Pre-print: "Variation in False Negative Rate of RT-PCR Based SARS-CoV-2 Tests by Time Since Exposure"

A new analysis of already published swab (NP swab?) PCR test data.

Goal: model & predict PCR detection probability starting d -5 of symptoms #COVID19 https://www.medrxiv.org/content/10.1101/2020.04.07.20051474v1

Data was from two previously published studies:
1. https://www.medrxiv.org/content/10.1101/2020.03.06.20031856v1
2. https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciaa344/5812996

Modeling assumptions*
- RT-PCR is 100% specific (no false positives)
- Exposure was 5 day before symptoms

*They aren't claiming or believe it, just setting parameters.

The pre-print is short and I'm just repeating the claims (statistical modeling is not my field).

First, model predicts, on day 1 of EXPOSURE 100% chance of PCR failure.

By day 4 post-exposure, 61% chance of failure "although there is considerable uncertainty in these numbers"

Second, prediction of SARS-CoV2 PCR false negative of #COVID19 patients based on time of SYMPTOMS

Median false negative PCR rate of 39% on day of symptoms.
By day 3 of symptoms, decreases to 26%
By day 19 of symptoms, back up to 61% false negative rate.

https://www.medrxiv.org/content/10.1101/2020.04.07.20051474v1

Third, the reverse question is if SARS-CoV-2 PCR is negative, what's the probability being infected?

Seems like this one figure short paragraph feel insufficient to the gravity of that question. Test predictive value is population dependent.

This probability is in who exactly?

Unsurprisingly since it's based on previously published #COVID19 data, model recapitulates the "60-70% clinical sensitivity" claim that gets repeated everywhere (got it in an email today)

But modeling change over time reiterates that best nasal swab result: day 3 post symptoms!