Using hospital administrative data to infer patient-patient contact via the consistent co-presence algorithm

Area of Science:

• Healthcare Informatics
• Network Science
• Epidemiology

Background:

• Patient proximity, or co-presence, in healthcare settings can influence health outcomes.
• Existing methods for measuring co-presence have limitations in complex healthcare environments.

Purpose of the Study:

• To introduce a novel method for measuring patient co-presence in healthcare settings.
• To address limitations of previous co-presence measurement approaches.

Main Methods:

• Developed a "consistent co-presence" algorithm that implicitly models patient scheduling and movement.
• Randomly perturbed patient entry times to generate interaction networks.
• Compared the novel algorithm with established methods using model comparison.

Main Results:

• The consistent co-presence method generates networks suitable for patient outcome modeling (e.g., 1-year mortality).
• This novel algorithm is preferred over existing alternatives from a model comparison standpoint.
• Consistent co-presence retains significant information about patient-patient interactions.

Conclusions:

• The consistent co-presence method offers a more appropriate way to measure patient interaction in complex healthcare systems.
• This approach can enhance the accuracy of models predicting patient outcomes.
• The method's generalizability allows application to diverse complex systems beyond healthcare.