Very low birth weight (VLBW) infants are among the highest-risk patient populations. Treated in the neonatal intensive care unit (NICU), these infants are closely monitored and provided with an array of life support measures and intensive interventions. While they account for only one percent of births, VLBW infants account for half of infant deaths in the US each year. Neither differences in patient characteristic nor hospital and unit characteristics can explain the large discrepancies observed in risk-adjusted patient outcomes. The Vermont Oxford Network (VON) was formed in 1990 as a not-for-profit organization with the goal of improving the quality and safety of medical care for newborn infants and their families through a coordinated program of research, education, and networking of NICUs around the world. Throughout its 20 year history, the VON has maintained extensive databases of its growing number of member hospitals'VLBW patients and hospital and NICU characteristics. In addition, it has maintained thorough records of a wide variety of VON facilitated interactions between member NICUs and their individual health care practitioners. With over 800 member hospitals and over half a million VLBW infants in its records, the VON data set is exceedingly rich. The long term goals of this project are to understand the structure and evolution of the VON, understand how this social network has influenced the care practices and health outcomes for infants treated at member hospitals, and determine ways the VON could enhance innovation and spread of innovative practices in the future.
The specific Aims of this project are:
Aim 1) To assemble data regarding heterogeneous types of interactions within the VON and create a unified database of VON interactions from 1990 through 2009;
Aim 2) To describe the structure and evolution of the multi-scale and heterogeneous interactions in the VON as the network evolved from 50 NICUs in 1990 to over 800 NICUs in 2009;
Aim 3) To identify associations between social network characteristics of hospitals in the VON and risk-adjusted morbidity and mortality of their patients;
and Aim 4) To generate hypotheses regarding the evolution of health care quality improvement network structure and its effects on patient outcomes, and identify enhanced data collection strategies that will support prospective studies to test those hypotheses. This is the first large-scale social network analysis of a collaborative improvement network between different hospitals. Novel computational approaches will be developed and used to explore the structure and evolution of the network using multiple views at both person and hospital scales, and over multiple scales of time. High-dimensionality feature selection and both linear and nonlinear association analysis will be used to identify relationships between network properties and patient outcomes. Fundamental questions about social networks and the spread of health care innovations will be addressed, including the importance of team diversity of personnel occupations and hospital attributes, geography, and connectivity.
Medical care is now provided by complex interdisciplinary teams, and health care institutions frequently join large networks where they work together to collect data, conduct multicenter clinical research, and participate in organized quality improvement collaboratives. Using data from a large network of over 800 hospitals worldwide, with a 20 year history of quality improvement initiatives and standardized patient data collection, the proposed work will study how interactions within the network have impacted health outcomes in a population of over half a million high-risk infants. The methods and results from the proposed work will have widespread application to managing collaborative health care networks that seek to innovate and spread quality improvement practices between hospitals around the world.
|Manukyan, Narine; Eppstein, Margaret J; Buzas, Jeffrey S (2016) Tunably Rugged Landscapes With Known Maximum and Minimum. IEEE Trans Evol Comput 20:263-274|
|Manukyan, Narine; Eppstein, Margaret J; Horbar, Jeffrey D et al. (2013) Exploratory Analysis in Time-Varying Data Sets: a Healthcare Network Application. Int J Adv Comput Sci 3:322-329|
|Manukyan, Narine; Eppstein, Margaret J; Horbar, Jeffrey D (2013) Team Learning for Healthcare Quality Improvement. IEEE Access 1:545-557|
|Eppstein, Margaret J; Horbar, Jeffrey D; Buzas, Jeffrey S et al. (2012) Searching the clinical fitness landscape. PLoS One 7:e49901|