Though this was a community randomized intervention study, it nevertheless has several limitations. It was difficult to control extraneous factors that may have a direct or indirect impact of the indicators of interest. For example, women in the intervention arms sometimes shared their IEEC materials with the women in the control arm. There were also some reports that several of the TBAs trained by the project were called to deliver babies in control villages. Also in the women's only intervention arm, husbands could always look at their wives' booklets, which might have diluted the effect of explicitly adding the husbands' IEEC in the second intervention arm. However, such 'contamination' would serve to decrease the observed difference between the study arms.
In spite of these limitations, we conclude that community-based interventions do have an impact on neonatal mortality in remote rural areas of Pakistan. First, the women in the women's only IEEC intervention arm reported fewer problems and visited a health facility less frequently than their sisters in the control arm, probably because they were relatively better equipped to identify a serious problem. In the husbands' IEEC arm, on the other hand, more women visited a health facility -- perhaps encouraged by their husbands to see a healthcare provider even for minor problems. Also a significantly greater percent of women in the two intervention arms visited the district hospital for problems related with pregnancy and immediately after delivery. This would have been unlikely without a better understanding of the problem and support from the husband. Second, while the overall percentage of hospital deliveries was low, significantly more women in the intervention arms delivered in the district hospital. The use of prenatal care was also significantly higher in the intervention arms. An added impact of the husbands' IEEC on prenatal care was not observed. Nonetheless, husbands' IEEC resulted in significant improvements in diet and reduced workload of their pregnant wives, more regular use of prophylactic iron and folic acid and more frequent visits to the hospital during pregnancy and delivery.
Historically, studies on maternal mortality have divided the causes of maternal deaths into patient-oriented and hospital-oriented categories. The distinction between deaths among 'booked' and 'un-booked' cases was sometimes used to suggest that pregnant women who did not register for prenatal care were more likely to have complications and die. It was only after the launch of the global safe motherhood initiative  that the complex nature of the multiple causes of maternal mortality was recognized. For example, the 'three delays' model  identifies the first delay in decision making to seek medical care for obstetric emergencies, which is attributed to lack of awareness. However, another cause of the first delay could be a lack of trust in the health system. Consequently, the strategies to reduce maternal mortality can be classified into community-oriented and health system-oriented categories. This, in part, is an application of the demand and supply theory in maternal health. There is an ongoing debate on which strategies work best to improve the demand and supply of maternal health services. Unfortunately, experimentation with different strategies has been disorganized and lacking an evidence base. Many projects end with only vague conclusions such as maternal mortality is a complex public health problem requiring broad-based improvements in health and social systems . On the other hand, there is little evidence for or against particular sets of interventions to reduce maternal and perinatal mortality at the community or at the health system level .
Bullough et al.  reviewed the literature on strategies to reduce maternal mortality and found that current strategies were based upon insufficient evidence. In particular, interventions of community mobilization and training of traditional birth attendants were not based upon sound evidence. They argue that safe motherhood interventions are complex public health approaches and quite different from single clinical interventions. The conclusion is that there is insufficient evidence to recommend universally effective interventions to reduce maternal mortality. While a consistent and sustained improvement in the quality of health services will almost certainly reduce maternal mortality, short-term interventions and vertical programs are less likely to achieve the millennium development goals in maternal health. Unfortunately, the maternal health agenda in many developing countries is driven by donor priorities and perspectives. Donors support vertical programs because their monitoring and evaluation is easier and their results are available quickly. But vertical programs may erode the general standards of health systems by diverting resources from overall quality and focusing on a single problem .
Delays in accessing medical care during emergencies, which are responsible for the bulk of maternal and neonatal deaths, can be addressed in a number of ways. Increasing awareness of pregnancy danger signs and birth preparedness and streamlining existing transportation systems are the most effective means to address the first two delays. However, their impact is conditional upon availability of good quality health services.
Birth preparedness programs are a way to increase awareness about the dangers associated with a supposedly normal pregnancy and delivery. During 2003-2004, a field trial of the Birth Readiness Package was conducted in the Siraha district of Nepal . This was primarily a program to increase demand, which resulted in a significant increase in the knowledge, and some improvement in maternal practices in the intervention site. However, use of skilled birth attendants and emergency obstetric care remained unchanged, indicating that the problem is not resolved merely by improving the demand for better health services. In fact, while demand is partly associated with supply (i.e. trust in the quality of health services leads to greater demand) it is essential to work on both. The importance of having a two-pronged approach for reducing maternal and neonatal mortality is emphasized by many studies. Darmstadt et al.  reviewed the impact of community-based interventions on neonatal mortality in a number of countries. They suggest that in settings with high neonatal mortality, community and family interventions can bring down that mortality effectively, but that a complementary health facility-based care model is also necessary. In Nepal, an educational program in neonatal care for the healthcare providers improved their practices in hospital settings, which led to increased utilization of neonatal services and better chances of survival of the neonates brought to the hospitals .
During 2003-2006, the Population Council conducted an operations research study similar to ours in two rural districts of Punjab province of Pakistan . The primary objective of that study was to disaggregate the impact of CBI from the impact of interventions directed at improving the quality of health services. The study had two intervention arms. In one arm, a CBI package similar to that of our study was implemented along with interventions to improve the quality of emergency obstetric care. In the other arm, interventions were directed to improve the quality of emergency obstetric care but no CBI was implemented. Their results indicate a significant reduction in perinatal mortality, attributed to community-based interventions (IEEC to women and husbands, training of skilled birth attendants and setting up transportation systems). Improvements were also found in other process indicators. The project report concludes that: "This intervention package led to declines in maternal and neonatal mortality... by increasing the proportion of mothers and neonates with serious emergencies who seek appropriate and timely help, and for neonates by improving home care at and after birth. While there is evidence that perinatal mortality declined in the CBI communities, the changes in knowledge and practice that would be expected to lead to this result are not convincing. There are several possible explanations, for example: The measurements of knowledge and practice were not sufficiently accurate or precise to pick up real and important changes where they occurred" [: p. 42, Report 1]. On the other hand, the study did not find an impact of health services interventions (such as training of doctors) on perinatal and neonatal mortality .
The primary purpose of our study was to test whether CBI to reduce the delays in seeking medical care for obstetric complications could significantly reduce perinatal and neonatal mortality. The project focused on the first two delays, although attempts were also made to strengthen and improve the EmONC services available at the district hospital, which served the entire project area. Our study provides evidence that the CBI package, along with the interventions addressed at the health facilities, led to a significant reduction in perinatal mortality. The study population was too small to detect significant differences in MMR; however, a number of process and output indicators were significantly improved.
While the role of husbands in family planning has been extensively studied, literature on their role in reducing maternal or neonatal mortality is scanty. There are very few randomized trials or quasi-experimental studies on the impact of involving husbands in IEEC interventions to encourage safe motherhood behaviours. An exception is a study in Mumbai, India, where one out of three pregnant women attending a maternity care clinic was encouraged to bring her husband and then both received education about pregnancy and child-rearing; in a control group, only women received such education . In that study, the intervention group had a significantly lower perinatal mortality rate -- 15 deaths per 1,000 total births versus 35 per 1,000 in the control group (P < 0.001).
In our study, lower levels of perinatal mortality in the intervention sites at the follow-up survey could plausibly be attributed to the interventions. A number of other indicators, including referral to district hospital, use of prophylactic iron and folic acid therapy during pregnancy and prenatal care, were also improved in the intervention sites although the effects became non significant after adjusting for clustering. Although some of these indicators were better in the husbands' IEEC intervention arm, most of the differences were not statistically significant. This could be explained by the fact that even in the women's only IEEC intervention arm, husbands could look into the IEEC materials provided to their wives. Therefore, the statistical effect estimated for the exposure to male IEEC materials might have been diluted.
Finally, the residual impact of the interventions was tested through a post-project survey, almost two years after closure of project. This survey provided evidence that the IEEC impact on perinatal and early neonatal mortality persisted in the village clusters where the CBI was implemented (information to women and husbands, training of traditional birth attendants in clean delivery and recognition of obstetric and newborn danger signs, and setting up transport and telecom systems). Moreover, the IEEC intervention carried out in other village clusters in Khudzar by the local NGO also resulted in a significant reduction in neonatal mortality and a significant increase in prenatal care and iron use during pregnancy.