The following is a summary of “Choosing the right treatment – combining clinicians’ expert knowledge with data-driven predictions,” published in the September 2024 issue of Psychiatry by Maekawa et al.
A Bayesian network model was proposed to aid mental health specialists in making data-driven decisions about appropriate treatments.
Researchers conducted a retrospective study to form a probabilistic machine learning model to assist psychologists in selecting a suitable treatment for individuals with 4 mental disorders: Depression, Panic Disorder, Social, and Specific Phobia.
They studied a dataset from 1,094 individuals in Denmark with socio-demographic and mental health information. A Bayesian network was operated using a purely data-driven approach, later refined with expert knowledge, creating a hybrid model that created probabilities for each disorder, with the highest probability indicating the appropriate disorder for treatment.
The results showed improved performance of the data-driven approach, with an AUC score of 0.85 compared to 0.80 on the test data. Some cases were reviewed where the model’s predictions did not match the treatment. Symptom questionnaires revealed that these participants had comorbid disorders, with the actual treatment aligning with the model’s second-highest probability.
They concluded the hybrid model ranked the actual disorder treated as either the highest (67.3%) or second-highest (22.8%) on the test data in 90.1 % of the cases. It provided probabilities for multiple disorders rather than a single disorder for treatment, enabling individuals seeking treatment or their therapists to use this information as an additional data-driven factor for treatment.
Source: frontiersin.org/journals/psychiatry/articles/10.3389/fpsyt.2024.1422587/full
The post Enhanced Performance of Hybrid Model in Treating Mental Disorders first appeared on Physician's Weekly.