– A natural language processing tool was able to analyze notes in the EHR and identify prostate cancer patients experiencing social isolation, revealed a study conducted at the Medical University of South Carolina (MUSC) and published in BioMed Central Medical Informatics and Decision Making.
Social isolation can have a detrimental impact on patients, the researchers said, with its effects equal to those of standard clinical factors such as smoking, obesity, and hypertension. Typically, providers collect data about social isolation from patient surveys and questionnaires, so it isn’t always easy to see if a patient is experiencing loneliness.
“Unlike other social determinants such as race/ethnicity, depression, alcohol use, and nicotine use, information about social isolation is not captured routinely and is usually not encoded in the EHR; however, social isolation might be documented in clinical notes where providers record the information as told by their patients,” the researchers said.
“Because those clinical narratives are available in electronic format, a potential alternative to identify and extract patients’ social isolation information from clinical narratives is natural language processing (NLP).”
Among 55,516 notes from 1057 patients, the NLP algorithm identified 40 notes with a likely mention of social isolation from 17 patients, or 1.7 percent. After manual review, researchers found that there were four false-positive mentions of social isolation, mostly due to ambiguities and alternate meanings of words.
The tool performed with 90 percent accuracy and 97 percent recall, further demonstrating the potential for the technology to extract insights from unstructured data. In a 2017 study conducted at Massachusetts General Hospital, researchers applied NLP techniques to EHR data and found that it helped providers identify search terms associated with the social determinants of health.
Most of the patients in the MUSC study were white and Medicare or Medicaid beneficiaries, revealing that social isolation is more prevalent among certain groups than others. Providers could use this information to target interventions for their patients.
“Our study found that prostate cancer patients with evidence of social isolation had similar demographic, social, and economic characteristics leading to multiple risk factors for mortality,” the team stated. “Thus, in addition to medical treatment, providing sufficient social support to those patients is important to improve their quality of life and survival rate.”
Future studies could focus on developing and testing interventions for socially isolated patients, the team said. For now, these patients can get referred to support services within their hospitals and communities.
The research also raises questions about whether providers are talking to their patients about social determinants – especially social isolation, which generally eludes easy documentation in the EHR.
“The algorithm performed well, but the problem remains that some physicians do not comment on these issues and so don’t leave a trail for NLP to follow,” said Lenet.
Her words echo findings from a 2018 study, which showed that although 68 percent of patients have experienced at least one of the social determinants of health, 60 percent have never discussed these issues with their providers.
To further improve the detection of social isolation in patients, and to encourage patient-provider communication about this problem, the team hopes to develop a machine learning tool that will identify which traits typically characterize socially isolated patients. The tool could then search for patients with these characteristics in the EHR, without providers having to mention social isolation in clinical notes explicitly.
The NLP study did have some limitations, the researchers noted. The research included only prostate cancer patients at a single academic institution, so the terms used to train the NLP tool could have excluded terms that are suitable for other populations.
Additionally, because there is no standard terminology for documenting social isolation in the EHR, the lexicon the researchers used may not include all possible variants of social isolation.
The results of the study demonstrate what many in the healthcare space already know: That comprehensive, effective care requires a focus on the social determinants of health.
“Sometimes physicians focus excessively on the ‘medical’ problems and don’t pay enough attention to the context that people live in and the social aspects that influence their health,” said Leslie Lenert, MD, MS, Chief Research Information Officer for MUSC and director of MUSC’s Biomedical Informatics Center (BMIC).
“Our study once again highlights the importance of knowing this information in order to provide patients our very best care.”