How to calculate 95% confidence interval for mortality rate?

Reply
Super Contributor
Posts: 543

How to calculate 95% confidence interval for mortality rate?

Dear All.

I have a calculated mortality rate (see code below) and would like to calculate the 95% confidence interval for the .

A little bit about the data:

We have end of study reason (eosreas = 1 means died), we have enrollment date, end of study date. So, I thought (and I could very much be wrong) to calculate the mortality rate, I needed to count how many deaths, what was the total person-time in the study, then used the counts*1000 / total time to get the mortality rate.

proc sql;

    create table cnt as

    select abs(enrdt - eosdt)/365.25 as time_in_study,

           (eosreas = 1) as death_event,

           sum(calculated death_event) as dth_cnts,

            sum(calculated time_in_study) as total_time,

           ((calculated dth_cnts * 1000) / calculated total_time) as mortality_rate,

           from my_data;

quit;

OK, so my question now is can I calculate a 95% confidence interval for this mortality rate? And if yes, how?

Thank you for reading.And for helping Smiley Happy

Anca.

Super User
Posts: 19,789

Re: How to calculate 95% confidence interval for mortality rate?

Posted in reply to AncaTilea

I wonder if you should be doing a survival calculation and looking for the opposite of the survival rate, the mortality rate and the corresponding 95% confidence of that, since I'm assuming not everyone in the study died.

Proc lifetest or phreg would then be things to look into.

Super Contributor
Posts: 543

Re: How to calculate 95% confidence interval for mortality rate?

Fareeza, thank you.

Would the opposite of survival rate be the hazard rate? which is not the mortality rate. I apologize for my ignorance, but 'rates' are hard for me to understand.

Smiley Sad

Super User
Posts: 19,789

Re: How to calculate 95% confidence interval for mortality rate?

Posted in reply to AncaTilea

1-survival rate is the mortality rate or failure rate Smiley Happy.

SAS has options to plot both the survival plot and failure plot and you might be able to tweak the proc to output the other, but I don't recall.

The other option is to reverse your coding (0=death, 1 = otherwise), quick and dirty but works!

Super Contributor
Posts: 543

Re: How to calculate 95% confidence interval for mortality rate?

Hi Fareeza.

My senior statistician stopped by and I was able to ask her and this is what I got:

proc sql;

    create table cnt as

    select abs(enrdt - eosdt)/365.25 as time_in_study,

           (eosreas = 1) as death_event,

           sum(calculated death_event) as dth_cnts,

            sum(calculated time_in_study) as total_time,

           ((calculated dth_cnts * 100) / calculated total_time) as mortality_rate,

            (sqrt((calculated dth_cnts *100) / (calculated total_time**2))) as stdErr_dth

     from my_data;

quit;

The bolded line is derived from the following:

Let # of deaths = D

and person time at risk = PT

estim_mort_rate = D / PT

so, var(estim_mort_rate) = (estim_mort_rate)^2 / D, which is

(D/PT)^2 * 1/D or

D/ PT^2

So, the StdErr is sqrt(# of deaths/ person time at risk squared)

Bam!

Thank you so much for the help!

Smiley Happy

Ask a Question
Discussion stats
  • 4 replies
  • 2693 views
  • 6 likes
  • 2 in conversation