148 Add EQ5D as a person attribute

[andrewbaxter439]

What

• Maps SF12 MCS and PCS values (dhe_mcs and dhe_pcs) to EQ5D utility index
• Adds PersonTest.java file as tests for this calculation (and placeholder for further Person tests)

Why

• Creates the EQ5D value for quality-adjusted life-years calculations
• This seems to show a logical route for getting testing working!

[dkopasker]

Why is the variable called deq5d? I expected this to be a health variable, so the naming convention would suggest heq5d. the d prefix would suggest a dummy or demographic variable. The same point is relevant to the mcs and pcs variables.

Do we want non-zero values on the diagonal of the VCM? This breaks the link to the SF-12 component scores where parameter uncertainty has already been captured. This process is a simple conversion of an estimated outcome, not a new estimation. It is worth keeping the VCM and the code structure to allow other users to make a different decision, but I would suggest a zero matrix as the base.

I would suggest that Erik and one of the Essex team are better placed to review the tests. Great that they have been added.

[andrewbaxter439]

Why is the variable called deq5d? I expected this to be a health variable, so the naming convention would suggest heq5d

Good point. I had copied other similar variables such as dhm for Demographic-Health-Mental. Can change it to suit, though it seems that there aren't any other variables (health included) which start with a 'h'?

Also, if it's helpful to be more verbose in internal variable and method naming, is a fuller name of HealthRelatedQolEQ5D or QolEQ5D helpful at various points?

[dkopasker]

I would suggest that dhm is poorly named and we should not repeat this convention throughout the code. It is likely that more health outcomes will be added in subsequent project, so dealing with this issue now seems prudent. Ideally, we would change the dhm, dhm_ghq, and dhe names, but this may be risky given how widely they are used in the code.

My suggestions would be:
he_ghq - for GHQ Likert score (currently dhm)
he_case - for GHQ caseness (currently dhm_ghq)
he_mcs - for SF-12 mental component score
he_pcs - for SF-12 physical component score
he_wellb - for life satisfaction
he_sr - for self-rated health (currently dhe and widely used in the code)
he_mort - for mortality risk

[dkopasker]

Review left in conversation tab.

[andrewbaxter439]

For renaming, my suggestion would be to then rename these as he_eq5d throughout this pull request and open another issue for renaming the rest (for deeper testing). Some of the complications would be redoing initial populations also, these would be breaking changes?

[igelstorm]

I've made some specific suggestions, but broadly this looks good!