Suggestions for modelling

[Priyanka Pandhi]

Hello everyone,

I am looking for suggestions on a very difficult question I think, I have.

Background: We have 5 types of drugs which are given in varying combinations over 6 hours. I have hourly data. All of the five drugs have been assigned standarized into 1 cumulative dose per hour (std_med).

Aim: To check if different combination of phenotypes based on 5 drugs exists that improve survival at discharge.


Following is the sample of the data

input long ID double(med1 med2 med3 med4 med5 std_med) byte Dead
27 0 0 0 2.4499999999999993 0 2.450000036507845 0
27 0 0 0 2.9999999999999973 0 3.0000000447034836 0
27 0 0 0 2.9999999999999973 0 3.0000000447034836 0
27 0 0 0 3.259999999999997 0 3.2600000388920307 0
27 0 0 0 2.9999999999999982 0 3.0000000074505806 0
27 0 0 0 1.7700000000000014 0 1.769999960437417 0
27 0 0 0 .02 0 .019999999552965164 0
28 7.999799999999998 0 0 0 0 7.999800145626068 1
28 7.999799999999998 0 0 0 0 7.999800145626068 1
28 7.999799999999998 0 0 0 0 7.999800145626068 1
28 7.999799999999998 0 0 0 0 7.999800145626068 1
28 7.999799999999998 0 0 0 0 7.999800145626068 1
28 7.999799999999998 0 0 0 0 7.999800145626068 1
28 .13333 0 0 0 0 .13333000242710114 1
35 0 0 .022499999999999996 0 0 .02249999961350113 1
35 0 0 .026999999999999986 0 0 .026999999536201358 1
35 0 0 .026999999999999986 0 0 .026999999536201358 1
35 0 0 .026999999999999986 0 0 .026999999536201358 1
35 0 0 .024659999999999987 0 0 .02465999950072728 1
35 0 0 .007109999999999992 0 0 .0071100000641308725 1
end


I don't know what kind of analysis would help. I would really appreciate if anyone can help me approaching this problem.

Thanks,
Priyanka Pandhi

[Marcos Almeida]

I believe your data set, being just a toy example, presents some lack of consistency. Exampple, no value for med2 and med5, the dosage seems to be the same for some IDs, etc.

That said, maybe you wish something like:

gen med = .
replace med = 1 if med1 !=0
replace med = 2 if med2 !=0
replace med = 3 if med3 !=0
replace med = 4 if med4 !=0
replace med = 5 if med5 !=0
xtset ID
xtlogit Dead std_med i.med

The output with this example with be preposterous, on account of the above-mentioned problems.

Hopefully this code works in the "real" data.

[Priyanka Pandhi]

Thanks, Marcos for your reply. Yes, it is not necessary that a patient has gotten all the drugs. Combinations could be varying so, therefore, you would find some missingness.