Show simple item record

dc.contributor.advisorGeorge Ochieng’ Otienoen_US
dc.contributor.advisorKenneth Ruchaen_US
dc.contributor.authorOlwendo, Amos Otieno
dc.date.accessioned2022-08-05T12:43:37Z
dc.date.available2022-08-05T12:43:37Z
dc.date.issued2022
dc.identifier.urihttp://ir-library.ku.ac.ke/handle/123456789/23903
dc.descriptionA Research Thesis Submitted for the Degree of Doctor of Philosophy (Health Information Management) in the School of Public Health and Applied Human Sciences of Kenyatta University, February 2022en_US
dc.description.abstractThe adoption of EHR in health care has resulted in the collection of lots of data thus the drive to utilize EHR data for secondary purposes such as research. However, EHR data are characterized by incompleteness and inconsistencies. Diabetes is a worldwide health problem and approximately 14 million persons had diabetes in Africa in the year 2015. In Kenya, type 2 diabetes makes 90% of all diabetes cases. However, a number of cases of diabetes in Kenya experience late diagnosis which result from prolonged experience of prediabetes thus the need for targeted case finding. The aim of this research was to examine the influence of EHR software design, use of a data dictionary, process of collection and entry of data into the EHR, and human resource development on the suitability of EHR data for computational phenotyping of diabetes. A retrospective study was conducted at the diabetes clinic at Nairobi Hospital located in Nairobi, Kenya. After obtaining relevant ethical approvals and informed consent from respondents, the study conducted interviews for 32 staff involved in the use of the EHR comprised of physicians (4), nurses (22), health records officers (HRIO) (4), and managers (2) for both sections. The study also sampled 652 historical records of confirmed cases of diabetes collected between January 2012 and December 2016. Data processing was conducted through outlier detection, smoothing and z-score normalization. Likert scaled results were collected using a usability questionnaire. Measures of central tendency and reliability analysis were calculated using SPSS version 2021 and inferential statistics were conducted through cluster analysis using density-based clustering algorithm. Results show that software design influenced the suitability of EHR data and usability of the EHR was acceptable with a USE mean score of 5.6/7. Regression analysis showed that software design explained 50.7% of the improvement in the suitability of EHR data. Participants 100% reported that the EHR had an inbuilt data dictionary and regression analysis showed that the use of the dictionary explained 32.3% of the improvement in the suitability of EHR data. Also, 98% of participants believed that process of data collection resulted in the collection of data of good quality and regression analysis showed that the process of data collection explained 23.5% of the improvement in the suitability of EHR data. Finally, 93% of participants believed that they had been provided with adequate training to use the EHR. Also, participants 90% reported that they were motivated to continue working at the same location due to opportunities for further training and specialization, good remuneration and working environment. Regression analysis showed that human resource development explained 16.6% of the improvement in the suitability of EHR data for computational phenotyping of diabetes mellitus. However, EHR data were determined to be unsuitable for computational phenotyping of diabetes mellitus given that the algorithm clustered 88% (574/652) of the records as noise. Nevertheless, the algorithm identified 23 meaningful clusters from 12% (78/652) of the diabetes data. However, the EHR software needs incorporate an attribute for the recording patient waist size, blood pressure, and patient-reported drug allergies. Likewise, this study recommends the need to understand the influence of HRIO on the suitability of EHR data for computational phenotyping of diabetes mellitus since they serve as mediators to the physicians for the entry of data into the EHR.en_US
dc.description.sponsorshipKenyatta Universityen_US
dc.language.isoenen_US
dc.publisherKenyatta Universityen_US
dc.subjectSuitabilityen_US
dc.subjectElectronic Health Record Dataen_US
dc.subjectComputational Phenotypingen_US
dc.subjectDiabetes Mellitusen_US
dc.subjectNairobi Hospital,en_US
dc.subjectKenyaen_US
dc.titleSuitability of Electronic Health Record Data for Computational Phenotyping of Diabetes Mellitus at Nairobi Hospital, Kenyaen_US
dc.typeThesisen_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record