Google Timeline is a great way to piece together a person's path over time. I've personally used it to back track where I've been before especially when filling out timesheets.
I may be wrong but I believe that contact tracing can take a quite a bit of time and manpower, especially when dealing with larger numbers of cases to track. Also it may not always possible to get a person's past movement history in a timely manner for a number of reasons:
- Cases may not remember where they have been
- Cases may not want to reveal where they have been
- Cases may not be in a physical state to share where they have been
- Other factors like language may slow down the tracing process
Being able to quickly recreate a version of a patient's "Google Timeline" could speed things up significantly and the underlying technology is readily available. Instead of waiting a few days to piece together a person's past movements, this could be done in minutes or maybe even seconds if the data is readily available.
What it does
- The solution will pull only the patient's past cell tower pings (tower id, position, signal strength) from the telco data warehouses.
- The patient's position, length of stay can then be overlayed onto google maps or maptive on a day by day basis which could help the contact tracing team quickly put together potential hazardous interactions.
How it can be built
- A query can be sent to the relevant telcos to extract past 2 weeks cell tower pings which can be ingested into a performant cloud data warehouse like google bigquery or amazon redshift. It should be less than 2 million rows of data which is tiny for cloud data warehouses to handle
- A standard query can be run to generate a summary of positions and duration of stay which can then be overlayed on a map interface for the contact tracing team to utilise.
- I've worked with a team which has implemented for commercial purposes aggregated user paths and movement through Jakarta using similar data sets.
Some potential challenges
- there may be privacy issues, but since it is limited to only 1 person's data which could be with consent or could be allowed by policy
- it may take time/money to extract the data from telcos
What's next for Speed up contact tracing using telco data for "Timeline"
- It can be expanded to identify/alert other people who may be spending time in the same point of interest
- if tracetogether is installed, it can be paired with the data from tracetogether to enrich it with location information only when a confirmed case has been identified.