Objectives

At the end of the conference workshop, participants should:

• be familiarized with two computer software (IMaCh and SPACE) for the computation of ‘health expectancy’;

• be able to prepare input data for the software;

• be able to  run the software; and

• be able to interpret the results

Description

The workshop is designed to combine both theoretical and practical sessions on the use of the software. There will be hands-on sessions that allow participants enable to run both software by themselves.

IMaCh and SPACE have been available for some time, and have been updated constantly. Both software apply multistate life table method to compute health expectancy and require longitudinal data.  Both software estimate transition probabilities among health states specified by users based on input data first. Then, they construct multistate life tables based on the estimated transition probabilities.

While IMaCh constructs multistate life tables and standard errors by using mathematical formulae, SPACE utilizes the simulation method to construct multistate life tables and the bootstrap method to compute standard errors.

The latest version of IMaCh is able to handle time invariant and time varying dummy variables as well as continuous explanatory variables to compute health expectancy and control for these variables at mean value.  IMaCh also can accommodate interaction terms in the model specifications.  However, there is one issue that remains unresolved with the latest version of IMaCh:  it requires that all transitions among living health states have to be observed. In other words, if there is no return transition between living health states, having a stroke as a living health state, for instance, IMaCh is unable to compute for health expectancy.

SPACE, on the other hand, is able to handle cases with no return transitions and other aspects of functions that IMaCh is also able to handle. In addition, the treatment of control variables is different from IMaCh. SPACE compute health expectancy by controlling for a variable or variables by computing the mean value of health expectancies computed for all possible values or all possible combinations of values of interest(s).  However, because SPACE is originally written in the statistical software package program SAS, users of SPACE are required to have access to SAS. In addition, SPACE takes a very long time to compute health expectancy partly because the bootstrap method is used to compute the standard errors. In order to overcome this weakness, the SAS version of SPACE was converted to C++ (partly by R) version. The C++ version of SPACE does not require access to SAS and shortens computation time 30 to 70 times faster compared to SAS-based SPACE.

Link

https://reves.site.ined.fr/en/resources/computation_online/imach/