What matters to magistrates when considering diversion into mental health treatment?

Figures & data

Figure 1. Flowchart of study sample for regression analyses. BOCSAR = Bureau of Crime Statistics and Research.

Table 1. List of included variables and their coding.

Variable	Coding	Source
Gender	1 = female 2 = male
Indigenous status	0 = not Indigenous 1 = Indigenous 2 = Unknown
Age group (years)	1 = 18–29 2 = 30–49 3 = 50+
Remoteness of area	1 = Major cities 2 = Inner regional 3 = Outer Regional/Remote	Obtained from ABS ARIA
Prior court	0 = no prior court 1 = one court appearance 2 = two to five court appearances 3 = six or more court appearances
Prior prison episode	0 = no prior prison 1 = prior prison
Seriousness of the principal offence	0 = less serious 1 = more serious	Obtained from BOCSAR MSR
Offence violence	0 = non-violent offence 1 = violent offence	Obtained from ANZSOC
Prior mental health dismissals	1 = no 2 = yes
Availability of SCCLS	0 = not an SCCLS court 1 = an SCCLS court.

Note: SCCLS = Statewide Community and Court Liaison Service; ABS = Australian Bureau of Statistics; ARIA = Accessibility and Remoteness Index of Australia; BOCSAR = Bureau of Crime Statistics and Research; MSR = median severity ranking; ANZSOC = Australian and New Zealand Standard Offence Classification.

Table 2. Sample description.

Variable	n^a (%)
Gender
Female	876 (29.98)
Male	2046 (70.02)
Age group, years
18-29	1096 (37.51)
30-49	1473 (50.41)
50+	353 (12.08)
Type of mental health dismissal
Not dismissed	1412 (48.32)
Section 32	1143 (39.12)
Section 33	323 (11.05)
Dismissal type not specified	44 (1.51)
Indigenous status
Non-Indigenous	2298 (78.64)
Indigenous	442 (15.13)
Unknown	182 (6.23)
Remoteness of area
Major cities	2089 (71.49)
Inner regional	631 (21.59)
Outer regional/remote	105 (3.59)
Unknown	97 (3.32)
Number of prior court appearances
None	592 (20.26)
One	352 (12.05)
Two to five	803 (27.48)
Six or more	1175 (40.21)
Prior prison
No prior prison	2088 (71.46)
At least one prison sentence	834 (28.54)
Prior mental health dismissal
Yes	1240 (42.44)
No	1682 (57.56)
Type of offence
Non-violent offence	1577 (53.97)
Violent offence	1345 (46.03)
Offence seriousness
Less serious	1457 (49.86)
More serious	1465 (50.14)
Availability of court liaison service
No court liaison service	751 (25.69)
Court liaison service	2172 (74.31)

^aThe full sample (n = 2922) before logistic regression is included, whereby those missing on remoteness of area and what specific section they were diverted under are included.

Table 3. Bivariate comparisons of variables associated with s32 and s.33 dismissal.^a

Variable	s32 Dismissal(%)	p	s33 Dismissal(%)	p
Gender		<.001		.280
Female	51.48		16.92
Male	41.78		19.22
Age group (years)		.001		<.001
18–29	46.36		11.24
30–49	41.53		22.38
50+	52.51		24.06
Indigenous status		<.001		.109
Non-Indigenous	46.60		18.73
Indigenous	33.15		20.51
Unknown	47.93		11.11
Remoteness of area		.120		.007
Major cities	45.94		20.72
Inner regional	45.57		14.01
Outer regional/remote	35.11		12.86
Number of prior court appearances		<.001		.001
None	63.30		15.25
One	51.41		12.43
Two to five	41.75		15.89
Six or more	34.42		22.50
Prior prison		<.001		<.001
No prior prison	49.17		15.44
At least one prison sentence	32.50		24.51
Prior mental health dismissal		.058		<.001
Yes	47.02		27.44
No	43.21		11.94
Type of offence		.031		.334
Non-violent offence	42.80		17.82
Violent offence	47.07		19.63
Offence seriousness		<.001		.478
Less serious	52.63		17.85
More serious	36.66		19.19
Availability of court liaison service		.742		<.001
No court liaison service	44.20		11.49
Court liaison service	44.93		21.00

^aBased on sample of 2781 observations.

Table 4. Models of Mental Health Dismissal via s.32 and s.33.

Given a treatment order under s.32 or s.33^a	s.32				s.33
	Odds ratio	p	95% CI		Odds ratio	p	95% CI
Gender
Female (reference)
Male	0.756	.002	[0.631, 0.906]		—	—	—
Indigenous status
Non-Indigenous (reference)
Indigenous	0.734	.013	[0.575, 0.937]		—	—	—
Unknown	0.881	.470	[0.624, 1.243]		—	—	—
Age group, years
18–29 (reference)
30–49	0.916	.339	[0.765, 1.10]		2.100	<.001	[1.531, 2.881]
50+	1.221	.144	[0.934, 1.596]		2.545	<.001	[1.641, 3.947]
Remoteness of area
Major cities (reference)
Inner regional	—	—	—		0.521	.070	[0.521, 1.026]
Outer regional/Remote	—	—	—		0.412	.663	[0.412, 1.758]
No. prior court appearances
None (reference)
One	0.637	.002	[0.479, 0.847]		0.677	.196	[0.375, 1.222]
Two to five	0.473	<.001	[0.373, 0.599]		0.749	.220	[0.472, 1.189]
Six or more	0.431	<.001	[0.328, 0.566]		0.685	.134	[0.418, 1.124]
Prior prison	0.787	.053	[0.617, 1.002]		1.530	.015	[1.086, 2.153]
Violent offence	1.00	.999	[0.847, 1.181]		—	—	—
More serious offence	0.591	<.001	[0.501, 0.698]		—	—	—
Prior mental health dismissal/s	—	—	—		2.565	<.001	[1.952, 3.370]
Court liaison service?
No (reference)
Yes	—	—	—		1.873	<.001	[1.332, 2.634]
Constant	3.433	<.001	[2.354, 5.004]		0.024	<.001	[0.013, 0.045]

Note: CI = confidence interval.

^aBased on 2781 observations.

Table 5. Multilevel logistic regression model (Level 2 only).

Dismissal vs. prosecution^a	Coef.	SE	z	p > |z|	95% CI
_cons	−.199	0.101	−1.98	.048	[−0.397, 0.002]
Magistrate var(_cons)	.340	0.105			[0.186, 0.923]

Note: CI = confidence interval. Likelihood ratio (LR) test vs. logistic model: χ² = 55.39, p > χ² < .001.

^aBased on 1590 observations.

Table 6. Multi-level logistic regression model (Levels 1 and 2).

S32 dismissal vs. prosecution^a	Odds ratio	p	95% CI
Gender
Female (reference)
Male	0.837	.144	[0.659, 1.063]
Indigenous status
Non-Indigenous (reference)
Indigenous	0.984	.924	[0.715, 1.356]
Unknown	0.794	.290	[0.518, 1.217]
Age group, years
18–29 (reference)
30–49	0.915	.463	[0.721, 1.160]
50+	1.061	.745	[0.741, 1.520]
No. prior court appearances
None (reference)
One	0.870	.468	[0.598, 1.266]
Two to five	0.493	<.001	[0.358, 0.679]
Six or more	0.424	<.001	[0.294, 0.612]
Prior prison	0.896	.505	[0.648, 1.238]
Violent offence	0.973	.814	[0.778, 1.218]
More serious offence	0.492	<.001	[0.392, 0.618]
Constant	2.944	<.001	[1.727, 5.019]
Magistrate	0.429		[0.241, 0.763]

Note: CI = confidence interval.

^aBased on 1590 observations.

Figure 2. Effect of magistrate on the likelihood of diversion. CI = confidence interval.