Date of Award

2021

Document Type

Dissertation

Degree Name

Ph.D.

Department

Quantitative Research Methods

First Advisor

Kathy Green

Second Advisor

Antonio Olmos

Third Advisor

Shimelis Assefa

Keywords

Statistics, Diabetes

Abstract

The advantage of using meta-analysis lies in its ability in providing a quantitative summary of the findings from multiple studies. The aim of this dissertation was first to conduct a simulation study in order to understand what factors (sample size, between-study correlation, and percent of missing data) have a significant effect on meta-analysis estimates and whether using univariate or multivariate meta-analysis would produce different estimates.

The second goal of this study was to evaluate the effect of clinical decision support systems CDSS on diabetes care management by conducting three separate univariate meta-analyses and one multivariate meta-analysis. CDSS are health information technology systems that analyze data within electronic health records (EHR) to help make decisions about a patient's care. Several studies reported inconsistent conclusions about how effective CDSSs are on diabetes care management based on three indicators. Low-density lipoproteins (LDL), glycated hemoglobin (HbA1c), and blood pressure (PB) have been used as indicators of diabetes care management according to the National Institute for Health and Care Excellence (NICE) guidelines. To combine the results from studies that evaluate the effect of CDSSs on diabetes care management, meta-analysis was used. The results of the two univariate and multivariate meta-analyses were compared.

The simulation study indicated that MVMA was less affected by missing values compared to UVMA. However, both methods performed equally when no missing data were present. The standard errors of the estimates in both methods were reduced by increasing the sample size with more reduction in standard errors found in MVMA. The results of UVMA and MVMAs of CDSSs’ effect concluded that CDSSs had a significant effect on reducing levels of HbA1c. CDSSs was only significant on LDL when UVMA was applied while pulse pressure (PP) was only affected by CDSSs in the case of MVMA with deleted missing values. CDSSs in general could have a potential effect on diabetes care management.

The results of the simulation and the meta-analyses of the CDSSs indicated that MVMA performed slightly better at different sample sizes and percent of missingness levels than did UVMA.

Publication Statement

Copyright is held by the author. User is responsible for all copyright compliance.

Provenance

Received from ProQuest

Rights holder

Abdelfattah Elbarsha

File size

140 pgs

File format

application/pdf

Language

en

Discipline

Statistics

Share

COinS