Use this URL to cite or link to this record in EThOS:
Title: The social and biological effects of patient-patient co-presence on health in hospitals using electronic medical records
Author: Lienert, Jeffrey
ISNI:       0000 0004 7653 1908
Awarding Body: University of Oxford
Current Institution: University of Oxford
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Co-presence, defined as two people being physically proximate to one another, is a ubiquitous and important phenomenon that remains understudied. There is strong reason to believe that co-presence may affect health, but it is likely that these effects are relatively small. Because of this, relatively large sample sizes are needed to reliably detect these effects, and the data to test such hypotheses has only recently become widely-available. In this thesis, I use electronic medical records and hospital administrative data to assess how patient-patient co-presence in a health care system may affect patient health outcomes. In Chapter 3, I examine the social effects of co-presence on 5-year survival in a group of 4,791 chemotherapy patients. Because no metric for measuring co-presence precisely addressed all the nuances of hospital administrative data, I create a method to detect when patients are co-present more often than expected by chance, terming this consistent co-presence. Consistent co-presence thus allows me to subset co-presence to only that which is likely systematic enough to elicit social influence. Using this, I construct a consistent co-presence network. I then model 5-year survival on 1) whether a patient had any consistent co-presence in the network, 2) the number of patients who survived with whom one was consistently co-present, and 3) and likewise the number patients who did not survive with whom one was consistently co-present. I find that being consistently co-present with at least one other patient increased one's likelihood of 5-year survival compared to being consistently co-present with no one. Being consistently co-present with patients who survived increased one's likelihood of 5-year survival, and being consistently co-present with patients who did not survive decreased one's likelihood of 5-year survival. In Chapter 4, I assess the ability to predict subsequent infection based on the number of hours a patient spends co-present with another patient suspected of infection. Across five nosocomial infections, I find that this tool has a sensitivity from 0.95 to 1.00, and a specificity from 0.90 to 1.00. If this metric were put in place prospectively, I estimate that it would lead to detecting infections between 4 and 32 hours earlier than the current standard operating procedure. I then use this information, along with biomarker information to detect subclinical infections in Chapter 5. Subclinical infections are those where the bacterial or viral load is below a test's threshold, meaning these infections go undiagnosed. I use a random forest model to perform the classification, and a variety of regression models to examine the validity of said model. I then show that subclinical infections have negative effects both on the affected patients and on the nosocomial disease dynamics, leading to increased infectious outbreak sizes. As a supplement to support my analyses in Chapter 5, I develop an efficient algorithm to be used in social networks analysis for the colored triad census in Appendix A. I apply this to the outbreak networks observed in Chapter 5 to understand the patterns of connections of subclinically-infected patients. In sum, I find that co-presence is a useful and informative construct which allows us to better understand patient health in hospitals. Additionally, the outcomes observed here are not exclusive to the health care setting; social influence and infectious disease spread both occur outside of hospitals. As a result, this research opens up a wide variety of future work, including studying these effects in more detail with hospitals and using similar data sources to examine these effects in other populations and settings.
Supervisor: Koehly, Laura ; Reed-Tsochas, Felix ; Marcum, Christopher Steven Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: epidemiology ; infectious disease ; Social networks ; health services research