Want to learn more about behavioral health management?
An increasingly common aim for NLP is improving mental health treatment. While adoption is still in the early stages, existing research shows its potential for two new applications.
1. Early detection of cognitive decline and mental disorders
NLP is opening up new possibilities for how mental health professionals can evaluate free speech to help identify and predict psychiatric illness in patients. As NLP and machine learning capabilities become more sophisticated, automated speech analysis can offer predictive power that exceeds today's standard clinical evaluation methodologies and scoring systems.
In one proof-of-concept study, researchers tested how NLP and machine learning could help predict psychosis onset in a group of teenagers and young adults. Each of the study participants had a baseline interview and was assessed quarterly for up to two and a half years. Using transcripts of the baseline interviews, researchers fed speech features into an algorithm and found that features such as semantic coherence and speech complexity had a significant correlation with eventual psychosis development— predicting psychosis onset with 100% accuracy.
Seniors in particular can also benefit from this type of NLP functionality. The Baby Boomer generation will greatly expand the senior patient population, putting additional stress on health care providers to incorporate new technologies for early identification of various geriatric health risks. One area where NLP can assist is with identifying speech alterations that signify cognitive decline (e.g., Alzheimer's disease), as impaired speech is often an early symptom of preclinical stages of dementia.
For instance, researchers from Johns Hopkins University and Atrius Health in one study used a NLP algorithm to evaluate unstructured EHR data from more than 18,000 senior patients. The researchers found that the algorithm's analysis of free-text clinical notes helped identify significantly higher rates of geriatric syndromes when compared with using claims and structured EHR data alone. Across multiple health conditions, the NLP use case outperformed structured data, ranging from identifying 1.5 times as many cases of dementia to 456 times as many cases of social isolation.
2. Predict the need for high-risk interventions
NLP and machine learning algorithms can also evaluate free text (e.g., EHR notes, patient portal messages) to help predict patients who may be at risk of self-harm or psychological distress, enabling care teams or other caretakers to intervene faster. For example, health providers can evaluate patients' mental health based on what they post on social media sites or to online blogs/message boards. Given the widespread adoption of social media and mobile devices, health care providers have a wealth of new data to work with. Feeding patient-generated text into NLP and machine learning algorithms, caregivers can create predictive models that signal when a patient is experiencing deep depression, undergoing an anxiety attack, or having suicidal thoughts.
This type of targeted intervention is already common in online marketing, where companies can evaluate consumer activity and preferences to tailor their advertising and steer consumers toward a desired outcome. Within health care, providers can leverage NLP to predict people's mental states to proactively reach out to patients through phone or text, or they can create "online pathways" that direct patients to digital peer groups, counseling, psychotherapy, or mental health educational websites.