DUF Item Analysis Task Final Report

• NCJ Number: 194921
• Author(s): Johathan Sushinsky, Thomas Gray, Erich D. Wish
• Date Published: January 1998
• Length: 40 pages
• Agencies: NIJ-Sponsored

Annotation

This report reviews the current Drug Use Forecasting (DUF) interview instrument, so as to identify items that could be eliminated from the interview due to low response or the need to clarify the content and wording of items.

Abstract

Data from all DUF collection sites which used the new DUF instrument in 1995 and 1996 were reviewed for missing data, consistency among items, and out-of-range data. Juvenile data were analyzed separately. The review anticipates that the findings may be useful to the National Institute of Justice staff as they plan the Arrestee Drug Abuse Monitoring (ADAM) program. Missing data were found to be a problem for all multiple-charge responses and for many items recorded from police records. Since all of the DUF sites had similar problems, this report suggests two changes for the ADAM instrument. It recommends consistently coding data on missing charges to a 77 (NA/SKIP) when there are no additional charges. Also, if police records at most sites do not contain the data requested in the instrument, it may be best to eliminate those variables from ADAM. More than 3 percent of responses for the HIGRADE variable were missing at most adult sites and all juvenile sites. This issue should be assessed to determine why so many missing responses are being reported. The types of inconsistent and out-of-range responses indicate that there are few differences between sites. UNDERCOC and EMPLOY were problems at all sites, and most sites had some inconsistencies between who a respondent was living with and that respondent's marital status. Other variables that were inconsistent at some sites were drug-use variables, comparing use in the past 30 days with use in the past 72 hours, and injection of a drug compared to use of that drug. There were few out-of-range problems and only ORGMONEY showed a consistent pattern between sites. The same types of variables were problems at the juvenile sites. Skip patterns were a small problem for both adult and juvenile sites. Although mistakes were infrequent, the data entry program should be modified to automatically skip over questions (code 77) that should not have been asked. If this were done, the final database would not have any valid responses for questions that should have been skipped. 6 tables