Derivation of a statistical model for predicting poor control of diabetes mellitus

Abstract

This study was conducted to identify the factors associated with poor control of diabetes and to derive a statistical model for predicting poor control of type 2 diabetes mellitus. The overall framework of the study comprised of five phases: A systematic review and meta- analysis, self administered questionnaire survey, database development, prospective study, and calculator development. In the first phase, a systematic review and meta- analysis was conducted, to summarize the newlinefactors associated with poor control of diabetes from various studies, conducted on type 2 diabetics. A comprehensive search was done by using various keywords to get all the newlineobservational studies, irrespective of region or languages. Those studies in which, it was newlineimpossible to discriminate data between the subjects with poorly controlled diabetes and newlinenot poorly controlled diabetes, and those conducted on type 1 diabetes or gestational newlinediabetes were excluded from the review. Meta- analysis was performed by pooling the results of selected studies. For metaanalysis newlineOdds Ratio (O.R.) with 95 % Confidence Interval (C.I.) and mean with Standard newlineDeviation (S.D.) were retrieved from each study. Odds ratios were combined and Mantel - newlineHaenszel Odds Ratio (M.H. O.R.) with its 95% C.I was calculated. For continuous newlinemeasurements mean difference with 95 % C.I. were calculated. A two phase strategy was adopted for study selection. In the first phase, the titles and abstracts obtained were scrutinized for inclusion and in second phase full manuscripts of all the qualified studies from phase one were obtained. Overall 7, 501 studies were identified during the initial search, of which 43 studies were newlineretrieved. Of these, 21 studies were included in the review and remaining was excluded newlinebecause of either not satisfying the inclusion criteria or insufficient information. Among the 21 studies, 10 studies were qualified for meta- analysis. Methodological qualities of each study were assessed and data were extracted.