Researchers thought they had a way to keep hard-to-treat patients from constantly returning to the hospital and racking up big medical bills. Health workers visited homes, went along to doctor appointments, made sure medicines were available and tackled social problems including homelessness, addiction and mental health issues.

Readmissions seemed to drop. The program looked so promising that the federal government and the MacArthur Foundation gave big bucks to expand it beyond Camden, New Jersey, where it started.

But a more robust study released this month revealed it was a stunning failure on its main goal: Readmission rates did decline, but by the same amount as for a comparison group of similar patients not in the costly program.

“There’s real concern that the response to this would be to just throw up our arms” and say nothing can be done to help these so-called frequent fliers of the medical system, said study leader Amy Finkelstein.

Instead, researchers need to seek better solutions and test them as rigorously as new drugs, said Finkelstein, of the Massachusetts Institute of Technology and the National Bureau of Economic Research.

Federal grants and research groups at MIT paid for the study, which was published in the New England Journal of Medicine.

Just 5% of the U.S. population accounts for half of health care spending, and hospitalization is a big part.

A decade ago, Dr. Jeff Brenner started working with hospitals in Camden, a city with high poverty and crime rates, to identify people who go to hospitals frequently and target them with special services. He won a MacArthur genius grant for the work, and federal grants expanded the program to Aurora, Colorado; Kansas City, Missouri; Allentown, Pennsylvania; and San Diego. The government also paid for a study to see if it truly worked.

The study enrolled 800 hospitalized Camden patients with at least two other admissions in the previous six months and at least two of these conditions: homelessness, drug use, a mental health problem, trouble accessing services, lack of social support or use of five or more medicines.

Half were given usual care when they left the hospital. The rest were enrolled in Brenner’s program with nurses, social workers and others coordinating their care for three months. Patients received seven to eight home visits and nine phone calls on average. The effort cost about $5,000 per patient.

Six months later, the readmission rate was 62% in both groups, and there was no difference in total health care spending.

Researchers don’t know what usual care was for the comparison group. If that suddenly improved, it could explain why the program failed to prove better.

However, two key goals of the program — a home visit and a trip to a health provider within a week of leaving the hospital — were met for only 28% of participants. Homelessness and lack of a telephone were big reasons, and the program more recently has worked harder to get people into stable housing.

“If you’re in a shelter or on the streets, it’s just not easy to help and that sometimes leads to more hospitalizations,” said the Camden project’s leader, Kathleen Noonan.

Three months of help also wasn’t enough, and there weren’t enough places to get help for mental health or substance abuse problems, she added.