The reliability of drug use indicators collected from a prisoner sample using the life events calendar method

Abstract

Some methods of retrospective data collection are known to produce reliable indicators of drug use, but the Life Event Calendar (LEC) method has not been convincingly tested. As well, previous assessments of reliability typically focus on treatment program participants or others that are not representative of criminally active populations, and address one or two substances and time frames ranging from just a few days to twelve months. Our aim is to address those gaps by examining the test-retest reliability of self-reported drug use among prisoners across a wider variety of substances and over a longer reference period than previous research. Accordingly, we analyzed data that were collected using the LEC method during 2005–2007 from a random sample of 110 minimum and medium security prisoners. Drug use over an eighteen month period preceding the arrest that led to their imprisonment was conveyed. On average, the main or first interview (test) was completed 162.4 days (SD = 62.47) following prison admission, and the retest occurred three weeks following the main interview (test). Reliability is assessed using gamma and kappa. Self-reported drug use was reliable across a variety of substances, although reliability did vary. Reliability was stronger among African American, more highly educated, and older subjects. Yet, self-reported drug use among white, less educated, and younger subjects was acceptably reliable. Considering the lengthy lag between admission and first interview, this study provides strong evidence that the LEC method is suitable to collect reliable self-reported data on multiple forms of drug use from prisoners.

Keywords:

• Life event calendar method
• reliability
• self-report
• timeline followback
• retrospective data collection
• substance abuse
• prisoners

Disclosure statement

The authors have no conflicts of interest to report.

Notes

1 Self-reports are perhaps most advantageous when utilized in conjunction with the results of urine testing.

2 We examine test-retest reliability of drug use that occurred during an eighteen month calendar period. However, readers should bear in mind that, because the first interview occurred, on average, six months after prison admission, the actual reference period of the study is closer to twenty-four months.

3 Readers should note that the mean age at interview (reported in Table 2) should be greater than mean age at admission (reported in Table 1). This discrepancy occurs because the data used to calculate mean age in Tables 1 and 2 are derived from different data sources. The former (Table 1) is derived from an ODRC database and is calculated based on birth year, month, and day. The latter is derived from the survey interview which, for confidentiality reasons, only collected birth year and month.

4 The Wilcoxon signed-rank test and the z-test statistic assessing kappa differences between groups in Table 2 do not account for clustering.