Program Profile: Staff Training Aimed at Reducing Rearrest (STARR)

This is a training program for federal community supervision officers providing direct service to persons convicted of a crime who are under supervision. The goal is to improve one-on-one officer-client interactions to reduce risk and recidivism. The program is rated Promising. Clients in the experimental group had statistically significant lower failure rates than control group clients at the 12-month follow up; however, there were no statistically significant differences in failure rates at

A Promising rating implies that implementing the program may result in the intended outcome(s).

Program Goals

Staff Training Aimed at Reducing Rearrest (STARR) is a training program for federal community supervision officers providing direct service to clients (that is, people under supervision). Its overall goal is to reduce clients’ failure rates and recidivism by training officers to use behaviorally based skills during client interactions. STARR teaches officers about a structured cognitive–behavioral supervision approach that seeks to address dynamic risk factors of clients by improving one-on-one officer–client interactions.

Target Population

The STARR training program has been made available to community supervision officers in 16 federal districts. Officers primarily have had moderate- or high-risk clients on pretrial supervision or postconviction supervision.

Program Theory

The STARR curriculum was developed with the risk–need–responsivity (RNR) model (Andrews and Bonta 2003; Andrews, Bonta, and Hoge 1990), which has three core principles:

• Risk principle: The level of services should be matched to the level of the individual. High-risk persons should receive more intensive services; low-risk individuals should receive minimal services.
• Need principle: Target criminogenic needs with services—that is, target those factors that are associated with criminal behavior. Such factors might include substance abuse, procriminal attitudes, criminal associates, and the like. Do not target other, noncriminogenic factors (such as emotional distress, self-esteem issues) unless they act as a barrier to changing criminogenic factors.
• Responsivity principle: The ability and learning style of the person should determine the style and mode of intervention. Research has shown the general effectiveness of using social learning and cognitive–behavioral style interventions.

Program Components

Officers in STARR participate in 3½-day classroom training sessions. The training sessions include a discussion of the theory and research supporting the development of the STARR curriculum (including the RNR model), a demonstration of each skill, exercises, and an opportunity for officers to practice each skill and receive feedback.

The STARR skills taught during training include specific strategies for Active Listening, Role Clarification, Effective Use of Authority, Effective Disapproval, Effective Reinforcement, Effective Punishment and Problem Solving, and for Teaching, Applying and Reviewing the Cognitive Model. Skill cards were developed that outline the specific actions and activities that officers should do to successfully deliver each strategy. Video examples of some of the skills were also presented, while other skills were demonstrated in person. During the training, officers are given the opportunity to practice each skill. For example, after listening to the discussion on reinforcement, officers are asked to identify a behavior and a reinforcement strategy for a specific person, and then role-play that interaction with another officer.

Officers in STARR also send in audiotaped interactions at designated intervals (one before the training sessions and up to 30 after training is completed). They were to send audiotapes from an initial meeting with a client, another audiotape of an interaction with the client 3 months later, and a third and final audiotape 3 months after that. The audiotapes were used to gain a better understanding of skill development and provide officers with feedback.

Four booster training sessions are held in the year following, to provide officers with additional training on skill deficits identified on the tapes. Booster trainings are delivered by phone and include discussion of specific skills, audiotape examples of the skill, and individual feedback and coaching.

Study

The study by Robinson and colleagues (2011) used an experimental pretest/posttest design to examine the impact of Staff Training Aimed at Reducing Rearrest (STARR). Federal probation officers who volunteered to participate in STARR were randomly assigned to the experimental (trained) and control (untrained) groups. The randomization procedure was completed so that 66 percent of the officers were randomly assigned to the experimental group and the remaining officers were assigned to the control group. Although the officers were randomly assigned to the two study conditions, the clients (people on probation) were not. However, as with most community supervision agencies, client assignment was based on rotation, caseload size, or other factors that were unrelated to officer status in the study.

The total number of officers in the experimental group was 41, and the total number of officers in the control group was 26. Clients were identified for inclusion in the study based on when their period of supervision began. The sample included clients on pretrial supervision and postconviction supervision. Pretraining cases were those cases that either 1) began pretrial supervision during 2007 and 2008 and terminated supervision before the training event date, or 2) began postconviction supervision between May 31, 2007, and May 31, 2008. This allowed for a follow-up time of at least 12 months. Posttraining cases were those that either 1) were assigned to study officers for pretrial supervision after May 31, 2009, or 2) were assigned for postconviction supervision after May 31, 2009, and up until Dec. 12, 2009.

In total, there were 446 pretraining cases and 295 posttraining cases assigned to the experimental officers and 345 pretraining cases and 218 posttraining cases assigned to the control officers. Most of the clients were on postconviction supervision. Clients assigned to the experimental officers were 15 percent female and 57 percent minority, with an average age of 35.0 years. Clients assigned to the control officers were also 15 percent female and 57 percent minority, with an average age of 34.6 years. There were no significant differences between the clients on demographic characteristics.

Audiotape recordings of interactions between the officers and their clients were reviewed to measure the use of skills taught during STARR. In total, 731 audio recordings were submitted for review, which included 491 recordings from the experimental group and 240 from the control group. Trained raters coded the audiotapes and concentrated primarily on behaviors consistent with the skills introduced during the training. The analysis concentrated on three intermediate variables: 1) the officer’s use of reinforcement and disapproval, 2) interactions where cognitions, peers, or coping skills were discussed, and 3) the officer’s use of cognitive techniques during interactions with clients. There were no significant differences between the experimental and control groups in the use of these skills found in the analysis of pretraining audiotapes.

There were two client outcome measures used in the study. For pretrial clients, the outcome measure was failure on supervision, including failure to appear in court, supervision being revoked, or being arrested for a new criminal charge while on pretrial supervision. The data for this measure was collected from PACTS (Probation/Pretrial Services Automated Case Tracking System), an electronic case management tool used by probation and pretrial service officers in all 94 federal districts to track federal defendants and convicted persons. For postconviction clients, the outcome measure was arrest for a new criminal behavior. Data on arrests was collected from the National Crime Information Center and the National Law Enforcement Telecommunications System, which together include data on federal and local charges.

Bivariate analyses were used to assess the change in officer behavior from pretraining to posttraining and across the experiment and control groups. Bivariate analyses were also used to assess the impact of training on client outcomes. Multivariate analyses were used to determine the interaction between individual and client characteristic and officer training, and their impact on client outcomes. Subgroup analysis examined the failure rates of clients by risk level.

Study

Lowenkamp and colleagues (2014) used the same sample as in Study 1 by Robinson and colleagues (2011), but extended the follow–up period to look at outcome measures at the 24-month follow up.

The first study (Robinson et al. 2011) included those on pretrial supervision and postconviction supervision. However, due to policies in place with regard to the collection of arrest information, the 2014 follow-up study only included persons on postconviction supervision. The new sample size comprised 999 people; there were 595 in the experimental group and 404 in the control group. The new sample was 14 percent female, 61 percent minority (no other race/ethnicity information was provided), and the average age was 35 years. A little over half of the sample (52 percent) was considered high risk, and the other half (48 percent) was considered moderate risk. There were no statistically significant differences between the groups on baseline characteristics. The study sample included 26 officers in the trained group (experimental group) and 15 officers in the untrained group (control group).

The outcome of interest was failure rates, which was measured as arrests for a new criminal behavior. The differences between the groups was examined using chi-square test statistics that account for the time period supervised (pre- and posttraining) and group assignment (control and experimental group officers). Subgroup analysis examined the failure rates of clients by risk level.

Subgroup Analysis

Robinson and colleagues (2011) looked at the failure rates of clients by risk level. At posttraining, the failure rate of clients in the control group was 32 percent, while the failure rate of clients in the experimental group was 16 percent (a statistically significant difference). Similarly, for high-risk clients, at posttraining the failure rate was 35 percent for experimental group clients and 37 percent for control group clients, a small but statistically significant difference.

Lowenkamp and colleagues (2014) also looked at the failure rates of clients by risk level. For moderate-risk individuals, the failure rate for the experimental group was 28 percent compared with 41 percent for the control group at the 24-month follow-up. Although the result favors the experimental group, the difference was not statistically significant. For high-risk at the 24-month follow-up, the failure rate was the same for both groups (55 percent).

These sources were used in the development of the program profile:

These sources were used in the development of the program profile:

Following are CrimeSolutions-rated programs that are related to this practice:

Pretrial Interventions for Ensuring Appearance in Court

During the pretrial process, defendants may be released on certain conditions. To ensure that released defendants show up to their court date, jurisdictions have used three strategies: 1) court-date reminder notifications, 2) bonds, and 3) supervision in the community. The goal of is to reduce the failure-to-appear rates of defendants. Across the three strategies, the practice is rated Promising for decreasing failure-to-appear rates but rated Ineffective for reducing arrest rates.

Evidence Ratings for Outcomes

	Crime & Delinquency - Multiple crime/offense types
	Justice Systems or Processes - Failure-to-Appear

Why might a practice's outcome ratings differ from the ratings of specific programs encompassed by that practice?