Social Networks of Healthcare Worker and Patient Interactions

Our first effort to model the spread of nosocomial infections was a traditional observational study, where we measured contacts between healthcare workers and patients at the University of Iowa Hospitals and Clinics (UIHC), a 700-bed comprehensive academic medical center and regional referral center in Iowa City. Data were collected by randomly selecting UIHC employees from each of 15 job classifications and then using graduate students to ‘‘shadow’’ the 149 selected employees and manually record their every human contact (within approximately three feet) over a total of 640 hours. A total of 6,654 contacts were recorded, with each contact indicating type of contact (patient or category of healthcare worker), location, length of contact time, whether physical contact was made, whether the contact took place in a patient room, and whether handwashing/sanitizing occurred prior to contact.

Using a generalization of the well-known Erdõs-Rènyi model for random graphs, we can generate contact networks from these observational data by making two important simplifying assumptions: first, healthcare workers of the same type are assumed to have the same contact probabilities, and, second, edges between different pairs of agents are placed independently. Notwithstanding these assumptions, the contact networks generated in this fashion are quite consistent with the observed data. Using simulation studies based on these contact networks, we proposed and evaluated? a targeted vaccination strategy, where vaccination priority is determined by type of healthcare worker. The general idea is that workers with relatively higher connectivity to other workers and patients (as defined by the underlying contact network) should have higher priority in situations where, e.g., vaccine is in limited supply. The results show that this particular vaccination strategy leads to a reduction in attack rate over a broad range of simulation parameters. Additional simulations were used to model related issues, such as the use of quarantine and associated policy decisions.