Read on Twitter
Preprint: NYC has been the epicenter of COVID-19 in the US. What is the prevalence of COVID-19 in NYC, and how does it vary by borough? To what extent does this variation correlate with mobility? https://dash.harvard.edu/handle/1/42665370
A recent report showed hospitalizations and mortality vary in NYC, but because we haven't had enough testing to assess population prevalence, it has not been clear whether this is due to differential access to care or to variations in prevalence. https://jamanetwork.com/journals/jama/fullarticle/2765524
Universal COVID-19 screening of pregnant women admitted for delivery offers a view of prevalence. We used data from multiple hospitals in NYC (1746 women) and found variation in prevalence, e.g., Manhattan substantially lower than the Bronx, that mirrors hosp & death rates
With several weeks of data, we found the suggestion that there was a COVID-19 prevalence peak in the full dataset around ~March 30, with prevalence then appearing steady for the month of April, and variation again by borough.
In keeping with prior work, a minority of those infected were symptomatic at the time of testing (22.5%). https://www.nejm.org/doi/full/10.1056/NEJMc2009316
What about the impact of mobility? To answer this, we used high-volume mobility data from Facebook to describe morning trips out of and evening trips into each borough to get a sense of commuting.
We chose commuting between boroughs, rather than within-borough movements, because movements within a borough may include activities consistent w social distancing, but commuting is likely associated with work & thus likely a good indicator of inability to engage in distancing.
We found mean estimated COVID-19 prevalence by borough was strongly inversely correlated with the reduction in commuting movements (Pearson R = –0.88, [–0.52, –0.99]) by borough.
![We found mean estimated COVID-19 prevalence by borough was strongly inversely correlated with the reduction in commuting movements (Pearson R = –0.88, [–0.52, –0.99]) by borough.](https://pbs.twimg.com/media/EXkd2kcWAAAp_Br.png "We found mean estimated COVID-19 prevalence by borough was strongly inversely correlated with the reduction in commuting movements (Pearson R = –0.88, [–0.52, –0.99]) by borough.")
A few thoughts about conclusions: 1. The pandemic is made up of hyperlocal epidemics. Just as the variable prevalence by borough would be masked by a NYC aggregate, there may be substantial heterogeneity even within boroughs by neighborhood.
2. Commuting mobility patterns predict COVID-19 prevalence in NYC boroughs. This highlights the need to provide greater support to neighborhoods with essential workers and others whose circumstances make it difficult to comply with social distancing recommendations.
3. While we looked at prevalence, these data suggest that large parts of NYC may remain at risk for additional COVID-19 outbreaks. More seroprevalence studies to assess cumulative incidence are needed.
4. Widespread testing remains key for assessing geographic disparities in prevalence, for more tailored interventions, and for a better assessment of the risk of additional outbreaks.
Great to work with many collaborators in NYC (Columbia, Cornell, Mt Sinai), Boston, and elsewhere. Excellent stuff as always from @StephenKissler, @nish_epi, @DanLarremore, @Caroline_OF_B.
