Dead people were sent letters days before Christmas about their transition to the National Disability Insurance Scheme, based on data provided by Queensland authorities that was not double-checked.
The state government provides names and contact details of people eligible for transition to the NDIS to the National Disability Insurance Agency.
"Is it true that several hundred such records that were provided by the Queensland government to the Commonwealth were for people who were already deceased, and if so, how did this happen?" LNP Disability Services spokesman Christian Rowan asked during state estimates hearings on Tuesday.
Queensland Disability Services director-general Tony Hayes confirmed the incident related to data "worked through" with the NDIA before Christmas 2017.
However, Mr Hayes said the NDIA did not complete a "filtration process" at the conclusion of the process, where it compares information to databases held by Centrelink and Births, Deaths and Marriages.
"On this occasion, they failed to do that particular step at the end so therefore what happened on the 22nd, or leading up to the 22nd of December 2017, is that letters were sent to people who had been deceased," he said.
Mr Hayes said the NDIA took immediate action to reach out to families and apologise, including in writing.
"This was an adverse event that occurred on the eve of Christmas and leading up to Christmas; the NDIA were very regretful of the situation but it came down to the final checks that were actually undertaken in response to what needs to happen prior to letters going out to people, encouraging the NDIA to make contact," he said.
Dr Rowan moved to clarify: "That was based on inaccurate data from the Queensland government being provided?"
Mr Hayes said the data provided met the appropriate data standards at the time but he said data standards changed.
"The approval of plans starts at a point in time — six months prior to that, data is loaded into their production system so people can access the NDIA," he said.
"On the basis of that, the data would have been at least six months old in terms of the feed but also it's reflecting periods that were prior to that in terms of cut-off dates for data cleansing."