Clarifying Questions

[andkov]

• Q1. One of the states of the outcome must be a dead state. Does it mean that studies without mortality data should be excluded? Ongoing studies?

[ampiccinin]

My guess is that all studies have mortality, just that some have more than others.

[andkov]

• Q2. Constructing the outcomes

We are considering the following outcomes to be the criterion (output) of our multistate models:

• ADL/IADL
• MMSE
• self-rated health
• dementia
• other cognitive variables ( to be specified)

For simplicity we assume that one outcome is measured by a single item in a study, although these items are not necessarily equivalent. Harmonization is not the focus of the project.

Each outcome we model must be a categorical variables with no more than 5 exhaustive and mutually exclusive categories. (see data format guidelines). As people progress in time they change membership in these categories, invariable ending in the dead state, which may not be yet observed. For example, focusing on dementia, our multi-state model will have a single categoriacal criteriondementia which can take three possible values:

• 0 - no dementia diagnosed, alive
• 1 - dementia dianosed, alive
• D - dead

While the outcome MMSE < 24 will also be reprsented by a variable that have three possible values:

• 0 - MMSE > 24, alive
• 1 - MMSE < 24, alive
• D - dead

Only one outcome will be modeled at a time. The possible values of a specific outcome DO NOT have to be the same across studies (e.g. categories for self-rated health in LASA may differ from categories of self-health in OBAS)

@ampiccinin , @GracielaMuniz , Please verify the following statements for accuracy. Do they reflect your thinking?