Project Overview The purpose of this study was to test the feasibility of implementing a traumatic brain injury (TBI) prognostic model as a clinical decision support tool at Kilimanjaro Christian Medical Center (KCMC) and amongst Duke-affiliated healthcare providers. Our prognostic model is a statistical tool that uses clinical information to predict a traumatic brain injury patient’s risk of a poor recovery. This study used the opinions of KCMC- and Duke-affiliated healthcare providers to inform how best to improve the user interface of our prognostic model application, as well as how best to implement our prognostic model application as a clinical decision support tool.
This study was a multi-site feasibility study using co-design sessions, an integral part of human-centered design (HCD) research. Two research assistants (RAs) from KCMC and one from Duke conducted the co-design sessions at their respective institutions. The RAs did not have any previous relationships to the participants. They were trained by the study investigators and conducted the co-design sessions using WebEx video platform (Duke Hospital) or in private rooms one-on-one (KCMC). Paige O’Leary, at the time an MSc student at Duke, and the formal author of this data deposit, was the Duke data collector and used all of the qualitative data for her Master’s thesis.
Co-design sessions, part of HCD research, were used to seek a deeper insight into the users’ cultural, emotional, and practical requirements of the TBI tool. The purpose of the co-design sessions was to create an implementation strategy focused on participants, thereby developing interventions catered to future users, while facilitating their enhanced understanding and ownership of the project.
Data Collection Overview This two-site study took place in Moshi, Tanzania, and North Carolina, USA. Purposive sampling and snow-ball sampling were used to recruit participants. Study participants were healthcare providers at KCMC (N=9) and Duke Hospital (N=12) involved in the care pathways of TBI patients. Participants included emergency medical doctors, nurses, physician assistants (PA) and medical residents. These subjects are representative of the future end-users of the TBI prognostic model. Participation was voluntary, and all participants provided written consent.
Data Analysis Interviews were recorded and then transcribed prior to the analysis by RAs. Once interviews were transcribed, the participants’ qualitative responses were structurally coded to identify themes. The structural codes were developed using the interview guide. To ensure that the codes were appropriate and reproducible, two RAs coded 25% of the Duke Hospital and KCMC transcripts and compared the results. The intercoder reliability score was Kappa = 0.91, meaning that coding by both RAs strongly agreed. The themes from the structural codes were identified using the coding reports generated in NVivo12. Subthemes within the structural codes were identified and collected in a data reduction table.
Shared Data Organization As per the original Duke, KCMC, and NIMR ethics approval requirements, even after de-identification, the data cannot be shared openly. De-identified transcripts are available under controlled access.
The deposit includes 21 de-identified transcripts and five documentation files including the co-design interview guides for both study sites, the two consent forms, an administrative README file, and this data narrative.
| 