Prediction

How much time will it take me to drive to work?

Prediction of healthcare outcomes

Question:

For a specific patient, how likely is it that they experience an outcome of interest?

How likely is it for a patient with:

• hypertension to experience a heart attack?
• COVID-19 to be admitted to the intensive care unit?

Prediction of healthcare outcomes

Prediction of benefit

Question:

For a specific patient, is treatment A better or treatment B?

Prediction of benefit

Problem:

We can never observe what would have happened, had the patient received the alternative treatment.

Prediction of benefit

Thought process:

Time machine ❯ Cloning ❯ Twin siblings ❯ Similar patients

Prediction of benefit

Question:

What makes two patients similar?

Literature review

Aim 1:

Categorize methods available in the literature for the prediction of benefit based on their definition of patient similarity.

Literature review

Risk-based methods:

Patient similarity is based solely on risk factors.

Risk-based methods

Aim 2:

Develop scalable risk-based approaches to the assessment of treatment effect heterogeneity.

Observational setting

Problem:

Patients receiving different treatments tend to be systematically different.

Observational setting

Problem:

Lack of interoperability between observational databases.

Standardized framework

Standardized framework

Continuous benefit prediction

Thank you!