Research/Medicine
State wastewater data confirmed ER surges but gave no earlier warning
In a pre-registered back-test across all 50 states and DC, a simple wastewater alert rule beat random timing for COVID-19, influenza and RSV, but it was not earlier than the same rule run on emergency department data. 11 of our 16 predictions held.
Oct 29, 2026 · 7 min read

Wastewater surveillance is often described as an early warning: virus shows up in sewage before people show up at the hospital. We tested that claim for the data most US health departments can see, CDC's published state-level wastewater viral activity, against state emergency department (ED) visits for COVID-19, influenza and RSV. The wastewater alerts were far better than chance. But for none of the three viruses did they come significantly earlier than the same alert rule run on the ED data itself, and for influenza and RSV they missed most surges. Under the reading we fixed in advance, state wastewater gave confirmation, not lead.
0.36 / 0.51 / 0.44
share of wastewater alerts that were right, COVID-19 / influenza / RSV
0.12–0.15
share expected by random timing
37% and 26%
of influenza and RSV ED surges caught by the wastewater rule
11 of 16
pre-registered predictions held
The question
Studies that found a wastewater lead were mostly single-state, or scored against hospital admissions or laboratory data, which arrive later than ED visits. We found no open, pre-registered back-test that ran one rule on CDC's state wastewater data for all three viruses in every state, and compared it both with random timing and with the same rule on clinical data. That comparison is the one that matters to a health department deciding whether wastewater buys time.
What we did
We took two public CDC datasets: wastewater viral activity by site from the National Wastewater Surveillance System, and ED visit percentages by state from the National Syndromic Surveillance Program. For each state and virus we built a weekly series, the median site value for wastewater and the ED percent for clinical data. One fixed rule fires when a series is above its own trailing 52-week median and has risen two weeks running. We ran it on wastewater, on ED data alone, and on both together, and scored it against ED surge onsets over 8,448 state-weeks per virus, with weeks ending 2022-01-01 to 2026-10-17.
The rule, the outcomes, six hypotheses and our predictions for each were published before the job ran. Every grade below applies that plan exactly as written. Nothing was re-run, re-tuned or dropped.
What we found
First, the alerts carry real information. For all three viruses, the share of wastewater alerts followed by a surge was far above what random timing would give (p = 0.0005 each, against 2,000 random draws).
Share of wastewater alerts followed by an ED surge (PPV)
Second, reach. For COVID-19 the wastewater rule caught 65% of ED surges, which met our bar. For influenza and RSV it caught 37% and 26%, well below the 50% we predicted and well below the ED-only rule. In 24 of 51 RSV series and 12 of 51 influenza series it caught no surge at all.
Share of ED surges caught
Third, and most important, timing. Among surges both rules caught, we counted which alerted first. For COVID-19 it was close to even. For influenza and RSV the ED-only rule was usually earlier. We had predicted a wastewater lead for RSV; that prediction was wrong. When the weekly publication schedule is counted, with wastewater a week behind, the ED rule was earlier in 78 of 104 untied COVID-19 pairs, 30 of 31 influenza pairs and 15 of 18 RSV pairs.
Which rule alerted first, among surges both caught (same release cycle)
| Virus | Wastewater earlier | ED-only earlier | Ties | Sign test p |
|---|---|---|---|---|
| COVID-19 | 42 | 37 | 41 | 0.33 |
| Influenza | 2 | 26 | 4 | 1.00 |
| RSV | 6 | 12 | 3 | 0.95 |
One pattern explains much of this. For influenza and RSV, 46% and 57% of wastewater alerts came when ED visits were already above their surge threshold, against 25% and 15% for the ED-only rule. At the state level, as built here, wastewater trailed ED for these two viruses more often than it led.
Follow-up studies
The state median may hide early sewersheds. A descriptive audit found that 58% of influenza A and RSV site-weeks sit exactly at the site's own baseline, so a state median moves only once more than half its sites rise. A second pre-registered study tried a population-weighted mean: it caught far more influenza and RSV surges (58% and 51%), but at about three times the false alarms, and still did not warn earlier than ED data. A third tried outlier-resistant aggregates and found none that kept the extra surges without the extra false alarms. A fourth counted how much the data change between weekly releases: between two releases, 21.9% of state wastewater values in the weeks both covered changed, while the ED values did not change at all.
What it means
This does not say wastewater surveillance is useless. Its alerts are informative, it covers people who never visit an ED, and sewershed-level or county-level signals may behave differently from state medians. It says that for a state health department already watching ED visits, the state-level wastewater series as published did not buy lead time in the seasons tested. The project's report page now states that plainly beside each virus. A prospective test on the 2026-27 season, scored on data as first published each week, is recording now.


