Machine learning highlights mental health challenges facing healthcare workers during COVID-19 pandemic

In a study published in JMIR IAResearchers assessed the anxiety and depression faced by healthcare workers (HCWs) in the United States during the coronavirus disease 2019 (COVID-19) pandemic.

Using machine learning methods, they showed how unique the issues faced by healthcare workers are and highlighted ways to more effectively support this essential workforce.

Background

Healthcare professionals are more vulnerable than the general population to mental health problems such as depression, anxiety and suicidal ideation. COVID-19 has further increased the stress and workload faced by healthcare workers. As the pandemic intensified, the number of patients exceeded available beds and hospitals were forced to operate beyond capacity.

Health care workers were working longer hours in unfavorable conditions, including a lack of equipment and resources, forcing them to ration care and make difficult decisions.

As frontline workers, they were more exposed to the virus and often had limited access to masks and other protections. Like many others, they also lost support from social and family networks due to strict quarantine guidelines.

Healthcare workers with depression and anxiety are more likely to make mistakes, inadvertently putting patient safety at risk. Improving their well-being is essential to strengthening the health system as a whole.

This requires more research to better understand the mental health challenges healthcare workers face and provide them with the support they need. Such interventions will help make the health system resilient to future pandemics and other disruptions.

About the study

Researchers obtained treatment transcripts from 820 health care workers who received psychotherapy from approved providers from March to July 2020. These transcripts have been anonymized to protect patient privacy.

Healthcare workers included providers such as physicians, residents, nurses, social workers, and emergency medical service providers. They were all self-referred and had active National Provider Identifiers (NPIs).

They received therapy as part of an initiative to provide free treatment to healthcare workers for a month. The telehealth platform that donated these services also treats non-health workers.

To identify how the challenges facing healthcare workers differed from those in the general population, researchers compared each provider to a non-healthcare worker based on similarities in symptoms, demographics, date of start of treatment and state of residence. Non-healthcare workers included in the study were English-speaking U.S. residents with access to the Internet.

Before receiving treatment, all patients were assessed for depression and anxiety by a licensed provider. The Patient Health Questionnaire-9 was used to measure depression symptoms, and a General Anxiety Disorder Scale-7 assessed anxiety symptoms.

They were excluded from the study if they (1) required hospitalization, (2) had suicidal thoughts, or (3) had bipolar disorder, substance abuse, and other disorders.

The researchers used a heuristic classification algorithm to obtain the occupation of each healthcare worker from the transcript. They then processed the anonymized transcriptions by converting the words to their root forms to create a “vocabulary.” They removed empty transcriptions and words from fewer than 50 documents.

This resulted in 1,208 terms from the transcripts of 820 healthcare workers and 1,259 from the 820 transcripts of non-healthcare workers. Structural topic modeling (STM) methods were then used to identify topics raised by patients and associations between topics and levels of depression and anxiety.

Results

Healthcare workers were predominantly women (91%) and had an average age of 31.3 years. New York and California accounted for more than a quarter of the sample. Just over half of healthcare workers were nurses, while less than 20% were doctors.

Notably, 35.2% of healthcare workers reported that this was their first experience with psychotherapy. Just over 56% of patient caregivers were diagnosed with anxiety disorders, while only 8.2% were diagnosed with depressive disorders. Before treatment, 601 out of 820 healthcare workers (73.3%) suffered from depression or anxiety.

STM showed that health care workers frequently discuss four topics related to health care delivery. Topics exclusively mentioned by them included (1) coronavirus fears, (2) their work in intensive care units (ICUs) and hospital floors, (3) masking and patients, and (4) their roles (such as attending or resident).

In contrast, non-healthcare workers mentioned only one topic related to pandemic anxiety and only one topic related to their employers.

Regarding mental health, healthcare workers and their witnesses covered five topics, discussing panic attacks, mood disorders and experiences of grief. Health care workers also frequently reported sleep disturbances.

Providers with moderate to severe depression or anxiety were more likely to discuss the hospital or areas such as the ICU. Compared to matched controls, healthcare workers were also more likely to report mood disorders or sleep disturbances.

Conclusions

By comparing 820 healthcare providers with 820 matched non-healthcare worker patients receiving treatment from the same platform, the researchers in this study used machine learning computational linguistics methods to show that healthcare worker patients had unique associations between psychiatric symptoms and their work. The findings show that the pandemic has increased the levels of work-related stress that healthcare workers regularly face, highlighting the need to prioritize their mental health.

The authors acknowledged the limitations of the study and identified avenues for future research. Because the patients were self-represented, the researchers were unable to include those with limited access to virtual therapy.

There was a clear disparity in the sample favoring female providers, particularly nurses, indicating the need to reach more male physicians and practitioners. Further studies could also include more complex linguistic models and allow analysis of non-English transcripts.

Despite these shortcomings, it is clear that the study provides concrete evidence of the unique challenges facing healthcare workers during the COVID-19 pandemic.

It also demonstrates how machine learning algorithms can be used to process and analyze large data sets and guide clinical interventions while maintaining the confidentiality of study participants.