Survival Analysis¶
Todo
Insert summary/introduction that explains the purpose of this section, how it relates to cause modeling, and what will be covered
Terms and Definitions¶
Todo
Add definitions and descriptions to terms
Term |
Definition |
|---|---|
Event |
|
Time |
|
Censor |
|
Observation period |
|
Closed cohort |
|
Open cohort |
Survival Function¶
Todo
Include disclaimer that we will first discuss closed cohorts without event recurrance until a later section
Include formula for survival function, interpretation, description, etc.
Include kaplan-meier plot
Hazard Function¶
Todo
Include formula for hazard function, interpretation, description, how it relates to survival function, etc.
Include graphic of observed failure rate made up of decreasing early infant mortality hazard rate, constant random hazard rate, and increasing wear out hazard rate along with corresponding survival curves
Include an example that contrasts population time (2016) with time relative to an event (years since diagnosis).
Implications for Cause Model Transition Rates¶
Todo
Link to cause model transition rates data source page
Briefly discuss how incidence is used as transition probability and assumption that it is constant over the timeframe it represents
Discuss how this assumption is valid for constant hazard rates, with examples
Discuss how this assumption is less valid for the following scenarios, with examples:
Increasing hazard rates
Decreasing hazard rates
U-shaped and bell shaped hazard rates
When the hazard rate varies between population timeframe and the individual timeframe
Impact on YLLs and YLDs¶
Todo
Include figures that demonstrate how YLLs and YLDs can be under/ overestimated with biased hazard rate approximation
Include examples of how this has been/can be handled (function to increase/ decrease incidence rate around the mean relative to time since infection?, etc.)
Open Cohorts¶
Todo
Discuss how assumptions/implications will vary with open cohorts (kaplan meier plot no longer applies)
Event Recurrance¶
Todo
Discuss how assumptions/implications vary with event recurrance