Estimating mortality rate (date of death not available)

[Hatim Tareq]

Hi All,

I really need your help.

I would like to estimate the mortality rate before and after specific period. The problem is I don't have date of death. I only have registration date (the date where the patient has been registered) and the treatment outcome date (the outcomes are: death, cured, completed and defaulted). I also have the age of each patient.

What I did, is simply divided the number of those who died by the total number of those who has treatment outcome (I did that by age group and sex), but I think that, I should first estimate the person-time at risk, but really don't know how to do that. available

Any help will be appreciate it.

Regards,

[Carlo Lazzaro]

Hatim:
is there any chance to calculate person-time as the difference between treatment outcome date and registration date?

[Svend Juul]

If you are studying the prognosis of a disease with high initial mortality, and if you have a reasonable (but not precisely known) follow-up-time, a statement like this may still be informative, estimating not the mortality rate, but the mortality proportion within a certain follow-up time:

"Mortality within an average follow-up time of 90 days was 5.3%".

The average follow-up-time of interest for this statement is calculated from the follow-up-time to censoring (cured, completed, defaulted). With the information you gave, this may, or may not, be reasonable.

[Hatim Tareq]

Hi Carlo,

Both registration and outcome date are in the following date format (for example: 1960m4), I think I will be able to calculate the difference between them. Do I have to multiply by 30 (as approximate) to have the persons-time in months.what my Numerator and Denominator will be?


Regards,

[Hatim Tareq]

Hi Svend,

I am studying the effect of new policy (ART usage depending on CD4 count). So, I would like to know if the new policy has any effect on mortality.

Regards,

[Svend Juul]

If both the date of entry and the date of outcome (death) or censoring (cure, completed, defaulted) is recorded as months, your can, by subtraction, calculate the time-at-risk with a precision of months, which may be, if not perfect, then fair enough. If you want to express the rates per year, just divide the time-at-risk by 12. Here is an example of month as values and as presented by Stata with the %tm format:

. gen month2 = month1
. format month2 %tm
. list, clean
       month1    month2 
  1.        0    1960m1 
  2.        1    1960m2 
  3.        2    1960m3 
  4.       30    1962m7 
  5.       40    1963m5 
  6.       50    1964m3 
  7.       60    1965m1 
  8.       70   1965m11 
  9.       80    1966m9 
 10.       90    1967m7

You will benefit from, for example, the stset, strate, and stcox commands. For people dying in the same month as they entered the study, I would set the time at risk to 0.5 months, to avoid these observations to be dropped form the calculations.

[Hatim Tareq]

Thank you Svend.

I would like to be able to estimate the mortality by age-group, any help will be appreciate it.

Regards,

[Svend Juul]

Hatim:
In post # 1 you write:

The problem is I don't have date of death.

and in post # 4 you write:

Both registration and outcome date are in the following date format (for example: 1960m4),

Do you or don't you have the month of death?

[Hatim Tareq]

Hi Svend,

I don't have the date of death, but the treatment outcome date is available. The treatment outcome include if the patient is died or not.

Regards,

[Svend Juul]

But if the patient died, this is one of the possible outcomes, and in that case the treatment outcome month is the month of death. Right?

I also assume that you have a variable that tells what the outcome was, for example, death. Right?

Do you have a birthdate or birth-month? If not, at which event was the age recorded?

[Hatim Tareq]

Hi Svend,

Yes, I do have a DoB.

Regards,

[Svend Juul]

My first two items in post # 10 were about this: If you know the outcome (death, cured, completed, defaulted) and you know the month of the outcome, then you have the information needed to know the month of death among those who died while in the study.

Where am I wrong? What didn't you tell?