﻿ Stratified Covariate Balancing

# Lecture: Multimorbidity & Survival

Applications to Analysis of Prognosis

• Review of findings in multimorbidity index PubMed►
• Prognosis of heart failure patients using their comorbidities PubMed►
• Prognosis of patients in ICU depends on their comorbidities PubMed►
• Comparison of Multi-Morbidity Index to variants of Charlson Index PubMed►

## Assignment

Please provide a summary page for this assignment, indicating whether you got the same answers as the ones posted.

Question 1, Distinguish between Comorbidity and Complication: Identify which diagnoses are complications of lung cancer and which ones are comorbidities.  Comorbidities typically occur prior to the disease and complications occur afterward.  If you do not have access to temporal data, comorbidities and complications can be distinguished by how they affect outcome of care.  Complications are progression of a disease.  Therefore, they are in the causal pathway from disease to death.  Comorbidities are not in the causal pathway from the disease to death.  Complications mediate the effect of the disease on death.  Knowing the complication is sufficient to know the risk the patient faces and there is no need to know the original occurrence of the disease.  In other words when you know that the patient has the complication you also know that they have the disease.  Given this definition of complication, one would expect inclusion of complication in the statistical model will reduce the coefficient for the disease in predicting outcome of care; but inclusion of a comorbidity in the same model will not do so.

• What are the top 5 most frequent comorbidities and 5 most frequent complications of lung cancer?
• What is the average probability that a patient with long cancer will survive another year?
• What is the probability that a patient with lung cancer and no other comorbidities will survive another year?

Resources for Question 1

Question 2, Feature Construction:  Construct a variable indicating progression of cardiovascular diseases.  In this variable, cardiovascular diseases are listed in order of their prognosis, Each cardiovascular disease is a marker in the variable.

1. What percent of variation in survival of patients is explained by the progression in cardiovascular diseases? in particular, what percent of variation in survival is explained by the most severe cardiovascular disease of the patient.
2. For patients, consider patients with arrest, shock, and CHF.   Does the interaction among these three variables increase or reduce the percent of variation explained by the model?

Resources for Question 2

## More

For additional information (not part of the required reading), please see the following links:

1. Complications and comorbidities Read►

This page is part of the HAP 819 course on Advanced Statistics organized by Farrokh Alemi, Ph.D. Home► Email►