trajectories of bp over time and annualized slopes

[javier rodriguez]

Hi all,

First, I would like to apologize because my question is probably a general question of biostatistics but I use to program on STATA and I absolutely need to have an advice. I am a little bit desperate I work on hypertension in a database with repeated measures of blood pressure every year.

I have been asked to "compute the trajectories of systolic blood pressure change using annualized slopes of the repeated measures of SBP together with the variability in SBP between visits".

I don't really understand what this sentence means by trajectories and annualized slopes.

Could you please help me to better understant what I am supposed to do?

Thank you so much +++ for your help.

JR

[Clyde Schechter]

Well, I can think of several things this might mean, and a few ways to approach it in Stata. You probably have to go back to the person who tasked you with this and ask him or her to clarify what the intent was. But if you want to go back to that person with something to show that you've tried, and perhaps get told that it isn't what was wanted, you could do something like this.

I assume your data contains a patient identifier variable, id, a systolic blood pressure variable, sbp, and a date variable, date (which must be a genuine Stata internal format date variable, not a string that looks like a date or a number like 20181220.)

by id (date), sort: gen annualized_slope = 365.25*(sbp-sbp[_n-1])/(date-date[_n-1])
mixed annualized_slope date and maybe some covariates here || id:
margins, at(date = (some interesting range of dates))
marginsplot

The variability of blood pressure between visits would require a somewhat different approach:

mixed sbp || id:

The output for var(Residual) will give you the within-patient (i.e. across visits) variance of sbp.

As I said, I can think of many variations of this. The trajectories might be based on age rather than calendar date, or perhaps on elapsed time since accession to the database. The relationship with whatever the time variable is might be non-linear. All sorts of covariates might be brought into the mix. But this at least is a starting point for you.

[javier rodriguez]

Thank you very much Clyde. I will start with thinking about that and I get back to you. Again, THANK YOU so much +++.

Javier