Skip to content

Demo Script

Dylan Phelan edited this page Oct 13, 2017 · 3 revisions

Flux Notes Demo Script

Intro:

  • My name is Dylan Phelan, I’m a developer here in the health lab.
  • Today I’ll be demoing a prototype we’ve been developing called FluxNotes, an application that facilitates the capture of patient data without interrupting common clinician workflows.
  • The inspiration for FluxNotes comes from conversations we’ve had with domain experts and clinical user groups. What comes through every time we talk? Clinical notes are, for most clinicians, the gold source of patient information.
  • Conversely, Forms are the most common approach EHR's take to capture patient data, which work well in theory.
  • Problem: Clinicians are overburdened as is, and these forms are additional steps to their workflow that are not as effective in practice.
  • FluxNotes enables clinicians to rely on the workflows they know, recording clinical notes directly into an editor, while simultaneously aggregating and extracting structured patient data from semi-structured data in their notes.
  • This data not only provides the long term benefits of a structured, machine-computable electronic patient record, but provides immediate value to clinicians in the form of condition-critical treatment summaries based on the patient’s structured data.

Set the stage, Clinical Note I:

  • Let’s imagine we’re a clinician sitting down to record one of our notes after an encounter with a patient.
  • In this scenario, Debra Hernandez just received a pathology report and we're writing an assessment of it. We can start the review with basic information regarding the patient.
  • (Start typing note: @name is a @age year old @gender presenting with @condition [select Lobular carcinoma]).
  • Structured patient information is pulled in from patient EHR, maintaining consistency and enabling template driven approaches to note generation.

Top:

  • Basic Demographics: At the very top we can see a small selection of basic demographic information for our patient.
  • Main Diagnosis: On the right, we can see which current conditions the patient has. In this case, we can see Debra has been diagnosed with Lobular carcinoma of the breast.
  • Condition Select: If the patient we were seeing had multiple care plans, we would be able to use the select box to update their current condition and the structured data relevant to treatment.

Left-hand side:

  • Problem Summary: Here we see a table of the structured data that is condition-critical for the patient’s current treatment. These data elements have been carefully selected by domain experts and clinical user groups as the information most necessary for effective and efficient treatment of breast cancer.
  • How: Flux notes plugs in with the patient’s EHR to retrieve their clinical information, and then filters this display to present only data that our clinical user group has identified as being critical to the treatment of the patient’s current condition – in this case breast cancer.
  • Different Backgrounds: Light blue for information retrieved from the EHR, light red to inform the clinician that some information is still missing from the patient’s structured data, without interrupting their current workflow.
  • Inserting Extracted Data: If there’s information in this summary we want in our notes, we can click these plus buttons to insert it directly into where we were typing last.

Clinical Note II & Right Hand Panel:

  • Filling Gaps: On the left-hand panel, we noticed that the patient staging information was missing for this current condition. Why don’t we go ahead and include that in our clinical note?
  • Context and Ambiguity: A common problem run into when it comes to structured data capture is that of context. ‘Staging’, for example, is a value that qualifies the type of cancer someone is being treated for. To establish this context, we rely on a cascading series of contexts established within the note.
  • Right hand panel: In this case, we use the prior mention of patient condition to establish context for a) what data we can enter next (as we can see in the right-hand panel), and b) what those data elements refer to. At the top of this panel we can see what our current contexts are and see what structured phrases are available for insertion.
  • (Continue typing and using right hand panel, or begin using dictation mic if available: Pathology report indicates patient has #staging values #T1 #N1 and #M0, corresponding to @stage) We can use structured phrases in our clinical note to record that information, and FluxNotes will aggregate that information into our patient EHR.

Timeline:

  • In addition to the data summary panel, we have a condition specific longitudinal timeline below our central panel, providing a single glance perspective of the patient’s treatment history.
  • Currently, we can see that the timeline extracted information regarding the patient’s current medication regiment and a history of key events.
  • Like the updates to our left-hand panel, the timeline will update when new key events are recorded in our clinical notes.
  • (Continue typing and using right hand panel, new paragraph: Based on pathology results, we’ve determined the patient #progression is #Stable based on #Imaging and #Pathology)
  • This, in addition to the data summary panel, provides immediate payoff to clinicians for their (minimally) added effort.

Moving forward:

  • These are some of the mechanisms core to the development of flux notes
  • Keep in mind – this is an early stage prototype designed to get feedback on what features work and where we can improve.
  • Feedback has identified that this application, while valuable, could run into issues when it comes to full installation and training clinicians to use our shortcuts.
  • This has prompted the development of FluxNotes Lite, a slimmed-down version of FluxNotes.
  • Enables clinicians to familiarize themselves with shortcuts, and copy-paste them into their clinical notes. Application also guides users through value selection with specific tool tips and clarifying information regarding possible selections.
  • No EHR integration means minimal to no installation problems.
  • Envisioned use case: Clinical trials where users are more adherent to study requirements, where additional burden is a part of the
  • Feedback also highlighted the need for an additional retrospective tool; what would also be great is the ability to batch process old notes using a similar extraction process.
  • We’re currently threshing out new features to this application that would enable a clinician to retrospectively curate clinical notes to extract the relevant structured patient data, while defining patterns for what the data looks like in notes.
  • In this way, we envision the system learning to pick up and identify these patterns in the notes, eventually moving towards a hybrid human-automation approach where the system pre-processes notes and marks sections that seem to be identifying certain pieces of clinical information.
  • With that I’d like to leave the remaining time open to you all for any comments and feedback you may have on our prototype or future directions we could take this data capture tool. Thank you.
Clone this wiki locally