A sensor episode and a calendar day are different units
Barnett and colleagues used transdermal alcohol monitoring in a small randomized study of incentives for reducing drinking. An episode that started on one date sometimes continued past midnight. The investigators reported their main outcome with the next day counted both ways, depending on the analysis.
The counting rule changed the percentage
During the intervention weeks, the contingent-incentive group averaged 54.3% of days with no drinking detected when crossover days counted as drinking days, and 65.7% when the second day was excluded as a drinking day. The corresponding percentages in the comparison group were 31.2% and 47.7%. The paper’s two rules change the day-level summary while using the same sensor observations.
Read the event record before counting incidents
For a monitoring review, distinguish the number of continuous episodes from the number of calendar dates touched by their curves. Check the recording time zone, the event start and end, and the rule used to summarize days. An episode that crosses midnight does not, by itself, establish a separate drink on the second date.
Leave a Reply