(196) PRE2DUP-R, open source program for assessment of PREscriptions to Drug Use Periods

Background: In pharmacoepidemiological studies one of the most important steps is to estimate time of exposure. PRE2DUP method to evaluates drug use periods from drug dispensing data, and free, shareable version would benefit researchers in the field.

Objectives: The aim was to translate and further develop the PRE2DUP method with Statistical Programming language R (PRE2DUP-R) and compare results against the original version

Methods: PRE2DUP estimates drug use periods for each person and for each drug (Anatomical Therapeutic Chemical (ATC) code) from drug dispensing data. Method uses information of purchased drug packages and calculates expected refill lengths based on sliding averages of daily DDDs and user’s personal variation for each ATC.
The main elements of PRE2DUP are used in PRE2DUP-R, main difference being that in the original program the whole drug use history for the person was taken into account in exposure estimation, whereas in PRE2DUP-R history is limited by researchers’ decision on maximum allowed gap between connectable purchases.
Methods were applied to Finnish drug purchases data. The agreement between PRE2DUP and preliminary version of PRE2DUP-R was tested regarding the number of periods and total exposure time for each person and ATC via correlation and correspondence plot. Differences were analyzed with Bland-Altman (B-A) plot followed with mean and standard deviation. The agreement was checked overall and by ATC-classes A, B, C and N.

Results: Dataset included 21,935,199 records for 102,197 persons from years 1994 to 2017, with 358 ATC codes and 6,115 different drug packages. Most common ATC class was N (nervous system) covering 74.5% of purchases, followed by C (cardiovascular system, 20.1%). PRE2DUP-R produced 6.3% more exposure periods.
The amount of the same number of drug use periods per person and ATC was 971,836 (79.8%), with correlation of 0.879 (95% CI 0.879, 0.879) and B-A difference of -0.118 (-2.507, 2.271). Correlation ranged from group N 0.897 (0.896, 0.897), and B-A -0.110 (-2.435, 2.215) to group C 0.799 (0.798, 0.801), B-A -0.152 (-2.915, 2.611).
For the total exposure time in days per person and ATC, correlation between versions was 0.998 (0.998, 0.998), B-A -16.105 (-212.985, 180.775). Correlation ranged from group B 0.999 (0.999, 0.999), with B-A -24.435 (-181.310, 132.440) to group A 0.998 (0.998, 0.998) with B-A -13.341(-205.960, 179.277).

Conclusions:

Conclusion: Agreement between PRE2DUP and PRE2DUP-R was good.