The delivery of high performance clinical care care that is reliable, efficient, and timely is difficult. Even under the best marriage of clinical workflow and algorithmic modeling, unpredictable adverse outcomes can occur in the modern clinical practice. Many analytics vendors address this through human oversight of their artificial intelligence (AI) algorithms.
Pieces takes this further with our Clinician-in-the LoopTM machine learning. Clinician-in-the-Loop enables Pieces DS to achieve extremely high levels of model performance in areas of clinical ambiguity. Clinician-in-the-Loop offers chart review and outcomes monitoring designed to efficiently identify areas for improvements in health system care pathways. Our vision is that this feature helps our health system clients to more rapidly improve, reinforce and modify their clinical pathways. For Pieces, the human-in-the-loop is a clinician.
Pieces-employed clinicians work with our models and outputs to ensure accuracy and reduce over-alerting seen in other products. This unique approach takes into account the highly nuanced and localized nature of clinical medicine.
The core technology that underlies Pieces Clinician-in-the-Loop is natural language processing (NLP). The tasks of NLP at Pieces Technologies are to extract medical concepts and their relations from clinical notes; integrate these NLP features into the machine learning models to identify disease conditions for early interventions; and highlight and index these NLP features to facilitate knowledge discovery in chart review. Pieces NLP implements symbolic NLP (rule-based and ontology-based), statistical NLP, and the mixture of both approaches for information extraction and retrieval tasks.
A chart review of a patient can typically take one to two hours. Pieces combines NLP and machine learning algorithms with an intuitive user interface to guide our clinicians to the important parts of the chart. These technologies allowed us to shorten chart reviews from hours to minutes.
This month, we released Clinician-in-the-Loop version 3.0, coupling algorithms identifying patients that can be discharged from the hospital earlier than planned with those identifying patients with low readmission risk. Together, the cohort of patients that is identified are those that can safely be discharged home early.
These algorithms also identify the particular clinical tasks required prior to the early discharge to help guide hospital clinicians and prioritize work. This allows hospitals to better predict real-time capacity and improve patient throughput.
Pieces Clinician-in-the-Loop enables continuous improvement of our artificial intelligence platform for better decision making by hospital caretakers.