Developing a clinical decision tool based on electroretinogram to monitor the risk of severe mental illness

Abstract

Background: We have shown that electroretinograms can discriminate between patients with severe mental illness (SMI) and healthy controls in previous studies. We now intend to enhance the development and clinical utility of ERG as a biological tool to monitor the risk of SMI. Methodology: A sample of 301 SMI patients (bipolar disorder or schizophrenia) and 200 controls was first split into a training (N = 401) and testing dataset (N = 100). A logistic regression using ERG was modeled in the training data, while external validation and discriminative ability were assessed in the testing data. A decision curve analysis was used to test clinical usefulness. Moreover, the identification of thresholds of uncertainty based on the two-graph ROC and the interval of uncertainty was used to enhance prediction. Results: The discriminative assessment of the ERG showed very high sensitivity (91%) and specificity (89%) after considering uncertainty levels. Furthermore, for prediction probabilities ranging from 0.14 to 0.95 in the testing data, the net benefit of using our ERG model to decide whether to intervene or not exceeded that of never or always intervening. Conclusion: The ERG predicted SMI risk with a high level of accuracy when uncertainty was accounted for. This study further supports the potential of ERG to become a useful clinical decision tool to decide the course of action for subjects at risk of SMI. However, further investigation is still needed in longitudinal studies to assess the external validity of the instrument.