Social Network Study Predicts Health Risks from Spouse Falling Ill

In findings that highlight how health effects can reverberate through a social network, a researcher at HMS and his colleague report that the serious illness of an elderly spouse increases the risk of death of a husband or wife. In fact, a few illnesses in a spouse, such as dementia, may pose more of a risk to the partner than if the spouse had died.

Photo by Leah Gourley

In a study of one half million married elderly couples led by Nicholas Christakis, the hospitalization of a spouse for more disabling diseases, such as hip or other serious fractures, chronic obstructive pulmonary disease, and congestive heart failure, increased the healthy partner’s risk of death. Cancer in a spouse, on the other hand, was relatively harmless to a partner’s health.

It was already known that disease in one spouse can harm the health of a partner. Many studies have shown that a spouse’s demise can be fatal to the other half. The new analysis extends this literature.

“We showed that you can die of a broken heart not just when a partner dies, but when a partner falls ill,” said lead author Nicholas Christakis, HMS professor of medical sociology in the Department of Health Care Policy and a sociology professor in the Harvard Faculty of Arts and Sciences. “This is a hard and unambiguous endpoint.”

The study, in the Feb. 16 New England Journal of Medicine, may be the largest study to quantify how a spouse’s illness or death, or both, affects the subsequent risk of death in a husband or wife. The researchers analyzed the health records of a half million married couples older than 65 enrolled in Medicare in 1993, and they tracked diagnoses, hospitalizations tions, and deaths through 2001.

“One of the key innovations of the paper is that it separates the estimation of illness from death,” Christakis said. “We show the separate effects of death and illness.”

Illness Breeds Illness
In the nine-year study, a spouse’s death increased the partner’s chance of premature death by about 20 percent. In a single year, this meant an increase to 6.7 percent from 5.6 percent in men and to 3.1 percent from 2.6 percent in women. A spouse’s illness increased the partner’s risk of death by one quarter, according to a standard statistical analysis that controlled for pre-existing diagnoses and other factors in the study’s nearly 400,000 couples older than 68.

But the danger to the partner can be much higher within the first month of a hospitalization that marks the onset of a serious or chronic illness in the spouse, according to another type of analysis in the paper. In most cases, the U-shaped curve dips about three to six months after spousal hospitalization and then starts rising again.

The initially high risk of death may reflect the early stress of adjusting to a spouse’s sickness, the researchers speculated. Then a rising risk months or a year later may reflect a drop in social support associated with the illness of the spouse. The partner’s health, for example, may be suffering from lack of healthy meals or assistance with managing their own illness, or it may decline for other practical reasons associated with the loss of a spouse’s support.“One of the big, important things about this article is that people will be more tuned in to the living [partner],” said Suzanne Salamon, a geriatrician at Beth Israel Deaconess Medical Center and HMS instructor in medicine. “There needs to be active intervention early, [ideally] when the couple is still a couple. Physicians really need to be educated in this. They need to pay as much attention to the spouse as the sick person.”

For Christakis, the results set the stage to explore the health effects of larger social networks, such as family, friends, and neighbors. Just as in systems biology, social networks “have emergent properties not explained by the constituent parts and not present in the parts,” Christakis wrote in an editorial in the July 24, 2004 British Medical Journal. “The existence of social networks means that people and events are interdependent and that health and health care can transcend the individual in ways that patients, doctors, policymakers, and researchers should care about.”

In other words, in health, as in much of life, it is not what you know, but whom you know that counts for and against you.

Risks from Partner Disability
Knowing how a medical intervention or risk for an individual can propagate through a social network may lead to changes in the calculus of health care costs and benefits, Christakis said. For example, treating postpartum depression in mothers may make it more likely that they will vaccinate their children. And in an earlier study, he and his colleagues estimated that taking care of a dying husband can save one out of every 200 women from premature death.

In the husband–wife dyad studied for the NEJM paper, the collateral health effects varied by the different diseases, by gender, and over time.

Dementia and psychiatric diseases in spouses were as lethal to their partners as the ill spouses’ death. Hip or other serious fractures, chronic obstructive pulmonary disease, and congestive heart failure also ranked high on spousal diseases most dangerous to partners. Cancer, on the other hand, was relatively harmless to the partners.

“Diseases can have two different properties,” Christakis said. “They can be disabling, and they can be fatal, and the two are not necessarily the same. We found that how disabling they are can predict mortality in partners.” Other studies have shown that a higher degree of disability leads to a greater burden on partners. “The bottom line is that cancer in spouses is more deadly to them, but it doesn’t impose as much hardship on their partners.”

The study did not explicitly focus on gender differences, but there is a hint that husbands suffer more from the loss of wives than wives suffer from the loss of husbands. Nevertheless, said Christakis, “I believe the kind of interpersonal health effects we are investigating are a fundamental property of human interactions that is not gender-specific.”

To rule out external factors that may be causing death in both husband and wife, co-author Paul Allison at the University of Pennsylvania adapted a case–time control statistical method to look within the couples and see how the partners’ risks of death changed before and after the spouses’ hospitalizations.

Two postdoctoral fellows in Christakis’s group—Lei Jin, a Robert Wood Johnson health policy scholar, and Kirsten Smith, a Program on the Global Demography of Aging postdoctoral fellow—are conducting follow-up analyses to tease out potential mechanisms for the increased mortality of partners following a spouse’s hospitalization.

In a related paper, Felix Elwert, a sociology graduate student at Harvard, found strong evidence that the health effects of social ties depend on the individual attributes of the people with whom they connect. Elwert and Christakis showed that the “widowhood effect” varies by race. Black couples and white men married to black women do not appear to suffer increased mortality, perhaps because the family and church support structure within the black community created during marriage survives the death of the spouse, Christakis said. The study is published in the February issue of American Sociological Review.

Next, Christakis and HMS colleagues James O’Malley and Alan Zaslavsky are constructing a social map of 10,000 people in the Framingham Heart Study to find the signature of more complex social network phenomena writ in the biology of health and disease.

—Carol Cruzan Morton with reporting from Leah Gourley

top