observe-interpret-act: an nyulmc learning analytics case ... - pusic - final.pdfmartin pusic md phd...

18
Martin Pusic MD PhD on behalf of… Institute for Innovations in Medical Education NYU School of Medicine Observe-Interpret-Act: An NYULMC Learning Analytics Case Study

Upload: others

Post on 12-Feb-2021

2 views

Category:

Documents


0 download

TRANSCRIPT

  • Martin Pusic MD PhD on behalf of…Institute for Innovations in Medical Education

    NYU School of Medicine

    Observe-Interpret-Act:

    An NYULMC Learning Analytics Case Study

  • LEARNING

    ANALYTICS“Learning analytics refers to the interpretation of a wide range of data produced by and gathered on

    behalf of students in order to assess academic

    progress, predict future performance, and spot

    potential issues”

    - DOE 2012

  • How is Big Data Different?Traditional Big Data

    Data Collection Purposeful Incidental / Opportunistic

    Intrusiveness High Low

    Data Acquisition Cost High Low

    Frequency e.g. Quarterly Continuously

    Temporality Static Reports Dynamic Dashboards

    Hypotheses Causal Associations

    Sample Size Small biopsies Large swaths

  • NYU Education Data Warehouse

    Education Data

    Warehouse

    LMS

    Curricular

    Content

    AdmissionsExams

    ePortfolioPatient

    Log

    Learning

    ModulesEvaluations

    SIS

    Reporting and Analytics

    Data Marts for

    Education Research

    Simulation

    Clinical Data

    Warehouse

    NYU Practice

    NetworkEpic EMR

  • iBeacons

  • New “Listeners”

    • iPads

    • Point-of-care digital forms

    • Immediate Lecture evaluations

    • Learning interactions with EHR

    • Anything a SmartPhone can do

  • NYU Education Data Warehouse

    Education Data

    Warehouse

    LMS

    Curricular

    Content

    AdmissionsExams

    ePortfolioPatient

    Log

    Learning

    ModulesEvaluations

    SIS

    Reporting and Analytics

    Data Marts for

    Education Research

    Simulation

    Clinical Data

    Warehouse

    NYU Practice

    NetworkEpic EMR

  • NYU Education Data Warehouse

    Education Data

    Warehouse

    LMS

    Curricular

    Content

    AdmissionsExams

    ePortfolioPatient

    Log

    Learning

    ModulesEvaluations

    SIS

    Reporting and Analytics

    Data Marts for

    Education Research

    Simulation

    Clinical Data

    Warehouse

    NYU Practice

    NetworkEpic EMR

  • Gartner Model

    © Gartner Group

  • Path Diagram for the Class of 2014

    • Path coefficients (with standard errors) were calculated using multiple regressions.

    • Each variable is set as the dependent variable and all variables to its left as possible predictors.

    • We only include the paths that are statistically significant.

    • Think of each number in the following way – adjusted for all the other incoming arrows, a one standard deviation change in the predictor results in an x standard deviation change in the thing being predicted.

    • For example, the regression equation predicting USMLE1 includes MCAT, College GPA and Medical School GPA. College GPA does not signfiicantly predict USMLE1 so there is no number. Adjusted for MCAT score, a one standard deviation rise in Med_GPA results in a 0.62 standard deviation rise in USMLE1 score.

    MCAT

    College GPA

    USMLE1

    0.26(0.067)

    0.37(0.085) 0.62(0.073) USMLE2SHELF

    0.36(0.09)

    0.42(0.09)

    0.29(0.09)

    0.57(0.08)Med_GPA

  • NYU Analytics Center

  • Clinical Curriculum

    High level

    view for

    chairs

    Curriculum Stage

  • LEARNING

    ANALYTICS“Learning analytics refers to the interpretation of a wide range of data produced by and gathered on

    behalf of students in order to assess academic

    progress, predict future performance, and spot

    potential issues”

    - DOE 2012

  • LEARNING

    ANALYTICS“Quality improvement refers to the interpretationof a wide range of data produced by and gathered

    on behalf of clinicians in order to assess progress,

    predict future performance, and spot potential

    issues”

    - DOE 2012

    Quality

    Improvement

  • The New CME

    • Un-Announced Standardized Patients

    • In-Situ Simulations

    • Patient Reported Outcome Measures

    • Process Metrics

    – E.g. Door to Needle Time in Stroke Activations

    – E.g. Surgical Video

  • How is Big Data Different?Traditional Big Data

    Data Collection Purposeful Incidental / Opportunistic

    Intrusiveness High Low

    Data Acquisition Cost High Low

    Frequency e.g. Quarterly Continuously

    Temporality Static Reports Dynamic Dashboards

    Hypotheses Causal Associations

    Sample Size Small biopsies Large swaths

    Feedback

  • Door to Needle Time

  • Discussion