Shortly after midnight on Jan. 7, 1998, a 34-year-old inmate died after a struggle with guards in the Jefferson County jail in Louisville, Ky. Within days, the county coroner ruled his death a homicide, setting off multiple investigations by the jail, the police, prosecuting attorney and The Courier-Journal.

Early on, we reported that three of the five guards involved had prior on-the-job disciplinary records, including one who had been accused of inmate abuse. That piqued my curiosity and a question that formed the basis of our inquiry: How many other officers in the department had been disciplined, and for what violations?

The answer came the following month when we reported that nearly 90 percent of the jail's 350 guards had been disciplined for violating workplace rules. One-fifth had been found guilty of major offenses such as inmate abuse, insubordination and sleeping on the job. Serious punishment, such as termination, was rare. One officer had been disciplined 48 times - more than any other for violations that included pummeling an inmate in the face and then trying to cover up the incident. Management had tried to fire him twice but failed after he appealed to an employee grievance board.

Open-records accommodations

To reach our conclusion, we relied heavily on computer analysis and Kentucky's open-records laws to review all the disciplinary records of current and former jail-house staff. With our findings in hand, we turned to the Internet to find jail-management experts to place the situation in a national context.

In the wake of our report and a follow-up piece that looked exclusively at inmate abuse, jail managers negotiated a new contract with the union representing guards that is supposed to make it easier to punish officers who don't follow the rules while keeping an eye on the jail's approximately 1,600 inmates.

When we first asked the jail for all employee disciplinary records, we were presented with a loose-leaf log that listed every employee offense, the date, a brief description of the offense, the recommended punishment, and the outcome. The log became our road map.

Unfortunately, the log wasn't computerized. Using Paradox, I entered each incident and eventually created a 2,236-record database. I included every scrap of information in the log because I wasn't sure what would be important until I was finished.

I detected patterns almost immediately: Some employees had been punished dozens of times; management seldom recommended termination; and employees often got their punishments overturned on appeal. For example, our analysis showed that nearly 34 percent of all recommended suspensions had been reduced by the employee grievance board. About 70 percent of the recommended firings taken to the board were reduced to lesser punishments.

In the case of the inmate who had died, three of the five officers involved had prior infractions that could have led to termination. But in only one instance was firing recommended, and that recommendation was tossed out during the grievance process. The employee grievance panel, it turned out, gave management and the Teamsters union an equal number of votes when deciding cases appealed to the board. Our analysis showed that virtually all discipline that involved a suspension or termination were appealed to the board. Anxious to avoid a prolonged appeal if they stuck to their guns, managers repeatedly struck bargains with the union.

Data entry demands

The lack of a pre-existing computer database was the biggest problem; virtually all the jail records were paper. I don't enjoy data-entry work, but there didn't seem to be any other way. It took about three solid days, but in cases like this I do the work myself because I can see trends early on, which helps me draft the story as I go. Inputting data myself also makes me feel more confident about accuracy.

The employee log was only a starting point; it included virtually no details from individual disciplinary cases. To learn more about inmate abuse and excessive force, for example, I had to isolate the names of all 34 officers accused of such violations in the past. Then I requested their complete personnel files. That led to a second story, which concluded that although inmate deaths at the hands of guards were rare, incidents of excessive force and abuse were not.

Lessons learned

What's true in traditional reporting also is true in computer-assisted reporting: A newspaper's greatest ally is the open-records law. In this case, however, the records told only part of the story. Although jail managers knew they were having a hard time enforcing the rules, they didn't know how serious the problem was because they had never taken a close look at their own records. That's where computer analysis proved invaluable.

Finally, database stories can be pretty dull stuff without concrete examples. Part of what made our two stories compelling was our focus on case studies, including the officer who had been disciplined 48 times. Using the database as a guide to inmate abuse cases, for example, I described in detail situations where guards had beaten inmates held down by other officers, falsified reports to cover up their actions, and gotten off scot-free.

Jim Hopkins can be reached by e-mail at hopkinsj@courier-journal.com