Mental illness in prison reduced by HarrisLogic predictive analytics

There are ten times more
mentally ill people in prisons than in psychiatric hospitals in
the US.
Stringer/Reuters

• HarrisLogic uses data to identify prisoners that
  require intensive treatment for mental
  illness. 
• The company’s analytics can predict who will
  return to prison within six months with a high degree of
  accuracy.
• By cutting back on treatment costs and recidivism,
  HarrisLogic has saved Dallas County $30 million over the
  course of four years. 

---

The majority of America’s prison population is
mentally ill, but there are few ways to effectively diagnose
and care for prisoners on a massive scale. 

In Dallas, the technology and clinical services
company HarrisLogic is attempting
to solve this problem using data-driven tools to determine the
appropriate behavioral health services for prisoners. By pooling
information from jails, police departments, emergency services,
mental health and social services, courts, and hospitals, the
company has saved Dallas County $30 million over the course of
four years.

This process of aggregating data across agencies, then using it
for predictive modeling and analytics, could have major
implications for prisons across the country. While mental health
care in US prisons is
notoriously inadequate, even prisons that offer decent care
tend to treat all prisoners the same, according to Hudson Harris,
the company’s chief engagement officer. 

It once took the Dallas County Criminal Justice Department
four to six weeks to identify mentally ill prisoners. HarrisLogic
now knows within 15 minutes when a prisoner is booked into
prison. The company then quickly contacts the prisoner’s public
defender and case provider to obtain consent for an evaluation.

On average, HarrisLogic evaluates 350 prisoners each month. Half
of these prisoners go through the company’s Jail Diversion
Program, which devises a care plan based on the prisoner’s
individual history and needs. These evaluations, combined with
the company’s robust database, help determine which prisoners
require higher levels of care, such as inpatient
services. By reducing the likelihood that a prisoner will
receive unnecessary treatment, HarrisLogic says it has reduced
higher-level care costs by 25% in the Dallas County prison
system. 

The company has also slashed costs by reducing the likelihood
that a prisoner will commit a future offense. Using predictive
analytic software from SAP, the company says it can predict who
will return to prison within six months with 72% accuracy, and
who won’t return with 99% accuracy. 

These are some big claims, but the idea that appropriate care can
reduce recidivism is
well-supported. “If you book into jail with a mental health
condition, your odds of returning are 67%,” says Harris. “We
worked on quantifying the variables and factors that were driving
people to come back.”

In many cases, he says, formerly homeless prisoners with
mental health conditions are more likely to return to the system.
But to avoid sweeping assumptions, the company combines its
knowledge of a prisoner’s criminal and mental health history
with other factors that might influence one’s behavior, such
as economic conditions.

This information is helpful in not only identifying individuals
that require extra care, but also in identifying people that may
be predisposed to more severe forms of mental illness. With this
knowledge at hand, the company has successfully reduced the
average daily jail population in Dallas County by 20%. 

There are reasons to be
skeptical of algorithms, which often fail to account for
human bias or changes in patterns and behavior. Criminals with
long arrest records, for instance, may come from neighborhoods
where
arrests are more frequent.

But there’s still a need to quantify mental health among
prisoners. “There are a lot of issues around the morality of
data, what you’re doing with it, how you’re helping people,” says
Harris. “[But] what we found is, even without the predictive
[analytics], having evidence-based treatment had a tremendous
impact on the quality of care.”