Large language model-assisted causal machine learning for identifying fatigue-related poor glycated hemoglobin in type 2 diabetes

We applied retrospective cohort design to collect data between January and June 2024. The settings were diabetes management centers in Indonesia. The eligibility criterion of participants in this study was individuals with type 2 diabetes. This study was approved by the Joint Institutional Review Board of the Ethical Committee of Medical Research at the Faculty of Dentistry, Universitas Jember (no.: 2683/UN25.8/KEPK/DL/2024).

We collected data for 10 variables by interviewing the participants at the time of survey and reviewing their medical records from the past 3 months. At the time of survey, we asked the participants to obtain 6 variables: (1) age (years); (2) sex (female/male); (3) marital status (unmarried/married/widowed or divorced); (4) educational status (no education/primary school/junior secondary school/senior secondary school/college graduate); (5) cigarette smoking(no/yes); (6) fatigue (no/yes). The remaining 4 variables were obtained from medical records: (1) HbA1c (%); (2) diabetes treatment (OHA/OHA + insulin); (3) diabetes duration (years); and (4) comorbidity (no/yes).