By Tony Gonzalez, The Tennessean
Three infants died in separate incidents in a Tennessee home that was well known to the Department of Children's Services, and after the third baby died, DCS flagged the family as being a high risk while allowing children to remain in the home.
In fact, dozens of cases show that DCS investigators allowed abusive and drug-using parents - sometimes in garbage-filled homes - to keep children who later died, according to department records obtained by The Tennessean late Friday.
In the home where the babies died, DCS later suggested that the regional office working with the family might want to find a different employee to handle any future concerns.
The documents were given to The Tennessean about 5 p.m. Friday by Molly Sudderth, DCS spokeswoman.
But most details of the complex decisions faced by DCS caseworkers remain hidden in records the department refuses to release. The Tennessean has sued to obtain those records.
Interviews and other department documents reveal an oversight process that fell behind on death reviews and ignored its own policies.
The workings of the internal Child Fatality Review Team, which is charged with figuring out what could have been done to prevent deaths - whether caseworkers need training or discipline and ways to improve investigations - is now the subject of a federal court filing in which a watchdog group charges that "grossly incomplete" reviews may be putting children at risk.
DCS admits the team fell behind in its work and spent 2012 scrambling to eliminate the backlog of fatality cases needing review.
In 2012, the team reviewed nine cases from 2010, 67 from 2011 and 58 from 2012, said Carla Aaron, DCS executive director of the office of child safety, which oversees the reviews. Some cases from late 2012 are yet to be reviewed.
She said Friday the team fell behind because information and autopsies were not always immediately available.
When the team of high-ranking staffers did gather, they followed a one-page "protocol" that Aaron said she wrote in January 2011.
At that time, she said the department wanted to do better death reviews, so the department discarded an existing four-page policy. That policy listed dozens of steps and set deadlines to complete the work.
Whether Aaron's protocol improved the way the department reviews child deaths remains in question.
It was criticized in the federal court filing last week, in which Children's Rights, a New York-based advocacy group, asked that a judge require the department to turn over more records.
The group said it had received partial "superficial" documents that made it "impossible to determine what transpired" in cases in which children died.
The department also failed to follow its own rules for changing the reviews.
At one point, DCS simply removed the old policy from its website - an approach to policymaking that fell far short of DCS procedures for such a change, according to DCS documents.
"The important thing is: We were still doing reviews and we had a protocol," Aaron said. "But the policy hadn't caught up with the work. We were doing (reviews) in a way, we thought, to reflect a more effective way."
Yet the four-page policy required a report for each fatality and an annual summary, but the team did not do that, Aaron said.
"There have never been annual (DCS) reports," she said.
After repeated requests in the past month from The Tennessean, DCS on Friday provided 53 pages of redacted fatality team meeting minutes.
The documents show the team did not record any recommendations about how to improve casework during four meetings from February to June 2012.
In a few instances, the team noted concerns about investigator decisions.
Although names, dates and locations were blocked out, documents show that in February, the team reviewed a fatal car crash, a beating, two medically fragile infants, the death of a boy and his foster father during a fishing trip, and others.
In March, the team reviewed a child's death due to malnutrition and neglect in a home that was "filthy," and in which the team found a lack of documentation about how siblings were cared for during previous investigations.
"It had numerous bags of trash inside the home. It was piled several bags high. It was also noted that there was not any food in the home," the team wrote in its summary.
Two other cases each had four home visits before children died.
Starting June 27, the team began writing recommendations, although for many cases, no recommendations were made.
In that meeting, the team heard details of a home in which a woman's three infant children had died.
A caseworker had visited the home before, at least twice, to recommend services, but some neglect allegations could not be confirmed. It's not clear if those visits were related to the first death - attributed to interstitial pneumonitis - or the second, for which the cause was undetermined.
The third death, which put the case in front of the team, was being investigated as unsafe sleeping, but homicide could not be ruled out, the records state.
The team decided to flag the case as "high risk" and to make contact with an unidentified person to make the person aware of the family's history.
"The regional folks may want to re-staff this case," the team concluded.
Earlier in the day, Aaron said reviews prompted other changes, such as providing staff members with more training on drug abuse cases and increasing responsiveness for cases in which any child under 3 was referred for a third time.
"If I saw something, I would not hesitate to pick up the phone and contact the regional administrator to discuss that," Aaron said.
New approach to data
Partway through 2012, fatality team minutes took on a new format, eliminating descriptions of the cases reviewed and focusing mostly on projects, including an effort to clarify how many children died in 2011, records show.
The team pursued what it believed would be a better way of looking at the deaths.
The department has had problems tracking fatalities and admitted in September it had broken the law by failing to notify lawmakers of deaths in their districts. DCS also has provided death data that included inaccuracies in response to The Tennessean records requests.
The team began recording regular updates about a new approach to collecting data.
Supporters of those reforms, according to the minutes, included executive directors Alan Hall and Debbie Miller, two experienced state government administrators whom DCS Commissioner Kate O'Day fired, without notice, on Tuesday.
A department spokeswoman did not attribute the firings to work being done on the fatality team, which was not central to their job descriptions.
In September, Hall led a discussion on a new Web-based database to make fatality information easier to access. A month later, Miller led a vote to adopt it, and as of Nov. 7, the new system was a week away from launching, the minutes state.
The team did not meet in December. Its status remains unknown.
Meanwhile, the team also stated in its November minutes that deaths in 2012 had been reviewed for the governor's office and by a different internal DCS team, so the team could skip reviewing those a second time.
"It was folks sitting here at Central Office almost Monday morning quarterbacking about things that happened months ago," Aaron said.
Separately, O'Day had something different in the works all along.
Faster reviews at the front lines
In mid-December, after more than a year in development in partnership with Vanderbilt University, a new type of fatality review was tried by DCS for the first time.
Leaders call the new approach "event analysis," a term derived from the nuclear power, aviation and health care fields, which have long conducted painstaking reviews of mistakes.
"We obviously think we need something else," O'Day said in December. "What we have been doing is accepted practice, and I think pretty typical. The event analysis does represent a new direction for us."
O'Day said the reviews should begin to happen immediately after deaths and will take place at regional offices and with the involved caseworkers present.
So far, DCS has examined one death under the new approach and plans to examine two more.
In some ways, the new reviews will match what department policy said should have been happening all along: step-by-step probes of how DCS interacted with families to create specific plans for improvements.
But a few things are new, O'Day and Michael Cull, assistant professor of clinical nursing at Vanderbilt, said during an interview in December.
For one, trained DCS staff members will ask a specific set of questions during sessions that could last longer - perhaps four hours for a single case, Cull said. And the new process includes face-to-face talks with the caseworkers involved.
O'Day said she used event analysis at the nonprofit she ran in Knoxville before coming to DCS in January 2011. She said the approach has a long track record in other industries and will make Tennessee a leader in child safety.
That's not guaranteed, experts said.
Tina Rzepnicki, a deputy dean at the University of Chicago and one of the main proponents of event analysis, said for the reviews to work, DCS will need caseworkers who trust they can be honest about the pressures of day-to-day work and about what they were thinking in the field.
"You're not going to get the goods without having a good relationship with the (caseworkers) you're interviewing," Rzepnicki said.
O'Day and Cull said DCS will continue to improve the new reviews as they go, but behind closed doors.
Although DCS had agreed to allow Tennessean reporters to observe an event analysis, officials later withdrew the offer, citing the newspaper's subsequent lawsuit seeking department records.
"I don't see it as experimental. I do see it as leading edge," O'Day said. "If it had never been tried anywhere, I would say it was experimental. If it had a spotty track record in other industries, I would say you could call it experimental. But when it is the standard tool in these other industries, then that's not experimental. It is an innovation to bring it into the child welfare system."