The most comprehensive survey of mental illness ever conducted suggests that at any given time about 29 million Americans -- nearly one in five adults -- suffer psychiatric disorders ranging from mildly disabling anxiety to severe schizophrenia.

But only about 20 percent sought medical treatment for their problem during the six months covered by the survey, going mostly to general physicians rather than to mental health specialists.

The study also overturns a long-held belief that mental problems are more common among women than among men and reveals a surprising frequency of anxiety disorders in the population, according to Dr. Darrel A. Regier, who headed the survey for the National Institute of Mental Health.

The $15 million door-to-door survey of nearly 10,000 people living in Baltimore, New Haven, Conn., and St. Louis provides "the first accurate assessment of the prevalence of specific mental disorders in our population," Dr. Donald MacDonald, head of the Alcohol, Drug Abuse and Mental Health Administration, said in a statement.

Regier said at a news conference yesterday that the continuing research ultimately will include about 20,000 subjects -- five times as many people as all previous major studies combined. Earlier studies, using less authoritative measures, found both larger and smaller rates of mental illness.

Regier cautioned that the three cities and two other communities involved in the ongoing study are not necessarily representative of the entire country, but he said the size of the survey provides a good estimate of national rates.

The new findings to date indicate that about 19 percent of adults over age 18 suffer from at least one mental disorder during a given six-month period. These range from a high of more than 23 percent in Baltimore to nearly 17 percent in St. Louis.

From 28 to 38 percent of those surveyed reported a psychiatric disorder at some point in their lifetimes.

The types of problems reported include:

*Anxiety disorders, such as phobias, panic disorders and obsessive-compulsive disorders. Afflicting about 8 percent of those surveyed, this appears to be the most common group of psychiatric problems -- not depressive illnesses as previously thought. It includes specific intense fears, such as fear of heights or animals, as well as "agoraphobia," a fear of leaving the familiar setting of home.

*Abuse or dependence on drugs, afflicting an estimated 6 to 7 percent of the population. About four-fifths of the cases are linked to alcohol.

*Affective or mood disorders, such as major depression and manic-depression, which struck about 6 percent of adults studied. Depression, a feeling of hopelessness that can sometimes lead to suicide, is often accompanied by problems in eating and sleeping and reduced activity. It may alternate with mania, involving increased activity and delusions of grandeur.

*Schizophrenia, the most severely disabling of mental illnesses. Found in about 1 percent of the population, it can involve psychotic disturbances in thought, accompanied by withdrawn or bizarre behavior and hallucinations. Another 1 percent are affected by antisocial personality disorders -- deeply ingrained behavior patterns, such as a low frustration level and an incapacity to feel guilt, that bring a person into conflict with others.

The overall rate of these mental illnesses appears to be roughly the same for men and women, said Regier. But while women more commonly suffer depression and phobias, men have more problems with drug and alcohol abuse and antisocial personalities. Regier said the notion that mental illness is generally more common in women came from previous surveys that concentrated on certain disorders, such as depression, rather than considering the breadth of illnesses striking both sexes.

The result of the new work "is a more balanced picture showing that rates in men and women are comparable but that the types of illnesses each characteristically develops are different," said Regier. In addition, the survey found that women sought professional help twice as frequently as men, he said.

The survey found about twice the rate of mental disorders in those under 45 compared to older individuals, he said. Anxiety disorders, substance abuse and antisocial personality problems were more common in the younger group.

Thus far, there appear to be no major racial differences in mental illness, he said, but preliminary findings suggest possible differences along socioeconomic lines, with those at the lower end who receive federal health assistance having more problems and seeking more treatment than other groups.

Regier said the massive data base will have "major implications" for the provision of mental health services and ultimately for prevention. It "may eventually prove to be as fruitful for identifying risk factors to prevent mental disorders as was the well-known Framingham Mass. heart study, which identified the role of cholesterol and obesity as risk factors for cardiovascular disease," he added.

The initial results are being published in six articles in the October issue of the Archives of General Psychiatry. In an accompanying editorial, journal editor Daniel X. Freedman called the study a "landmark" effort "that was not conceivable even 10 years ago."

The project has been under way since 1977, with the first interviews in 1980. But the study and its analysis is far from complete, said Regier. Surveys in two other communities in North Carolina and California are still under way, as are follow-up interviews with all of the study participants. In addition, findings from institutionalized mental patients in each community are still to be reported.

The diagnoses used in the survey are based on the American Psychiatric Association's latest Diagnostic and Statistical Manual of Mental Disorders, which spells out the criteria used by psychiatrists to diagnose patients.

To use these criteria on a massive survey basis, researchers translated them into an elaborate questionnaire with hundreds of questions that could be administered by a lay interviewer, much as census surveys are conducted. The results were then fed into a computer, which calculated the findings.

Regier said the accuracy of this approach withstood rigorous testing.

In the past, he said, smaller surveys in individual cities have used less precise definitions of mental illness.