rdingwell/flux

Flux: Data Capture Prototype for Clinical Oncology

Flux Notes™ Prototype

Features

This prototype is designed to demonstrate the concept of a clinician entering or dictating a clinical note that includes structured data embedded within it. A patient summary is shown for reference during the authoring of new clinical notes and provide incentive to identifying structured data during note authoring and review. The clinician can leverage templates for note creation. When authoring a note and within a structured field, Flux Notes™ displays a simple "point-of-sale" inspired "form" for choosing values that are part of the current structured field.

In addition, a no-patient version of the above concept has been created for situations where EHR integration is not possible and to facilitate use at hospitals without requiring installation. In no-patient mode, the clinician choosing the type of data being authored and then uses the point-of-sale form from above to author the content which produces the equivalent structured field text that can be pasted into their EHR's note entry area as part of authoring a clinical note.

Getting Started

• If you are using a Mac, follow our guide for setting up your Mac for Flux Notes development

• Ensure that the following are installed:

  • Git;
  • Node.js (LTS version v10.14.1) ;
    • If Node Package Manager was not installed with Node, make sure to also install NPM.
  • and Yarn.

• Git clone the repository

  • If proxy issues are encountered, make sure to set HTTP_PROXY and HTTPS_PROXY environment variables.

• Check out the master branch (or the branch you want to work in)

• In a terminal navigate into the flux folder (stay at the project root)

• Enter yarn install

  • If proxy issues are encountered, make sure to set npm config proxy variables. Type these commands into your terminal:

    npm config set https-proxy <proxy-server-url>:<port>
    npm config set proxy <proxy-server-url>:<port>

• Enter yarn start to launch the development web server and open a browser to view the application

• To view patient mode, append patient to the end of the default url (result url would be http://localhost:3000/pilot1)

Technical Details

For more details on the technologies used see the wiki page at https://github.com/standardhealth/flux/wiki/Flux-Capture-Prototype-%28Flux-Notes%29

Parsing notes

A note parser can also be executed (local command line - not a web application) by (assuming the above Getting Started steps have been executed successfully):

• Enter yarn parse <path to note txt file or directory of files>

For example:

• Enter yarn parse ./src/noteparser/samples

will parse all the sample notes in the samples directory. To only parse note 4:

• Enter yarn parse ./src/noteparser/samples/note4.txt

If a directory is specified, every file in that directory will be treated as a note and parsing will be attempted.

Patient Date Update

This patient date updater is designed to change all of the dates in each of the entries in a patient JSON. One of the parameters passed to the app is the entryid of an encounter. The tool will calculate the difference of the current date and the date of the encounter and will add that difference to all of the dates. The tool can also output a list of all the dates in the record.

The patient date updater can be executed (local command line) by running the following command:

yarn patient-date-update <path-to-patient-json> [encounter-entryid] [options]

Where the the available options are:

-o -- Output a list of all the dates in the record -O -- Output a list of all the dates in the record in chronological order

• If no encounter-entryid is provided, the program can still run and output dates, but will not update any of the dates in the record.
• The program expects the JSON file provided to be an array of SHR entries.
• The program will overwrite the file that was passed in and will create a backup using the same path but adding '.backup' to the end.
  • For example, the backup file that will be created for HardCodedPatientMidYearDemo18.json will named HardCodedPatientMidYearDemo18.json.backup.

For example:

yarn patient-date-update ./src/dataaccess/BreastMainTreatmentDiabetesHypertensionJaneV09.json 33

will search the patient entries for an ReferralRequest with an entryId value of 33. It will then use the ExpectedPerformanceTime.Value.BeginDateTime.Value to calculate the difference to add to the dates.

The dates in the file can be easily listed out using the patient-date-list script. To run the script:

yarn patient-date-list ./src/dataaccess/BreastMainTreatmentDiabetesHypertensionJaneV09.json

This command will only output a list of dates found in the record, it will not update the patient record. It will list the dates in chronological order.

Testing

For docs on writing new tests, see here in our wiki. To run tests, go into the central project directory and run:

yarn test

This command will determine your machine's OS and run all possible tests. When in development, individuals can run front end tests with yarn test-ui and backend tests with yarn test-backend.

Required Extra Steps: Due to quirks in the libraries we're using, there are few things to do in order for all tests to pass:

1. Run yarn start in another terminal before running yarn test. Some ui-tests will fail because it takes longer for the site instance to spin up than it takes for the tests to start.
2. Ensure that all browsers' testing windows are open while the tests are running. TestCafe has been known to have some issues if those windows are minimized, possibly causing some tests to fail when they shouldn't.
3. Zoom out ridiculously far out on the ui-test's browser windows once they open. There are currently some issues we have with the our fixed copy-button that cause some tests to fail if the

Other Known Issues:

• Using Git Bash on Windows? It may look like the test aren't working on your machine. Hit enter a second time after running your yarn test command. The specifics of why this happens are still unknown, but this should help when running your tests.

Using the REST API

To use the REST API to retrieve application data, you need to clone the shr-node-server repository and follow the instructions there to run the server.

While the server is running, you can access a version of the Flux Application that uses the REST API by navigating to /p2.

NOTE: The REST API is still in development and will continue to change. Processes for changing the API and its implementations are outlined in the respective repositories, which can be found at the Standard Health Record Collaborative GitHub project, here.

Mock Data Update

This script will update the mock data in the mcdoe-pilot folder to be compliant with the transformedDataModel. The mock data updater can be executed with the following command:

yarn mock-date-update

Running the script will alter the mock patients so they match the transformedDataModel, which is a JSON file that defines the structure of a mock patient. The script will only add new properties to the patients, it will not remove or change existing ones. To add a new property to the patients, simply add a new property to the model and then run the script. The model takes three types of properties:

choice - Chooses a single value from a list at random

date - Chooses a year from between a range and then appends a random month

multi-choice - Chooses multiple values from a list at random, can return with an empty list

range - Chooses a number from between the provided bounds

date and range assume a uniform distribution, but choice and multi-choice can be given a list of weights that correspond to each choice to provide a weighted selection. All types are given a values property which defines either the numerical range that they can choose from, represented as an array with two members and used for date and range, or an array of values to select from, used for choice and multi-choice. For choice and multi-choice, the weight array is matched with the value array by index, so the value at values[0] will have weight weight[0]. Since choice only selects a single value, the probabilities in choice need to add up to 1. The multi-choice type weights do not have this restriction. Probabilities should be represented as a decimal number between 0 and 1.

The structure of the model is directly reflected in the mock patients. The patients will copy any nesting and key values present in the model.

Support Tools

BrowserStack is supporting Flux Notes™ through use of their service to test the code in this repository across browsers.