SEMANTIC SEARCH MICHAEL HOSKING CHIA, MACHI | JULY 2018 CLINICAL - PowerPoint PPT Presentation

CLINICAL DOCUMENT SEMANTIC SEARCH MICHAEL HOSKING CHIA, MACHI | JULY 2018 CLINICAL PRODUCT SPECIALIST

Time

“The brute force approach” “Click, view, click, view…” 6

Research

Contextual Inquiry ▸ Understanding the user problems ▸ Observational study conducted at North Shore Hospital in NZ ▸ Supported foundations for design sprint

Clinicians rely heavily on previous “summary documents” “[Anaestheti c report] is good because you know someone has trawled through the history and pulled this together” – ICU Registrar

Clinicians rely heavily on previous “summary documents” “A good past medical history can be hard to come by ” – ICU Registrar

Finding patient history can be tedious "Its a pain, have to go clicking through the documents in the left side and sift through everything ” – Medical Registrar

Complex patients can be overwhelming “Look at all these clinic letters – there is just too much to do anything with them” – Critical Care Outreach Nurse

Often there is no time to do it… “You end up just giving up as it takes too long” – ICU Registrar

Current User Journey Confident for Scan for Manually look New Patient Clinical Context for specifics Decision/Treatment 14

Design Sprint

Design Sprint Day 1 Day 2 Day 3 Iterate Understanding the Discuss range of Consolidate ideas Create a testable problem ideas into a basic prototype prototype for users

Day 1: Understand

Day 1: Understand ▸ Present user research from contextual enquiry ▸ Decide and align on the understanding of the problem

How might we make it faster and easier for clinicians to find information about a patient? 19

Day 2: Diverge

Day 2: Diverge ▸ No idea constraints ▸ Multiple ideas for solving problem ▸ Short sketching exercises ▸ Withhold judgement

“ How might we" Questions ▸ “HMW enable clinicians to search /explore across a patient’s record?” ▸ “HMW show related clinical concepts present in a patient’s record?” ▸ “HMW highlight relevant segments of a clinical document?”

Day 3: Converge

24

Combine Sketches ▸ Review solution sketches and votes ▸ Which ideas are similar? ▸ Which are really different? ▸ As a group, combine ideas into one solution

Day 3: Converge 26

27

Combine Sketches Converge ▸ Review solution sketches and votes ▸ Which ideas are similar? ▸ Which are really different? ▸ As a group try and combine the ideas into one solution 28

Day 3: Key outcomes ▸ Shared understanding that we want to enable a clinician: ▸ to quickly understand a patient’s medical history ▸ the ability to drill into relevant sections for more details ▸ Basic Prototype

30

Day 4 & 5 … : Prototype & Test 31

Prototyping

Feedback “[Timeline] Useful for seeing that nothing happened between these years. If I see a "picket fence" here, there’s a problem I will check out” – ICU Consultant 33

Feedback "Really good way of giving initial big picture idea of what this person's past medical history is... it sifts through everything for you ” – Endocrinology Registrar

Feedback “…this would mean not needing to flick through three or four discharge summaries … – Critical Care Outreach Nurse

Challenges ▸ Copy paste paradigm ▸ Capturing/surfacing context ▸ Data quality ▸ Availability of clinicians

Next Steps ▸ Continue iteration and clinical validation ▸ Develop a Proof of Concept with local data and clinical scenarios ▸ Measure success – compare time, accuracy and user satisfaction ▸ Develop product

Takeaways ▸ Design Sprints can be an effective tool for working with clinical end users ▸ SNOMED CT can support intuitive clinical interfaces ▸ Natural Language Processing can be limited by the original data

Michael Hosking CHIA, MACHI William Campbell Clinical Product Specialist Senior UX Designer Associate Investigator Michael.hosking@orionhealth.com