The product is a patient management system that currently supports a single data source, think heart rate monitor. But in the future we plan to support multiple data sources, so we are looking to support arbitrary clinical parameters.

We currently use an SQL DB as we have all clinical parameters defined, but I am exploring moving to a nosql DB to support arbitrary parameters as we currently have a column for each parameter, like heart rate. So im wondering if this use case would justify use of a nosql DB like dyanmoDB or mongoDB or if storing parameters as a json blob would be enough, ive heard postgres has good json querying support. We currently use about 18 tables to store data. But only 2 of these tables really store clinical parameters and trends. It sounds like the team would prefer to not do a partial migration and maintain 2 DBs, but I could be convinced.

The only consideration is we have data filters such as filtering patients whose heart rate trend is above 50 bpm along with the other parameters. Would postgres be able to hand these kinds of filters efficiently. We also might add tags later down the road.

I have a bit of choice paralysis between the different approaches and would really appreciate some suggestions and feedback as I’ve not done much DB design previously.

While nosql DBs would definitely fit my use case, im afraid of losing the benefits of the rest of relational data benefits. Ive also seen people reccomend a single table for dynamoDB, though mongoDB seems to have good querying capabilities. Im also wondering if storing the parameters as unstructured json in an SQL DB would cause inefficiencies when applying the data filters I talked about earlier. I would really appreciate any advice :), thanks. I would appreciate if you recommended a database service you think would suit this use case. Please let me know if i could provide any more information that would inform your decision.

Im currently leaning towards using mongodb because it has good querying capabilities. But having amazon handle DB encryption and scaling for dynamoDB is enticing as it is HIPPA compliant out of the box. But if we can use postgres json and keep out relational data efficiently that would also be nice. Thanks again!! It seems people are hesitant to use nosql DBs and im wondering if my use case justifies it.