Introduction Management of quality of life (QOL) is important for patients with cancer. The critical issue in evaluating QOL is the low adherence to recording patient reported outcomes (PROs). Heart rate variability (HRV), which is associated with the autonomic nervous system, is easily measured. This study aims to develop an artificial intelligence (AI) algorithm to evaluate QOL using HRV.

Methods 180 data from 50 patients and 50 data from 15 patients with gynecological cancer were used as test and validation datasets, respectively. HRV and PROs (EORTC qlq-C30, FACT-G, PHQ9, PRO-CTCAE) were collected daily and weekly, respectively. A binary AI classification model that generates SHAP values was developed to predict whether symptoms related to QOL were severe using HRV. A clustering model was developed by clustering the SHAP values into three groups using Parametric Umap. Serum metabolites that contribute to HRV variation were identified.

Results Clustering derived from HRV indicated high, middle, and low QOL groups (Group A, B, and C, respectively). The total score of FACT-G was 82.4, 72.7, and 67.3 for Group A, B, and C, respectively. The scores of fatigue and other symptoms were also worst in Group C and best in Group A. Metabolites in serum contributing to HRV variation are Arachidonic acid and Dopamine, which are associated with inflammation and depression.

Conclusion/Implications Monitoring QOL over time using HRV may allow us to detect early deterioration in QOL, such as side effects of chemotherapy.