Role of Natural Language Processing (NLP) in Patient Experience

Role of Natural Language Processing (NLP) in Patient Experience

Over the last few years, the healthcare industry has been moving into an era of transparency and consumer awareness. The regulatory bodies have been publicly displaying the performance of healthcare organizations to enable patients to make informed decisions about the organizations they wish to receive care from. The major categories that many healthcare organizations are being evaluated include quality, safety, and cost. Patient experience, an integral component of quality, is gaining a lot of attention over the last decade as organizations strive to improve their customer service. To move towards patient-centered care, understanding the patient experience is a critical step. Patient experience matters as it not only has a direct impact on the bottom-line, value-based reimbursements, and reputation, but also there is an increasing evidence on the correlation with positive outcomes.

The Center for Medicare and Medicaid Services (CMS) and Agency for Healthcare Research and Quality (AHRQ) developed a national standard for reporting patient satisfaction called the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS). When it comes to collecting and understanding the patient responses, structured responses to check-box questions lack both context and the details. A lot of survey vendors also collect the patient verbatim comments and this part of the data is underutilized. To understand the root causes of patient dissatisfaction and identify potential opportunities to improve, some quality improvement teams review each patient-generated comment. This review process is very labor-intensive, expensive, and most of the times may not yield the output that we would be interested to make precise decisions to better the experience. Technologies like Natural Language Processing (NLP) can analyze and extract insights from narrative text like patient comments or any other unstructured data.

Using NLP to Analyze Patient Comments

The team at KPI Ninja worked with a healthcare organization to analyze the patient comments that have been received over the last couple of years using NLP techniques. NLP tools were used to extract the noun phrases and adjectives from each sentence of the patient comments. Parse trees and dependency trees were generated for each sentence. The comments have been classified into topics (topic classification) and sentiment analysis i.e., positive or negative has been performed.

All the patient comments have been categorized into four general broad categories – positive, negative, mixed, and neutral. These four categories further come with sub-categories which represent themes within each category – as an example, for positive responses, some of the themes include – care, staff, hospital, nurse, etc. These sub-categories could be further drilled down into specific comments under those themes to get to the bottom of the story.

The team also correlated this unstructured data (patient comments) with structured data (Age, gender, department, service, location, length of stay, discharge day/month, room number) to further identify insights on the patterns within this dataset. An example of this analysis is shown in the chart below.

Finally, a word cloud with the themes has been created (for both positive and negative comments) to provide a visual representation of patient comments data. The importance of each theme is represented by its font size as shown in the picture below.

Conclusion

Healthcare organizations are beginning to use technologies like NLP, big data, and machine learning to translate large volumes of verbatim patient comments into actions for a better patient experience. We cannot simply rely on results from check-box surveys, and technology is now available to make sense of verbatim patient comments in aggregate and turn this valuable feedback into actionable insights to take targeted improvement actions.

Tags: , , , , , , , , , , , , ,