Friday, August 28, 2015

Data Mining and the Electronic Health Record: A land mine or a gold mine for nursing research?

Linda H. Eaton, PhD, RN, AOCN, Post-Doctoral Fellow
T32 Interdisciplinary Training in Cancer, Aging and End of Life Care
University of Utah College of Nursing                                                                         

Documentation of patient data is a necessary and essential component of patient care. In the past, patient data was documented in the paper medical record and used by the healthcare team to communicate and track a specific patient’s health status, treatment, and outcomes. Today, due to the electronic health record (EHR), an enormous amount of patient data is readily accessible to providers, healthcare organizations, researchers, and insurers. Through data mining, important research questions can be addressed such as identifying best practices to improve quality of care and reduce health care costs.


https://www.flickr.com/photos/11139043@N00/1439804758 

What is data mining?
Data mining is a methodology used in healthcare and other fields for analyzing large databases and summarizing the data into useful information. Although causality is not established by data mining, patterns, associations, or relationships among the data can generate new information for predicting the likelihood of future events. Data mining is based on statistical concepts and developments from several disciplines including machine learning, artificial intelligence, data visualization, and pattern recognition (Berger & Berger, 2004).

How is Data Mining Changing the Research Paradigm? 
Data mining is changing the paradigm for how clinical research is conducted. The current paradigm is based on a carefully controlled study which takes years for findings to be implemented in practice. With data mining, data are readily available from a large numbers of participants. These may be patients who would not routinely participate in a traditional study due to lack of interest or time, or because they are too sick, i.e., multiple chronic diseases, critical health condition, end of life. Thus, data mining can address research questions that are not easily addressed by a traditional study.

The potential implications of this paradigm change for nursing research are several. Data mining can support the discovery of important relationships among clinical data, nursing interventions, and patient outcomes. It generates knowledge beyond what can be learned in a carefully controlled study. Data represent the “real world” which strengthens the applicability of study findings to practice. With the use of data mining, nursing knowledge generation may be accelerated.

Characteristics of Old and New Research Paradigms
Traditional Research Model
Data Mining Research Model
Requires monitoring by a human subjects committee
May be exempt to monitoring by a human subjects committee
Doesn’t require team science
Requires team science
Often difficult to enroll patients
Patients are readily available
May have high patient burden
Low patient burden
Expensive
Expensive
Very controlled
“real world”
Limit to number of research variables
Unlimited number of research variables
Limited data
Big data
May take years to complete
Can be completed in a shorter amount of time

What Challenges Exist with Data Mining?
Potential challenges with data mining need careful consideration. Some of these challenges are unique to data mining and others are typical challenges in any research study.

Quality of Data
  • Missing data, misspellings of medical terms, and redundant data all impact data quality. As with any research study, if data entry is poor, it will have a negative impact on the study findings. Cleaning of data is an essential step of data mining to ensure accurate findings, but it does not fix missing data.
  • Clinical nurses need to be educated about the critical need for accurate and complete documentation. They need to understand that their documentation in the EHR may be data for a research study that can inform and change nursing practice. 
Standardized of Language
  • Multiple labels are often used to represent the same concept.
  • Standardized language needs to be used in the EHR in order to generate meaningful findings from data mining. It also better facilitates the communication and exchange of data between different information technology systems.
  • Nursing needs to be involved in developing the standardized language; otherwise, data mining will not provide answers to nursing research questions.
Patient Privacy
  • De-identification of EHR data is essential. 
  • Any information that may identify the patient must be removed. This requires careful attention from the research team.
Interpretation of findings
  • o Due to the large sample sizes provided by the EHR system, statistical significance is often attained. 
  • o It is essential that a nurse expert determine if the statistical significance has clinical merit.
   ..”a difference, to be a difference, must make a difference”  (Sacristan, 2013)
Summary
Data mining is changing the research paradigm. As the National Institutes of Health Big Data to Knowledge Initiative is achieved, we will see more research studies using data mining of the EHR. This methodology has a huge potential for generating nursing knowledge in a timely manner, but must be carefully implemented by a research team that is experienced in the technological, statistical, and clinical practice aspects of data mining. As the potential for data mining is realized, we will see an explosion of nursing knowledge that will improve patient outcomes and clinical practice.

Links to Hyperlinked Text:


Other Links:

Key References:
Berger, AM & Berger, CR (2004). Data mining as a tool for research and knowledge development in nursing. CIN: Computers, Informatics, Nursing, 22(3), 123-131.

Goodwin, L, VanDyne, M, Lin, S & Talbert, S (2003). Data mining issues and opportunities for building nursing knowledge. Journal of Biomedical Informatics, 36 (2003), 379-388.

Sacristan, JA (2013). Patient-centered medicine and patient-oriented research: improving health outcomes for individual patients. BMC Medical Informatics and Decision Making, 13(6), 1-8.

Windle, PE (2004). Data mining: an excellent research tool. Journal of PeriAnesthesia Nursing, 19(5), 355-356.


This work is by Linda H. Eaton is licensed under a Creative Commons Attribution 4.0 International License





No comments:

Post a Comment