A poisson-based prediction model and warning system for MRSA daily burden

Area of Science:

• Infectious Disease Epidemiology
• Statistical Modeling in Healthcare
• Public Health Surveillance

Background:

• Methicillin-resistant Staphylococcus aureus (MRSA) poses a significant threat in healthcare settings.
• Effective surveillance is crucial for early detection of potential MRSA outbreaks.
• Traditional methods may not always provide rapid identification of increased MRSA incidence.

Purpose of the Study:

• To demonstrate that daily MRSA isolate counts in a community hospital follow a Poisson distribution.
• To validate the use of a one-sided Poisson control table for rapid detection of elevated MRSA cases.
• To provide a tool for early warning of potential MRSA outbreaks.

Main Methods:

• Retrospective analysis of 934 MRSA isolates collected over three years (2005-2007).
• Comparison of observed MRSA frequencies with Poisson frequencies using chi-square goodness-of-fit tests.
• Development and presentation of a one-tailed Poisson control table (mean + 2 sigma).

Main Results:

• Goodness-of-fit tests confirmed close agreement between actual MRSA isolate counts and Poisson frequencies for all three years.
• Statistical analysis (p > 0.05) supported the Poisson distribution model for daily MRSA isolates.
• A regression equation was used to predict future trends.

Conclusions:

• The Poisson probability model is theoretically and empirically supported for daily MRSA isolate counts.
• The Poisson control table is an effective and efficient tool for identifying unusual increases in MRSA isolates.
• Findings suggest generalizability to other healthcare settings for MRSA surveillance.