googlecloudplatform/deviceconnect

https://deviceconnect.readthedocs.io/

Device Connect for Fitbit uses Google Cloud and Fitbit web apis to enable developers to build clinical health applications that leverage data from users' Fitbit activity trackers. The tools provided in this repo are open source and build upon the Google Cloud services. These tools provide:

• end-user enrollment, consent management and Fitbit device linking,
• a data connecter that ingests data from the Fitbit web apis and pushes to Cloud BigQuery,
• looker dashboards for visualizing participants data in specific or in aggregate.

Deploy the infrastructure to capture your patient's fitbit data in minutes using the provided terraform scripts. Simply create your GCP project, install terraform, and run terraform apply.

The infrastructure uses Cloud Run and Fire Store to manage user enrollment and consent; and Cloud BigQuery to store data from the Fitbit webapis. Once the data is in BQ, build dashboards with Looker (or other dashboarding tool), and develop ML models with AI Notebooks or Vertex AI (tutorials coming).

Dashboards

The Fitbit data provides a unique 360-degree view of the patients' lives and can be clinically valuable for physicians or health coaches. It can also be integrated with other datasets like EHR data, or other device data.

Here are some dashboard views that might be valuable:

• the first view shows overview of stats. In this case, the patient's Fitbit data has been enriched with the date of a past surgury. It shows the percent of time she is wearing the Fitbit device, along with overall statistics.

  

• Furter details about steps can also identify potential problems to recovery.

  

• and more details about types and range of activity show more details for the clincial teams:

  

• or more details on sleep:

  

Machine Learning

While the dashboards can provide valuable insights about the patient's day to day, more detailed models can be developed that provide both population-level clustering across a patient population, and more predictive models that can help clinical teams identify address problems early.

Fine Print

• First and foremost, this is not a product! there is no support. This is provided as is, with no warrantees, as an example. Use at your own risk. post issues if you run into any, fork and develop if it is useful for your needs.
• By default, the code uses OIDC for user identities (seperate from Fitbit identities). The deployment instructions use Google OIDC but any OIDC can be configured.
• Data is ingested daily by default (you can configure the Cloud Scheduler component, see Terraform scripts). This is not intended for real-time use cases.

deviceconnect's Issues

add additional fitbit endpoints for SPO2 and skin temp

new webapis were added last year:

• https://dev.fitbit.com/build/reference/web-api/spo2/
• https://dev.fitbit.com/build/reference/web-api/temperature/

Data not ingested when language is not English

Description

My Fitbit application was set for Portuguese (Brazil) in the "Language by region/country", so the Device Connect was trying to capture the fields in English and it returned nothing. Changing this setting and registering myself again worked fine.

Screenshots

The screenshot below shows the empty fields, even with data collected by the device and the difference when the Language setting was changed.

To Reproduce

• Go to the Fitbit application (mobile) of the account you are collecting data;
• In your Account > Third-party apps > Manage third-party apps;
• Go to Settings > Personal Info > Language by Region/Country and change it to other than English (in my case, it was Portuguese)

add terraform for deployment

This code currently requires CloudRun, Firestore, Bigquery, and CloudScheduler to be configured to run. Having terraform scripts would make it easier to deploy into GCP.

exception occured: (missing_token) Missing access token parameter.

The token exists on firestore, but the following error occurs when the weekly data cannot be retrieved. The data cannot be retrieved even after the day when it cannot be retrieved.
Please let me know how to deal with this.

error message
exception occured: (missing_token) Missing access token parameter.

data can be missed

With the current approach -- with daily ingestion of data from the previous day -- data may be missed if the data has not been synced to the fitbit backend when the processes pull the data. One problem is that in general, the data may be missing

• because it does not exist, for example, if the battery has died no data would be generated.
• or the data may have been generated, but until the mobile phone app syncs the data, it won't show up on the backend web apis.

Similarly, if data is ingested multiple times, does the data get overwritten, or appended only?

fix for this issue: “application not populating the profile table”

fix for this issue: “application not populating the profile table”