Identifying Multi Factor Risk Combinations for IVF Failure in PCOS Patients Using Association Rule Mining

Background: Polycystic ovary syndrome (PCOS) is a major indication for in‑vitro fertilisation (IVF). Previous studies typically evaluate candidate risk factors in isolation, obscuring the multi‑factor interactions that often govern clinical pregnancy outcomes. Methods: This study retrospectively analyzed EMR data from PCOS patients undergoing IVF. Key clinical variables (age, body mass index, years of infertility, hormonal/metabolic disorders, tubal or uterine abnormalities, ovarian conditions including luteinized unruptured follicle syndrome [LUFS], and IVF treatment details) were one-hot-encoded. Apriori association rule mining was then used to identify patterns associated with clinical pregnancy failure, with thresholds of support ≥ 0.05, confidence ≥ 0.60, and lift > 1 to ensure robust rules. This novel approach enabled the detection of multifactorial risk associations that were not apparent in a traditional single-factor analysis. Results: The overall clinical pregnancy success rate in the cohort was ~40%. Association rule mining uncovered several clinically meaningful patterns; notably, maternal age > 35 years was a recurrent component of high-risk combinations, often alongside other factors (e.g., metabolic or anatomical abnormalities). For example, a combination of LUFS and tubal obstruction was strongly associated with failure, indicating a synergistic negative effect. Many of these multifactorial associations would have been missed by analyzing variables individually. Conclusions: Apriori rule mining effectively identified complex risk factor combinations for IVF failure in PCOS, informing individualized treatment strategies. Clinically, recognizing patients with advanced age coupled with specific reproductive or metabolic abnormalities can guide tailored interventions to improve IVF success. More broadly, this work demonstrates the potential of integrating association rule mining with EMR data for clinical decision support, enabling the discovery of hidden patterns to enhance personalized medicine.