Hive Mind: Harnessing Collective Intelligence for Predicting Clinical High-Risk for Psychosis Outcomes - The lifetime prevalence of psychotic disorders is 3%, with devastating consequences to the individual's well- being. Many patients with such conditions never recover from their first psychotic episode and end up deteriorating into a chronic phase. As a result, psychotic disorders are considered to be one of the most burdensome of illnesses. Due to the chronic trajectory of the disorder, attention is now focusing on the clinical high-risk (CHR) phase, before the onset of full psychosis, during a critical period when the brain is vulnerable and intervention is key. CHR benefits immensely from early intervention which could delay or even prevent conversion. Thus, predicting CHR outcomes such as identifying those who will convert early on would be of immense interest to the field and is crucial for tailoring and optimizing therapy. Current advances in artificial intelligence (AI) present valuable opportunities to develop and refine predictive analytical methods to tackle such challenges. Due to the high interest in applying AI in psychiatry, several risk calculators, and prediction algorithms, have been created to improve outcome prediction. However, such efforts lack a crucial component, a “Human- in-the-Loop” approach in which human direct involvement is included in the AI process. In order to diminish this gap, it is essential to establish a hybrid approach between humans and AI in making patient-related decisions. This can be achieved using Collective Intelligence (CI). CI in this case refers to the field of research in which the knowledge and wisdom of human groups are harnessed to achieve more accurate decisions, predictions, and insights than those produced by individual members or AI alone. It differs greatly from prior methods because it enables networked human groups to form real-time dynamic systems that quickly converge on unified decisions, assessments, and evaluations with significantly higher accuracy and insight compared to traditional methods. In this proposal we seek to 1) predict the outcome of CHR by a group of psychiatrists using clinical data individually and as a CI. We hypothesize that the psychiatrists' CI will have superior predictive accuracy compared to the average of each of them making the predictions separately; 2) predict the outcome of CHR by the CI of a multidisciplinary group of evaluators (including psychiatrists, and neuropsychologists) using clinical, and cognitive data as well as risk calculator predictions. When comparing the CI of the above-mentioned groups, we hypothesize that the integrated evaluation of the multidisciplinary group having access to the combined clinical and cognitive data will outperform the group of psychiatrists making decisions based on the clinical data alone. We also hypothesize that those having access to the risk calculator predictions will make more accurate predictions than both aforementioned groups. This work will determine and quantify the benefits of using CI, the value of using multimodal data, and risk calculator predictions, as well as the advantage of harnessing the knowledge, wisdom, and expertise of a multidisciplinary group of professionals for predicting outcomes in the CHR population.
