Key findings
- Given the importance of reporting systems such as the IDRS in detecting trends over time it is vital that valid trend inferences are drawn from the data.
- The impact of excluding previously participating respondents in statistical analyses of trends in drug use over time is minimal when analysing the full national data. However, it is substantial when analysing trends in drug use at the jurisdictional level.
- A method is available (and syntax is provided) to correct for the correlated nature of the data which provides a more conservative and accurate estimate of the variance around estimated trend coefficients.