Introduction
The purpose of this paper is to propose a nursing informatics project based on the use of natural language processing (NLP) to analyze adverse events and incidents in the hospital setting. The implementation of the proposed project will allow for monitoring risk factors for adverse events and incidents, which will help take timely preventive measures, thus improving patient safety and efficiency of healthcare. This paper will describe the project and identify stakeholders. Further, it will describe patient outcomes and patient-care efficiencies that the project aims to improve, as well as technologies required for the project. Finally, the paper will identify the necessary project team members and discuss the nurse informaticist’s role in the team.
Description of the Project
To reduce the amount of paperwork and improve the healthcare efficiency, healthcare organizations have transferred to electronic health records (EHRs). While this is a well-developed system allowing for clinical data storage and sharing, it presents some issues to nurses who complete most of the EHR documentation (Glassman, 2017). The EHR system allows for entering structured and unstructured data. As McGonigle and Mastrian (2017) point out, about 75% of a healthcare organization’s data is unstructured and resides in text files. Although unstructured, freetext clinical notes often contain valuable healthcare information, a large amount of them makes it easy for healthcare professionals to overlook useful data.
The proposed project aims at addressing this problem by means of NLP, which is a branch of an emerging technology of artificial intelligence (AI). In particular, the project suggests using NLP for the analysis of free-text clinical notes and incident reports. The purpose of this is to gain insights into what incidents and adverse events occur in the healthcare setting and why they take place. This information is necessary for identifying patterns leading to negative patient outcomes and developing measures to prevent incidents and adverse events, thus improving patient safety.
Stakeholders
Like any other change in healthcare processes, the proposed project will affect multiple stakeholders. According to Schwalbe and Furlong (2017), healthcare project stakeholders usually include the project sponsor, the project manager, the project team, patients, support staff, regulatory bodies, third-party payers, and opponents to the project. In the proposed project, the project sponsor, the project manager, and the project team will be stakeholders because they will be directly involved in the project development, implementation, and coordination. Nurses, clinicians, support staff, and other healthcare personnel will be affected because they will need to use the insights gained from the application of NLP in their practice. Patients will be stakeholders because the implemented project is likely to improve the rate of incidents and adverse events, which will affect patient outcomes and cost of care.
As for regulatory bodies, they will need to make sure that the project implementation does not violate any healthcare regulations and does not take a toll on patient safety and the quality of care. Third-party payers, such as insurance companies, will also be influenced since the introduction of NLP for analyzing clinical notes is supposed to reduce the cost of care as compared to manual review (Nakatani et al., 2020). The project may meet with opposition from healthcare professionals who are accustomed to manual review of clinical notes, and these opposing parties will also be stakeholders in this case. Finally, it is also necessary to mention the entire healthcare organization as a stakeholder because the project should align with the overall organization’s strategy and goals.
Patient Outcomes or Patient-Care Efficiencies
In the article written by Bacchi et al. (2020), it is hypothesized that the length of stay could be efficiently reduced with the help of an improved planning process and utilization of adequate technologies. The role of NLP, in this case, should be to predict possible patient outcomes and help care providers navigate through the existing regulations of consumer wellbeing. The researchers also claim that NLP could be beneficial due to its close association with machine learning and rather accurate predictions related to different patient groups (Bacchi et al., 2020). Even if some of the data are collected retrospectively, it would still be important to generate consecutive data sets to address admissions and the feasibility of improving patient outcomes. According to Bacchi et al. (2020), a reduced length of stay is a decent indication of the benefits associated with NLP, especially under the condition where the majority of care provision models currently do not support advanced technology. This study supports the idea that NLP could be used to predict patient outcomes and develop a treatment program that would resonate with their values as well.
On the other hand, there is research completed by Dreisbach et al. (2019), who performed a systematic review of the literature to gain more insight into NLP, data mining, and the impact of health informatics on patient outcomes. Their effort allowed them to analyze the role of online technologies in the development of care provision instruments. The researchers were able to validate the idea that patients who share their symptoms and provide feedback regarding treatment significantly contribute to improved quality of care as well (Dreisbach et al., 2019). After the implementation of such digital instruments as NLP, the majority of researchers found patients gaining more knowledge regarding essential medical terms and symptom taxonomy. It allowed them to view their health issues more realistically and contribute to the treatment process with the help of their feedback (Dreisbach et al., 2019). Even though the grade of reliability of these findings is questionable, it may be concluded that electronic instruments for outcome prediction and improvement represent a crucial variable in the process of clinical care provision. Text mining and processing cannot be ignored, as they show that patient outcomes improve when they get a chance to participate in the treatment process.
Technologies Required for the Project
The main technology necessary for the project is AI, namely, its branches such as machine learning and NLP. According to McGonigle and Mastrian (2017), machine learning is a subset of AI that allows computers to apply deductive or inductive learning to come to specific conclusions. Machine learning is required for the proposed project because it is necessary to teach the computer what data points it should identify in unstructured clinical notes. For example, in their study, Topaz et al. (2016) used machine learning when they supplied the machine with a training set containing data about wounds to teach it to identify the required information. NLP is the method that aims at understanding, processing, and interpreting human language (Young et al., 2019). This technology is necessary for the project because, to extract information from free-text clinical notes, the machine should be able to deal with unstructured textual data.
Another important technology that could improve the outcomes of the project significantly is electronic health records. In the article written by Kulshrestha et al. (2020), it is stated that natural language processing could be utilized in association with EHR software to improve the output of machine learning (NLP) and assess the context of patient outcomes as linked to the concept of complexity of care systems. It is crucial here that the value of EHRs could be seen through their impact on data output and provider-patient communication, with numerous health issues prevented based on the data collected via NLP (Kulshrestha et al., 2020). This would become an important medical intervention, which allows the providers to forecast the possible patient outcomes and suggest treatments that appeal to the consumers the most. The value of EHRs also relates to the ability to improve reporting standards and ensure that all patients adhere to the proposed methods of data collection and processing. In the future, the role of EHRs is going to become even bigger, so it should be crucial to view their impact as groundbreaking.
The Project Team
The project team is part of the stakeholders responsible for planning and executing the project. The project sponsor will be the primary team member, whose task will be to provide support for the successful implementation of the project and remove barriers. This role will be assigned to the Health Information Systems Director. The Nursing Informatics Manager will be the project manager, whose duty will be to oversee the implementation of the goals of the project and identify the necessary improvements. According to Schwalbe and Furlong (2017), healthcare project teams often include medical experts to ensure that the project is consistent with medical practices and will not harm patients. Since this project aims at identifying factors related to adverse events and incidents, clinicians and nurses should be included in the project team. Their role will be to determine whether the data extracted from unstructured clinical notes align with the medical practice.
The nurse informaticist will be an essential member of the project team. The nurse informaticist’s role will be to implement the discussed technologies in the healthcare practice, evaluate the success of the project, and suggest possible improvements. Glassman (2017) points out that for clinical data to be meaningful, the nursing staff needs to capture patient information correctly. Therefore, the nurse informaticist will also need to educate nurses and physicians on appropriate ways of making clinical notes to make the proposed NLP method provide more accurate results.
Conclusion
The proposed project involves establishing a healthcare information system that will implement NLP to analyze free-text clinical notes. The project will aim to extract information related to adverse events and incidents in healthcare settings from clinical notes. Since manual
reviewers can easily overlook this information, NLP, combined with machine learning, will help healthcare professionals to identify patient safety issues and make informed decisions to prevent them. The project is likely to improve patient safety and healthcare efficiency.
References
Bacchi, S., Gluck, S., Tan, Y., Chim, I., Cheng, J., Gilbert, T.,… & Koblar, S. (2020). Prediction of general medical admission length of stay with natural language processing and deep learning: A pilot study. Internal and Emergency Medicine, 1-7.
Dreisbach, C., Koleck, T. A., Bourne, P. E., & Bakken, S. (2019). A systematic review of natural language processing and text mining of symptoms from electronic patient-authored text data. International Journal of Medical Informatics, 125, 37-46.
Glassman, K. S. (2017). Using data in nursing practice. American Nurse Today, 12(11), 45–47.
Kulshrestha, S., Dligach, D., Joyce, C., Baker, M. S., Gonzalez, R., O’Rourke, A. P.,… & Afshar, M. (2020). Prediction of severe chest injury using natural language processing from the electronic health record. Injury, 1(59), 1-8.
McGonigle, D., & Mastrian, K. (2017). Nursing informatics and the foundation of knowledge (4th ed.). Jones & Bartlett Learning.
Nakatani, H., Nakao, M., Uchiyama, H., Toyoshiba, H., & Ochiai, C. (2020). Predicting inpatient falls using natural language processing of nursing records obtained from Japanese electronic medical records: Case-control study. JMIR Medical Informatics, 8(4), 1–14.
Schwalbe, K., & Furlong, D. (2017). Healthcare project management (2nd ed.). Schwalbe Publishing.
Topaz, M., Lai, K., Dowding, D., Lei, V. J., Zisberg, A., Bowles, K. H., & Zhou, L. (2016). Automated identification of wound information in clinical notes of patients with heart diseases: Developing and validating a natural language processing application. International Journal of Nursing Studies, 64, 25–31.
Young, I. J. B., Luz, S., & Lone, N. (2019). A systematic review of natural language processing for classification tasks in the field of incident reporting and adverse event analysis. International Journal of Medical Informatics, 132, 103971.