Instructions for Running SAS Program

Modify the MACRO variables at the top section of the program.    From here on, you should be able to run the rest (step by step) without further modifications.  

a.       Clean/format the input dataset according to CDC Opioid Overdose Indicator Support Toolkit (limit to residents with acute care hospitalizations that did NOT have discharge diagnosis of death). 

b.      Check data quality

c.       Identify/tag drug poisoning cases according to ICD9/10CM definitions for H1/H2/H3.  Restrict dataset to drug poisoning cases only.  

d.      Create 3 time series datasets (1-combined, 2-stratified by age group, 3-by intent), generate quarterly counts for H1/H2/H3 à export Quarterly count excels

e.      Using the COMBINED time series dataset: Forecasting (perform regression using the pre-transition data points and predict the post-transition values of H1/H2/H3 based on pre-transition trend)

f.        Using the COMBINED time series dataset: Interrupted time series analysis (Fit a regression model that quantifies slope and level changes in the expected number of H1/H2/H3 per quarter) à export Trend excels

 

As a side note: steps 5 and 6 may require using PROC GENMOD / PROC PLM if a jurisdiction has a very low count for H3: Heroin overdoses, such as Montana. (This alternative approach is included in the code but commented out. Jurisdictions with low H3 counts will have to take extra care to run the commented code and NOT run the %create_forecast or %exporttrends macro’s for H3). Also, When exporting PROC GENMOD parameter estimate tables to excel, it is impossible to include the variable labels as a column. So I apologize but I couldn’t figure out another way to get the labels included in the excel. Take a look at the difference between H1TRENDS parameter estimate tables and H3TRENDS and you will see what I mean.