Corrections_Today_January_February_2019

Office of Correctional Health

Predictive analytics show the future Predictive analytic studies of this project are key in determin- ing sustainability. By applying the policy and procedure rules from the expanded AS era to the current popu- lation of the NDDOCR, the analytics program can project the number of admissions, census and bed days that would be needed in absence of the BIU program redesign. Comparing this to the actual admission data from the operational analytics study shows this program effectively manages a large proportion of these at-risk residents in a less restrictive setting by maintaining the BIU at the current staffing levels (Figure 5, page 69). This type of information is critical in defending cost estimates and man- power allocations during budget and legislative sessions. In absence of program changes, unit size and bed- day utilization can be forecasted by applying operational analytic results to projected populations. One ex- ample is the prediction of admission rate to BIU based on analysis of the demographics and data of residents currently incarcerated and those newly admitted to the NDDOCR (Figure 6, page 70). Prescriptive analytics can impact the future Predictive analytics is key in developing prescriptive analytic programs. In this study, the Querent

Querent computes Significant Variables & Risk Stratification Figure 1 — Querent computes Significant Variables & Risk Stratification

Traditional AS: Significant Variables

17%

18%

20% 15% 10%

11% 10% 9%

6%

4%

4%

3%

2%

5% 0%

Sentence Duration

Age

First Arrest Age

Criminal History Count

Drug Crime Count

Previous AS

Custody Max

Custody Med

Custody Min

Violent Offense Count

Variable Importance

Variables

Expanded AS: Significant Variables

17% 15%

18%

20% 15% 10%

11%

11%

2%

3%

3%

3%

3%

5% 0%

Race (Caucasian)

Sentence Duration

Age

Criminal History Count

Drug Crime Count

First Arrest Age

Offense Type Count

Not Employed

Custody Min

Violent Offense Count

Variable Importance

Variables

BIU Programming: Significant Variables

13%

13%

15% 10%

11% 9%

8%

8%

6%

6%

5% 5%

5% 0%

Sentence Duration

Custody Max

Sentence Duration

Drug Crime Count

Criminal History Count

No Education

Violent Offense Count

Race (Caucasian)

Custody Min

First Arrest Age

Variable Importance

Variables

a significant factor in determining placement into restrictive housing dur- ing any of the time periods. (Figure 1). Operational analytics prove program success and safety Operational analytics information yielded several insights. Establishment of the BIU system has led to an 81 percent decrease in census (Figure 2, page 67) and an 83 percent decrease in bed days per month (Figure 3, page 68). Monthly admission rates into AS/

BIU have dropped to single digits. Important information for line staff is that the incidence of violent and disorderly behavior as measured by analysis of disciplinary reports did not increase. Staff can see that providing positive feedback by issuing positive behavior reports clearly improves the long-term behavior of residents. Pro- gram success is clear: admission rate is 14 percent of the previous rate and readmission rates are at 20 percent, which is less than half of the histori- cal rate. This translates in to less than one readmission per month to the BIU (Figure 4, page 69).

66 — January/February 2019 Corrections Today