1/ I'm upset at how little information is being generated from the #COVID19 lab testing that is finally starting to roll out
Key Qs that public health, policymakers, and the public need to know aren't being answered-CAN'T be answered-unless a modicum of *design* is put in place
Key Qs that public health, policymakers, and the public need to know aren't being answered-CAN'T be answered-unless a modicum of *design* is put in place
2/ In 25 years of working with public health and clinical data, I've learned one thing.
The answer to every data problem is: "what's the question?"
(usually repeated several times)
SO... what's the question we are trying to answer with COVID testing, and are we getting answers?
The answer to every data problem is: "what's the question?"
(usually repeated several times)
SO... what's the question we are trying to answer with COVID testing, and are we getting answers?
3/ Here's how it works when you drill down into
What Is The Question (WITQ)
Here's example:
WITQ? "How many people tested positive yesterday?"
WITQ? "I mean, is outbreak getting better or worse?"
WITQ? "Is incidence of new cases decelerating? (given incub, testing lag)"
What Is The Question (WITQ)
Here's example:
WITQ? "How many people tested positive yesterday?"
WITQ? "I mean, is outbreak getting better or worse?"
WITQ? "Is incidence of new cases decelerating? (given incub, testing lag)"
4/ Here's another one:
WITQ? "How many cases do we have in our city?"
WITQ? "How many cases without known exposure to a cluster we already know about?"
[Really "Is containment working, or do we have to go to suppression?"]
Notice- you need more data to answer deeper Qs
WITQ? "How many cases do we have in our city?"
WITQ? "How many cases without known exposure to a cluster we already know about?"
[Really "Is containment working, or do we have to go to suppression?"]
Notice- you need more data to answer deeper Qs
5/ Let me put it bluntly-
There's very little value to getting a bunch of positive test results without being able to understand the person's clinical condition, their source of exposure, and perhaps most importantly - how those compare to people testing negative
There's very little value to getting a bunch of positive test results without being able to understand the person's clinical condition, their source of exposure, and perhaps most importantly - how those compare to people testing negative
6/ (before you yell at me- there's obviously value to that individual person and the doctor tryin to treat them- I'm talking about value to public health workers, policymakers trying to make decisions that will affect the lives of millions)
8/ The CDC does not know the hospitalization status for 36% of the reported positive cases of COVID19.
ICU admission (53%)
death (47%)
age (9%)
why? Because we've allowed state-based public health system to equate fragmented information system
ICU admission (53%)
death (47%)
age (9%)
why? Because we've allowed state-based public health system to equate fragmented information system
9/ but that's not all of it
Even in South Korea's vaunted national response, we don't know if the unusual spike in diagnosed 20-29 year olds is just due to increased testing and spread in the church that was at the epicenter of the outbreak there
Even in South Korea's vaunted national response, we don't know if the unusual spike in diagnosed 20-29 year olds is just due to increased testing and spread in the church that was at the epicenter of the outbreak there
10/ So what do we have to do, right now?
First, as a condition of receiving approval and test kits, laboratories should be required to submit some aggregate information on every case *tested* not just the positives
Age group
Zip/county
cnn.com/2020/03/10/pol…
First, as a condition of receiving approval and test kits, laboratories should be required to submit some aggregate information on every case *tested* not just the positives
Age group
Zip/county
cnn.com/2020/03/10/pol…
11/ Second, CDC needs to fund & provide TA for public health to set up sentinel testing for #COVID19 where comprehensive clinical and exposure information is collected on a systematic sample of patients.
*Drive-through testing
*ILI sentinel practices
*Severe Acute Resp Illness
*Drive-through testing
*ILI sentinel practices
*Severe Acute Resp Illness
12/ Third, We need to conduct a serosurvey.
Maybe in New York City or Seattle
Where a random sample of households are serologically tested and symptom-surveyed to understand the true case-fatality rate and infection rate of this virus
That's an effort I led for West Nile Virus
Maybe in New York City or Seattle
Where a random sample of households are serologically tested and symptom-surveyed to understand the true case-fatality rate and infection rate of this virus
That's an effort I led for West Nile Virus
13/ Fourth, we need to investigate syndromic surveillance clusters to understand how much of the (terrifying) increase in ED visits for cough/fever/flu-like symptoms is actually COVID-related.
This could have huge implications for understanding the extent of under-ascertainment
This could have huge implications for understanding the extent of under-ascertainment
14/ I spoke recently to a state health official who was agonizing over incredibly difficult choices they have to make around the best use of limited lab testing resources
we need to have a planned design around who we test, what info is collected, and why we are testing.
WITQ?
we need to have a planned design around who we test, what info is collected, and why we are testing.
WITQ?
15/ This is not just about lack of available testing resources- it's about how we make use of those resources- what's the DESIGN?
It's as if we have a small amount of a novel drug, and instead of an RCT, we are using it on a haphazard set of people, collecting info haphazardly
It's as if we have a small amount of a novel drug, and instead of an RCT, we are using it on a haphazard set of people, collecting info haphazardly
