Intensive fostering: an independent evaluation of MTFC in an English setting.

Biehal, Nina ; Ellison, Sarah ; Sinclair, Ian 等

Introduction

This article reports on the first independent study of Multidimensional Treatment Foster Care (MTFC), the evaluation of the MTFC programme for serious and persistent young offenders in England. The MTFC programme, known as Intensive Fostering (IF) in this incarnation, was piloted in three areas with initial funding provided by the Youth Justice Board. MTFC is a community based intervention grounded in social learning theory, which was developed by the Oregon Social Learning Centre (OSLC) during the 1980s. As its name suggests, MTFC differs from routine foster care in that it offers treatment as well as substitute care. It provides young people with a short-term foster placement, followed by a short period of aftercare, and is delivered by a professional team and highly trained and supported foster carers. MTFC was initially designed for work with boys with serious criminal behaviour and was later extended to delinquent girls, but it has subsequently been adapted for use with other groups of children and young people (Chamberlain and Reid, 1991, 1998; Chamberlain et al, 1996; Fisher et al, 2005).

The programme developers have conducted a number of evaluations of its use with young offenders. An early study of 32 'severely delinquent youth', of whom half who received Specialised Foster Care (SFC, a precursor of MTFC) and half who were placed in group care or on intensive parole supervision, found that the SFC group were less likely to enter custody over a two-year follow-up. However, the group difference diminished during the second year of follow-up (Chamberlain, 1990). A subsequent randomised control trial of 79 male young offenders compared a group placed in MTFC with others in community-based residential programmes. By follow-up, one year post-discharge (or expulsion) from the programme, the MTFC group were found to have had fewer criminal referrals and to have spent 60 per cent fewer days incarcerated. A further trial with 81 female young offenders also reported positive effects for MTFC. At 12 months post-baseline, girls in MTFC had spent significantly fewer days in locked settings than those placed in community-based residential care, but although they also had fewer criminal referrals, this difference was not statistically significant (Leve et al, 2005).

In Sweden, MTFC has been used as an intervention for young people with a diagnosis of conduct disorder and who were at immediate risk of out-of-home placement. A randomised controlled trial followed up 35 young people for two years, comparing a group placed in MTFC with others placed in foster or residential care or remaining at home (Westermark et al, 2011). In this study MTFC was found to have a positive effect on externalising behaviour, as reported by the young people. Parent reports also indicated a group difference in externalising behaviour in the same direction, but this did not reach statistical significance (Westermark et al, 2011). However, we cannot compare patterns of reoffending with those in England or the USA because in Sweden all anti-social behaviour by young people is defined as a child welfare problem and is not, therefore, categorised as a criminal offence.

The independent replication of studies of MTFC in different locations was recommended by the Cochrane review of MTFC (Macdonald and Turner, 2009). All previous evaluations have been conducted either by the programme developers or, as in Sweden, by the team responsible for the implementation of MTFC in that country. This study there fore represents the first independent evaluation of the intervention. (1)

The intensive fostering programme

The English IF programme closely followed the MTFC programme developed by the OSLC. Fidelity to the Oregon model was monitored both by the national co-ordinator of the IF programme and through distance super vision provided by a member of the OSLC, who viewed videos of clinical meetings and provided telephone advice. The key elements of MTFC include: the provision of a consistent reinforcing environment in which young people are mentored and encouraged; a clear structure, with clearly specified boundaries and consequences for behaviour; close supervision of young people's activities and whereabouts at all times; diversion from associations with anti-social peers; and help to develop positive social skills and promote the formation of relationships with more positive peers (Chamberlain, 2003). This daily programme of positive reinforcement is delivered by specially trained foster carers, who are provided with intensive support. Behaviour is closely monitored and positive behaviours are reinforced using a system of points and levels. The intention is that after six to nine months, young people will accumulate sufficient points to move to the highest level on the programme and then return to their families or move to alternative carers.

MTFC teams are led by programme supervisors, who act as case managers and co-ordinate the intervention. Individual therapists work with young people to develop problem-solving skills and help them change identified behaviours. Skills workers help them improve and practise their social skills and try to involve them in positive recreational activities. Birth family therapists undertake work with parents or alternative follow-on carers during the foster placement and a three-month aftercare period, to ensure that desired behaviours continue to be encouraged and reinforced in a positive manner after they complete the programme. The IF teams also aimed to find appropriate education or training for the young people, help them settle into school or college and to encourage regular attendance. The English teams made minor adaptations to programme delivery by creating additional posts for programme managers and family placement social workers, to allow programme super visors to focus on clinical work.

Design and methods

Design

A prospective, quasi-experimental design was used to compare outcomes for 47 serious and persistent offenders facing an imminent custodial sentence, of whom 23 entered IF placements and 24 were either sentenced to custody or, in four cases, sentenced to an Intensive Supervision and Surveillance Programme (ISSP), a community-based initiative for persistent young offenders. The primary outcome measures were reconviction and the use of custody. Secondary outcomes were also assessed, including living situation, participation in education, training or employment and peer relationships.

Baseline for all the young people in the study was the date of their index sentence. This was the date that the IF group moved to their foster placements and that the comparison group entered custody (or began their ISSP sentence). Outcomes for the IF group were measured at two points in time: Stage 1, one year after the date of entry to the IF placements; and Stage 2, one year after the date of exit from IF placements.

The second follow-up after the young people left their IF placements was essential to ensure a valid comparison of outcomes, since while in placement they were under very close supervision. Their opportunity to reoffend therefore only became equivalent to that of young people released from custody once they had left their IF placements.

For the comparison group there was only a single point of follow-up. For those sentenced to custody, follow-up was one year after the date of their release as they only had the opportunity to reoffend once they returned to the community, but for those sentenced to ISSP follow-up was one year after the sentence began as they remained at liberty throughout. (2) Average time in custody was 5.8 months.

Sample recruitment

Young people were sentenced to IF if they were serious or persistent offenders who were at imminent risk of a custodial sentence and who also met certain eligibility criteria, namely severity scores of three or more (on a four-point rating scale) on the 'family and personal relationships' and 'lifestyle' sub-scales of the Asset assessment tool. The Asset is a structured assessment tool used by Youth Offending Teams (YOTs) with all young people who come into contact with the criminal justice system. It is used to identify the factors which may have contributed to an individual's offending behaviour and to inform decisions about intervention and the assessment of the risk that they will reoffend (Youth Justice Board, 2011). Only 15 IF placements were available across the three pilot sites and, due to the small numbers entering IF and the lengthy duration of the placements (which were intended to last for nine months), it was possible to include only 24 young people in the intervention group during the evaluation timescale (one of whom died prior to follow-up). The intervention group included all young people who entered the programme during our recruitment period.

Young people in the IF group were compared to 24 others who met the eligibility criteria for the IF programme, that is they (a) were sentenced to custody or ISSP (the 'service as usual') and (b) had scores on the 'lifestyle' and 'family and personal relationships' domains of the Asset tool which indicated serious difficulties in these areas. We worked closely with local YOT teams to identify young people who had just been sentenced to custody or ISSP and who also met the additional eligibility criteria for entry to the IF programme in respect of their Asset scores. There were no significant differences in Asset mean scores between the IF and comparison groups. (3) All but three of the young people who met the criteria for our comparison group during the study time frame agreed to participate in the study.

Data collection

Official data on dates and types of all offences and sentences ever received were available for the measurement of baseline sentencing histories and primary outcomes for the total sample, so there was no sample attrition in relation to the primary outcome measures. All offences were systematically coded for offence gravity using the official government scoring system, the 'counting rules,' which were developed to ensure consistency between police forces in the classification of crime (Home Office, 2011). This standardised system was used to assign a gravity score to each crime committed by the young people up to and including baseline and subsequently between baseline and follow-up.

In addition, semi-structured interviews were conducted with young people at baseline (n = 36) and one year after entry to IF placements/release from custody (n = 37), and with the parents of the IF group, although not many parents agreed to participate in these (11 at baseline and nine at follow-up). Although sample attrition meant there were some missing interview data, a complete dataset was available for all measures of reoffending (reconviction, number and nature of offences, use of custody), as this information was drawn from official administrative sources. The interviews explored the young people's histories, circumstances, behaviours and views of MTFC. The young people's YOT workers also completed questionnaires at the same points in time and provided their standard (Asset) assessment records and updates. IF teams and foster carers completed questionnaires six weeks after the foster placements began and again at Stage 1 follow-up, or sooner if the placement had already ended.

This evaluation did not have the resources to independently assess treatment fidelity, but we were aware that problems with staff recruitment and turnover may have had an impact on programme delivery at times. However, staffing or other operational difficulties may occur with the wider implementation of any evidence-based programme, so our results are likely to reflect the effects of 'real' implementation, as opposed to implementation by programme developers in more controlled conditions.

Analysis

Analysis of data at baseline and Stage 1 drew on all data sources, but only official data on sentencing histories were collected at Stage 2. Quantitative data from official records and postal questionnaires were analysed using SPSS 17. Bivariate analyses were used to compare the characteristics, histories, behaviour and offending of the two groups, using Chi-square tests for the analysis of nominal variables and Mann-Whitney U tests for interval data. Logistic regression was used for multivariate analysis of patterns of reconviction. For reasons of space, the results of our qualitative analysis will be reported in another article, although at times we briefly refer to findings from it in this one.

Results

Characteristics and histories of the young people and families

The IF and comparison groups were well matched in terms of their demographic characteristics and criminal histories. The two groups were also equally likely to have committed a range of index offences. The only significant difference at baseline was that the comparison group was more likely to have committed an index offence of 'violence against the person'. No significant differences were found in mean scores for the gravity either for the most serious index offence or in relation to the most serious previous offence ever committed. Table 1 compares the characteristics and histories of the two groups.

Professional reports indicated that many of the young people were highly vulnerable. One in six had a learning disability and one in ten had a diagnosis of Attention Deficit Hyperactivity Dis order (ADHD). One-fifth were known to have previously self-harmed or attempted suicide, and six had been diagnosed with mental health problems. Family and peer-related risk factors for youth offending were common among both groups (Farrington, 1995; Rutter et al, 1998). Nearly half (46%) had experienced disrupted family relationships, 40 per cent had experienced family conflict, and one-quarter had parents with criminal convictions of their own. According to professional and self-reports, over three-quarters of those in both groups were associating with pro-criminal peer groups at baseline.

Over half had experienced maltreatment during the course of their lives, with emotional abuse the most common form, followed by physical abuse and neglect, both of which have previously been found to be associated with juvenile offending (Runyan and Gould, 1985; Widom, 1991; Thornberry, 1995; Thornberry et al, 2001). Around 40 per cent of those in each group had spent some time in out-of-home care during the course of their lives, including 15 per cent who were in care at the time of the conviction for the index offence. Only half were living with a parent at the time of conviction and, again, the groups were well matched in this respect.

One-third of the young people had difficulties with basic literacy or numeracy. Over half of those who were still of compulsory school age had truanted from school during the three months prior to conviction and only 23 per cent of them were known to have spent any time at all in mainstream schooling during this period. Among those who were beyond the minimum school leaving age, half were not in education, employment or training at baseline.

Programme completion

Placement duration ranged from one week to nearly 17 months. Two-thirds of the young people remained in placement for nine months or more, including four who remained for 12-16 months.

Overall, the mean number of days in placement was 260.48, higher than the mean of 207.91 days found in a study of MTFC with young offenders in the USA (Chamberlain and Reid, 1998). Two-thirds (15) of the IF group were 'graduates' who formally completed the programme, a completion rate reason ably close to that of 73 per cent reported in the above US study. A further three young people left the programme when their placements disrupted after nine to 15 months. Over three-quarters (18) of the young people, therefore, either graduated from the programme or remained on it for the intended period.

For the remaining five, placements disrupted at an early stage. Four absconded and the fifth was arrested for criminal damage two weeks after placement. In total, therefore, eight placements disrupted at some point, a disruption rate of 35 per cent, which is not dissimilar to the disruption rate of around 40 per cent (during the first year of placement) found in other studies of adolescents in out-of-home care (Farmer et al, 2004; Sinclair, 2005; Sinclair et al, 2007; Biehal, 2009).

Reconviction during Stage 1

During Stage 1 follow-up (that is, the year after the young people entered their foster placements and the comparison group was released from custody), the IF group was significantly less likely to be reconvicted. Thirty-nine per cent of the IF group were reconvicted compared to 75 per cent of the comparison group. These figures include all young people who entered IF placements during the study's recruitment period, including the five whose placements disrupted at an early stage. There was no significant relationship between placement duration and the likelihood of reconviction (Mann-Whitney U test p = 0.431). Importantly, only three young people (17% of the IF group) committed any recorded offences during the time they were living in their IF placements.

Placement on the IF programme was not only associated with a lower rate of reconviction per se, but also with a lower frequency and gravity of offending during the Stage 1 period. On average, the IF group took roughly three times longer to reoffend than the comparison group, and the comparison group were convicted for five times as many substantive offences as the IF group. Even when we took account of the young people's opportunity to offend, that is the total number of days not spent in custody during Stage 1, the mean number of recorded offences per day at liberty remained significantly lower for the IF group. The mean gravity score for the most serious recorded offences committed during this period was also significantly lower for the IF group. Patterns of reconviction at Stage 1 are summarised in Table 2.

Overall, young people with a higher number of previous convictions were more likely to be reconvicted during Stage 1 (Mann-Whitney U test p = 0.042). The reverse was true in relation to offence gravity. Those who had committed more serious offences during the 12 months prior to baseline (including the index offence) were less likely to be reconvicted than those who had committed less serious offences during this period. Only 43 per cent of those with higher baseline gravity scores [6-7] were reconvicted, compared to 76 per cent of those with lower gravity scores [3-5] (Chi-square Test p = 0.020). (4) Offences with a high gravity score included both violent offences and other serious crimes. Committing a violent offence at baseline was also associated with reconviction. We saw in Table 1 that the comparison group was more than twice as likely to have an index offence of violence. At Stage 1 follow-up, 77 per cent of those who had committed an index offence of violence were reconvicted, compared to 47 per cent of those who had not committed a violent offence at baseline (Chi-square Test p = 0.047).

To further test our finding that the IF group was less likely to be reconvicted, we used logistic regression to determine the factors which predicted reconviction by Stage 1 follow-up. A series of logistic regressions were run, initially without controlling for other factors. The only independent variables which showed a significant bivariate relationship with reconviction were the intervention group and the offence gravity score in the 12 months prior to baseline. These two variables were then entered into a multivariate logistic regression (enter method). Table 3 presents the results of the bivariate analyses and (in the column on the right) the final model.

Multivariate analysis confirmed that during the year after entry to the placement, Intensive Fostering significantly lowered the odds of reconviction once the gravity of previous offences was taken into account. The model explained 31.5 per cent of the variance in reconviction. The inclusion of these variables in the model increased the percentage of outcomes correctly predicted from 57.4 to 72.3. The regression was run again with conviction for a violent offence at baseline as a covariate, as this came close to significance in the bivariate analysis. However, intervention group and previous offence gravity remained the only two significant variables, so conviction for an index violent offence was not related to reconviction once intervention group and offence gravity were taken into account (p = .202).

Reconviction during Stage 2

In contrast to our findings on reconviction during Stage 1, during the year after the young people left their IF foster placements they were as likely to be reconvicted as the comparison group. The proportion of the IF group who committed recorded offences during this period (74%) was virtually identical to that for the comparison group (75%).

During the Stage 2 follow-up period, the number of offences committed by the IF group rose sharply, from a mean of .61 (a group total of 14) during Stage 1 to a mean of 2.7 (62 offences in total) by the end of Stage 2. Although this was still lower than the mean of 3.29 for the comparison group (78 offences in total), the group difference in the number of offences committed was no longer significant. Nor was there any longer a significant difference in the mean number of offences per day at liberty during Stage 2, or in the gravity scores for the most serious offences committed during this follow-up period, as shown in Table 4.

As we saw in Section 4.3, during the year after entry to the IF placement, the mean number of recorded offences committed by the IF group per day at liberty (that is, out of custody) was 0.0024. However, during the year after exit from the placement, this increased nearly fourfold to 0.0086 offences per day at liberty, a statistically significant increase (Wilcoxon Signed Ranks test significant at p = 0.005).

The variables presented in Table 3 were again entered into logistic regressions to test for any bivariate relationships. This showed that intervention group no longer predicted reconviction. The only variable which predicted reconviction during this follow-up period was the gravity score for the most serious offence in the 12 months prior to baseline. As at Stage 1, a lower baseline score for offence gravity increased the risk of (recorded) reoffending, raising the odds of reconviction by 0.080 (significant at p = .022).

Entry to custody

Only 22 per cent of the IF group entered custody during Stage 1, whereas 50 per cent of the comparison group had (re)entered custody by follow-up. In the year after entry to the foster placements, therefore, the IF programme appeared to have a protective effect on many young people it served, as the IF group spent just under 60 per cent fewer days in custody than the comparison group (an average of 32 days compared to 75 days). This was similar to the difference in incarceration rates found in studies of MTFC in the USA (Chamberlain, 1990; Chamberlain and Reid, 1998; Leve et al, 2005).

All five of those in the IF group who entered custody during this period were young people whose placements disrupted in the early stages of the programme, but only one of them did so as a consequence of a substantive offence (criminal damage). The other four were sentenced to custody following convictions for the technical offence of breach of sentence after they absconded from their foster placements.

During the year after the young people in the IF group left their foster placements four additional young people entered custody, making a total of nine (39%) sentenced to custody by the end of Stage 2. By this point, the difference between the groups in the use of custody was no longer significant. Neither was there a significant difference in the mean number of days in custody, although this remained lower for the IF group (44) than for the comparison group (75), as shown in Table 5.

Results were more positive for the 18 young people who remained in their IF placements for a minimum of nine months (the intended duration of the placements). These had significantly fewer days in custody by the end of Stage 2 than the comparison group, with a mean of 14 days in custody for the IF group and 75 for the comparison group (Mann-Whitney U test p = .018.).

Secondary outcomes

Secondary outcomes were assessed at the end of Stage 1, at which point four of the IF group were still living in their IF placements. One young person in the comparison group had also been sentenced to IF by this time and was in placement at this point. One year after entry to the IF placement or, for most of the comparison group, release from custody, the IF group were more likely to be living with parents and less likely to be in custody (Chi-square test p = .002), as shown in Table 6.

The IF teams successfully reintegrated many of the young people into education. This was often difficult, as the teams struggled to find any suitable schooling or training places. Due to the large geographical areas covered by two of the IF teams, some young people moved to new schools close to their foster placements but were unable to remain in them once they returned to their home area. However, schools willing to offer a place could not always be readily found in their home areas.

Having an anti-social peer group is a known risk factor for offending (Leve and Chamberlain, 2005; Wilson et al, 2006). Research on MTFC in the USA reported that it was significantly better at reducing delinquent peer association than group care, but this was not the case in our study (Leve and Chamberlain, 2005). As we saw in relation to characteristics and histories of the young people and families, at baseline the majority of young people in both groups were associating with pro-criminal peers. Data from YOT workers on 43 (91%) of the sample at Stage 1 indicated that the two groups remained equally likely to have pro-criminal peers (67% of each group). By this point most of the IF group had left their foster placements. Young people reported to be associating with anti-social peers at Stage 1 were significantly more likely to have been reconvicted (72%) than those not reported to have anti-social peers (21%) (Chi-square test p = 0.002). They also committed roughly five times as many offences per month at liberty in the community as those not reported to have anti-social peers (0.33 offences per month compared to 0.06 offences per month at liberty for those without an anti-social peer group (Mann-Whitney U test p = 0.007).

Three-quarters (17) of those sentenced to IF were interviewed at follow-up, 13 of whom had been reported to have antisocial peers at baseline. Follow-up inter views with these young people indicated that nearly half (six) of those with antisocial peers at baseline had been diverted from them while they were living in their IF placements. Two of the IF teams covered a wide geographical area, so diversion from former peers was some times reinforced by placement many miles from home. Consistent with our quantitative findings reported above, none of these had been reconvicted by Stage 1 follow-up, whereas all seven who had continued to associate with antisocial peers had reoffended by this point.

Drug use has also been identified as one of a number of interacting risk factors that may contribute to anti-social behaviour (Farrington, 1996; Hine and Celnick, 2001; Wilson et al, 2006). Professional reports of drug misuse may underestimate the full extent of the problem.

Nevertheless, these indicated that just under two-thirds of the young people (30) were known to be using drugs at baseline, in most cases cannabis rather than hard drugs. Drug use was associated with a higher frequency of offending (Mann-Whitney U test p = .034) and with higher scores for the gravity of offences (p = .003). All but one (93%) of the young people reported to be using drugs were also reported to be associating with anti-social peers (Chi-square test p = .016). (5) Our data therefore suggest that drug use, in combination with a return to associating with antisocial peers, may have contributed to the risk of reoffending.

Discussion of results

Our findings suggest that Intensive Fostering may be effective in containing persistent young offenders in a community setting, reducing the use of custody, delaying reconviction and reducing the number of offences recorded during the time they are on the programme. However, the dissipation of the positive effects of the programme after the young people left their foster placements indicates that it was difficult to sustain positive changes in maladaptive learning and relationships once they were re-exposed to the risk factors in their local environment. For a number of the young people therefore, learned pro-social behaviour did not appear to be fully internalised, with the result that situational factors had a powerful effect on their behaviour once they left their foster placements. The implication of this is that the changes that occurred when young people entered their IF placements and after they left them may have been due, at least to some extent, to changes in their environment.

Our evidence suggests that the change of environment resulting from entry to the IF placements had a protective effect. The majority of the IF group spent much of the year after entry to the programme living in the closely supervised environment of their foster placements, and most of those reconvicted began to reoffend only after they left these placements. With training in MTFC and intensive support from a clinical team, IF carers provided care, a mentoring relationship, supervision, clear boundary-setting and the consistent reinforcement of positive behaviours. Although some of the young people's parents were clearly caring too, they had not provided the consistent, authoritative parenting and limit-setting that the IF carers were able to offer. The effects of the IF behaviour management programme delivered in a new environment were reinforced by the IF team's work to promote the young people's participation in education and other structured activities. Apart from any other beneficial effects, this further reduced the young people's opportunity to offend as they had very little unstructured and unsupervised time, particularly in the early stages of the placement.

The close supervision provided while the young people were living in their foster placements helped to reduce deviant peer influences while they were in placement and this, too, contributed to the reduction in antisocial behaviour during Stage 1. Our follow-up inter views with young people on the IF programme indicated that, for those who committed only one or two minor offences or did not reoffend at all during Stage 1, a reduction in contact with negative peer groups was often a feature of the change in patterns of offending, particularly while they were in their foster placements. This is consistent with research in the USA, which found that a reduction in deviant peer association and foster carers' family management skills mediated the effect of placement in MTFC (Eddy and Chamberlain, 2000). Through diversion from anti-social peers, the IF programme may interrupt the mediational chains whereby criminal identities and behaviours are reinforced by association with negative peer groups, at least while the young people are in their foster placements. In contrast, for most of the comparison group the use of custody de facto reinforced their congregation with anti-social peers. Interventions which congregate anti-social peers may reinforce negative behaviours rather than reduce them, as research suggests that young people at high risk of antisocial behaviour may be particularly vulnerable to negative peer influences when placed in group settings (Dishion et al, 1999). However, some have suggested that the iatrogenic effects of negative peer influence in group settings have been overstated (Weiss et al, 2005).

Environmental effects also appeared to be powerful after the young people left their foster placements. Although we were unable to investigate the supervision provided in the environments to which the young people returned, it is highly unlikely that their parents or other carers/lodgings providers would maintain as high a level of supervision as that provided by IF foster carers and clinical teams. Many other studies have similarly found that a wide range of difficult behaviours may abate when children or young people are removed from troubled families, but that the behaviour is reinstated when the children return home (Allerhand et al, 1966; Sinclair, 1971; Zimmerman, 1982; Minty, 1987; Taussig et al, 2001; Sinclair et al, 2005). In particular, those who return to homes that are disharmonious or marked by poor parenting have been found to do less well (Sinclair, 1971; Quinton and Rutter, 1988; Sinclair et al, 2005). For some, these environmental effects may be reinforced by broader neighbourhood effects, as a concentration of deprivation in the surrounding environment may have a significant impact both on parenting style and on young people's offending behaviour (McVie and Norris, 2006).

The distance of some IF placements from the young people's home areas, which was due to the wide geographical area covered by two of the pilot programmes, had both positive and negative consequences. For some young people, diversion from former anti-social peers was accomplished when they moved to foster placements many miles from home and, as previous research has found, young offenders placed away from their home environment are less likely to be pressurised to engage in criminal behaviour than those placed close to home (Sinclair and Gibbs, 1998). However, at follow-up young people in the IF group were as likely to have anti-social peers as those in the comparison group. Our interviews with young people indicated that once they returned home, they often re-engaged with their former anti-social peer groups. Also, as we have seen, after entry to the IF programme some young people moved to new schools close to their foster placements, but schools willing to offer a place could not always be found once they returned to their home areas. For some, therefore, the distance of the foster placements from home meant that the positive effects of re-engagement in education dissipated once they returned home.

Deficiencies in the follow-up support provided may also have contributed to the poor outcomes after young people left their foster placements. The IF programme includes aftercare by the IF teams, provided for a period of three months. However, in a number of cases recurring staffing difficulties appeared to compromise the aftercare support pro vided by the teams. The accounts of the parents interviewed at follow-up suggested that there were sometimes serious inadequacies in the nature and intensity of aftercare support provided by the IF teams. Parents were also unhappy about the low level of support provided by local YOTs, which are also expected to provide follow-up support, as this often amounted to little more than a fort nightly interview with the young people. This lack of continuing support further increased the risk that young people would fail to make constructive use of their time and would drift back to their former pro-criminal peers, increasing the likelihood that they would reoffend.

Studies of youth offending have consistently found that the effects of treatment on young people's behaviour are weaker once they return to their communities and have argued that more attention should be paid to the environments to which they return and to post-treatment social adjustment (Rutter et al, 1998; Bottoms, 2006). The duration of the intervention's effects may depend, to some extent at least, on the nature, intensity and duration of follow-up support provided once the young people leave their placements, including support with education or training and diversion from peers. What happens after placements end is clearly crucial. Local services need to undertake proactive and intensive work with these young people and families once they return to their communities, since without this there seems little chance that any gains made during IF placements will be sustained.

Limitations to the study

The main limitation to the study was the small sample size, which was determined by the small number of IF placements available and the research funding available. However, the fact that the results at Stage 1 were highly significant despite the small sample size (which would normally make it difficult to detect an effect) and were, moreover, consistent with previous research on MTFC, suggests that the group difference identified at Stage 1 was unlikely to be a false positive. Equally, the almost identical proportions of the two groups reconvicted by Stage 2 suggest this finding was unlikely to be a false negative. Furthermore, the study's repeated measures design, which showed that the effect of the intervention dissipated once it was removed, reduces the threat to causal inference. Nevertheless, it remains a possibility that, due to small sample size, there may have been a small to moderate effect at Stage 2 which we were unable to detect.

Acknowledgements

With thanks to Jonathan Green, Catherine Randerson, Sharon Mallon, Catherine Kay and Linda Cusworth.

References

Allerhand M, Weber R and Haug M, Adaptation and Adaptability, New York: Child Welfare League of America, 1966

Biehal N, 'Foster care for adolescents', in Schofield G and Simmonds J (eds), The Child Placement Handbook, London: BAAF, 2009

Bottoms A, 'Desistance, social bonds and human agency: a theoretical exploration', in Wikstrom P-O and Sampson R (eds), The Explanation of Crime: Context, mechanisms and development, Cambridge: Cambridge University Press, 2006

Chamberlain P, 'Comparative evaluation of foster care for seriously delinquent youths: a first step', Community Alternatives: International Journal of Family Care 2:2, pp 21-36, 1990

Chamberlain P, 'The Oregon multidimensional treatment foster care model: features, outcomes, and progress in dissemination', Cognitive and Behavioral Practice 10:4, pp 303-12, 2003

Chamberlain P, Ray J and Moore KJ, 'Characteristics of residential care for adolescent offenders: a comparison of assumptions and practices in two models', Journal of Child and Family Studies 5:3, pp 285-97, 1996

Chamberlain P and Reid J, 'Using a specialized foster care community treatment model for children and adolescents leaving the state mental hospital', Journal of Community Psychology 19, pp 266-76, 1991

Chamberlain P and Reid J, 'Comparison of two community alternatives to incarceration for chronic juvenile offenders', Journal of Consulting and Clinical Psychology 6:4, pp 624-33, 1998

Dishion TJ, McCord J and Poulin F, 'When interventions harm: peer groups and problem behaviour', American Psychologist 54:9, pp 755-64, 1999

Eddy J and Chamberlain P, 'Family management and deviant peer association as mediators of the impact of treatment condition on youth anti social behaviour', Journal of Consulting and Clinical Psychology 68:5, pp 857-63, 2000

Farmer E, Moyers S and Lipscombe S, Fostering Adolescents, London: Jessica Kingsley Publishers, 2004

Farrington D, 'The development of offending and antisocial behaviour from childhood: key findings from the Cambridge study in delinquent development', Journal of Child Psychology and Psychiatry 360:6, pp 929-64, 1995

Farrington D, Understanding and Preventing Youth Crime, York: Joseph Rowntree Foundation, 1996

Fisher P, Burraston B and Pears K, 'The early intervention foster care program: permanent placement outcomes from a randomized trial', Child Maltreatment 10:1, pp 61-71, 2005

Hine J and Celnick A, A One-year Reconviction Study of Final Warnings, Sheffield: University of Sheffield, 2001

Home Office, Home Office Counting Rules for Recorded Crime, 2011; www. homeoffice.gov.uk/publications/ science-research-statistics/research-statistics/ crime-research/ countingrules/count-intro?view=Binary

Leve L and Chamberlain P, 'Association with delinquent peers: intervention effects for youth in the juvenile justice system', Journal of Abnormal Child Psychology 33:3, pp 339-47, 2005

Leve L, Chamberlain P and Reid J, 'Intervention outcomes for girls referred from juvenile justice: effects on delinquency', Journal of Consulting and Clinical Psychology 73:6, pp 118-184, 2005

Macdonald G and Turner W, 'Treatment foster care for improving outcomes in children and young people (review)', The Cochrane Database of Systematic Reviews, Art No: CD005649, DOI: 10.1002/14651858.CD005649.pub2, 2009

McVie S and Norris P, Neighbourhood Effects on Youth Delinquency and Drug Use, Edin burgh: Centre for Law and Society, University of Edinburgh, 2006

Minty B, Child Care and Adult Crime, Manchester: Manchester University Press, 1987

Quinton D and Rutter M, Parenting Breakdown: The making and breaking of intergenerational links, Aldershot: Avebury, 1988

Runyan D and Gould C, 'Foster care for child maltreatment: impact on delinquent behaviour', Pediatrics 75:3, pp 562-68, 1985

Rutter A, Giller H and Hagell A, Antisocial Behavior by Young People, Cambridge: Cambridge University Press, 1998

Sinclair I, Hostels for Probationers, London: HMSO, 1971

Sinclair I, Fostering Now: Messages from research, London: Jessica Kingsley Publishers, 2005

Sinclair I, Baker C, Lee J and Gibbs I, The Pursuit of Permanence: A study of the English care system, London: Jessica Kingsley Publishers, 2007

Sinclair I, Baker C, Wilson K and Gibbs I, Foster Children: Where they go and how they get on, London: Jessica Kingsley Publishers, 2005

Sinclair I and Gibbs I, Children's Homes: A study in diversity, Chichester: John Wiley & Sons, 1998

Taussig HN, Clyman RB and Landsverk J, 'Children who return home from foster care: a 6-year prospective study of behavioral health outcomes in adolescence', Pediatrics 108:1, p 10, 2001

Thornberry T, 'The prevention of serious delinquency and violence: implications from the program of research on the causes and correlates of delinquency', in Howell James C (ed), A Sourcebook: Serious, violent and chronic juvenile offenders, London: Sage Publications, 1995

Thornberry T, Ireland T and Smith, 'The importance of timing: the varying impact of childhood and adolescent maltreatment on multiple problem outcomes', Development and Psychopathology 13, pp 957-79, 2001

Weiss B, Caron A, Ball S, Tapp J, Johnson M and Weisz J, 'Iatrogenic effects of group treatment for antisocial youth', Journal of Consulting and Clinical Psychology 73:6, pp 1036-44, 2005

Westermark P, Hansson K and Olsson M, 'Multidimensional treatment foster care (MTFC): results from an independent replication', Journal of Family Therapy 33:1, pp 20-41, 2011

Widom C, 'The role of placement experiences in mediating the criminal consequences of early childhood victimization', American Journal of Orthopsychiatry 61:2, pp 195-209, 1991

Wilson D, Sharp C and Patterson A, Young People and Crime: Findings from the 2005 offending, crime and justice survey, London: Home Office, 2006

Youth Justice Board, ASSET Young Offender Assessment Profile, 2011; www. yjb.gov.uk/en gb/search?LinkClick=%2Fcgi-bin%2FMsmGo. exe%3Fgrab_id%3D0% 26page_id%3D146% 26query%3Dasset%26hiword%3DASSETS%25 20asset%2520

Zimmerman R, 'Foster care in retrospect', Tulane Studies in Social Welfare 14, pp 1-119,

* This article was first published in Children and Youth Services Review 33, pp 2043-49, October 2011, and is reproduced here with the kind permission of the Editor.

(1) The authors will shortly be reporting the results of a larger companion study of MTFC in England (n = 219). This focuses on an MTFC programme targeted at adolescents in out-of-home care.

(2) The four young people sentenced to ISSP were included in the comparison group at the request of the government department which funded the evaluation.

(3) Mean Asset scores were 2.7 for IF group and 3 for comparison group (family and personal relationships domain) and 2.9 for IF group and 3.1 for comparison group (lifestyle domain). Both differences were non-significant.

(4) Since the offence gravity variable had only five values and the data were not normally distributed, this was recoded as a binary variable.

(5) The reverse relationship was less marked, as drug misuse was reported for just 45 per cent of those reported to have anti-social peers.

Nina Biehal is Research Director, Sarah Ellison Research Fellow and Ian Sinclair Emeritus Professor, Children and Young People's Social Work Team, Social Policy Research Unit, University of York, UK

Table 1
The samples at baseline (n = 48)

                                                Comparison
                                   IF group     group        Sig. p

Male (a)                           83%          83%          ns

Mean age (years) (b)               14.9 years   15.5 years   ns

Experience of maltreatment (a)     52%          63%          ns

Mean age at first conviction       12.7 years   13.1 years   ns
(years) (b)

Mean number of previous recorded   15.26        17.38        ns
offences (b)

Mean number of recorded offences   7.54         7.7          ns
in past 12 months (b)

Previous custodial sentence (a)    26%          33%          ns

Mean gravity score for most        5.22         5.21         ns
serious previous offence (b)


Mean gravity score for most        4.7          5            ns
serious index offence (b)

Index offence of 'violence         21%          50%          .035
against the person' (a)

(a) Chi-square test.

(b) Mann-Whitney U test

Table 2
Patterns of reconviction at Stage 1 follow-up

                                                 Comparison
Outcome                               IF group   group        Sig.p

Reconvicted n = 47% (n) (a)            39 (9)    75 (18)      0.019

Mean days to first recorded offence   287        89           <0.001
n = 47 (b)

Total number of offences n = 45 (b)    14        78           0.003

Mean number of offences n = 45 (b)       .61      3.29        0.003

Mean number of offences per day at       .0024    0.0128      0.002
liberty n = 45 (b)

Mean gravity score n = 45 (b)           1.87      3.65        0.004

(a) Chi-Square Test: X2 6.182, df1

(b) Mann-Whitney U test

Table 3
Odds of reconviction during Stage 1 year of follow-up n = 47

Variable                 Not controlling    Sig. p   Multiple
                        for other factors            variables
                           Odds ratio                Odds ratio (p)

Age at baseline                .778         0.262
Total number of               1.057         0.072
  offences
Number of offences            1.162         0.066
  in 12 months pre-
  baseline
Days in IF placement           .988         0.416

Index offence of
  violence
No                             .269         0.053
Yes                           1

Highest offence
  gravity in 12
  months pre-baseline
Higher score(6-7)              .229         0.023*   .184 (p = 0.019)*
Lower score (3-5)             1

Intervention group
IF group                       .214         0.015*   .174 (p = .013)*
Comparison group              1

Constant = 9.002 ** Nagelkerke R2 = .315.
Significant differences are shown in bold type.

Note: Significant differences indicated with *.

Table 4
Patterns of reconviction at Stage 2 follow-up

                                                 Comparison
Outcome                               IF group     group      Sig. p

Reconvicted n = 47% (n) (a)            74%         75%         .1
Total number of offences n = 45 (b)    62          78          .789
Mean number of offences n = 45 (b)      2.7         3.29       .789
Mean number of offences per day at       .0086       .0128     .589
  liberty n = 45 (b)
Mean days to 1st offence n = 47 (b)   129          89          .401
Mean gravity score n = 45 (b)           3.26        3.65       .761

(a) Chi-square test.

(b) Mann-Whitney U test

Table 5
Entry to custody during Stage 1 and Stage 2 follow-up

Outcome                             Stage 1
                            IF     Comparison   Sig. P
                           group     group
Entered custody (a)         22%       50%       .044*
Mean days in custody (b)    32         75       .038*

Outcome                             Stage 2
                            IF     Comparison   Sig. p
                           group     group

Entered custody (a)         39%       50%        .454
Mean days in custody (b)    44         75        .253

(a) Chi-square test.

(b) Mann-Whitney U test. Significant differences shown in bold type.

Note: Significant differences indicated with *.

Table 6
Accommodation at one-year follow-up (n = 43) per cent (n)

                                           IF group    Comparison group
Accommodation                            per cent (n)    per cent (n)

With parent (19) or other relative (1)      56(13)         29 (7)
In custody                                  --             38 (9)
IF placement                                17 (4)          4 (1)
Other foster placement                       9 (2)          4 (1)
Supported lodgings or own accommodation      9 (2)         12 (3)
Sleeping rough                              --              4 (1)