Validity and utility of community health workers' estimation of kava use

OBJECTIVE: Estimating illicit substance use in epidemiological studies is challenging, particularly across ethical, cultural and language barriers. While developing the methods for a case-control study of the effects of heavy kava consumption among Aboriginal people in remote Northern Territory (NT), we examined the validity and utility of alternative methods for estimating exposure. METHODS: We assessed the level of agreement between a consensus of Aboriginal health workers in two different communities using interviews conducted with community members and health workers and individuals' self-reported kava consumption. Exposure measures included history of kava use, current kava use and history of heavy use. Agreement between a health worker consensus classification and individuals' self-report was analysed and agreement among several health workers in a consensus classification without self-report was assessed. RESULTS: Health workers concurred about an individual's history of kava use (k=0.83), current use (k=0.43) and also level of use (k=0.33). There was very good agreement between health workers' consensus and self-reported history of use (k=0.77). Agreement among health workers about current kava use was poor (k=0.08), while there was fair agreement between health workers and self-reported heavy kava users (k=0.36). Data from review of clinic patient notes supported agreement between consensus classification and self-reported history and level of use (k=0.39). CONCLUSIONS: Self-reported kava use may be a poor estimate of current use especially when obtained from interviews away from a confidential clinic setting. Consensus classification by knowledgeable Aboriginal health workers provided comprehensive coverage, efficiently and with greater reliability and assisted to identify 'excessive' kava use.