Risk of maternal mortality using relative risk ratios obtained from poisson regression analysis

This paper highlights the importance of carrying out a meta-analysis in epidemiologic research of a non-common outcome without the loss of study/ies due to insufficient data. A good number of researches have adequately covered common disease outcomes in a meta-analysis, however, non-common disease outcomes have not been well determined. This paper proposes the use of relative risk ratios as effect sizes for the meta-analysis of non-common disease outcomes, obtainable from the Poisson regression analysis which is usually reserved in statistics for count data, where response outcomes are rare or non-common. Many users of meta-analysis are inclined to health-Science; hence they lack the Statistical competence with which to tackle the draw backs encountered while meta-analyzing non-common disease outcomes. Some draw backs include: Small number of studies available for the meta-analysis; Presence of heterogeneity; Insufficient data to be able to collate effect sizes. The literature used for this paper encountered loss of data in the Loudon, 1992 study, following the use of Poisson regression analysis, a relative risk ratio (RR) of 2.83 was obtained with a confidence interval of (2.62, 3.06). Literature was expanded to Google Scholar, Cochrane database, jstor website, MEDLINE, PUBMED and relevant journals of maternal healthcare. Twenty studies were meta-analyzed altogether, and results were in favour of mortality with a relative risk ratio and confidence interval of 1.66 and (1.32, 2.09) respectively. There was a high presence of heterogeneity from the result of I-squared = 77.1%, p-value <0.001. Sequential use of sensitivity analysis reduced heterogeneity and the risk of mortality by 12%, p-value<0.001; relative risk ratio, 1.91 [confidence interval (1.53, 2.39)] for fifteen studies that were not excluded but were meta-analyzed.