When is a person just a little different, and when is his peculiarity a symptom of a disabling disorder?

An expert panel has taken a new crack at that question, proposing a redefinition of autism and related conditions for the manual used by U.S. mental health professionals. The plan has aroused fears that it may strip many people of a diagnosis and thus the insurance and government benefits that can go with it.

So far, it’s not certain whether this would be the result. But until the consequences are known, it’s premature for the American Psychiatric Association to change the autism diagnosis in the fifth edition of its Diagnostic and Statistical Manual of Mental Disorders.

The stakes are high. About 1 percent of American children are affected by autism and related disorders, according to the Centers for Disease Control. This represents a dramatic increase, which has been attributed to changes made to the diagnostic manual in 1994. That version, the DSM-IV, carved out a separate diagnosis for Asperger’s disorder, typified by social awkwardness and an all-absorbing interest in specific topics.

For the DSM-5 (the manual will use Arabic numerals from now on), the APA’s autism working group has proposed combining classic autism, Asperger’s and two related conditions into autism spectrum disorder and has tightened the requirements for arriving at the new diagnosis. For example, to diagnose an Asperger’s case under the proposed changes, a clinician would have to identify impairment in five of seven categories rather than only three in eight, as under DSM-IV.

The autism working group is confident that its language would not disqualify those who are truly autistic. It cites still-unpublished research indicating that the DSM-IV criteria, if literally applied, lead to diagnoses that two clinicians cannot agree are warranted.

Panelists cite additional analysis showing that those whose diagnoses are confirmed by two clinicians also meet the criteria of the proposed DSM-5 rules.

Unfortunately, the APA doesn’t plan to publish this research until the DSM-5 rules are final. That’s backward. Far too many people are potentially affected by the planned changes for the working group to say, in essence, “Trust us on the evidence.”

Psychiatrists know there is a relatively simple way to clear all this up. In addition to releasing the research, the APA could order a modest new study that uses a stronger means of determining who is truly autistic than clinician judgment, a method so weak the APA gives it no validity. The study would screen children with autism characteristics using both DSM-IV and the proposed DSM-5 criteria. To check the validity of the diagnoses, researchers could rely on family history, a strong validator because the disorder runs in families.

If a portion of children lose the diagnosis in the switch to DSM-5, it should be relatively easy to tell how many of them were falsely diagnosed under DSM-IV. They would be the ones without a family history of autism. If significant numbers had such a history, however, it would suggest that the proposed new standards would deprive children who genuinely have autism disorder of the care they need. Then the APA would have an opportunity to change the rules before they did any harm.

It’s conceivable that such a study could be done in time for a full airing of the issues so that the APA could still make its Dec. 31 deadline for the DSM-5. If not, the organization should resist pushing through, without sufficient evidence, a change that would mean the end of insurance, education and social support for children who need it. Although this might save some tax dollars in the short term, it would cost society more in the long run, because without support these children would be less likely to succeed as independently functioning adults.

The reason the APA dropped the Roman numerals this time around was to allow for interim revisions of its manual. The matter of autism disorder could be a perfect subject for DSM 5.1.

Read more opinion online from Bloomberg View.

To contact the Bloomberg View editorial board: view@bloomberg.net.