57 Social Network Influence on Vaccination Uptake Among Healthcare Workers

Friday, March 19, 2010: 11:45 AM
Centennial III-IV (Hyatt Regency Atlanta)
Donald E. Curtis , University of Iowa, Iowa City, IA
Chris S. Hlady , University of Iowa, Iowa City, IA
Sriram V. Pemmaraju , University of Iowa, Iowa City, IA
Alberto M. Segre , University of Iowa, Iowa City, IA
Philip M. Polgreen, MD , University of Iowa, Iowa City, IA

Background: Influenza vaccination is one of the most effective measures for preventing the transmission of influenza within healthcare settings.  However, in many facilities, influenza vaccination rates of healthcare workers remain unacceptably low (<50%). Several recent investigations have shown the impact of social networks on health-related behaviors and outcomes (e.g., smoking, eating habits).

Objective: Motivated by other health-behavior-related, social-network research, we construct social networks for hospital-based healthcare workers and examine the impact of neighbors' vaccination status on the vaccination status of healthcare workers.

Methods: By combining de-identified EMR login data with data regarding the hospital space, we construct social contact networks of varying densities for University of Iowa Hospital and Clinics (UIHC) healthcare workers (10,596 healthcare workers). UIHC vaccination data for the 2007 (6302 vaccinations) and 2008 seasons (6616 vaccinations), when overlaid on the contact networks, allow us to determine statistics such as the number of vaccinated neighbors an individual has at the time of being vaccinated.

We use these data to determine if (i) we can reject the hypothesis that random vaccination, oblivious to social context, explains our observed data and (ii) whether healthcare workers with greater number of vaccinated neighbors are more likely to get vaccinated relative to healthcare workers with fewer vaccinated neighbors.

Results: Using a Chi Square test we compared the observed number of healthcare workers who got vaccinated when a given number of their neighbors were vaccinated with the expected number of such healthcare workers resulting from the random vaccination process.  The resulting p-values ranged from 0.0614 to <0.0001 for the set of contact graphs we considered, providing strong evidence that our observed data cannot be explained by the random process. This result is complemented by results showing that on average, a vaccinated healthcare worker tends to cluster with other previously vaccinated healthcare workers much more in the observed data than in a random process. We use a maximum likelihood estimator to estimate from our data the (i) probability that a healthcare worker with no vaccinated neighbors is vaccinated and (ii) the probability that a healthcare worker with one or more vaccinated neighbors is vaccinated. We show that the latter probability is 30% higher.

Conclusions: Our results suggest that there is a strong association, that cannot be explained by a random vaccination process oblivious to social context, between higher vaccination rates and how “connected” healthcare workers are to other healthcare workers.