A Systematic Review of the Efficacy of Alcohol Interventions for Incarcerated People

Dorothy Newbury-Birch; Jennifer Ferguson; Sarah Landale; Emma L. Giles; Grant J. McGeechan; Charlotte Gill; Kelly J. Stockdale; Aisha Holloway

Disclosures

Alcohol Alcohol. 2018;53(4):412-425. 

In This Article

Results

The search yielded 10,298 papers, of which 28 papers were fully assessed for eligibility (Figure 1). Eleven papers from nine studies were included in the final analysis (Chance et al., 1990; Baldwin et al., 1991; Peters et al., 1993; Davis et al., 2003; Stein et al., 2010, 2011a, 2011b; Begun et al., 2011; Bowes et al., 2012, 2014; Owens and McCrady, 2016) (Table 1 and Table 2). Seven of the included studies were from the USA (Chance et al., 1990; Peters et al., 1993; Davis et al., 2003; Stein et al., 2010, 2011a, 2011b; Begun et al., 2011; Owens and McCrady, 2016) and two from the UK (Baldwin et al., 1991; Bowes et al., 2012, 2014). The included studies consisted of 2435 participants (range 27–729). Most of the studies included either all male participants (Chance et al., 1990; Baldwin et al., 1991; Bowes et al., 2012, 2014; Owens and McCrady, 2016) or majority male (Peters et al., 1993; Davis et al., 2003; Stein et al., 2011a, 2011b). Only two of the included studies examined women only (Stein et al., 2010; Begun et al., 2011). Because of the heterogeneity of the studies meta-analysis was not possible.

Figure 1.

Flowchart of data. High risk of bias was recorded if 'no' or 'unsure' was recorded for 6 or more of the 11 questions on the tool. Medium risk of bias was assigned if 'no' or 'unsure' was recorded for 4–5 questions and low risk for 1–3 questions.

Screening for inclusion for five of the included studies was carried out by researchers (Davis et al., 2003; Stein et al., 2010, 2011a, 2011b; Begun et al., 2011; Owens and McCrady, 2016); one by social workers (Baldwin et al., 1991) and one by programme counsellors (Peters et al., 1993). Two studies did not include this information (Chance et al., 1990; Bowes et al., 2012, 2014). A range of tools was used to screen participants into studies. Two studies used the Alcohol Use Disorders Identification Test (AUDIT; Babor et al., 2001) screening tool to screen for risky drinking (Stein et al., 2010; Begun et al., 2011); one used the ASSIST (Owens and McCrady, 2016); one used the Alcohol-Related Aggression Questionnaire (McMurran and Baldwin, 2006; Bowes et al., 2012, 2014); one the Form-90 alcohol tool (Davis et al., 2003); one the Addiction Severity Index (Peters et al., 1993); one the risks and consequences of drinking questionnaire (Stein et al., 2010, 2011a, 2011b); one used the question 'more than half of their total offences being drink related' (Baldwin et al., 1991) and one did not give this information (Chance et al., 1990).

Brief Interventions

Five studies (six papers) from the USA examined the efficacy of brief interventions for incarcerated participants (Davis et al., 2003; Stein et al., 2010, 2011a, 2011b; Begun et al., 2011; Owens and McCrady, 2016). The length of the brief interventions ranged from 45 to 150 min (Davis et al., 2003; Stein et al., 2010, 2011a, 2011b; Begun et al., 2011; Owens and McCrady, 2016). One study (two papers) was from the same authors and included relaxation training as the control condition (Stein et al., 2010, 2011a, 2011b). One study included educational videos as the control condition (Owens and McCrady, 2016). The other studies all included treatment as usual as the control condition (Davis et al., 2003; Stein et al., 2010; Begun et al., 2011). Studies did not give information on what treatment as usual was. Four of the studies were conducted with adults (Davis et al., 2003; Stein et al., 2010; Begun et al., 2011; Owens and McCrady, 2016) and one study with juveniles (Stein et al., 2011a, 2011b) (Table 1). In terms of quality assessment, one of the studies was classified as having a low risk of bias (Stein et al., 2010), three as medium risk of bias (Begun et al., 2011; Stein et al., 2011a, 2011b; Owens and McCrady, 2016) and one as having a high risk of bias (Davis et al., 2003) (Table 3).

The five studies all used different outcome measures (Table 2), meaning results were unable to be synthesised by meta-analysis, yet despite this, some significant results were found. Davis et al. (2003) found that those that were given a brief intervention were significantly more likely to schedule follow-up appointments for treatment (66.7 vs. 40.5%; X 2 5.01, P = 0.025) (Davis et al., 2003). Stein et al. (2010) found that those in the intervention group had reported significantly more days abstinent at follow-up (OR = 1.96) (Stein et al., 2010). Begun et al. (2011) found that for the intervention group the mean reduction in AUDIT score from baseline to follow-up were greater in the intervention group (F(1,148) = 6.336, P< 0.001; Begun et al., 2011). The Stein et al. (2011a, 2011b) study found no significant results related to alcohol. Owens and McCrady (2016) was a small feasibility study and although they found the study to be feasible they did not find any significant differences between groups.

Extended Brief Interventions

Four studies (five papers) examined the efficacy of alcohol interventions with adults in the prison system using extended brief interventions (Chance et al., 1990; Baldwin et al., 1991; Peters et al., 1993; Bowes et al., 2012, 2014). Two of the studies were from the USA (Chance et al., 1990; Peters et al., 1993) and two from the UK (Baldwin et al., 1991; Bowes et al., 2012, 2014). The interventions lasted from six sessions to 18 months in duration (no actual times given). Three of the studies were conducted with adults (Chance et al., 1990; Peters et al., 1993; Bowes et al., 2012, 2014) and one study with juveniles (Baldwin et al., 1991) (Table 1). In terms of quality assessment, one of the studies was classified as having a medium risk of bias (Bowes et al., 2012, 2014) whilst three had a high risk of bias (Chance et al., 1990; Baldwin et al., 1991; Peters et al., 1993) (Table 3).

The four studies all used different outcome measures (Table 2) meaning results were unable to be synthesised. Chance et al. (1990) found no significant results related to alcohol (Chance et al., 1990).

Baldwin et al. (1991) found that the control group increased the average number of alcohol units per week compared with the intervention group F(1,19) = 4.546 (P < 0.05); The control group also increased average alcohol units per drinking session compared to the intervention group F(1,19) = 6.753 (P < 0.05). In comparison, the intervention group reduced the average number of offences against property compared to the control group F(1,13) = 6.489 (P < 0.05) (Baldwin et al., 1991).

Peters et al. (1993) found that those that were randomised to the intervention group had a significantly longer time period before being arrested again t(418) = 3.0 P < 0.01, significantly fewer arrests t(418) = 2.7 P < 0.01, and served significantly less jail time t(418) = 2.4 P < 0.05 compared with the control group.

Bowes et al. (2012) found significantly lower scores for individuals in the intervention group compared to the control in relation to alcohol-related aggression (P < 0.05) as well as the different components of the Controlled Drinking Self-Efficacy Tool and the State-Trait Anger Expression Inventory (Bowes et al., 2012; Table 2). However, a follow-up study by the same authors found no statistically significant differences in relation to recidivism (Bowes et al., 2014).

TIDieR Rsults

Results relating to how interventions were described are shown in Table 3 and Table 4 using the TIDieR checklist (Hoffman et al., 2014). We found that for some categories detailed information was not given in the included papers.

TIDieR Rsults—brief Iterventions

All included studies described the brief intervention as being based on the motivational interviewing work of Miller and Rollnick (2002) with all papers giving some indication of what the components in the interventions were (Davis et al., 2003; Stein et al., 2010). All studies reported that interventions were given one-to-one and were based on the results of clients' individual screening (Begun et al., 2011; Davis et al., 2003; Stein et al., 2010, 2011a, 2011b; Owens and McCrady, 2016). All studies were delivered by trained research staff, which calls into question how pragmatic the studies are and whether they could be implemented with fidelity in real life situations by existing programme staff.

None of the included studies gave information about modifications during the study and only two gave information related to fidelity (Stein et al., 2011a, 2011b), with one giving in-depth information in relation to the intervention development (Stein et al., 2010).

TIDieR Rsults—extended Brief Interventions

The four studies involving extended brief interventions were very different in content from the brief intervention studies (Chance et al., 1990; Baldwin et al., 1991; Peters et al., 1993; Bowes et al., 2012, 2014). Intervention details in these studies were sparse, meaning that they would be unable to be replicated. The total amount of time spent in intervention varied in length from a total of 12 h (Baldwin et al., 1991) to 20 h (Bowes et al., 2012, 2014) to 54 h (Peters et al., 1993). The remaining study stated that the time frame was 6–18 months but did not say how many sessions (Chance et al., 1990). Very little detail was provided about the information given during or as part of the intervention. According to the TIDIeR checklist authors, this is the question that is least likely to be answered (Hoffman et al., 2014).

None of the included studies gave any information relating to where in the prison the interventions took place or of any fidelity checks. However, unlike the brief intervention studies, all of the extended brief interventions were delivered by trained individuals employed within the services (Table 5).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....