In response to PA-12-241, Large Research Projects for the Prevention and Management of Healthcare- Associated Infections, this project will examine an important clinical problem, healthcare-associated infections (HAI), using available electronic data already collected for other purposes, greatly reducing the costs and increasing the efficiency of the research. While nursing staff are on the front line of preventing and controlling HAI, studies of the relationship between nursing care and risk of HAI have been limited primarily to assessment of single factors such as staffing or adherence to specific evidence-based guidelines. There are, however, numerous system- and unit-level factors (e.g., staffing levels, distractions, competing priorities within the busy healthcare environment) that may impinge on the ability of nurses to provide care and adhere to practices which affect patient safety. Development and expansion of health information technology holds great promise for health services research, but even within the same healthcare system, data sources are often `siloed' and unlinked. We have developed an electronic database of ~1 million patient discharges over 9 years (2006-2014) from four urban Manhattan hospitals which makes it possible to examine complex relationships within entire health systems. After expanding the database to include 2015-16, the goal of this proposed project is to identify factors within the acute care health systems which increase the risk of HAI among hospitalized patients.
Our specific aims are to assess the relationship of HAI with (1) intensity of nursing care demands and staffing levels and (2) outbreaks of emerging/re-emerging community-onset infectious diseases such as influenza A H1N1, Ebola, and measles as well as hospital-based exposures to epidemiologically important community pathogens such as TB, pertussis, meningitis, scabies, and norovirus. To assess the impact of intensity of nursing care demands on risk of HAI, we have developed and tested a Nursing Intensity of Care Index comprised of specific patient characteristics, technologic and procedural demands, and staffing factors available from electronic records. The outcome variable of interest is days from admission to HAI, and we will apply survival analysis methods using a Cox proportion hazard model with time-dependent covariates. The predictor variables include (1) demographic and clinical characteristics such as patient acuity/severity of illness at admission, etc.; (2) daily patient intensity score; and (3) unit daily level variables: nurse staffing, patient movement for tests and procedures, and intensive care unit. To test the impact on HAI of emerging/reemerging community-onset infections, we will identify the time periods during emerging infections and/or outbreak investigations or preparations to prevent and control outbreaks within the hospital have occurred. The outcome variable of interest for this aim is the number of HAI, and the analysis is at the nursing care unit-weekly level. We will apply a generalized linear time series model to examine the impact of community-onset infections on HAI incident rate using Poisson or negative binominal models.
Health information technology holds great promise for improving coordination and standardization of clinical care and health outcomes for individual patients and across health care systems, but its promise is yet to be fulfilled. Using data from ~1 million discharges from four hospitals we propose to examine the impact on HAI of (1) intensity of nursing care demands and staffing and (2) outbreaks of emerging/re-emerging community infectious diseases and hospital-based exposures to epidemiologically important community pathogens.
| Hessels, Amanda J; Liu, Jianfang; Cohen, Bevin et al. (2018) Severity of Illness Measures for Pediatric Inpatients. J Healthc Qual 40:e77-e89 |
| Larson, Elaine L; Cohen, Bevin; Liu, Jianfang et al. (2017) Assessing Intensity of Nursing Care Needs Using Electronically Available Data. Comput Inform Nurs 35:617-623 |
