Jump to main content

A judges gavel sitting on a table

Collaboration to Improve Pretrial Risk Assessment Earns National Recognition

UNC researchers and Weld County Justice Services team up to improve popular tool used by judges to determine if and when individuals should be detained or released on bail

The CPAT-R is a list of 9 survey questions that are asked of an arrestee after they are booked into jail. These questions capture information about their stability within the community and prior involvement with the court and justice system.

Two University of Northern Colorado (UNC) faculty members collaborated with Weld County officials on a criminal justice reform project to improve bail release decisions in 2018. They recently received national recognition for their work.

The National Association of Counties awarded Weld County Justice Services and UNC’s Criminal Justice Program and UNC the Achievement Award for their research project to improve bail release decisions. The group created a risk assessment tool called the CPAT-R (Colorado Pretrial Assessment Tool – Revised), a highly-validated tool that has made the process of pretrial release conditions less biased.

The Need for the CPAT-R 

After a person is arrested, a judge must decide the next steps prior to the trial —allow a person to post bail, which would give them the chance to pay a fee and secure their freedom before their next court appearance or remain in custody to ensure they appear at their next court date. In Colorado, the law requires judges to use a risk assessment tool to inform bond decisions. The tool provides consistent criteria judges can use to help make those decisions. It’s intended to help balance the goals of lowering pretrial jailing while preventing high-risk individuals from being arrested again and not showing up for their next court date.  

Weld County Director of Justice Services Doug Erler says such a tool is essential in helping balance the interests of public safety with the individual rights of arrestees before the court, otherwise, irregular decision-making could occur.  

“Implicit bias would likely become concerning,” Erler said. “Someone could say, you look like a bad dude and that would be enough for a decision. There would also be longer pretrial jail detention, because, without a tool, judges may get skittish and keep defendants in longer just to be safe. Added to that, there would be increased costs to taxpayers and criminal justice systems, possible increased public safety concerns and constitutionality issues would emerge more.” 

Before 2020, most counties were using a tool called the CPAT (Colorado Pretrial Assessment Tool), however, since it was created in 2012, Erler thought the tool could use some upgrades.  

“The CPAT was working well, but there were a few elements that were consistently being challenged,” Erler said. “For example, the tool asked arrestees if they have a working phone, and in today’s age most people do so that’s not a top risk factor.” 

To analyze the data and develop an updated version, Erler turned to Criminology and Criminal Justice Associate Professors Kyle Ward, Ph.D., and Victoria Terranova, Ph.D. Terranova worked as a pretrial officer in Texas while earning her doctoral degree, so she had good perspective on how tricky bond decisions can be. 

“No one is the same. It’s a hard decision that judges and bonding officers have to make, and it’s a high-volume decision,” Terranova said. “In the correctional process at that stage, you’re seeing the largest amount of people, and you know the least about those people because they’ve been in jail for maybe a day. So you’re making a lot of decisions on the fly.” 

CPAT-R Success 

The goal of the CPAT-R is to help leaders feel confident when making objective and fair decisions about whether to release individuals accused of crimes. The researchers say this happened two-fold with the CPAT-R. Since implementation, more people have been scored as low risk, potentially making it easier for a judge to decide on bail/bond, and the process has lessened ethnicity, gender and race bias. 

“When we say bias what we’re really talking about is if the tool works differently in its accuracy, better or worse, based solely on being a member of one of those subgroups, so totally outside of any legal factors,” Terranova said. “But we also looked at residential status, which was the first time anybody had looked at that.” 

Other questions related to violence and drug use were also altered to limit bias. 

“If there was a history of violence, that didn’t quite work when we ran the analysis. It didn’t end up being an item that needed to be there, both in its predictive ability, but also it might have led to some of the bias contributors,” Ward said. 

According to Weld County, this is the most comprehensive bias assessment conducted of a pretrial risk assessment tool to date. 

“The goal with all pretrial decision making is to ensure that it’s fair and it doesn’t compromise community safety,” Terranova said. “What we’re learning more about pretrial detention every day is that being detained in jail can have detrimental effects while you’re going to court. So, minimizing that without increasing any risk to public safety is good for all.” 

According to the National Association of Criminal Defense Lawyers, arrestees jailed for 48 hours can lose their employment, housing and custody of their children. Studies have also shown that being jailed directly increases an individual's likelihood of being convicted, and once convicted, a higher chance of a harsher sentence.  

How the CPAT-R was Created 

Generating the new risk assessment tool took two years of Terranova and Ward conducting thorough analysis in a three-phase approach. 

“The first phase was looking at the current version of the CPAT, which was an empirically based tool but not so much of a validated tool,” Ward said.  

To better understand how the tool was being used, Ward and Terranova sorted through 3,386 cases from 2015 and 2016 and sent an online survey to 382 criminal justice stakeholders including pretrial executives, judges, defense attorneys and prosecutors, to verify the accuracy and effectiveness of the CPAT. 

“We wanted to see the outcomes of some of the people who were arrested,” Ward said. “Did they FTA (fail to appear in court)? Then we followed up with them to find out what happened in the cases and why.” 

During phase two of the project, Ward and Terranova went on the road to observe pretrial services agencies in person and conduct focus groups with private and public attorneys, prosecutors and public defenders. Finally, phase three began construction of the alternate tool based on the information gathered in the previous phases. 

“We included all the survey questions that were in the CPAT originally and added additional types of information focusing on community stability. We included factors on employment, where they lived and for how long, and court and criminal history, if there is a history of not showing up to court or if this is their first arrest, things of that nature,” Terranova said.  

Terranova says they took a list of 90 questions and selected the most predictive ones using machine learning variable selection techniques and ended up with 11 risk factor questions that make up the CPAT-R. 

The CPAT-R was then piloted in 2019 in seven Colorado counties starting with Weld County. During the multiple-year analysis of the tool, Terranova and Ward had the counties use both the CPAT and CPAT-R in their assessments, so they could compare the results. Overall, the new tool worked and validated higher than any other being used throughout the country, sparking most counties in Colorado to use it.  

Seeing that success and now receiving recognition for it, Ward and Terranova said shows they helped make something meaningful.  

“As researchers, we have been very fortunate to have such a successful collaboration with all the agencies we worked with, and it really was a kind of all-hands-on-deck effort of listening to the needs of these different agencies and then acting on them. I think that’s the strength of the study,” Terranova said. “Besides that, it’s been really cool to hear how the tool is working and hopefully making the process better and more accurate.” 

Though the CPAT-R is updated and working successfully now, Ward says, there is a chance they may have to start this process all over again in the coming years.  

“Things change, policies change and risk assessment tools should be looked at every five to 10 years to make sure the data still holds up,” said Ward. “So, maybe it will be us, or maybe somebody else will take it on, but we’re ready, we know what we’re doing.” 


Several agencies provided funding for the research project. The Colorado Governor’s Office of State Planning and Budgeting supported financing of $125,429, and the University of Northern Colorado provided internal funding of $22,500. 

More Stories