부산대학교 도서관

Introduction Maternal morbidity and mortality during pregnancy, childbirth or the puerperium puts in serious difficulties the goals of the Millennium Development Goal 5 (MDG 5). Given that the majority of maternal deaths are preventable and avoidable and occur in contexts of poverty, we need to address this problem. One factor that may influence the burden of mortality is the limited use of research in health decisions. This project aims to build a map that reflect the gaps of scientific knowledge on interventions aimed at directly or indirectly reducing maternal mortality, also identify priority research questions and, evaluate the quality of the studies included in the systematic reviews of the Cochrane Collaboration through risk of bias tool and, its variation over time. Each stage was published and they are part of a compendium of three sequential and interrelated publications. Methods A systematic search for systematic reviews in the Cochrane Library published or updated from January 2006 to March 2011, except for the EPOC group. We selected all those that had to do with maternal health. The 'Implications for research' section was evaluated to identify research gaps and research questions were developed with PICO format, which were classified in determinants of maternal mortality. An online survey was elaborated, 319 questions were mapping and we built 12 groups of domains and 12 surveys. Four selected groups of the Cochrane Collaboration have participated, which directly or indirectly have some relationship with maternal health. Two rounds of consultation were made both to refine, to add new questions and prioritize. Likert scale was used for prioritization. We survey to 155 participants between authors, referees and consumers and we sent four reminders. The first round was between 5 and 31 August 2011. The second round was held between November 11 and December 22, 2011 with three reminders. Prioritized questions in the first round were sent to 2,121 participants in the second round. To assess the quality of randomized clinical trials (RCTs) included in systematic reviews of Cochrane database, the number 12 of the year 2012 was used and the RCTs were selected and included independently. The inclusion criteria was had at least one domain of risk of bias (RoB) tool informed. All RoB domains were assessed for differences in time into four strata according to date of publication in a) 2006-2012 b) 2000-2005, c) 1990 - 1999, d) ≤ 1.989. Considering the amount of data to be handled, were determined a priori evaluate four relevant domains: sequence generation, allocation concealment, blinding and incomplete outcome data. Statistical modeling was also done using logistic regression to assess the association between the presence of low risk of bias according to domain and year of publication of the trial, the type of intervention, sample size and country of ECA. Results 204 SR were located and included 178 RS. We mapped a total of 319 questions. Finally 62 research questions were prioritized as highly relevant. The majority was of policies and health systems, unintended pregnancy and abortion, postpartum hemorrhage and hypertensive disorders. The types of intervention most commonly implicated were drugs in 31%, followed by policies and health systems interventions (27%). The overall response rate was 47% in the first round and 17% (363/2121) in the second. In the second round only 12% (253/2121) of participants completed the questionnaire. More women (235) than men (128) participated in the survey. Women however provided incomplete answers more than men (49% versus 35%, p = 0.01). Statistically significant differences (ES) were found when comparing the group of very relevant questions by sex of participant on six questions; half of them related to diabetes. No differences were found when comparing the ES group of very important questions by type of respondent, or country or number of round. For the assessment of risk of bias we included 1,732 RCTs from 97 SR. The ECA judged low and high RoB increased significantly over time, while rates RoB unclear decreased over time in several domains. RCTs published after 2007 had the highest rates of low RoB accompanied by a decrease in the rates of the categories of high and unclear RoB like most domains. RCTs that included drugs as intervention were more likely to show a lower risk of bias for 4 of the 6 domains assessed. The RoB for blinding domains when compared only drugs versus other types of interventions was significative. These differences were not significant for the other domains RoB. Discussion The use of scientific evidence in healthcare decisions is increasingly common. For effective and ethical use first it needs to know that evidence there are and its quality. Under the main hypothesis of this thesis that not everything is known about maternal mortality, both on their causal pathways and interventions to address them, our first task was to map these gaps. A second step was to prioritize through experts. Finally, if the evidences produced in the same source were improving in quality over time. Part of the explanation of why maternal mortality did not achieve the target of MDG 5 could be explained by the deep gap and was brought to light through more than 300 research questions that need answering. One feature to note is that health systems were important determinants of maternal health; both in the mapping as well in the prioritization process. One advantage to highlight these two stages is that they use the same source and the same standards. The main disadvantage was the low response rate. We found no gold standards to prioritize maternal health research, including this thesis to our knowledge, was the first comprehensive attempt. We believe that our simple design but complex in implementation process allowed us to have a ranking of more than 60 questions. Some of these responses are also shown in the annexes of this thesis. The author herself is participating in the development of two relevant systematic reviews of maternal health after obtaining these results. Finally also conclude that some international efforts to improve the quality of the studies are paying off. We have also shown that clinical trials included in systematic reviews improved their quality over time. Conclusions It is posible identify research gaps in maternal health through the use of systematic reviews as a source of them. It may also prioritize these gaps with the participation of different actors belonging to the same source. It is important that those who conduct research in this field not only have a map of prioritized questions but also take into consideration the health systems as determinants. It is noteworthy that the RCTs taking place in different areas of health (not just in maternal health) have improved reporting and enhanced in its validity.