Processing transdermal alcohol concentration (TAC) data to detect low-level drinking.

<bold>Background: </bold>Several studies have objectively quantified drinking through the use of Alcohol Monitoring System's (AMS) transdermal alcohol concentration (TAC) device known as SCRAM CAM. Criteria that AMS uses to detect drinking are known to be conservative and only reliably detect heavy drinking equivalent to 5 or more standard drinks. Our group has developed Research Rules used to process TAC data in a manner that will detect low-level and moderate drinking even though it is below the AMS criteria for detection.<bold>Methods: </bold>Sixteen male and 14 female paid research volunteers wore TAC monitors for 28 days in their natural environments and responded daily to text message prompts to self-report the previous day's drinking. Current analyses describe the Research Rules that we developed and how use of those rules impacts the detection of self-reported drinking treated as the standard in sensitivity/specificity analysis.<bold>Results: </bold>We observed 606 occurrences of positive TAC events over a total of 867 days and processed the TAC data to retain 345 as possible drinking events, even though AMS criteria confirmed drinking for only 163 of these events. The kinds of TAC events removed or retained by our rules are illustrated as cases of low and moderate drinking days that were detected by our rules but not by the conservative AMS criteria. AMS-confirmed TAC events have a high specificity (99.8%) to detect primarily heavy drinking, but have a poor sensitivity to detect lower-level drinking and a poor specificity as an indicator of alcohol abstinence. In contrast, our Research Rules detected 100% of TAC events detected by AMS but also detected 31% of the lower-level drinking events not detected by AMS, with 91% specificity.<bold>Conclusions: </bold>Reliance upon the AMS criteria for alcohol detection affords a high specificity for detection of heavy drinking but is a poor indicator of abstinence rates. In contrast, use of our Research Rules provides more sensitive means to quantify either any drinking or low-moderate levels of drinking while still maintaining good specificity. [ABSTRACT FROM AUTHOR]