Unsupervised machine learning approach to interpret complex lower urinary tract symptoms and their impact on quality of life in adult women.

PURPOSE: To identify clinically meaningful clusters of lower urinary tract symptoms (LUTS) in adult women using an unsupervised machine learning approach and to examine their associations with patient-centered outcomes, including quality of life (QoL), willingness to pay (WTP) for treatment, and physician visits.

METHODS

We analyzed data from 3,088 women aged ≥ 20 years in a 2023 nationwide Japanese internet-based epidemiological survey on LUTS. Principal component analysis reduced dimensionality across 16 LUTS, including storage, voiding, and post-voiding symptoms, as well as bladder pain and discomfort. Hierarchical clustering grouped participants based on symptom severity. QoL, WTP, and physician visits were compared among clusters.

RESULTS

Three symptom-based clusters were identified: the mild/asymptomatic group (n = 1,635, 52.9%), the moderate storage symptom-dominant group (n = 1,224, 39.6%), and the severe complex group (n = 229, 7.4%). The identified groups had substantially different demographic characteristics and LUTS risk factors. The severe complex group, experiencing the highest symptom burden, was most affected in daily life, emotions, sleep/energy, personal relationships, and role, physical, and social limitations. The median monthly WTP was ¥500 ($3.5) in the mild/asymptomatic group and ¥1,000 ($7) in the other two groups. Physician visit rates were highest in the severe complex group but remained low at 11.8%.

CONCLUSION

An unsupervised machine learning approach identified three distinct LUTS clusters in adult women. Even in the most severely affected group, WTP was limited, and physician visit rates were low, highlighting significant unmet medical needs.