Use of big data to generate nursing science is sparking much discussion among nursing scholars (Harper & Parkerson, 2015; Clancy & Gelinas, 2016). Defining the value of nursing through analysis of big data is a contemporary area of focus (Pruinelli, Delaney, Garcia, Caspers, & Westra, 2016). Big data is a termed coined by the media to describe vast amounts of data now available in huge electronic repositories of healthcare data (Clancy, 2016). Most nurses are novice at understanding the potential use of big data. The science of “big data” provides a great enhancement and new methodological approaches to understand the complex question of nursing costs associated with patient care and patient outcomes. The purpose of the primary study was to explore variability of nursing cost per acute care episode for patients with similar DRGs using patient level data and to investigate the relationships among nurse characteristics and patient characteristics on nursing cost. Some results of the primary original study have been previously reported (Jenkins & Welton, 2014). The purpose of this presentation is to provide an overview of the complexities of data acquisition/management and outline the extensive collaborations among multiple stakeholders.
Methods:
The sources of data included three large databases; 1) patient assignment software, 2) medical management system, 3) human resources. The software company staff wrote a query so data could be extracted including multiple variables required for the study. Hospital information technology staff de-identified the data. Forty-nine total variables were extracted or constructed including patient characteristics such as DRG, length of stay, age; plus, nurse characteristics such as years in organization, actual wage, education level. Nursing intensity by shift was a key variable extracted from patient assignment software. The data acquisition process required multiple iterations between the PI, the hospital based nurse scientist and the software company.
A four step data management model was used (Long, 2009); 1) planning, 2) organizing, 3) computing, 4) documentation. A systematic plan was organized using Stata do-files to record all code written for data management and analysis, so the study is replicable. Seven excel spreadsheets were merged using a patient and nurse common identifier. The sample included 3111 patients and 150 nurses in 44,771 total shift observations. Stata v.12 software was used for data analysis (Cameron & Trivedi, 2010). Shift level nursing intensity multiplied by actual nursing wage was summed for all nurses caring for a given patient resulting in direct nursing cost per shift. Shift level data was aggregated to construct nursing cost per day and acute care episode (defined as admission to discharge on the study unit).
Results:
The nursing cost model generated using secondary data from big data sources provided direct nursing cost per patient shift, day, and acute care episode (i.e. DRG 192, nursing cost per day range $4.87-322.66, nursing cost per acute care episode range $54-1570). Direct nursing cost per patient data is not available or viewed by nursing leaders today.
Conclusion:
Data is available in current big data sources that can be extracted through collaborations between interprofessional healthcare and software company staff and used in scientific inquiry to answer patient level research questions and advance the science of nursing. Managing large amounts of data is a research skill nursing scientists can learn and use to generate knowledge to better understand nursing cost as well as quality patient outcomes. Big data acquisition and management is an exciting vista and the possibilities for evolving nursing science are vast.
See more of: Research Sessions: Oral Paper & Posters