Syndromic surveillance can be used to assess change in drug abuse rates and to find regions in which abuse is most common. This paper compares the power of three syndromic surveillance procedures (a paired-sample test, a process control chart, and a conditional autoregressive model) for detecting change in opioid drug abuse patterns, using data from two reporting systems (the OTP and PCC datasets). We find that the conditional autoregressive model provides good power and geographic information and that the OTP data carry the strongest signal.