If you found non-aggregate data based on latitude/longitude, then you'd still need some algorithm to map your point to the nearest measurement coordinates - which is not trivial. It's unlikely to find US-wide "incident-level" data, whereas most data will be aggregated to some geographical region and time frame.
For that reason, I think most demographic, economic, environmental and crime data will be only mappable from aggregates like zip code, city or municipality, voting district, county and state. You can create a simple mapping table between all your latitude/longitude combinations and then which zip code, municipality, etc they belong to. For 200k records, you probably can't use Google Maps Reverse Geocoding (also due to license):
But OpenStreeMap provides reverse geocoding as well
Geonames is another resource for reverse geocoding.
Whatever API service you use, make sure you understanding the quotas, and that you also have a good strategy for importing all the data to a local database (or file archive). This way you can do the reverse geocoding once for all your records, over a period of days or weeks, and then not need any more queries.
Once you have your latitude/longitude mapped to geographical regions, there are tons of resources at data.gov to start to join to your original dataset. (Don't forget about season differences!)
Sidenote - for those using zip codes - check out this research that shows how zip codes masked the contaminated water crisis in Flint, Michigan.
Their ZIP code data included people who appeared to live in Flint and receive Flint water but actually didn't, making the data much less accurate than it appeared.