Improving Social Services, Privacy and Project Management Top 2008 Watch List

Most of the projects highlighted here were selected because they involved strategies many state and local governments are still deciding whether or not to try. Industry observers are watching for the success or failure of the Indiana Family and Social Services Administration's (FSSA) attempt to outsource food stamp eligibility. All states are seeking ways to cut Medicaid costs, making Tennessee's Medicaid fraud analytics process critical to watch. Most state and local governments are investigating best practices for redacting Social Security numbers from online public documents. The Richmond (Va.) Police Department's (RPD) crime-predicting system will certainly lead to clones elsewhere if it further reduces crime, and the District of Columbia's "stock portfolio" approach to IT project management is already making waves.

A Step Ahead of Crime

During 2008, many government IT observers took special notice of an analytics-based tool for predicting crimes that the RPD developed.

The system, which gives police a color-coded map showing where crime is likely to be most intense, features a new function for 2008 that classifies the crime it measures into six categories -- burglary, robbery, auto theft, theft from auto, violent crime and other larceny.

"If you're predicting increased robbery activity versus auto theft or a burglary, we have different tactics to prevent those types of crimes," said Stephen Hollifield, information systems manager for the RPD.

Hollifield uses SPSS Predictive Enterprise Services to run the analytics. The system tries to improve its accuracy each week by automatically creating a new crime-predicting algorithm and testing it against the one used the previous week. Hollifield calls this the "champion-challenger" process. At the end of each week, the analytics tool uses the crime information collected to create a new "challenger" algorithm; the existing algorithm is the "champion." The analytics system applies the challenger algorithm to crime records from the past five years, and it tests how accurate the challenger would have been at predicting the crimes that actually happened. If the challenger algorithm would have been more accurate than the champion algorithm, then the challenger becomes the new champion algorithm. If not, the champion predicts crime for another week.

The RPD introduced the project in 2006, and in one year, murders dropped 32 percent, rapes went down 20 percent, robberies decreased 3 percent, aggravated assaults and burglaries dropped 18 percent and auto thefts fell 13 percent. Observers are waiting to see if the system's new functionality -- performing the same analytics on six, narrower categories of crime - drops the crime rate further.

Indiana Overhaul

State IT departments that are considering whether or not to consolidate and outsource welfare eligibility processes are watching how the strategy works for the FSSA. The agency has slowly rolled out its program -- outsourced to IBM and its partners -- across the state, enabling any caseworker answering the phone to handle all clients' issues, no matter who they are. In the past, clients often had to wait and call back if their assigned caseworkers were unavailable.

The project aims to accelerate the eligibility filing process, but that goal has been slow in reaching fruition. In the early stages of the rollout, which started in November 2007, the automation slowed processes in the participating pilot counties. In June 2008, the U.S. Department of Agriculture's Food and Nutrition Service (FNS) sent FSSA Secretary Mitch Roob a letter of concern because of the project's troubles.

"Indiana's most recent monthly reports indicate a decline in the timeliness of application processing that has occurred in the pilot region since the transition to the modernized service delivery model," said the letter from Ollice Holden, regional administrator of the FNS.

The automation program was deployed in 59 of Indiana's 92 counties. Before the rollout, participating counties approved 85 percent of benefits applications within 30 days of receiving them. With the new system, those