Conditional mean

[wj2]

Hello Community,

I have a data analysis problem that I was wondering if I could get some help with. I am analyzing some healthcare data, which includes the dates that a certain medication was prescribed (Date), the number of days of each prescription since the initial prescription for each patient (Days), and the number of days between subsequent prescriptions (Days_bwn).

I need to calculate the average number of days between prescriptions received within 180 days of the first prescription (Days le 180). HOWEVER, for patients who only had ONE prescription within 180 days, the first prescription received after 180 days of the first prescription will be used to calculate the average time between prescriptions (see patient_ID 3 for example). Could someone please help me with the code to produce the desired output as shown below? Please let me know if I can provide any clarification or additional information. Thank you!!

Have:

Patient_ID	Date	Days	Days_bwn
1	15DEC2019	0	.
1	30JAN2020	46	46
1	01MAR2020	77	31
1	15APR2020	122	45
2	26DEC2019	0	.
2	10JAN2020	15	15
2	03MAR2020	68	53
2	18MAY2020	144	76
2	26JUN2020	183	39
2	18JUL2020	205	22
3	30AUG2017	0	.
3	29MAY2018	272	272
3	07JUN2018	281	9

Want:

Patient_ID	Mean
1	40.667
2	48
3	272

[Reeza · Accepted solution]

This does not deal with the case of having multiple prescriptions on the same day - not sure how you'd want to account for that. 

data inclusion;
set have;
by patient_id date;
if first.patient_id then counter=0;
counter+1;
if counter<=2  then inclusion=1;
else if days <180 then inclusion=1;
run;

proc means data=inclusion;
where inclusion=1;
class patient_id;
var Days_bwn;
run;