top of page
Search

Causal Disease Management – Explainable Risk Modeling and Intervention for Diabetes

  • Writer: Vizuro team
    Vizuro team
  • Mar 25
  • 3 min read


Empowering Clinicians and General Populations with Actionable Insights


Diabetes is one of the world’s fastest-growing health crises, impacting over 38 million individuals in the United States and placing an additional 97.6 million at risk due to prediabetes [1-3]. As the global burden of diabetes escalates, healthcare providers are seeking deeper, more actionable insights to improve prevention and treatment. Enter Karma360, a causal discovery and intervention optimization platform that empowers clinicians and general populations with data-driven strategies to mitigate diabetes risk.


The Challenge: Making Sense of Diabetes Complexity


Diabetes is a multifactorial disease shaped by a high-dimensional and integrated network of genetic, lifestyle, and environmental factors. Traditional analytics often fall short in untangling these complex relationships, making it difficult to pinpoint the most impactful interventions.

Karma360 addresses this challenge with its advanced causal discovery engine, enabling clinicians to unearth hidden connections and actionable levers for early intervention and personalized care.


The Solution: Karma360’s Causal Discovery Workflow


Karma360’s process is rooted in scientific rigor and domain expertise, combining machine learning with causal inference. Here’s how it works:


  1. Data Preparation: Cleaning and standardizing datasets by removing missing values. For this case, 260 samples (130 positive, 130 negative) were used, with 75% allocated for training.


  2. Initial Causal Discovery: The platform identifies causal relationships between variables using a cutting-edge inference model.


  3. Domain Knowledge Refinement: Prior research and expert knowledge guide adjustments to ensure logical, explainable connections (e.g., ensuring variables like age and race are upstream without in-degrees).


  4. Model Re-inference: Karma360 iteratively refines the causal graph, delivering a robust, explainable model. See Figure 1 for the constructed Causal Graph using Karma360.



Figure 1. Causal Graph from 194 diabetes samples
Figure 1. Causal Graph from 194 diabetes samples

Key Findings


Karma360 identified the top three contributors to diabetes risk:


·       Pregnancies (0.225)

·       BMI (0.165)

·       Glucose (0.157)


These insights align with established medical research [4], reinforcing Karma360’s credibility.


From Insight to Action: Causal Prediction in Real-World Scenarios


Beyond discovery, Karma360 enables causal prediction to simulate real-world interventions. By applying the model to unseen test samples (33 positive, 33 negative), Karma360 demonstrated predictive strength even with limited data.


Actionable Simulation


The platform tested multiple intervention scenarios aimed at reducing risk factors. By focusing on glucose (selected as the key variable), simulations explored how changes in upstream variables (BMI, Insulin, and SkinThickness) could impact diabetes outcomes. See Figure 2, which demonstrated an actionable “sub–Causal Graph” for Diabetes.




Figure 2. sub-Causal Graph for actionable planning
Figure 2. sub-Causal Graph for actionable planning

Real-World Scenarios Tested:


  1. Medical Approach: Reduce Insulin through medication.

  2. Lifestyle Approach: Lower BMI and SkinThickness via exercise and nutrition plans.

  3. Combined Approach: Integrate both medical and lifestyle strategies.

  4. Extended Approach: Add blood pressure management.


Results


The most comprehensive strategy—targeting Insulin, BMI, and SkinThickness—reduced diabetes risk in 97% of cases (32 out of 33 patients). Notably, even a purely lifestyle-based intervention achieved an 82% improvement. See Figure 3 below for real evidence.



Figure 3. Reduction of outcome value on 33 samples
Figure 3. Reduction of outcome value on 33 samples

Conclusion: Transforming Disease Management with Explainable, Causal AI


Karma360 empowers healthcare professionals and patients with explainable, data-driven pathways to combat diabetes. By providing a holistic view on how different factors interact and impact the disease risks, clinicians and patients can jointly determine the most suitable intervention strategies that fit each patient’s needs, hence improving compliance and the outcomes.


References:


Comments

©2025 VIZURO LLC. ALL RIGHTS RESERVED.

bottom of page