association between two variables

[Kyra]

Hi,

I am trying to see if there is association between number of pills prescribed after surgery (exposure) and number of pills consumed (outcome).

My dataset looks like below. I want to adjust for age, gender, prior medicine use.

Please let me know hoe should i do that.

I cannot use linear regression and and am not sure if i should use logistic regression or not considering 9 levels of my outcome variable. Thank you very much in advance.

Age	Gender	Prior medicine Use	Number of pills consumed	Number of pills prescribed
69	0	Yes	1 to 4	15
79	0	Yes	0	20
47	0	No	5 to 9	20
70	1	Yes	1 to 4	30
51	1	No	0	10
64	0	No	5 to 9	10
81	0	Yes	0	20
60	0	No	15 to 19	30
86	1	No	1 to 4	8
70	0	Yes	30 +	30

[ballardw]

Do you have the actual number of "pills consumed" available instead of the category? Is that actually a character variable or an ordinal (0,1,2 etc.) with a format to display the range each value represents?

How many records do you have? Since Age potentially has a range on the order of 1 to 100 then adjusting for age, gender and prior medicine use potentially has about 400 combinations. So there could be some issues about having enough data in each combination. Or do you intend to group age by age ranges?

Could you provide an example of the sort of question(s) your analysis should answer? The type of answer you want may indicate a choice between approaches.

[Kyra]

Thank you very much for the reply.

I do not have the actual number of pills consumed. It is in categories. If patient consumed 3 pills he answered 1-4.

Our question, number of pills consumed had 9 levels - The patient picked from them.

I have 100 surveys right now and we are still collecting. I can group age into categories if required.

The question i intend to answer is:

Does increased prescription size leads to increase in medicine use? Is there a pattern between prescription and usage of medicine.

NumberPills consumed

0

1 to 4

10 to 14

15 to 19

20 to 24

25 to 29

30+

5 to 9

unknown