New statistical methods advance digital health research

The digitalization of medicine is transforming prevention, diagnosis, and therapy. Data from wearables, health apps, and mobile sensors offer valuable insights into parameters such as heart rate, movement, or blood pressure. They hold great potential for patient-centered and continuous medical care, especially in the field of cardiovascular diseases.

However, the analysis of such data presents significant challenges. Datasets are often highly complex, incomplete, or irregular, and in some cases, patient numbers are limited, for example in studies on rare diseases. To address these challenges, Eleonora Carrozzo, a researcher at Salzburg Research, is developing statistical tools specifically designed for digital health data.

“The aim of my work is to arrive at statistically sound and clinically relevant conclusions despite small case numbers, high data complexity or erroneous values,” explains Carrozzo.

Carrozzo’s research focuses on non-parametric statistical approaches, which do not rely on strict assumptions about data distribution. These methods are ideal for high-dimensional or incomplete data and can better reflect the variability found in digital health monitoring.

The goal is to close methodological gaps and provide medical professionals with tools to evaluate and apply digital health interventions in a scientifically sound and clinically relevant way. This can improve data-driven decision-making in both research and medical practice.

A key element of the project is practical implementation. The newly developed methods will be made available as R software packages, allowing for simple and transparent use in research and clinical settings.

These tools can handle:

• Small or incomplete datasets
• Complex study designs with multiple groups or time points
• High-dimensional data from digital sources

This ensures reliable statistical analysis even when conventional methods reach their limits.

MEDICA-tradefair.com; Source: Salzburg Research