Reflection on Recent Research Presentation and Feedback

Presenting my research on the feasibility of developing a predictive model to forecast the probability of South African antiretroviral therapy (ART) patients falling out of HIV care was an enlightening experience. The feedback I received from my supervisors during and after the presentation significantly altered my perspective on the research topic and guided the future direction of my research.

Initially, my research problem focused on the high rates of ART discontinuation among HIV-positive patients over the age of 15 in South Africa and the known determinants contributing to this issue. My supervisor's feedback emphasised the need to explore determinants that are not typically linked to ART discontinuation in the literature. I also realised that I need to consider the granularity of the determinants that would be incorporated into a predictive model. That it, the extent to which I broke down the determinants into distinguishable parts that still speak to the specific outcome. This feedback prompted me to delve deeper into the literature to capture a more nuanced understanding of each determinant's weight and interaction with each other and the specific outcome.

However, my efforts towards this revealed that the analysis of meta-level causes is considerably broad, and it would need me to focus my learning. I managed to learn about Construct Identity Fallacy. Larsen & Bong (2016) define Construct Identity Fallacy as when the same construct/concept is incorrectly assumed to represent distinct real-world phenomena or when two or more constructs/concepts are incorrectly assumed to represent the same phenomena. Being cognizant of this will be invaluable to finding the right balance of granularity in my proposed determinants.

Feedback from my co-supervisor highlighted that I also need to consider what my definition of feasibility is in relation to a predictive model and the predictive outcomes for this model. This led to a re-evaluation of my initial criteria for feasibility. Instead of merely considering technical feasibility, I now understand the importance of incorporating aspects such as data availability, model accuracy, and the practical implications for healthcare providers and patients.