Our findings suggest that the nurse mentoring program, which was designed to improve the quality of care for women and children during the labour, delivery and post-partum periods, was associated with a significant decline over time in neonatal mortality. There was a three-fold reduction in the neonatal mortality rate over a three-year period observed among deliveries in primary heath care facilities with NM programs, with the greatest reduction in mortality seen on the first day of life. This corresponds to an average annual rate of reduction in NMR of about 23%. No significant changes were observed in the estimated NMR among newborns delivered in facilities not supported by the NM program or at home. Differences between place of delivery groups were also observed in the late postnatal period, and no neonatal deaths were reported among infants who delivered in NM facilities from day 8-28 post-birth.
Analysis of known risk factors for neonatal death that may have acted as confounders suggests that they were unlikely to have accounted for the large change in neonatal mortality in our study, as there were no significant differences in socio-demographic characteristics between the two survey rounds. Interestingly, a greater proportion of more marginalized women utilized services at the NM PHCs, suggesting that the NM program was able to reach a population where a greater burden of neonatal mortality has been documented.17-19 Significant improvements were seen in pregnancy-associated characteristics between survey rounds, but this was similar in both the NM and non-NM facility groups, and is similar to trends in utilization of health services from other studies.20,21 With the exception of an increase in Caesarean sections and assisted operative vaginal deliveries, there were also no significant differences in delivery characteristics between baseline and endline. Encouragingly, there was a significant improvement between the two survey rounds in the proportion of women who received postnatal care, either at home or in a facility (53% to 80%).
Although the study was not designed to establish causal linkages between the NM program and reductions in neonatal mortality, several aspects of the NM program could have contributed to the positive effect on newborn survival. These would include: a focus on improving diagnosis and management of MNCH complications through enhanced skills, knowledge and practices; use of self-check lists to identify system level gaps and develop action plans to respond appropriately; and use of case sheets as job aids to prompt appropriate treatment and referral to ensure follow-up once complications are identified. We have previously shown that this multipronged approach resulted in improved quality of care at institutions supported by the NM program within the project area, through improvements in facility readiness as well as provider preparedness, and improved knowledge and skills around essential obstetric and neonatal care.10,11
From 2009 to 2011, the average annual rate of reduction in NMR for all of India was 4·6%.22 The reduction in NMR observed in our study was much higher than this, and several factors may have contributed to this finding. One is timing, as the greatest reduction in mortality observed in our program occurred on the first day, when in India more than one-third of newborn deaths occur.19,22 Inability to prevent early neonatal death is a known contributor to slow progress in newborn survival22, and the nurse mentoring program was designed to address MNCH care during this critical period. Secondly, the context of being embedded in the Government of India’s NRHM likely facilitated the accelerated mortality decline, as the program built on gains already achieved by the NRHM. Demand for delivery in health facilities was already high in the state (although lower in our program districts), and key infrastructure, supplies, personnel and emergency transport were already in place when our nurse mentoring program began.23 Furthermore, faster rates of decline in NMR have been documented in more marginalized and vulnerable populations,17,24 and our data indicate that a large and increasing proportion of marginalized women utilized services at the health facilities where the nurse-mentoring program was implemented; and this may have also contributed to an accelerated rate of decline.
We also attempted to identify predictors of newborn survival for the overall sample population and sub-populations by place of delivery. When survey round and place of delivery were controlled for, single births and births to mothers who had received two tetanus toxoid injections during pregnancy were associated with increased newborn survival. This is not surprising, as multiple gestation is recognized as a risk factor for neonatal morbidity and mortality,25,26 and the protective effect of maternal vaccination for TT on neonatal mortality in India is well-documented.27,28. Among newborns delivered in NM facilities, factors that were significantly associated with neonatal mortality included: consumption of 100 IFA tablets during pregnancy; meeting with an ASHA at least once a month during pregnancy; use of an ambulance to reach the delivery point; and presence of any delivery complication. We hypothesize that these factors may represent proxies for high-risk pregnancies, leading to adverse outcomes. Use of an ambulance, for example, may be due to a pregnancy or neonatal medical emergency. From 2012 to 2104, 40-43% of the medical emergency trips in Karnataka were pregnancy related.29 A hospital-based study in Karnataka observed that one-fifth of the patients arriving in hospitals using government ambulance services had pregnancy-related complications.30
Recently, the Government of India has made provisions for nurse mentors in several states and in high priority districts. For instance, the Government of India has been supporting, since 2014, a set of 100 on-site nurse mentors in 25 high priority districts of Uttar Pradesh, a large north Indian state. Various elements of the NM intervention were incorporated in the National Guidelines on Quality Assurance31, strengthening competency-based training of health care providers32, and strengthening pre-service education for nursing33.
We used household surveys to determine neonatal mortality, which are often considered better sources of data on neonatal mortality than either the civil registration data or the routine facility data.8,34 The civil registration data is often incomplete.35 Facility data are limited by substantial selection bias since births and deaths may occur outside the health system.36 Data from household surveys where calculations are based on full birth history, date of birth of each child, whether each child is alive and if not the age of death are hence used for neonatal mortality estimates. In India, the Sample Registration System (SRS) and the household surveys have been the major sources of data on neonatal mortality. India has completed 4 rounds of National Family Health Survey (NFHS), designed and implemented on the models of Demographic and Health Surveys. And the SRS data on neonatal mortality has been made available since 1971. However, both the NFHS and the SRS do not provide the neonatal mortality estimates for the project area. And hence special surveys, at baseline and end line were designed and implemented, similar to the NFHS, in the project districts.
Our study has several limitations. We used a non-randomized design to compare the health outcomes between different types of facilities and home deliveries, and so we cannot directly attribute the observed decline in neonatal mortality to the NM program. Moreover, we have used population-level data to measure neonatal mortality, rather than facility-level data, and thus there are limitations in terms of assessing the effectiveness of the NM program, which is a facility-based intervention. The facilities have significant heterogeneity in terms of institutional capacity and numbers of complicated cases. While a randomized controlled trial among the PHCs would have been ideal, this was not possible due to the desire of the state government to institute rapid program scale-up. Nevertheless, we believe that the comparisons made between NMR at baseline and endline as a measure of program effect are suggestive, because participants in the surveys were randomly recruited; they had no information about the NM intervention per se, and thus had no bias for use of PHCs over non-PHCs as delivery points; and the NM intervention was implemented in rapid succession in all PHCs in the project districts, using similar processes and content. Another limitation is that we do not have information on referral pathways , so cannot know if those who delivered at non-PHC facilities were first seen or referred from NM PHCs. However, as higher-level institutions also showed some improvement in their NMR, it is unlikely that the rapid decline in NMR at PHCs was only due to referral of complicated cases, and shifting of place of newborn death.