JMIR Publications

ABSTRACT

Background:

Non-communicable diseases (NCDs) have become the leading cause of mortality worldwide. NCDs account for 89% of all deaths in the United States and cost the US economy over 47 trillion dollars in direct and indirect expenses. NCDs also account for the main cause of disability worldwide and the incidence is increasing. The leading NCDs include diabetes, cancer, cardiovascular disease, chronic respiratory disease, and mental health conditions. Outside of aging, NCDs are caused by modifiable behavioral risk factors that include smoking, drug and alcohol abuse, unhealthy diet, obesity and inadequate physical activity and treatment must be directed to all of these domains.

Objective:

A ‘digital twin’ is a term defined as a detailed virtual representation of a physical object. When used in healthcare a digital twin is used to model a real person allowing for more individualized care and to support clinical decision making. Unlike clinical simulations that are based on population-based research and clinical trials, a digital twin is constructed from an individual’s presentation and is continuously updated from real-time tracking data. It uses statistical-based machine learning and artificial intelligence to provide evidence-based guidance and contextual motivational messaging throughout each episode of care [1]. We present a methodology to validate this concept which would provide a new clinical approach towards addressing the leading cause of disability and mortality worldwide today. Randomized clinical trials (RCTs) have been the gold standard for evidence-based medicine. Yet, most published RCTs using homogenous populations provide only sufficient data to answer a clinical question being studied as either effective or ineffective. These studies are then generalized to heterogeneous populations that make up clinical care. In addition, concerns have been raised that clinical trials have historically not included sufficient minorities and certain specific demographics making generalizing the results to diverse populations problematic. [2-5].

Methods:

Methods:

This study proposal will use delta scores between treatment arms to ascertain whether that distribution was normal for each of the study variables. Parametric (e.g., ANCOVA) or nonparametric analyses will be to examine the variables to determine the impact of digital twin efficacy over normal treatment paradigms.

Results:

The expected results will show that digital twin modeling using the biopsychosocial characteristics of each patient will be statistically significant supporting this using this approach for personalized medical care.

Conclusions:

Our position paper proposes a methodology and structure that will be used to evaluate the concept of a digital twin that can personalize treatment regimens through analysis of data that allow for AI-based decision making. The goal is to identify significant clinical characteristics to help mitigate the impact of NCDs through biopsychosocial treatment paradigms. This paper proposes a statistical framework to evaluate the validity of the platform’s modeling in support of clinical decision making.