Institutional and Provider Networks
Neighborhood Effects on Health
Statistical Methodology for Social Network Analysis
Below are active and recent areas of research in the Christakis Lab. Past efforts in other areas that are no longer an active part of our agenda are reflected in our publication list.
Our research typically involves new datasets, new methods, and new ideas. We focus as much as possible on pushing forward what is known about the role of social factors in determining health. A major focus of our work at present is to create and expand a basic understanding of how social networks affect health behaviors, health outcomes, and health care, and on how networks are biologically and genetically encoded.
New directions for the lab include investigations of the pro-social function of networks, experimental work manipulating network connection and contagion, explorations of social networks in the developing world, and genetic studies of social network features.
Our work is interdisciplinary and collaborative, and our lab group meetings involve students at all levels (undergraduate, graduate, medical, and post-doctoral), faculty from numerous disciplines (medicine, sociology, economics, political science, physical, anthropology, evolutionary biology, health policy, statistics), and staff with diverse expertise (data management, information technology, statistical analysis). Currently, we have active research agendas in the following areas:
Our main effort at present is directed at understanding the structure and function of social networks and how this structure and function relate to health and health care. By taking advantage of archival data from the Framingham Heart Study that we computerized in a novel way (and with the collaboration of our colleagues there, Joanne Murabito and Emelia J. Benjamin), we have developed a longitudinally resolved network of 12,000 people followed for 32 years (the "FHS-Net"). We have used this data set to explore numerous topics, including the spread of health behaviors in the network (e.g., obesity, smoking, drinking, eating, health screening); the spread of other health-related phenomena in the network (e.g., depression, happiness, loneliness); the influence of health events (such as a heart attack or a stroke) in one person on the health of others to whom he or she is connected; the socio-biological determinants of network ties; and the role of egocentric reference groups in individuals’ health. We have also recently linked the FHS-Net to genetic data and are investigating the role of genetics in social networks. We are also actively developing other data sets that allow further investigation of such topics, including experimental studies with online an offline networks. Finally, we are exploring ways that social networks could be used as sensors to get early warning regarding biological and behavioral epidemics. Much of this work has been done in close collaboration with James Fowler, a political scientist at UCSD. Click here for published papers on these topics.
With respect to our work on the spread of obesity, we discovered that weight gain (and weight loss) appears to spread along social network ties, thus contributing to the rise in obesity seen in the US in the last 30 years. Click here for a three-minute video interview with Nicholas Christakis regarding this work. Click here for a copy of the paper in the New England Journal of Medicine. Click here for a narrated animation showing the spread of obesity in the network available at the New England Journal of Medicine. Click here to view and download the original animation.
Other previous work in this area focused on a very simple type of social network, namely, dyads of husbands and wives. This work investigated the health benefits of marriage, the widowhood effect, and caregiver burden. Click here for a list of published papers on marriage and health and here for papers on widowhood and caregiving. Collaborators on this work have included: Felix Elwert, Paul Allison, and Jack Iwashyna.
The team of faculty investigators working on health and social networks includes: James Fowler, James O’Malley, Alan Zaslavksy, Sam Arbesman, Coren Apicella, Jukka-Pekka Onnela, Mark Pachucki, Sae Takada, and Jessica Perkins. This work is supported by a program project grant from the National Institute on Aging entitled "Networks and Neighborhoods" and by a Pioneer Grant from the Robert Wood Johnson Foundation.
INSTITUTIONAL AND PROVIDER NETWORKS
We are developing new projects that examine how health care institutions and providers are inter-connected, and how such patterns of inter-connection affect the diffusion of innovation, the flow of patients through health care systems, and the quality of health care. For example, we are interested in the determinants of referrals among physicians and how these informal networks arising from referral-making might affect the delivery of health care or the quality of that care. This work involves the assembly of large-scale, longitudinally resolved, bipartite networks of hospitals and doctors, or of doctors and patients.
The team of faculty investigators working on this topic include: Bruce Landon, Nancy Keating, James O’Malley, Alan Zaslavksy, Michael Barnett, and Peter Marsden.
In collaboration with László Barabási's group, we are using healthcare claims databases and other sources of data (about genetic and metabolic networks) to study the role of network phenomena in disease initiation and progression. We have constructed a large-scale phenotypic "diseasome" that uses co-expression of diseases in patients to ascertain the links between diseases. This network map can then be combined with information on metabolic and genetic networks to help discern the determinants of illness co-expression. Papers regarding this topic can be found here. [bio networks pubs]
NEIGHBORHOOD EFFECTS ON HEALTH
We are currently investigating geographic variation in health outcomes and the effect of attributes of place (e.g., local wealth, local infrastructure) on individual health. These investigations encompass studies that examine how attributes of both micro-neighborhoods (e.g., Census block groups) and macro-environments (e.g., counties, or larger 'health service areas') affect outcomes as diverse as cancer treatment, cancer survival, depression, cardiovascular disease outcomes, obesity, and other health-related phenomena. These investigations also involve analyses of how geographic distance between people, and between people and services (e.g., hospitals, fast food restaurants), affect various aspects of health and health care. Click here for a list of publications to date in this area. Future work will investigate the concurrent role of social networks and neighborhoods on individual health outcomes and the effects of specific neighborhood social features on health.
The team of faculty investigators working on this topic include: James O’Malley, Alan Zaslavksy, Rob Sampson, Elizabeth Lamont, Jason Block, SV Subramanian, Felix Elwert, Jukka-Pekka Onnela, and Sam Arbesman.
STATISTICAL METHODOLOGY FOR SOCIAL NETWORK ANALYSIS
As part of our Program Project grant on "Networks and Neighborhoods," we are actively seeking to develop new statistical methods that allow for more robust causal inference with complex network data. The work focuses on exploiting the longitudinal change in network topology and in nodal covariates to better identify causal effects, and on developing instrumental variable methods for network applications. In addition, we are developing methods of missing data imputation in networks. This work involves numerous datasets, including the FHS-Net, various online datasets, and others. Click here for some papers in this area. [methodology]
The team of faculty investigators working on this topic include: James O’Malley, Alan Zaslavksy, Peter Marsden, James Fowler, Joe Blitzstein, Jukka-Pekka Onnela, Sam Arbesman, and Neils Rosenquist.
This area was formerly a substantial focus of our research group. We investigated the determinants and outcomes of hospice use and ICU decision making and conducted an extensive set of investigations regarding the role of prognosis in medical care. We have a few ongoing research projects in this area, involving mathematical models of physician decision making with respect to prognosis, and involving studies of the distribution of hospice care around the country. Click here for papers about prognosis, here for papers about bereavement and caregiver burden, and here for papers about hospice care and palliative medicine.