Why are positivity rates in/outside of NYC schools so different? My conclusion: though multi-causal, main driver is that in-school tests are less sensitive, which produces divergence in + rates over & above differences in pop sampled. We are missing a lot of + cases. 1/
Let’s start with an example we know all too well: standardized reading and math tests. Decisions about what counts as “proficient” are arbitrary, and can lead to widespread confusion when tests use wildly different thresholds for passing. (Think NAEP/state; pre- and post-CC.) 2/
We face a similar challenge in interpreting Covid test results. Currently administered tests vary dramatically in their ability to identify positive cases. The tests used in NYC schools miss more cases because both the type & quality of tests differ. 3/
In-school tests are self-admin “anterior nares” (short swab); at City MD, you’re more likely to get a nasopharyngeal swab (NPS). Ant nares tests miss more positive cases; if ant. nares at City MD, it's a diff testing provider, tho now DOE moving some tests over to same (PRL). 4/
Think of the limit of detection like a proficiency cut score. The NYC DOE @FulgentGenetics tests detect 5 virus copies per microliter; best-in-class NPS are .1-.5. [@BioReference declined to provide its LOD, despite 1 call & 3 email requests.) 5/
For sure: the tested populations differ in at least 2 ways: we should expect folks getting tests outside to have higher positivity because they’re more likely to have symptoms, & outside of @NYCSchools tests include kids > & < grades 1-5. 6/
Data time: + rates based on NYS-reported labs & NYC tests, broken out separately for kids & staff. (Note: included last week's tests that are >50% returned.) Reporters, let's stop using the cumulative positivity rate in schools that @NYCMayor cites when we have weekly data. 7/ 
In addition, talking about the city rate is misleading, since student and staff rates vary so much by borough. 8/
Summing up: @NYCSchools is one of few districts to roll out any testing, and they deserve kudos for doing so. But we have echoes of the NYC Great Test Score Wars of 2004-2010 here @DanKoretz, when tests were being used to legitimate policy decisions rather than inform them. 9/
We should seek not only to implement testing, but to implement tests sensitive enough to identify the overwhelming majority of asymptomatic cases, and have humility about how we use these data to inform and justify policy decisions. 10/10
• • •
Missing some Tweet in this thread? You can try to
force a refresh
