| Description |
Patient safety is a top priority for healthcare system leaders, providers, and patients, but data can often fall through the cracks. An increasing number of healthcare systems are adopting patient safety reporting systems, yet leveraging this data to improve safety remains a challenge, particularly with large data sets composed of thousands of events reports (Ratwani & Fong 2014). Realizing the value of a patient safety reporting system is largely dependent on the number and quality of events being reported (Ratwani & Fong 2014). Researchers and practitioners have primarily focused on understanding and developing methods to increase reporting (Ratwani & Fong 2014). However, an equally important consideration that receives less attention concerns the need to develop strategies to effectively analyze the reports and ‘make sense of the data (Ratwani & Fong 2014). Currently, most healthcare facilities use adverse event reporting systems to gather information about safety problems and require healthcare providers to initiate safety reports (Daniels, et al., 2010). The problem with having all of this data is trying to analyze it. While some patient safety reporting system software systems include an analysis component, these capabilities are often limited to basic static graphs of the event data with limited availability to view the data based on variables of interest to the organization (Ratwani & Fong 2014). Reporting systems focus on data collection, rather than result reporting, and are not tailored for unique healthcare systems where results are commonly documented in narrative form. While narrative text can be challenging to summarize, natural language processing (NLP) can be used to semi-automate the analysis of large sets of reports. However, NLP is not simple. NLP consists of making sense of vast amounts of data by reducing the volume of raw information, followed by identifying significant patterns, and finally drawing meaning from the data and subsequently building a logical chain of evidence (Wong, 2008). While NLP may seemlike a good strategy for analyzing a dataset, there are things to consider. As Wong indicates, ‘Computers don't analyze data for the researchers, they still have to create categories, code, decide what to collect, identify patterns, and draw meaning from the data' (Wong, 2008). Getting started with NLP may be the hardest part because one must analyze data manually before anything else can happen. |
