Much of this data is stored in unstructured formats which used to be difficult to analyze on a large scale. However, organizations are now leveraging tools such as natural language processing (NLP) text mining to glean better insights from this wealth of data.
NLP text mining gives users the ability to transform unstructured data, such as free-text clinical notes or transcribed voice-of-the-customer calls, into structured data to provide insights that can improve the health and well-being of patient populations. As we move into 2018, look for the following NLP trends to make an impact on healthcare and the life sciences over the next year.
1) The empowerment of end users
Self-services is everywhere, from retail to hospitality. Self-service is also becoming the norm in healthcare with patients demanding better online tools to access their medical records, communicate securely with their physicians, and schedule appointments.
In healthcare and the life sciences, clinicians and scientists are seeking self-service tools to obtain information. Traditionally users have had to turn to information specialists to exploit NLP. In response to end-user demands for faster, self-service alternatives, companies are now delivering broader access to easy-to-use NLP text-mining tools and putting the power of these technologies into the hands of everyday clinicians and researchers.
2) Wider use of machine-learning technologies
Organizations are leveraging the use of machine-learning technologies to drive better health outcomes and to advance the discovery and commercialization of new drugs. In the past, machine-learning projects were based only on existing structured data. However, with 80% of information within unstructured text, machine-learning projects need to use features from unstructured as well as structured data.
For example, in healthcare there is a need to access fine-grained information from the unstructured fields of the medical record to better-understand patient populations, including the social determinants of health such as social support network, food insecurity and ambulatory status. With streamlined access to critical information stored as unstructured text, providers can build machine-learning models to more easily identify potentially at-risk patients and take proactive measures to improve outcomes.
NLP-based text analytics are also helping biopharma companies to extract key details from both internal documents, and scientific literature. Use of NLP combined with relevant ontologies turns the unstructured into structured to drive analytics or machine-learning models. They’re then able to identify treatment patterns, side effects, drug interactions, and more. Based on findings, drug companies are better-equipped to advance the innovation and delivery of their treatment offerings.
3) Large-scale population screening
NLP text-mining technologies are helping with the identification of at-risk patients on a large scale. For example, clues about potential opioid abuse are often hidden in unstructured clinical notes. By analyzing data extracted using NLP text-mining, providers can examine content from both structured and unstructured sources to identify patterns that reveal possible abuse, such as symptoms consistent with opioid withdrawal, patient histories that include numerous “accidents” and the need for pain medications, or the presence of certain psychiatric disorders.
Similarly, the analysis of unstructured clinical records can help identify patients with early signs of cancer, diabetes, or other conditions that require closer monitoring or additional care coordination. By identifying at-risk patients before their health is severely compromised, providers can appropriately dedicate resources and better address the needs of their patient populations. These patient safety nets will increasingly rely on automated NLP screening, frequently combined with a further layer of machine learning. This automated screening will identify high risk patients in large, real-time feeds of EHR data to improve care and reduce risk of litigation.
Healthcare and the life sciences have access to a treasure trove of useful data, but much of it is stored as unstructured text. NLP text-mining is helping to make this data readily available to providers and researchers in structured formats that are easily consumed for machine-learning technologies or other applications, or for broader access and decision support. As we move into 2018, look for wider use of these technologies at the clinician and researcher level, as well as the expanded use of machine learning applications that incorporate NLP text-mining. Finally, NLP text-mining will play a larger role in the identification of at-risk individuals, allowing providers more opportunities to proactively address their health needs.
By David Milward, PhD, CTO, Linguamatics
