Corrections_Today_January_February_2019

Correctional Health Perspectives

seven-fold increase in the staff-to- resident ratio for this population. Custody staff have also become essential team members in delivering behavioral health training and educa- tion to residents. Staff acceptance of such a disruptive change in day- to-day operations has been mixed. Sustainability of this new program depends heavily on staff buy-in. Ob- jective measurements of success and safety are needed. Analytics to the rescue! The NDDOCR has developed and implemented a dramatically differ- ent and very progressive behavioral intervention program that has cut monthly bed-day occupancy in restric- tive housing to less than one-seventh of the previous usage. The NDDOCR needs to show that the program elimi- nates isolation of seriously mentally ill and minimizes use of restrictive housing while maintaining safety and improving long-term behavior of the residents. The organization needs to determine if the program is sustainable and needs to be able to predict pro- gram volume. Leadership wants to be able to identify individuals at high risk for placement in restrictive housing within 30 days of admission. The goal is to target pre-emptive programming to change behavior before crises occur. The NDDOCR needed data to satisfy these needs and goals. The or- ganization reached out to the private sector to identify a partner for this pilot analytics project. Guy Scalzi,

CEO at Vantari Health, responded and offered to undertake the project at no cost to the NDDOCR. Scalzi has extensive expertise in correc- tions and in health care information technology (IT) best practices. He engaged Vamsi Chandra Kassivaj- jala, CEO at Enlightiks, which is an analytics company that is famous in the e-healthcare globally. 2 Together they assembled an international team of experts in analytics, corrections and health care IT. Off-site analytics access critical data safely One requirement for this proj- ect is that NDDOCR data remains within its infrastructure. The analyt- ics team achieved this by installing their data extraction and process- ing tools and their analytics suite on servers within the NDDOCR IT department. NDDOCR stakeholders and the analytics team met to deter- mine specifically what the questions and insights the NDDOCR wished to answer. The team then met with NDDOCR IT and business analysts to learn where and how data was stored in the different NDDOCR IT systems. Natural language processing techniques were used when needed to extract data from free-form documents and reports. Enlightiks’ Querent ana- lytics suite then modeled the data to develop meaningful insights. Finally, reports and dashboards were created to convey information in a meaningful format to NDDOCR personnel.

Historical analytics study clarifies NDDOCR restrictive housing history To determine which variables influenced admission of residents into restrictive housing, NDDOCR line staff, case managers, unit managers, wardens and parole and probation staff were interviewed. A list of more than 75 potential indicators were identified. After preliminary discus- sion and modeling of the information available, 15 potential variables influencing placement into restrictive housing were identified. A historical analytics study was performed on admissions into restrictive housing during three eras, based on changes in policy and procedure: traditional AS (42-month period prior to June 2013), expanded AS (26-month period in 2013-2016) and BIU (27-month pe- riod beginning in January 2016). The analysis compared adult male resi- dents admitted into restrictive housing to all adult male residents housed in NDDOCR prisons during the time periods under study. This study combines data from mental health and medical records with data from the offender management system. This analysis ordered the variables to show their influence on the decision to ad- mit. Age and sentence duration were two of the top three determinants in each of the eras. The analysis proves that the existence of a serious mental illness (SMI) diagnosis at any time in the resident’s history was never

Corrections Today January/February 2019 — 65