first implimentation of rsv

[aysim319]

Description

add RSV for NHSN

Associated Issue(s)

• Addresses Adding RSV for nhsn #2116

[nmdefries]

This looks good.

We should do a brief statistical review on the new RSV signals. At minimum, we need national and state signals plotted over time, and RSV for some of those geos compared to flu and covid signals.

[nmdefries]

Statistical review plots: rsv_analysis.zip

Using one of those plots as an example,

RSV is shown as recently (in the current season) increasing maybe 50x above past values. This trend is visible in all plots.

Such a large increase is not believable and doesn't match other data sources, e.g.,

shows that hospitalizations this year are comparable to the last two years (CDC RSV-NET, go to "Patient Characteristics" -> "Case Counts").

Emergency department admissions per capita show the same trend.

(CDC RSV-NET, select "All seasons" on the side bar).

Percent of emergency department admissions attributable to RSV (not shown) also have the same trend.

I'm assuming something went wrong with our handling of the data. It's also possible that the source started reporting new streams (e.g. previously only adults, now added in children), but honestly the increase is so large that this seems unlikely. Edit: possibly a plotting bug?

We need to

• Look at the raw data from the source to check values
• Compare those values to values reported by CDC RSV-NET -- I think they have national and state
• Same for our processed values

[aysim319]

I'm assuming something went wrong with our handling of the data. It's also possible that the source started reporting new streams (e.g. previously only adults, now added in children), but honestly the increase is so large that this seems unlikely. Edit: possibly a plotting bug?

We need to

• Look at the raw data from the source to check values

not quite the same but mentioned below

• Compare those values to values reported by CDC RSV-NET I think they have national and state

from the download link (https://data.cdc.gov/Public-Health-Surveillance/Weekly-Rates-of-Laboratory-Confirmed-RSV-Hospitali/29hc-w46k/about_data) i'm not quite sure how to compare? do I use rate? Cumulative Rate? there is rate column

• Same for our processed values

We don't do any transformation besides filtering for "US" or "region" value for the geo columns

one of the main issues is that at least in NHSN data, people didn't report RSV until 2023/2024 that's the main reason for zero, and then peaking in the tail end many zeros in the csv
non-zero query (starts 2024-03-23)
NHSN CDC link: https://data.cdc.gov/Public-Health-Surveillance/Weekly-Hospital-Respiratory-Data-HRD-Metrics-by-Ju/ua7e-t2fy/explore/query/SELECT%20%60weekendingdate%60%2C%20%60jurisdiction%60%2C%20%60totalconfrsvnewadm%60%0AWHERE%20%60totalconfrsvnewadm%60%20%3E%200/page/filter

non-null query (starts 2023-10-23)
https://data.cdc.gov/Public-Health-Surveillance/Weekly-Hospital-Respiratory-Data-HRD-Metrics-by-Ju/ua7e-t2fy/explore/query/SELECT%20%60weekendingdate%60%2C%20%60jurisdiction%60%2C%20%60totalconfrsvnewadm%60%0AWHERE%20%60totalconfrsvnewadm%60%20IS%20NOT%20NULL%0AORDER%20BY%20%60weekendingdate%60%20ASC%20NULL%20LAST/page/filter

there's also spearman through time

Once I filter out the dates the comparison does seem more reasonable -- minus the plotting bug
not sure how to fix it

national level looks fine (i think?)

[nmdefries]

in NHSN data, people didn't report RSV until 2023/2024 that's the main reason for zero

Ah, I see, it was because of the reporting mandate changes:

All RSV data through October 31, 2024 represent voluntarily reported data; as such, all voluntarily reported data included in this dataset represent reporting hospitals only for a given week and might not be complete or representative of all hospitals during the specified reporting periods.

from the NHSN data description. We'll need to note this in the documentation, because the numbers prior to this season are not useful.

The national numbers match RSV-NET well for the most recent season. (RSV-NET is a representative sample covering only 9% of the population, so I multiplied it by 10 to compare values.)

This is good to go.