Lead author Hui Zheng of The Ohio State University in Columbus and Patricia Thomas of the University of Texas at Austin used data on about 789,000 people who participated in the National Health Interview Survey from 1986-2004.
In the survey, participants rated their own health on a five-point scale -- excellent, very good, good, fair or poor. Zheng and Thomas then used follow-up data to identify the nearly 24,100 people who died from 1986-2006.
Overall, the researchers confirmed the volumes of previous research that has found that, overall, being unmarried -- including never married, separated, divorced and widowed -- significantly increases the risk of death within three years. For example, a never-married person who lists his health as excellent is two times more likely to die within three years than a similar married person in excellent health, Zheng said.
However, as self-rated health declined from excellent to poor, the mortality advantage for married people diminished.
"These results suggest that marriage may be important for the prevention of disease, but not as helpful once people become seriously ill," Zheng said in a statement. "That's why we see a protective effect of marriage when people are in excellent health, but not when they are in poor health."
In addition, the married don't seem to report their health as being poor until they've already developed much more severe health problems, Zheng said.
"They have a different threshold for what they consider to be bad health compared to unmarried people," Zheng said.
The study was published in the Journal of Health and Social Behavior.