Final report on the North Queensland Drug Court

Jason Payne
ISBN 0 642 53886 7 ; ISSN 1445-7261
Australian Institute of Criminology, 2005
(Technical and background paper series, no. 17)

Section 3 - Drug court outcomes

• 3.1 Reducing crime
  • The data
  • Measuring re-offending
  • The comparison groups
  • The follow-up period
  • Time to first offence (post-entry)
  • Adjusting for covariates
  • Time to first offence (post-exit)
  • Offending frequency
  • Comparative analysis
  • Issues and limitations
• 3.2 Reducing drug use
• 3.3 Improving participant health and social function
• 3.4 Post graduation issues
• 3.5 Discussion

The North Queensland Drug Court pilot program aims to deliver a number of important objectives. The Australian Institute of Criminology was specifically asked to examine whether the drug court program was successful in reducing:

1. the level of criminal activity associated with drug dependency;
2. the level of drug dependency in the community; and
3. the health risks to the community associated with drug dependency.

This section examines these important issues.

3.1 Reducing crime

Before examining whether the North Queensland Drug Court pilot program was successful in reducing crime, a number of important, albeit technical, elements must be established.

The data

Re-offending estimates used in this evaluation have been calculated from data extracted from QPS's centralised criminal history records. Although re-arrest measures are believed to underestimate the true frequency and volume of offending, there is a long tradition in criminology that relies on officially recorded offence data as a measure of offending (Coleman & Moynihan 1996). Despite its imperfections, official police records are the closest and most consistent measure of re-offending available to this evaluation. It is this same data that were used to measure the pre-referral criminal history in chapter one. Typically, for each offence committed the police record will indicate;

• offence type;
• number of counts - times charged with that offence on any one day;
• offence date - the date on which the offence was committed; and
• court date - the date on which the offence was heard in court.

Using the recorded offences and relevant offence dates, it is possible to determine which of the new recorded offences were committed during or after participation on the drug court program. Moreover, the arrest dates are used to calculate the number of days between events so that we may determine how long it took for an offender/participant to be re-arrested.

Finally, as a matter of consistency, offences occurring on the same date are coded and referred to as an offence episode. Episodic measurement is used for a number of reasons, not the least of which is the possible inconsistency in how individual offences are recorded. Aggregating to an arrest episode allows us to measure the unique intervals in which a criminal transaction occurs, and controls possible bias in police recording practices.

Measuring re-offending

Recidivism is the term used to define the act of re-offending and its measurement is complicated by a number of factors, not the least of which is the availability of suitable data. For the purposes of the present study, two measures have been developed:

1. the time taken to re-offend; and
2. the frequency of offending.

Note that the raw prevalence of re-offending is not included as a measure of recidivism in this study. Although prevalence rates are provided, they are reasonably uninterpretable without the consideration and application of a metric of time. Given that some participants were on the drug court program for longer than others, the opportunity to re-offend is therefore different for each participant. As such, longitudinal analytic techniques such as survival analysis are used to measure the occurrence of an event (in this case re-offending) as a function of the time an individual was available to be re-arrested.

Moreover, time to re-offending measures recognise that some drug dependent criminal offenders are significantly enmeshed in a criminal lifestyle prior to their contact with the drug court program. As such, total abstinence from offending, while ideal, is not a realistic outcome for all drug dependent offenders. Prolonging an offender's re-entry into the criminal market is, therefore, a positive outcome. Simply increasing the time it takes for an offender to re-engage in a criminal lifestyle can have significant financial and social benefits for the community.

The comparison groups

In order to determine whether drug court participants were less likely to re-offend, we must compare the re-offending results with a comparison group1. In North Queensland, random allocation to the drug court program was not attempted nor considered as a viable option for the delivery of drug court services. Such an option would have denied access to a viable population of offenders whom the program was designed to target and benefit. Because of this there is no randomised control group with which we can compare the recidivism results. In its absence, however, we can compare the relative recidivism rates of drug court participants with the group of offenders who were referred to the drug court program but refused to participate (n=36). For the most part, these offenders were otherwise eligible to participate but chose not to. Earlier descriptive statistics indicated little difference between these refused offenders and those who agreed to participate (see section one).

While the refusal group is the only comparable group of offenders, it is important to note that by refusing to participate an offender is either sentenced by the drug court magistrate or remitted back to the normal magistrate's court for sentencing. This final sentence may not lead to a period of incarceration if the magistrate does not believe it is necessary. Despite this, the pre-referral offending rates illustrated later in this report indicate very similar offending profiles in the 12 months prior to referral for both the drug court participants and the refusal comparison group.

Note that unlike in South East Queensland, this evaluation does not have prisoner comparison group - a group of prisoners identified as similar to drug court clients, but who did not participate in the drug court program (Makkai & Veraar 2003). This was because:

• the sample of participants on the North Queensland Drug Court program was too small to facilitate any detailed comparative analysis with adequate controls; and
• the drug court program was already experiencing difficulty in achieving its quota of participants, and any attempt to identify a reasonably sized comparison would prove futile.

Not having a prisoner comparison group limits our capacity to conclude that the drug court program has achieved 'real and comparable' reductions in offending. However, using a comparison group that is not comparable would also have the potential to bias the results. In this case, the probability of detecting a large enough sample of prisoners with the same characteristics was considered too low to warrant the identification of this comparison. Any attempt to widen the net would have resulted in the identification of a non-comparable comparison group and biased our interpretation of any resulting benefits. In this case, we err on the side of caution.

The follow-up period

For the purposes of measuring recidivism, recall that re-offending is calculated against a function of time. This is because each participant spent a varying amount of time on the drug court program and their re-offending rates can only be measured against the length of time in which they were observed. In the present evaluation, the follow-up period is measured as the number of days each subject was observable - for drug court participants this was the number of days between being admitted to their program and the evaluation date of 31 December 2004. For the refusal comparison group, the follow-up period is measured from the date of refusal to the evaluation date. Both the refusal date and the program commencement date represent the date at which each subject was determined by the court.

Table 3.1 illustrates the average number of observable days in the follow-up period for both the participant group and the refusal comparison group. Graduated participants were observed for an average of 503 days, terminated participants for 473 days and the refusal comparison for 352 days. The final category in Table 3.1 provides the average follow-up time for all participants of the program, including graduates, terminates and those currently participating. This group is representative of the general category of drug court participants and is consistent with Blenko (2002, p.183), who argues that 'it is important for drug treatment evaluations to analyse the outcomes for all drug treatment court clients, not just graduates'.

Note that the total number of observable days in the follow-up period is not indicative of the number of days each offender was free and able to re-offend. Some participants of the drug court program are incarcerated for varying periods of time, during which they are technically unable to engage in recidivist activities. An adjustment is needed to the follow-up period to account for the time an offender was incapacitated and unable to re-offend. To do this, the number of days spent in prison is subtracted from the total number of observable days. This results in a new estimate of the follow-up time, which is herein referred to as the number of free days each offender was able to re-offend.

To illustrate, we track two hypothetical offenders for 100 days, over which time each committed 10 new offence episodes. Without the adjustment for prison time, each offender is taken to have re-offended at a rate of 0.10 (or one offence episode every ten days). Imagine, however, that Offender A was apprehended and sentenced to 50 days imprisonment during this time, while Offender B was not. For Offender B, the offending rate remains at 0.10, while for Offender A, the 10 offences committed in the 100 observable days actually occurred within the 50 days that the offender was not in prison. The offending rate is therefore adjusted to 0.20 (or one offence episode every five free days), suggesting that Offender A's rate of recidivism is twice that of Offender B's.

Table 3.1 illustrates the average time each comparison group spent in custody during the observable follow-up period. For each group, prison time is calculated as the sum of:

• refusals - any post-refusal custody;
• graduates - any IDRO custody sanctions and any post graduation custody;
• terminated - any IDRO custody sanctions and the final sentence at termination, and any additional post termination custody; and
• current participants - any IDRO custody sanctions.

By comparison group, the final calculations reveal the averages for each group were:

• refusal comparison - 40 days in custody resulting in 312 free days;
• graduates - four days in custody resulting in 499 free days;
• terminates - 160 days in custody resulting in 328 free days; and
• for all IDRO participants combined - 66 days in custody and 313 free days.

Table 3.1: Post-entry follow-up times by comparison groups as at 31 December 2004

	Refused (n=36)	Graduated (n=24)	Terminated (n=45)	All participants (n=120)
Post-entry (mean days)
Days since program start (a)	352	503	473	373
Prison days since program start (b)	40	4	160	66
Free days since program start (a-b)	312	499	328	313
Source: Australian Institute of Criminology, North Queensland Drug Court Database [computer file]

Time to first offence (post-entry)

The first measure used to assess re-offending is the time to first offence. Table 3.2 provides the relative re-offending estimates for each of the comparison groups by offence type. The estimates indicate that at 31 December 2004, 48 per cent of drug court participants (all participants placed on an IDRO) had been apprehended and charged by the police for at least one episode of offending across all offence types. This compares with 53 per cent of offenders in the refusal comparison group. By offence type, the relative prevalence of post-entry offending was:

• any property offence - 21 per cent of drug court participants compared with 14 per cent of the refusal comparison;
• any violent offence (including robbery) - five per cent of drug court participants compared with zero for the refusal comparison;
• any drug offence - 14 per cent of drug court participants compared with 28 per cent of the refusal comparison; and
• any breach offence - 26 per cent of drug court participants compared with 42 per cent of the refusal comparison.

While it is possible for an offender to be apprehended by the police on a breach of bail, probation or other community corrections order, it is difficult to ascertain from the official police records whether this breach was in relation to a previously recorded offence, or whether the breach itself indicates a new offending episode. In addition, should a participant abscond from the drug court program or fail to appear as scheduled, the drug court magistrate may issue a bench warrant for their arrest. This bench warrant is the official mechanism through which police may lawfully apprehend a drug court participant without needing evidence or reasonable suspicion of fresh offending. The fail to appear does not constitute a new offence, but the recording of the bench warrant as a breach by the police will be coded as a new offence episode. The implications of this are twofold:

1. should a breach be related to a previous offence but given a different date, it will, if occurring alone on any single day, be recorded as a new episode of offending and may contribute to the double counting of offending episodes; and
2. should the police record a breach to facilitate the apprehension of a participant who has failed to appear, it will, if occurring alone on that day, also be recorded as a new episode of offending.

To overcome this possible bias, this evaluation presents two any offence episode measures in Table 3.2. The first measures any day where any offence (including a breach) is recorded. The second calculates all offence episodes, less the number of episodes where a breach was the only offence recorded on that day. The resulting differences in the prevalence estimates of re-offending are significant, reducing by 14 per cent for the refusal group, 13 per cent for the graduates and eight per cent for all participants of the drug court combined.

On a final note, it is impossible to say which of these estimates is the most accurate in determining the level of actual re-offending, so we use the point estimates to conceptualise the upper and lower limits of re-offending across each comparison group.

Table 3.2: Post-entry re-offending as 31 December 2004 (per cent)

	Refused (n=36)	Graduated (n=24)	Terminated (n=45)	All IDRO participants (n=120)
Post-entry
Any offence episode*	53	42	78	48
Property offence episode*	14	13	36	21
Violent offence episode*	0	0	13	5
Drug offence episode	28	17	24	14
Breach offence episode	42	33	42	26
Any offence episode (excluding breach only)*	39	29	69	40
*Statistically significant at p<0.01 between refused, gradated and terminated. Source: Australian Institute of Criminology, North Queensland Drug Court Database [computer file]

So far the re-offending prevalence rates indicate little or no difference between the drug court participants, and the refusal comparison group - with re-offending estimated at between 40 and 50 per cent. By offence type, it appears that drug court clients were slightly more likely to have committed a property or violent offence, and less likely to have committed a drug offence.

These raw prevalence rates are provided for illustrative purposes only and are relatively meaningless unless measured against time. The longer an offender is observed, the greater the number of days in which their offending can be measured. Further, while the prevalence rates indicate that 52 per cent of drug court participants had not re-offended by 31 December 2004, it is impossible to determine whether they would have offended should the evaluation have followed them for longer.

Survival analysis is a statistical technique used to model event occurrence (in this case re-offending) from longitudinal data. The technique measures the risk of re-offending at any time point, as experienced by a randomly selected offender should they not have previously re-offended. The technique allows us to account for two important factors: that the actual survival time (time to first offence) for those who had not re-offended before 31 December 2004 is unknown; and that the follow-up times for each offender can vary (Singer & Willett 2003).

Figure 3.1 illustrates the survival function for the number of free days to first offence (any offence). The comparative curves illustrate the survival experience for the 120 IDRO participants and the refusal comparison group. The two curves can be interpreted as the proportion of each sample that will not have re-offended (survived) at any given time interval. A line drawn at 50 per cent helps to further quantify and understand the survival experience between the two groups. Where each survival curve crosses the 50 per cent mark, it is at that point in time (or after that number of free days) when half of the sample will have re-offended. In this case, 50 per cent of the refusal comparison group will have re-offended within 100 free days. This compares with the IDRO participants of whom 50 per cent will have re-offended between 300-350 days.

Although the survival cures appear to over-estimate the prevalence of re-offending, their values at any time point are drawn from a statistical calculation of the probability of re-offending. For example, the prevalence rates in Table 3.2 estimate the prevalence of re-offending for all IDRO participants at 42 per cent. However, the survival function estimates that after 650 days, more than 80 per cent of IDRO participants will have re-offended. At any time point, the survival function represents the estimated proportion of each group who will have survived. As we move further forward in time, the number of observable subjects decreases and our estimate becomes less reliable. Further discussion about the reliability of the survival estimates is provided later.

Tests of significance are used to determine whether the differences seen in Figure 3.1 are the result of real differences between the IDRO participants and the refusal group. The value of a statistical test (which varies between zero and one) tells us the probability that the observed differences are due to chance or error. The higher the test value (the closer it is to one), the greater the probability. Conventionally, the maximum probability level for determining a significant difference is set at p=0.05 (DeVaus 2002). In other words, we can only be confident that the differences observed are real differences, and that should the probability of error or chance be less than five per cent.

For the present survival analyses, two tests of significance are used - the Log-Rank Test of Equality and the Wilcoxon Statistic. Further information about the differences between these tests can be found elsewhere (Hosmer & Lemeshow 2003), but in brief, the Log-Rank test weights the survival data to the end of the survival curve while the Wilcoxon Statistic weights closer to the beginning. The Log-Rank test will, if significant at p<0.05, indicate that the differences in the final lifetime survival rate between the two groups are significant. Alternatively, the Wilcoxon statistic will indicate whether the initial survival experience between both groups is significantly different, despite the possibility that the final lifetime survival rate may not be. For the time to first offence data presented in Figure 3.1, both tests are insignificant, suggesting that we cannot confidently say that the survival experience of IDRO participants was better (or worse) than the refusal comparison group.

Similar survival analyses are presented in Figures 3.2 and 3.3 for both time to first property offence and time to first drug offence. In terms of property offending, similar results were indicated with no significant difference in the survival experience of IDRO participants and the refusal comparison group. For drug offending, however, there appears to be a significant short term effect, with fewer IDRO participants being re-arrested for a drug offence in the early stages of their participation on the drug court program (Wilcoxon X²=5.24, df=1, p=0.02). Note that overall lifetime survival for IDRO participants is not significantly different (Log-Rank X²=3.26, df=1, p=0.0629), suggesting that the benefits of the drug court program in reducing drug related offending are short lived. This is indicated where the survival curves intersect at approximately 650 days.

So far we have examined the relative survival experience of all 120 IDRO participants placed onto the drug court program by 31 December 2004. These results have indicated that, overall, there is no discernable difference between those who participated (or were participating) and those who refused. While Blenko (2002) argues that drug court evaluations should look at the aggregate results for all program participants, it is also important to consider the relative survival experience of sub-groups of participants - namely comparisons between terminated and graduated participants. This is for a number of reasons, but most importantly both the South East Queensland and New South Wales Adult Drug Court evaluations illustrated a termination effect, where terminated participants performed less favourably than both the graduates and the comparison groups.

Analysing the comparative survival experience for graduates and terminates has received some criticism. The critics argue that the mere fact that graduates do better than terminates is self evident and does not indicate that the drug court program is effective. On the other hand, the drug court program was not designed to work for every offender, and the termination criteria are designed to remove participants whose drug use and criminal behaviour are not amenable to community based rehabilitation. The court clearly states that should a participant fail to meet the requirements of the drug court rehabilitation plan, they will be terminated and returned to prison. In this light, others argue that by aggregating the survival experience of all participants together, one might conclude that there was no program effect. However, these results are based on the combined survival experience for terminated participants for whom the drug court was never expected to succeed.

Despite these continuing debates, comparative analysis between graduates and terminates allows us to examine whether the drug court program is effective in reducing crime for those who are compliant. Recall that the bi-variate analysis (see Table 2.3) indicated very few differences between graduates and terminates on the demographic, drug use or criminal history variables. This indicates that at the time of referral there were few discernable factors which could reliably predict whether a participant would or would not be compliant. The comparative analysis helps us to quantify the impact of the program should a participant comply with the requirements of their IDRO.

Figure 3.4 illustrates the comparative survival experience for the time to first offence (any offence). It demonstrates that graduates of the drug court program were not only less likely to have re-offended by 31 December 2004, but that they also took significantly longer to do so. By comparison group, the number of days taken for half of all offenders to re-offend (median lifetime) was:

• 634 days for graduates;
• 71 days for terminates; and
• 91 days for those who refused the drug court program.

Consistent with South East Queensland, terminated participants were more likely to re-offend, and re-offend sooner, than the graduates or the refusals (Makkai & Veraar 2003).

By offence type, the comparative survival functions illustrate that graduates of the North Queensland Drug Court program also took longer than terminates to commit their first property offence and drug offence. In terms of drug offending, the early survival experience of graduates is markedly better than both the terminated and refusal comparison groups, despite no significant difference in the long term survival probability (Log-Rank X²=4.57, df=2, p=0.1018).

Adjusting for covariates

The survival analysis presented thus far has assumed that the comparison groups are not biased in any way. That is, we assume that no other covariate factors have influenced the survival experience of each group, and that the actual differences in the survival functions are attributable to the offender's participation or non-participation on the drug court program. It is easy to conceive, however, that drug court participation is not the only factor which may result in more favourable re-offending outcomes.

To examine this, we use a Cox Proportional Hazards Regression Model to predict whether any other offender characteristics are linked to a reduction in re-offending. More importantly, we use the regression modelling techniques as a means to confirm that participation on the drug court program was at least one of these linking factors. For each of the survival functions presented in Figures 3.1 to 3.6, we posit a Cox Proportional Hazards Model with the following predictor variables:

• age at referral to the drug court (a continuous variable centred at the mean age of 29 years);
• gender (where 1 is male and 0 is female);
• court location (where 1 is Cairns and 0 is Townsville);
• prior imprisonment (where 1 is prior imprisonment and 0 is no prior imprisonment);
• prior offences (a continuous variable centred at the mean number of 40 prior offences);
• referring offences (a continuous variable centred at the mean number of 14 referring offences); and
• refusal (where 1 is for the refusal group and 0 is the IDRO group).

The regression results indicate that for the time to first offence (any offence):

• IDRO participants as a combined group were no more or less likely to re-offend than the refusal comparison group;
• by sub-group, graduates, was less likely to re-offend than terminates and refusals. Terminated participants were four times more likely to re-offend than graduates; and
• the only other predictor of an increased likelihood to re-offend (any offence) was the age of referral, with younger offenders being at greater risk.

In terms of property offending:

• there was no discernable difference between all IDRO participants and the refusal comparison group;
• by sub group, there was no difference in the survival experience between graduates and refusals, but terminated participants were four times more likely to re-offend with a property offence; and
• the only other predictor of re-offending was the number of prior lifetime offences, with the risk of re-offending increasing for offenders with a greater number of prior offences.

In terms of drug offending:

• there was no discernable difference between all IDRO participants and the refusal comparison group;
• by sub group, graduates were less likely to commit a post-entry drug offence than were terminates and refusals. Terminated participants and the refusal comparison group were four times more likely to re-offend than were graduates; and
• a number of other factors emerged as being linked to re-offending. The risk of re-offending was decreased for females, but increased for younger offenders with a history of prior imprisonment.

Time to first offence (post-exit)

The follow-up period used in this report so far has been measured from the date of IDRO commencement to 31 December 2004. This means that our measurement of re-offending has included any offence episode which occurred during the time in which a participant was actively engaged with the drug court program. While it is expected that re-offending will be reduced during this time, it is also expected that the drug court program will deliver cumulative benefits to its participant population. In other words, the impact of the drug court in reducing re-offending is maximised at the time of successful graduation - some 300 days after program placement.

Table 3.3 illustrates the average number of days each comparison group was observable after the completion their IDRO, as well as the prison and free days calculations. Note that all time estimates for the refusal comparison group remain the same as the post-entry estimates used earlier. This is because the refusal comparison group did not enter a program. For graduates, the average number of free days was 173, while for the terminated participants it was 157.

Table 3.3: Post-exit follow-up times by comparison groups as at 31 December 2004

	Refused (n=36)	Graduated (n=24)	Terminated (n=45)	All participants~ (n=69)
Post-exit (mean days)
Days since program exit (a)	352	174	279	243
Prison days since program exit (b)	40	1	122	34
Free days since program exit (a-b)	312	173	157	163
~Sample includes offenders having terminated or graduated (n=69). Source: Australian Institute of Criminology, North Queensland Drug Court Database [computer file]

Table 3.4 illustrates the post-exit re-offending estimates for each of the comparison groups in this study. Twenty-nine per cent of the graduated participants had re-offended since leaving the program until 31 December 2004. The majority of these offence episodes (13%) were for drug offences. Of the terminated participants who spent at least one free day in the community, 34 per cent had recorded a fresh offence episode since terminating from their IDRO. The majority of these offence episodes were for a property offence (19%).

Survival analyses were similarly conducted for the post-exit re-offending measures. The results indicated no significant difference between the 69 IDRO participants who had exited the program and the refusal comparison group, but similar beneficial outcomes for graduates.

Table 3.4: Post-exit re-offending as at 31 December 2004 (per cent)

	Refused (n=36)	Graduated (n=24)	Terminated (n=32)	All IDRO participants~ (n=69)
Post-exit
Any offence episode	53	29	34	30
Property offence episode	11	4	19	10
Violent offence episode	0	0	3	1
Drug offence episode	28	13	3	7
Breach offence episode	42	21	25	23
Any offence episode (excluding breach only)	39	17	31	20
~Sample includes offenders having terminated or graduated (n=69). *Statistically significant at p<0.05 between refused, gradated and terminated. Source: Australian Institute of Criminology, North Queensland Drug Court Database [computer file]

Offending frequency

The second measure of re-offending used in this evaluation is offending frequency, which is measured as the number of offending episodes recorded per 365 free days. In doing so, the total number of offence episodes recorded for each offender is divided by the number of free days in which they were available to re-offend. The result is a re-offending rate per one free day. Multiplied by 365 yields the estimated number of offence episodes recorded for that offender over 365 free days. Table 3.5 illustrates the pre-entry offending rates for each of the comparison groups. For all 120 IDRO participants, the average number of offence episodes per 365 free days prior to their program entry was 5.1. This compares with an average of 5.9 offending episodes for the refusal comparison group, a difference which was not statistically significant. In addition, the pre-entry offending episode rates are also provided for both the graduates (4.8) and terminates (6.6) of the drug court program. The results indicate that prior to program placement, graduates committed fewer offences on average, but this difference was not statistically significant. All in all, there was not a discernable difference in the pre-entry offending rates between graduates, terminates, and those who refused the program.

A similar calculation is used to estimate the post-entry and post-exit re-offending rates. Standardised to 365 free days, this calculation allows us to compare the relative improvement in re-offending as a result of the drug court program. Of all 120 IDRO participants, the re-offending rate per 365 free days drops from 5.1 pre-entry to 1.8 post-entry. For each of the comparison groups, the reductions were:

• graduates - from 4.8 pre-entry to 0.6 post-entry;
• terminates - from 6.6 pre-entry to 4.1 post-entry; and
• refusals - from 5.9 pre-entry to 1.6 post-entry.

Table 3.5: Offending episodes per 365 free days

	(n)	Pre-entry (for previous 24 months)	Post-entry	Post-exit^
All IDRO Participants	(120)	5.1	1.8	n/a
Graduated	(24)	4.8	0.6	0.4
Terminated	(45)	6.6	4.1	3.3
Refusal Comparison	(36)	5.9	1.6	1.6
^Post-exit estimates are calculated only for offenders who had spent at least one free day in the community post-exit from the IDRO program. Source: Australian Institute of Criminology, North Queensland Drug Court Database [computer file]

Comparative analysis

So far in this report we have included several sections to examine the differences and similarities between the North Queensland and South East Queensland drug courts. These comparisons were made using the final evaluation data from the South East Queensland evaluation (as at 31 December 2002) and the current evaluation data for North Queensland. In this section, we examine the re-offending outcomes using survival analysis to compare the survival experience of comparable offenders in each region.

In terms of re-offending, Figure 3.7 illustrates the survival experience of graduates from both the North and South East Queensland courts. It indicates that up to about 650 days, graduates of the North Queensland Drug Court program had equal probability of re-offending as did the graduates from the South East. From 650 days onwards, graduates in North Queensland experienced a marked drop in their survival rate. This result should be considered with caution, as only three graduates in North Queensland remained under observation beyond this time. The fewer the number of subjects under observation, the less confidence we can have that the results displayed are not the result of error or chance. The tests of significance illustrate that there is no significant difference between graduates of the North Queensland Drug Court program and graduates in the South East.

Figure 3.8 plots the survival functions for terminates of the North and South East drug court programs. The results indicate that at all time periods, participants terminated from the North Queensland Drug Court program experienced less favourable survival rates. Both the Log-rank and Wilcoxon tests of equality suggest that despite the observed differences between the two regions, we cannot be confident that these differences are not the result of error or chance. As such, we must conclude that the relative survival experience of terminated participants in North Queensland is equal to that of the terminated participants in South East Queensland.

Finally, Table 3.6 provides the comparative pre- and post-entry re-offending rates for both graduates and terminates of the North and South East Queensland drug courts. Of particular note is the obvious disparity between the courts on the average number of offence episodes recorded prior to program entry. In the North, graduates recorded an average of 4.8 offence episodes per 365 free days, while graduates in the South East recorded an average of 3.6 offence episodes. Similarly, terminated participants in the North Queensland pilot program recorded an average of 6.6 offending episodes prior to program entry, compared with 4.4 episodes for terminated participants in the South East.

In terms of post-entry offending, graduates in the North recorded a similar episode rate per 365 free days to their counterparts in South East Queensland (0.6 episodes compared to 0.5 episodes). For the terminates, the post-entry re-offending rates remained higher in North Queensland (4.1 episodes compared with 1.9 episodes). The final two rows in Table 3.6 illustrate the relative change in re-offending from pre- to post-entry. It illustrates a number of important findings:

• while graduates in North Queensland appeared to be offending more regularly prior to their IDRO, they experience a greater decline in re-offending rates (-4.2) than their counterparts in South East Queensland; (-3.1); and
• despite higher pre- and post-entry offending rates for terminated participants in North Queensland, the relative decline in re-offending (-2.5) was the same between the regions.

Table 3.6: Offending episodes per 365 free days, by region

	North Queensland	South East Queensland
Pre-entry (a)
Graduated	4.8	3.6
Terminated	6.6	4.4
Post entry (b)
Graduated	0.6	0.5
Terminated	4.1	1.9
Difference (b-a)
Graduated	-4.2	-3.1
Terminated	-2.5	-2.5
Source: Australian Institute of Criminology, North Queensland Drug Court Database [computer file]

Issues and limitations

This discussion of the North Queensland Drug Court's capacity to effectively reduce re-offending is not complete without an examination of its limitations. First and foremost, this evaluation has relied on estimates from statistical applications derived from particularly small sample sizes. Many of the comparison groups used in this analysis consisted of fewer than 50 subjects. When measured over long intervals of time, this significantly impedes our capacity to draw substantive statistical conclusions. In survival analysis, the primary statistical technique used in this report, the probability of re-offending is calculated as the number of subjects who re-offended at any one time, proportional to the number of subjects who were observed at that time. For example, suppose that at 500 days post-entry only three graduates were observable and that one of these graduates re-offends on that day. In survival analysis, the risk of re-offending experienced by any graduate on the 500th day is estimated at 33 per cent. Compare that with risk of re-offending attributable to any one offender in a time period when one of 10 observed graduates had re-offended (equalling a 10 per cent risk of re-offending). Clearly, as the number of subjects declines the relative weighting applied to one episode of re-offending increases, but our confidence in the estimate decreases.

To illustrate this important point we provide the comparative post-entry survival functions for the North and South East Queensland graduates. This time, however, we also illustrate the computed confidence intervals for the North and South East Queensland graduates. A confidence interval can be calculated for any statistical estimation and represents the upper and lower limits of the estimate, for which we can be 95 per cent confident that the true value falls. As the number of subjects under observation declines, the distance between the upper and lower confidence intervals increases. Moreover, at times when the confidence intervals of two survival functions overlap, we cannot conclude with confidence that they are different. Note in Figure 3.9 that despite the apparent drop in the survival function for North Queensland graduates beyond 650 days, the confidence intervals between the North and South continue to overlap. This leads us to conclude that there is no significant difference in graduate re-offending between the regions. This issue of sample size presents the same types of issues for all analysis presented in this report.

Furthermore, the pre-entry offence rates between the North and South East Queensland Drug Courts appeared to be disparate. The inference was that participants in North Queensland were more active offenders (recording a greater number of offending episodes) than their counterparts in the South East. Given that the method of calculation was consistent in both cases, the possibility that drug court participants in the North were more active offenders prior to their referral presents some important implications for this evaluation. Should the average participant offending profile have been worse in North Queensland, we can only expect similar, if not less favourable, program outcomes when compared with those derived in the South. The fact that the North Queensland Drug Court program has delivered greater post-entry reductions in offending for graduates and the same level of reduction for terminates suggests that despite higher pre-entry prevalence rates, the North Queensland program is maintaining similar levels of effect as those illustrated in the South.

The reasons participants in the North appear to be more prolific offenders is not completely understood, although two possibilities have been suggested. First, that the probability of arrest, particularly for drug related offences, is higher in a small regional city than in a large capital city. Drug markets are less dispersed and often centrally located in areas well known to the police. Moreover, an individual's criminal activity may be more amenable to detection within a small population, and where crime is more concentrated. In any case, should offenders in North Queensland face a greater probability of detection, any measurement that examines police records of crime is subject to this bias - a bias that may artificially distort our re-offending estimations. In other words, it may not be that graduates of the North Queensland Drug Court are more likely to re-offend, but that graduates in the South East are less likely to be apprehended.

The second possible explanation is that, as a by-product of the eligibility criteria, many participants referred to the North Queensland Drug Court had not previously been in prison, or if they had, the average time already served was below six months. Moreover, the eligibility criteria stated that in order for a referred offender to be eligible for the drug court program they must be facing a term of imprisonment. Placing these criteria side by side suggests that successful referrals in North Queensland may have needed to be more serious offenders in the 12 months prior to their referral to warrant the application of imprisonment.

To illustrate, we use the example of two offenders, A and B. Offender A has spent more than one year in prison prior to their referral while Offender B has not been in prison before. To justify a term of imprisonment, Offender A needs only to commit a small number of fresh offences, while Offender B will need to demonstrate a significant and problematic offending profile consisting of multiple offences or offences of greater seriousness. It is possible that by granting access only to those offenders with less than 12 months prior imprisonment, the North Queensland Drug Court program has inadvertently increased the number of more prolific offenders being placed onto the program. If this is the case, the drug court program in North Queensland may have needed to do more to achieve the same results as demonstrated in the South East.

Finally, re-offending must be considered in relation to other program factors having the potential to result in re-offending. Throughout this report, a number of sections have been dedicated to program implementation and operational issues which should be considered when interpreting the recidivism results presented here. For instance, drug court team members highlighted a number of issues having the possibility of increasing re-offending. For example:

• limited after-hours residential treatment supervision;
• treatment centres and non-residential treatment programs being offered in and around local drug market activity; and
• limited emergency accommodation.

While it is not possible to determine from this evaluation whether any of these factors can be empirically attributed to re-offending, the possibility needs to be considered when interpreting the recidivism results for any program. Some factors amenable to further policy and program development might help to explain higher re-offending prevalence rates and shorter time to re-offending.

3.2 Reducing drug use

The second objective of the North Queensland Drug Court pilot program was to reduce the levels of drug use within the community. It was noted in chapter two that the number of positive drug tests submitted by participants declines in each phase. As such, this general result indicates a significant reduction in the level of drug use amongst those participants who progress through each consecutive phase. Not surprisingly, those participants who progress to final graduation submit fewer positive drugs tests than those who are subsequently terminated.

Figure 3.10 illustrates both the average test positive rate and the average number of positive drug tests submitted for drug court graduates and terminates in the first six months of participation. The results indicate that for participants who graduate from the drug court program, the average number of positive drug tests declines from four in the first month to zero after four months of participation. Despite being terminated from the drug court program (presumably for non-compliance) Figure 3.10 also illustrates that the average number of positive tests submitted by terminated participants also declined (from four in the first month to one on the fourth, fifth and sixth months).

It should be noted that no information on drug use is collected for participants after termination or graduation. Therefore, it is difficult to ascertain whether these reductions in drug use are sustained after a participant leaves the program. In any case, these results indicate a decline in the average number of positive drug tests for both graduates and terminates and suggest that, at least for the duration of participation, the average levels of drug use decrease for all drug court participants.

3.3 Improving participant health and social function

The North Queensland Drug Court pilot program was established with the aim of improving the health and social functioning of its participants. In an effort to measure health outcomes, North Queensland Drug Court participants are required to complete the 36 question Short Form Health Survey (SF-36) at the time of referral, and at each phase graduation. The survey consists of 36 questions which are scaled to population norms and collapsed into eight specified areas of general health. These are:

• physical functioning - assessing limitations in physical activity because of health problems;
• role physical - assessing limitations in usual role activities (e.g. employment) because of health problems;
• body pain - assessing the presence of pain and limitations due to pain;
• general health - assessing the self evaluation of personal health;
• vitality - assessing levels of energy or fatigue;
• social function - assessing limitations in social activities because of physical or emotional problems;
• role emotional - assessing limitation in usual role activities (e.g. employment) because of emotional problems; and
• mental health - assessing psychological distress and wellbeing.

Table 3.7 indicates the number of SF-36 surveys that were administered in North Queensland at 31 December 2004. In total, there were 286 surveys completed - the majority of which were administered to referred participants during the preliminary assessment phase. In addition, nine participants had undertaken the SF-36 in phase one, 28 participants in phase two and 19 participants in phase three.

Table 3.7: SF-36 administration episodes and participant numbers (n)

	Episodes	Participants/offenders
SF36 administration
Preliminary Assessment	217	211
Phase 1	11	9
Phase 2	32	28
Phase 3	26	19
(Total)	286	178
Source: Australian Institute of Criminology, North Queensland Drug Court Database [computer file]

The individual responses to the SF-36 are collapsed and transformed according to population weightings determined by the Australian Bureau of Statistics in their National Health Survey in 1995 (ABS 1995). This transformation was conducted using the SF-36 programming command developed for the statistical software package Stata (Ryan 1999). This transformation allows us to compare the general health profile of the drug court participants with the general Queensland population (see Figure 3.11). Across all eight items, participants referred to the North Queensland Drug Court program reported lower scores than the Queensland population averages. These differences ranged between three (physical functioning) and 39 (role emotional). Compared with the Queensland population averages, it appears that drug court referrals were much more likely to report that their physical and emotional health was causing problems with their work and other daily activities. They were also more likely to report issues of psychological distress.

The important question for this evaluation is whether the general health of drug court participants improves as a result of their participation. To answer this, we compare the SF-36 results for a group of 16 graduates in North Queensland at two time intervals - the preliminary assessment phase and phase three, just prior to final graduation. Note that not all 24 North Queensland Drug Court graduates are included in this analysis. This is because the remaining eight graduates did not have either a preliminary assessment result or a phase three result for the SF-36 survey.

Overall, the results indicate improvements across all eight general health areas, with the greatest improvement noted for the role physical and role emotional scales. These scales measure how an individual perceives their physical and emotional health as impacting upon their general daily activities. An improvement illustrates the drug court's capacity to prepare participants for community re-integration - a core goal of phase three. Moreover, the phase three SF-36 results reported by these graduates are equivalent to the Queensland population averages (ABS 1995).

It is not possible to examine whether the drug court program in North Queensland also delivers improvements in the general health and wellbeing of terminated participants. This is because the administration of the SF-36 is undertaken commensurate with each phase progression, and the vast majority of terminated participants did not progress beyond phase one. The results thus far present a compelling argument that upon entry into the North Queensland Drug Court program, participants' general health is significantly lower than for the Queensland population as a whole. Moreover, the general health of those participants who completed the program was vastly improved by the time of final graduation.

3.4 Post graduation issues

Throughout the process of this evaluation, it was possible to interview six graduates of the North Queensland Drug Court. At the time of these interviews, half (n=3) had only recently graduated while the remaining three graduates had spent up to 12 months in the community. All graduates were overwhelmingly satisfied with the drug court program, most reporting their overall drug court experience positive and worthwhile. The graduation experience was also highlighted as being an important component of the program. One North Queensland Drug Court graduate stated that:

'I had never really achieved anything in my life until now. My graduation [from the drug court program] gave me something to be extremely proud of.'
- Female, graduated.

While the majority of participants reported a heightened sense of achievement at the time of graduation, some participants also reported being nervous and unsure about their life after graduation. In these cases, participants appeared concerned about how well they would cope without the daily support of the drug court program. This indicates that some participants may become highly reliant on the support and assistance of the drug court team. Some participants reported that graduation marked a difficult transition to a period of 'dealing with life by myself' and were concerned about their ability to manage themselves without the support of the drug court team. Some examples included:

'I was a little scared [to graduate]. Being told that you are graduating means that you have to deal with things on your own. There are huge expectations that [drug court graduates] will be able to cope in the real world without the help of the team. I was worried that I might not be able to - I guess I just wasn't confident in myself.'
- Female, graduated.

'I was frightened to graduate - I didn't know what I was going to do with all my spare time. I was always so busy [with the drug court], but I knew that would end when I graduated.'
- Male, graduated.

Despite some graduates' concerns about their ability to cope after graduation, all interviewees reported feeling comfortable to approach DCS or ATDOS should any post-graduation issues arise. This is a significant finding, as it demonstrates that drug court graduates develop good support networks that they can rely on, should their personal situation deteriorate.

This evaluation has so far noted significant improvements in the health of participants by using the quantitative data collected in the SF-36 health survey. These conclusions are supported through the qualitative interviews, with all graduates reporting major improvements in their mental and physical health. Graduates also reported improvements in their family and personal relationships as a result of their participation on the program. Some examples include:

'Before the program my son was scared and cagey around me. I was always on drugs and I don't think he knew what I was doing. Since I stopped using drugs, my relationship with my son has really improved. I spend much more time with my son that I ever did when I was using drugs.'
- Male, graduated.

'My eldest son is now going to school - when I was on drugs I didn't really think much about my kids.'
- Female, graduated.

'[The drug court] gave me so much more confidence. I don't think my kids actually respected me when I was on drugs, but now they do. I think they can see that I am a different, more confident person. They make me feel proud to have done this - that makes me very happy.'
- Female, graduated.

'I lost a lot of friends and family when I started using drugs. I am slowing getting them back.'
- Male, graduated.

The benefits of the drug court program in improving personal and family relationships are not universal. The experience of one graduate highlights the continuing personal issues faced by graduates beyond program completion:

'The relationship with my kids is much better, but not with my partner. He still uses drugs and thought that the drug court program was a waste of time for me. I am trying to kick him out, but he won't go.'
- Female, graduated.

A number of the long term graduates (those whose interview was conducted some time after their graduation) had re-offended or used drugs since their graduation. At the time of interview, two of these graduates were facing fresh criminal charges. When asked to explain why they had engaged in further offending the responses, while varied, illustrated a common theme - that the use of drugs and subsequent criminal offending was an unwelcome reaction to increasing levels of stress and pressure in their personal lives. All graduates reported significant personal disappointment in what they termed their failure of the drug court program. Despite having graduated, interviewees saw post-graduation drug use and offending as their inability to maintain the standards and expectations of the drug court program for the long term. While this demonstrates that the drug court was successful in teaching graduates the ideals of drug abstinence, the flipside is that with these expectations comes the pressure to succeed, which, as one graduate stated, was one of the causes for their continued use of drugs and subsequent offending. For example:

'[When I started to re-use] I felt so disappointed in myself. I could see myself slipping back to my old habits, but I was so embarrassed that I didn't know what to do. I felt that I had let the [drug court team] down and I think I continued to re-use drugs because I felt so bad.'
- Female, graduated.

Finally, graduates were asked whether the drug court program might have done anything else to help them after their graduation. All graduates reported that the drug court program was a positive and worthwhile experience. Nonetheless, a number of graduates stated that since becoming drug-free several other personal and psychological issues had surfaced that were difficult to manage in the community and without the support of the drug court team. These issues highlight some of the very important underlying reasons drug-dependent persons use drugs, and that drug-use may not be the only factor contributing to offending. If drugs are used to mask other personal or psychological issues, these issues will also need to be addressed by the drug court program. This relies primarily on the drug court team's capacity to identify and assist participants in dealing with these factors, as well as the participant's honesty and willingness to disclose other information related to their personal circumstances.

In all, the interviews with graduated participants provided an opportunity to understand some of the important social and personal factors that have an impact on the life (and long term success) of a participant after graduation. While most graduates reported significant improvements in family and personal relationships as a result of their drug court participation, changes to these relationships after graduation can exert additional pressure, resulting in a return to drug use.

This information is based on the self-reported experience of only six graduates in North Queensland, and further research is needed to isolate and understand these post-graduation issues so that strategies can be developed to help overcome them.

3.5 Discussion

This section has examined the North Queensland Drug Court pilot program's achievements in reducing offending and drug use, and improving the health and social function of its participants. It commences with a word of caution - at the time of this evaluation the North Queensland Drug Court pilot program had accepted only 120 participants, of whom only 24 had successfully completed the program and graduated. The relatively small number of participants at this stage limits our capacity to draw substantive conclusions based on statistical analyses. These limitations must be first and foremost in our interpretation of the results in this section. In terms of re-offending, it is with some caution that this evaluation finds that:

• there was no significant difference in re-offending (any offence and any property offence) between the North Queensland Drug Court participants and the comparison group;
• drug court participants took significantly longer to commit a drug offence, although the lifetime probability was not significantly different from the comparison group;
• graduates were significantly less likely to re-offend than those who were terminated;
• of those graduates who did re-offend, the time taken to re-offend was significantly longer than for the terminates; and
• the frequency of offending was reduced for both graduates and terminates - although the reduction in offending was greatest for the graduates.

Comparative analysis was used to examine differences in re-offending between the North and South East Queensland Drug Court pilot programs. These analyses suggest that:

• prior to referral and admission, participants in North Queensland were, on average, arrested at a higher frequency;
• the post-entry re-offending estimates for graduates were similar in North and South East Queensland;
• the post-entry re-offending estimates for terminated participants were also similar in North and South East Queensland; and
• despite differences in the pre-referral arrest frequency, the post-entry offence frequencies were improved in both the North and South East Queensland and graduates in North Queensland demonstrated a greater overall reduction in offending frequency.

In terms of drug use and health outcomes, this evaluation finds evidence that:

• both graduates and terminates reduced the level of their drug use for the duration of their participation, as indicated by the declining number of positive drug tests; and
• graduates of the drug court program report significant improvements across the range of physical and mental health measures of the SF-36 Health Survey.

Interviews with drug court graduates highlight that:

• the drug court program was reported as a unique and worthwhile opportunity for dealing with drug use and offending;
• post-graduation offending and drug use appears linked to changes in personal and family relationships;
• drug re-use was often reported as a mechanism for dealing with stressful situations; and
• other areas, such as psychological and mental health issues, may also be linked to post-graduation re-offending.