"Overall, extremely few interaction terms were significant, broadly indicating that drug courts were comparably effective for all types of drug-involved offenders. Three exceptions were as follows. The drug courts were especially likely to produce a reduction in criminal behavior among offenders with a history of violence—indicated by a self-reported prior violent conviction (p < .001). On the other hand, drug courts were especially unlikely to produce a reduction in criminal behavior among offenders with narcissistic personality (p < .05) and among black offenders (p < .05). None of 14 other tested interactions was significant in either direction. As a follow-up analysis, we had contemplated developing a 'risk' classification, enabling us to draw general conclusions as to whether drug courts work particularly well with high- or low-risk offenders. However, because so few of the interaction terms were significant in our initially parsimonious three-variable models—and of those that were significant, only prior violence was readily classifiable as connoting “risk,” we considered it pointless to take that next step."

Source

Rossman, Shelli B., et al., "Final Report, Volume 4: The Multi-Site Adult Drug Court Evaluation: The Impact of Drug Courts" (Washington, DC: Urban Institute, June 2011), p. 75.
http://nij.gov/topics/courts/…
https://www.ncjrs.gov/pdffile…