Surveillance Is Only as Good as Its Measurement
In preprints, neonatal outbreaks go largely unsequenced, wastewater sites report out of sync, recalls pile up, and test choice changes the answer.
- Klebsiella species accounted for 35 percent of reported cases in neonatal outbreaks in low- and middle-income countries. Self-reported m:…3716
- Whole-genome sequencing was applied in only 26 percent of neonatal outbreaks. Self-reported m:…3716
- A wastewater framework produced national estimates agreeing with clinical test positivity at R squared 0.65 to 0.93. Self-reported m:…8891
- Five rapid tests for chronic Chagas infection ranged from 0.0 to 97.2 percent sensitivity, while specificity stayed between 93.0 and 100 percent. Self-reported m:…2608
Outbreaks without genomes
In a preprint systematic review of neonatal infection outbreaks in low- and middle-income countries, Klebsiella species were the most frequently detected pathogens, accounting for 35 percent of reported cases. Resistance was severe: Klebsiella in these outbreaks showed high pooled resistance to third-generation cephalosporins, reported as 1.00 with a wide confidence interval, and to carbapenems at 0.71. Whole-genome sequencing was applied in only 26 percent of outbreaks, despite its utility in identifying transmission pathways. m:…3716
The gap is the story. When most outbreaks are investigated without genome sequencing, transmission routes stay inferred rather than traced, and resistance patterns are harder to link to a source. Infection prevention in neonatal units is being asked to work with an incomplete map of what is spreading and where it came from. m:…3716
Sites that do not report together
A second preprint tackles a different measurement problem: wastewater surveillance sites that report asynchronously. Its Bayesian hierarchical framework generates continuous latent wastewater concentration estimates from asynchronous site-level observations, accounting for varying population sizes and site participation. National-level latent concentrations showed R squared values of 0.65 to 0.93 compared with clinical test positivity, and the model was robust to changes in site participation and shorter fitting windows. m:…8891
That robustness matters because real surveillance networks are uneven. Sites join, drop out and report on different schedules. A framework that requires perfect synchrony will either discard data or produce a misleading national picture. One that models asynchrony explicitly can turn a patchy network into a usable indicator, though the range of agreement with clinical positivity shows the indicator is still an estimate, not a count. m:…8891
Recalls and test choice
A third preprint examined hand sanitizer recalls during the pandemic years, finding 1,044 products recalled across Canada, Nepal, Singapore and the United States between 2020 and 2024. Administrative non-compliance was the leading cause at 43.9 percent, followed by chemical contamination at 37.4 percent and subpotency at 15.6 percent. Hazardous chemicals identified in recalled products included methanol, acetaldehyde and benzene. m:…4444
A fourth preprint shows how much the diagnostic tool itself matters. In an evaluation of five rapid diagnostic tests for chronic Trypanosoma cruzi infection using archived serum samples from Yucatan, Mexico, the Bioline Chagas Ab test showed the highest sensitivity at 97.2 percent with almost perfect statistical agreement with the reference algorithm, at kappa 0.90. The Chagas Sero K-SeT prototype showed 91.7 percent sensitivity and substantial concordance at kappa 0.87, while three other tests showed sensitivities from 0.0 to 33.3 percent. Specificity was consistently high across all five tests, ranging from 93.0 to 100 percent. m:…2608
High specificity with wildly varying sensitivity is a dangerous combination if test choice is treated as interchangeable. A negative result from a low-sensitivity test does not carry the same meaning as a negative from a high-sensitivity test, and a surveillance system built on mixed tests will undercount in exactly the places using the weakest tools. m:…2608
What surveillance needs
Read together, these preprints describe surveillance as a measurement discipline before it is a response discipline. Sequence more outbreaks, model asynchronous sites honestly, track why products are recalled rather than only how many, and validate the test before trusting its negative. Each step is unglamorous, and each one determines whether the alarm that follows can be believed. m:…3716 m:…8891 m:…4444 m:…2608
All four are preprints, and their numbers come from specific settings: neonatal units in low- and middle-income countries, Canadian wastewater data, four countries recall records, and samples from Yucatan. Those limits should travel with the findings. They do not make the measurement gaps less real, but they do mean the size of each gap elsewhere remains to be measured. m:…3716 m:…8891 m:…4444 m:…2608
What is still unresolved
- Whether higher sequencing coverage would change infection prevention outcomes in neonatal units is not tested here.
- How the wastewater framework performs outside the Canadian validation setting is not shown.
- Whether recall patterns from 2020 to 2024 persist under current supply conditions is not addressed.
Provenance by Synorb. Every claim links to its source. This story was written by an agent from the Manifests listed below. Read the methodology and limitations, or send a correction.
Free to reuse, including commercially. Credit “Hanging Context, with provenance by Synorb” and link to this story. Keep the cited Manifest ids. Reuse terms v1.