Background and introduction: Maternal deaths are highest in low resource countries. Skilled attendance at birth (SBA) and the availability of emergency obstetric care (EmOC) are key strategies to improve maternal health and achieve the millennium development goal number 5. In-service emergency obstetric and newborn care (EmONC) training has been used for many years to improve the quality of skilled attendance at birth and availability of EmOC, however few packages have been properly described and evaluated. There is no published comprehensive evaluation of EmONC in-service training packages in low resourced countries. An evaluation of the effectiveness of an EmONC training intervention in 10 comprehensive EmOC Kenya hospitals was carried out from 2010-2011. Methods: A systematic review was performed based on grading of recommendations assessments development and evaluation (GRADE) guidelines to identify the various EmONC training packages in low and middle income countries, identify literature on the effectiveness of these packages or effectiveness of various components of EmONC training globally. The components of the intervention were training in EmONC, provision of EmOC equipment and supportive supervision. The objective of the intervention was to improve the recognition and treatment of emergency obstetric and newborn complications at all study sites by trained maternity care providers (MCP). A before/after study design and an adapted four level Kirkpatrick framework (level 1: reaction to training, level 2: learning, level 3: behaviour/practice, level 4: EmOC availability, health outcomes and ‘up-skilling’) was used to evaluate the effectiveness of the training package. Mixed research methods (quantitative and qualitative approaches) were used to collect data 3 months before the intervention and at 3 monthly intervals after up to 12 months after the intervention. Quantitative data were analysed using SPSS version 20 and qualitative data was analysed using Nvivo 9. Descriptive statistics and analysis using t-tests were performed for quantitative data (significance in mean difference at 95% confidence) while framework analysis was used for qualitative data. Results: 20 EmONC in-service training programmes implemented in low and middle-income countries were identified. The content of 85% (17) of the programmes identified included EmOC signal functions and 7 programmes were 7 days or more in duration. 50% (10) of the EmONC training packages identified had training reports of which only two studies were evaluated at Kirkpatrick level 3 (behaviour) and there was no evaluation at level 4 (health outcomes) identified. Over 70% of all identified maternity care providers from all 10 hospitals were trained. 83% (328) of the 400 health care workers trained were midwives, 6% (26) were medical doctors, 2% (8) were clinical officers and 3% (11) were obstetricians. At 12 months post training the proportion of MCP trained in each hospital was at least 83% except for Nakuru PGH (23%) and Mbagathi GH (50%). Kirkpatrick level 1: About 95% (380) participants responded to level 1 assessment questionnaire. Trainees reacted positively to all lectures (n=11, mean score was 9.38/100, SD: 0.12) and breakout sessions (n=25, mean score was 9.33/10, SD: 0.14). Kirkpatrick level 2: There was a statistically significant difference between the pre and post training knowledge scores in all modules except preventing obstructed labour 0.10 CI (0.06-0.26) p=0.201. The mean difference between pre and post-test skill scores was statistically significant 3.5 CI (3.3-3.8) P<0.001, n=284. Kirkpatrick level 3: 153 data sources (FGDs, paired interviews, KIIs) were collected over 12 months and analysed. 49% (184) and 129 (34.5%) of health care workers and managers participated. They reported a positive impact of the intervention on communication and teamwork, pre-service midwifery education, reduced treatment time, improved knowledge, skills, improved confidence to perform EmOC, organisation of care and supportive supervision. Availability of EmOC equipment post training and supportive supervisors were factors that facilitated change in practice post training. Barriers to availability of EmOC identified were poor staff deployment and retention policy post training, lack of equipment to perform EmONC, lack of support from obstetricians, senior midwives and nurse/midwifery administrators, lack of training for all MCP (including medical interns, medical officers and staff from lower level health care facilities) and lack of clarity on the scope of practice for nurses/midwives. Kirkpatrick level 4: 16, 764 and 17, 404 deliveries were conducted at baseline and at 12 months post intervention respectively. There was 66.8% increase in obstetric complications recorded and managed at 12 months post training compared to baseline. Health outcome indicators: There was an expected increasing trend for number of complications recorded and treated, availability of SBA and EmOC. There was also an expected decreasing trend in the proportion of newborns admitted to NBU for birth asphyxia, direct obstetric case fatality rate (DOCFR) and stillbirth rate (SBR). There was no change in caesarean section (C/S) rate or Fresh stillbirth rate (FSBR). For the health outcome indicators (DOCFR, SBR, FSBR), when PGH Nakuru was excluded from the analysis, a non-statistically significant reduction but greater effect at 12 months compared to baseline was observed for complications recorded and treated (87.9% vs. 66.8%), DOCFR (47% vs. 35%), SBR (66% vs. 34%) and FSBR (14 vs. 10%). There was 34%, 48%, and 35% mean reduction in the SBR, proportion of newborns admitted to newborn care unit and DOCFR at 12 months post intervention compared to baseline respectively. “Up-skilling” indicators: There was a 53.8%, 80%, 100% mean increase in the proportion of all breech vaginal deliveries, proportion of all vacuum extractions performed and proportion of vacuum extractions performed by non-physician clinicians, at 12 months post intervention compared to baseline. Assisted vaginal delivery by vacuum extraction was the least available EmOC signal function (SF) and medical doctors only performed this SF at baseline. At 12 months post intervention, non-physician clinicians performed this as well, in all study sites. Overall the EmONC training intervention resulted in improved ‘up-skilling’ of maternity care providers, a trend towards improved availability of SBA and EmOC and improved health outcomes. Implications for policy and practice The results of this study are important for designing and implementing evidence based EmONC programmes in resource poor countries. None medical doctors can be ‘up-skilled’, the recognition and management of obstetric and newborn emergencies and the availability of quality EmOC can be improved using similar packages and implementation methods in other resource poor settings. Future research: Evaluation designs that include control groups are needed. Studies to assess the relative importance of supportive supervision for behaviour change after training, the knowledge and skills retention with time post training in resource limited settings should be undertaken.