EVIDENCE LIBRARY
Benzodiazepine Variety Predicted Diagnostic Conversion in a Model, Not a Drug-Caused Bipolar Disorder
A claims-based prediction study included benzodiazepine prescribing patterns among variables associated with later bipolar diagnosis.
A claims-based prediction study included benzodiazepine prescribing patterns among variables associated with later bipolar diagnosis.
The modeled outcome
Among 2,820 patients initially diagnosed with major depression, 536 later received a bipolar diagnosis. A classification-tree model used several treatment and service-use variables, including the number of benzodiazepine types within a visit. The study sought risk stratification rather than a causal drug effect.
Prediction can encode illness severity and recognition
More complex prescribing may mark anxiety, insomnia, treatment resistance, or already emerging bipolar features. Diagnostic codes and unmeasured family or environmental factors also limit interpretation. The analysis cannot show that benzodiazepines produced bipolar disorder or that changing prescriptions would prevent conversion. It also does not establish that the model predicts equally well in another population. The useful distinction is between a marker in a statistical model and a biological cause of disease.
Source paper
Original analysis of the supplied journal article, identified as PMC7165312.