Predictive health index: integrating body composition and heart rate variability metrics with artificial intelligence to predict chronic disease risk and specific chronic non-communicable diseases.

The Predictive Health Index (PHI) was developed and validated using data from 35,405 clinically monitored individuals collected over approximately 5 years.

• Data were obtained from 35,405 clinically monitored individuals over about 5 years.
• The study targeted cardiovascular, metabolic, psychological, neoplastic, and chronic inflammatory diseases.
• Metrics were obtained from non-invasive instrumental diagnostic tests.
• The dataset was used to train and test an artificial intelligence and random forest machine learning algorithm.

The PHI demonstrated highly significant predictive performance in distinguishing healthy subjects from those at high risk of chronic non-communicable diseases.

• Results demonstrated 'highly significant (p-value < 0.0001) predictive performance concerning the PHI.'
• The PHI successfully distinguished healthy subjects from those at high risk of disease.
• The PHI was validated in the context of cardiovascular, metabolic, psychological, neoplastic, and chronic inflammatory diseases.
• The tool was characterized as 'fast, non-invasive, easy-to-use, highly accurate' for assessing health and predicting NCD risk.

Body composition analysis using advanced bioimpedance techniques (BIA-ACC®) was employed as a key data source for the PHI.

• Body composition analysis utilized advanced bioimpedance techniques identified as BIA-ACC®.
• This constituted one of the two primary non-invasive instrumental diagnostic test modalities used.
• The approach was described as non-invasive and suitable for broad application in predictive health contexts.

Heart rate variability (HRV) analysis employing a photoplethysmography (PPG) system was integrated into the PHI alongside body composition metrics.

• HRV analysis was performed using a photoplethysmography (PPG) system.
• HRV metrics were combined with body composition data as inputs to the machine learning model.
• Both measurement modalities were non-invasive in nature.

A random forest machine learning algorithm was selected and trained to create the Predictive Health Index from the multimodal health metrics.

• Artificial intelligence and a random forest machine learning algorithm were trained and tested to create the PHI.
• The algorithm was applied to data from 35,405 individuals.
• The model was designed to assess health status and predict disease risk contextually across multiple NCD categories.

The study positions the PHI as a tool to support the transition from reactive to proactive healthcare through lifestyle modification interventions.

• Predictive health 'emphasizes proactive lifestyle modifications to reduce the risk or to potentially reverse the progression of chronic non-communicable diseases.'
• The acquired information 'could be extremely helpful for strengthening lifestyle measures and intervening early to prevent or reverse disease development.'
• The approach 'redirects the focus of medicine from treating NCDs to preventing them through lifestyle modifications, marking a fundamental shift toward disease prevention and long-term well-being.'
• The PHI has the potential to 'extend the duration of good health and reduce the incidence, prevalence, and costs of NCDs.'