Thread by @PardisSabeti

Altruism is not only good for society––in an outbreak, it can also be self-serving. A thread on our modeling paper medrxiv.org/content/10.110… w/ @ivan_specht @KianSani @codeanticode and thanks @EmilyAnthes piece @NYT nytimes.com/2021/03/24/hea… 1/16

COVID-19 made clear that diagnostic testing is a highly effective tool to curb viral spread. A positive test result informs infectious individuals that they & their close contacts must self-isolate, reducing transmission. #covid19 #testing 2/16

As well-endowed institutions––e.g. schools, workplaces––sought to revive operations, they turned inwards, deploying vast resources to test their own members regularly, often w/clinical-grade tests, while surrounding communities suffered testing shortages & delays. 3/16

For example, the NFL spent $100m on daily testing w/24 hour return time, burning through ~1M mostly negative tests. In spite of this, they had outbreaks across a number of teams. They were not alone. 4/16

The White House outbreak & many others show failure of confined testing programs. By not testing the surrounding community, institutions miss a chance for social good & remain blind to community cases that could breach the institution & spread like 🔥🔥 5/16

We hypothesized that institutions that only test themselves would be less safe than those that supported community testing. 6/16

We constructed an #epidemiological model to test this hypothesis, taking an agent-based approach & generating contact network for how agents interact. Beyond typical SEIR compartments, we have quarantine ones that agents enter when they or close contacts test positive. 7/16

We found that the deployment of diagnostic tests to members’ close contacts outside the institution offered the best protection (overall lowest transmission) in practically all tested scenarios. #protection 8/16

Under baseline parameters––modeled on our affiliated institution, @ColoradoMesaU––using about 45% of tests outside the institution was the best way to stay safe. This proportion depends on variable model parameters including interaction and contact tracing levels. 9/16

Going beyond @ColoradoMesaU, our results were robust to a wide range of factors, such as local community prevalence, social mitigation efforts, testing capacity, and contact tracing adoption 10/16

For best performance, institutions would ask members to report close contacts & know who is critical to test. But even w/o perfect reporting, our results show overall transmission w/in institution & beyond is lower when testing is deployed inside & outside the institution. 11/16

Mathematically, our results are hardly surprising. It’s intuitive that if institutions allocate most diagnostic tests to members’ contacts, they will likely preempt the virus’s emergence within their walls by establishing a ‘barrier of defense.’ 12/16

By leveraging a larger % of tests and resources towards strategic testing in surrounding communities, institutions can create circles of safety around them, & limit viral transmission both within and outside their institutional walls. 13/16

Our research justifies #altruistic––and more effective––testing strategies. Early in the pandemic, many institutions chose to devote considerable resources to testing themselves -and themselves only. This policy is neither the most ethical nor the most safe. 14/16

Epidemics are one of those rare instances where a seemingly selfless approach is, in fact, the most self-serving: institutions must test beyond their walls to stay safe behind them. 15/16

Thanks to our authors, full list: @ivan_specht, @KianSani, @BottiYolanda, @mhughes6510, @DrKHeumann, Amy Bronson, @MesaVeep, Emily Baron, @eparrie, @orglennon, and @ben_fry, @codeanticode, @PardisSabeti. 16/16