Sunday, March 21, 2010
Grand Hall (Hyatt Regency Atlanta)
Infection control teams and ward managers often face difficult decisions in the face of outbreaks of hospital acquired infections (HAIs).
Objective: We focused on a model of the spread of norovirus to help quantify the probabilistic consequences of such decisions Methods:
We used perl 5.10.0 to run Monte Carlo simulations of outbreaks on a 62-bedded UCLH geriatric ward (stratified into 16 bays) that was affected by a norovirus outbreak in January 2009. The demography of the model was based on observed staff turnover and patient admission and lengths of stay. The infectious process assumes seeding by a single admission. A fraction of staff and patients are assumed immune, and we used available distribution data for incubation periods and durations of illness. We estimated transmission parameters using local outbreak data from 171 functional care units of the Avon region of England that experienced 227 outbreaks. A parameter allowed the ratio of spread within vs. between bays to vary allowing for the importance of patient geographic distribution on the ward. Infection control was modelled on the UCLH experience: 7 isolation units and a 4-day delay before closing the whole ward for 6 days followed by a fragmented closing/opening of affected bays. Model outcomes include apparent prevalence and incidence (dependent on local PCR testing strategy), effective basic case reproductive ratio, Re, outbreak size (defined as time from the detection of the index case to 7 days after the last case) and % seeding event resulting in an outbreak. We used 200 simulations for each scenario and used medians as summary statistics for each outcome.
Results: In the Avon outbreaks norovirus resulted in relatively high attack rates (53% patients, 25% staff) for a short outbreak duration (16 days) on average. This observation was consistent with the subset of possible models where most staff were immune (65-75%) and where a focus of infection puts all patients at equal and simultaneous risk of infection (i.e. geographic location cannot be important or longer epidemic sizes are expected). Plausible alternatives are the presence of very mobile super-spreader patients and staff members or a contaminated toilet used by all patients (not explicitly considered in the model). Re varied between 4 and 10 units.
Conclusions: The use of isolation units failed to prevent outbreaks from 85% of all seeding events. The fragmented opening and closing of affected wards did nothing to reduce outbreak duration and attack rates, irrespective of the delay before intervention took effect. Closing the whole ward was more successful in terms of reducing outbreak duration but is accompanied with a temporary reduction in admissions that could affect ward management financially