional facility. If they do not need to be incarcerated they should be supported in the community. There are no clinical trials which compare imprisonment to enhanced community support programmes like one of the PRISON++ programmes in this review [16,23,31], and it is possible that there would be better perinatal and long-term outcomes for women and children who are supervised in the community. There is some limited evidence that antenatal care programmes targeting specific vulnerable groups are effective [89]. Specific models of care appeared to confer benefits on particular vulnerable groups such as drug users, socioeconomically deprived and teenage pregnant women [90][91][92][93][94][95][96]. Again however, most data related to short-term outcomes, and it is also important to investigate long-term outcomes given the potential benefits not just to the mother and child but also wider society.
We used an extensive search strategy and were able to locate relevant studies that had not been published in scientific journals. However, most studies were located in the US, limiting generalizability, and we particularly note the absence of any information from prisons in low and middle income countries, and the absence of studies in juvenile facilities. This study was also limited by the poor quality of the component papers. There was little consistency in the reporting of outcomes and very few measured beyond the postnatal period. Outcomes were not always defined consistently and another limitation is that we have pooled some data comparing outcomes across studies where the outcomes have been defined slightly differently, for example for recidivism and inadequate prenatal care. The largest number of studies reporting any one outcome was only six and many outcomes were reported in only one study. This systematic review provides evidence of the need for a minimum set of outcomes to be reported in future studies looking at the perinatal health care of imprisoned pregnant women.
For many outcomes there were small sample sizes and only a few cases in each group, which limits our power to detect significant differences between groups. Studies with low power have a reduced chance of detecting a true effect but also a reduced likelihood that a statistically significant result reflects a true effect. Future studies should use adequate sample sizes to detect significant differences between groups. For example, to differentiate between stillbirth rates of 0.5 % in one group and 1 % in another, 4,600 in the cohort and 4,600 in the comparison group would be needed. To detect a reduction in recidivism rate from 50 to 30 %, 100 in the cohort and 100 in the comparison group are needed.
There was some heterogeneity regarding the intervention groups, particularly in the PRISON++ group. One PRISON++ intervention was a nursery programme within the prison, and the other was a secure community-based residential facility for women to reside during pregnancy and after birth with their child. However, they both provided alternative accommodation to the usual prison accommodation, enabled mother and child to reside together, linked to community resources on release, and were compared to women in prison who were given no specific support during their pregnancies. We therefore decided to pool the data across the two studies measuring recidivism.
Perinatal care in prison is an important opportunity for health professionals to engage this vulnerable yet accessible population with potentially significant impacts on the long-term health of both mother and baby. Of the main modifiable risk factors during pregnancy for future child health (tobacco, alcohol, obesity, diet, illicit drug use, mental illness, low socio-economic status and psychosocial stress) [97], most could be targeted through a comprehensive perinatal care programme for pregnant prisoners. There are some interventions designed specifically for pregnant prisoners but very few of these support women beyond the immediate postnatal period, and they are not being adequately evaluated. Despite WHO's 2009 declaration that current arrangements for dealing with the health of women in prison fall far short of what is required by human rights [2], we are currently missing the opportunity to improve both the short and longer-term health of these women and their children.
Properly designed programmes with rigorous evaluation are needed as a matter of urgency so that those commissioning and providing services to these high risk women can ensure the delivery of evidence-based comprehensive perinatal care for imprisoned pregnant women. The design of such studies should be informed by the UK's Medical Research Council guidelines on complex interventions [98], should not be confined to the US, the UK and Germany, and should include longterm follow up with outcomes agreed by an expert panel and supported by the literature, as outlined by the COMET initiative to develop Core Outcome Sets [99]. Action is also needed to highlight and encourage political action on this important public health issue.

Conclusion
This is the first systematic review examining perinatal health care services for imprisoned pregnant women. Our findings suggest that increased perinatal care services for these women can improve both short and long-term outcomes. However, there is a paucity of data on the perinatal outcomes of imprisoned women and models of care, where they exist, are not being evaluated. Properly designed programmes with rigorous evaluation are needed so that we can better address appropriately the health needs of this vulnerable population. The costs to mothers, children and to society of failing to address these important public health issues are likely to be substantial.
Institute for Health Research (NIHR) Professorship. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.

Funding
This work was supported by a National Institutes of Health Research Professorship to MK; no other external funding.