PICTURE: Pediatric General Floor Deterioration Prediction

Written By Phil Jacokes

PICTURE is a suite of machine learning algorithms that utilizes electronic health record (EHR) data to predict patient deterioration in hospital settings. This iteration is modeled specifically for pediatric inpatients in a general floor.

Early detection of patient deterioration has been found to lead to reduced mortality risk, reduced length-of-stay and decreased hospital costs, yet identifying patient deterioration is a challenge for clinicians. PICTURE is a suite of machine learning algorithms utilizing electronic health record (EHR) data to passively and accurately predict ICU transfer or death as a proxy for patient deterioration. PICTURE also provides explanations for every single prediction, adding transparency to the model and how it calculated its output. This transparency is invaluable to clinicians who can use these explanations to guide decisions around patient care.

The Pediatric General Floor Analytic is modeled after a separate, pediatric data set and takes into account several features determined by Michigan Medicine pediatricians.

Unique Features

  • Provides an explanation of the main factors contributing to its prediction in each instance

  • Uses data that is collected during routine medical care such as common lab values

  • Will not disrupt clinician workflows or practices, even if policy changes alter routine tests

  • Defined thresholds allow each hospital unit to specify their desired precision

  • Designed to model patient’s physiology as opposed to clinician behavior, ensuring the alarms provide novel information to care providers

  • Pediatric model-specific: evaluates capillary refill time as a factor in the model

  • Pediatric model-specific: limited to ages 1-22

Digital Health Innovation Support

Digital Health Innovation is working with the Weil Institute and the Michigan Institute for Clinical and Health Research (MICHR) to conduct 10 human-centered design sprints, engaging with over 40 Michigan Medicine nurses, clinicians, and hospitalists to inform the optimal workflow, MiChart designs and requirements. After the designs are implemented by the Clindoc team, Digital Health Innovation and the Weil Institute will conduct a 3 month randomized control trial (RCT) to test the effectiveness of the model.

 

Principal Investigators
Sardar Ansari, PhD
Kevin Ward, MD

Intellectual Property
Invention Disclosure # 2019-094, 2021-244, 2021-245, 2021-250, 2021-251
Patent Issued: #11587977

Phil Jacokes
Previous

PICTURE: Adult General Floor Deterioration Prediction
Next

PICTURE: Rapid Response Team (RRT)