Measles is among the most infectious diseases known: more than 90% of susceptible close contacts of a case catch it. In February–April 2024 an outbreak linked to a Chicago shelter for newly arrived migrants, mostly from Venezuela, produced 57 confirmed cases — 52 residents, 3 staff and 2 community members. At the Chicago Department of Public Health's request, CDC built a dynamic model of the outbreak among shelter residents, updated as new cases came in, to forecast its size and length and to measure what the public health measures achieved.
The model
CDC adapted a model of measles in crowded settings first built during Operation Allies Welcome in 2021. It sorts people by infection state (susceptible, exposed, infected, removed), age and pregnancy, and adds active case-finding and the delay in detecting cases.
- The shelter held 1,877 people on March 8, when active case-finding began and a three-day vaccination drive gave 882 MMR doses, taking coverage to 93%.
- Immunity by age was estimated from Venezuela's vaccination record. One MMR dose was assumed 84% effective at 6–11 months and 92.5% from 12 months.
- Disease parameters came from the literature, calibrated to the data; each scenario was run 10,000 times over 365 days from February 1, and the 100 runs closest to the observed cases formed the forecast.
The outbreak fit a basic reproduction number (R0) of 25 — each case would infect 25 others in a fully susceptible population — far above the 12–18 usually cited, a sign of how fast measles can move through a dense shelter. It also fit active case-finding cutting the infectious period by 25%, from 5 days to 3.8.
Forecasts as the outbreak unfolded
| Data as of | Cases among residents | Forecast final size (median, IQR) | Forecast last rash onset |
|---|---|---|---|
| March 11 | 7 | 29 (20–39) | April 16 |
| March 18 | 18 | 38 (31–41) | April 18 |
| March 25 | 47 | 60 (57–65) | April 20 |
| April 1 | 51 | 60 (58–63) | April 20 |
| April 8 | 52 | 58 (56–60) | April 18 |
The final count among residents was 52. Early forecasts were uncertain and low; they sharpened as data accumulated, giving the city an expected size and end date weeks before the last case.
What the interventions did
The model re-ran the outbreak under different choices (not calibrated to the observed cases):
| Mass vaccination | Active case-finding | Chance of 100+ more cases | Median last rash onset |
|---|---|---|---|
| never | never | 69% | May 26 |
| March 15 (a week late) | never | 15% | April 21 |
| March 15 | March 8 | 8% | April 17 |
| March 8 (as done) | never | 1% | April 14 |
| March 8 | March 8 (as done) | 1% | April 9 |
| March 1 (a week early) | never | 0 | April 3 |
| March 1 | March 8 | 0 | March 30 |
- Doing nothing gave a 69% chance of 100 or more cases, a median of 235, and an outbreak lasting into late May.
- What Chicago did — vaccination and case-finding from March 8 — cut that chance 69-fold, to 1%, and shortened the outbreak by a median of about 7 weeks.
- A week's delay in vaccination would have raised the chance eightfold, to 8% (15% without case-finding).
- A week earlier, had the first case been reported when measles was first suspected, the chance of 100+ cases would have been zero, and of 50–99 cases 2–3%.
Modelled total cases among shelter residents under each combination of vaccination and case-finding start dates. CDC.
How it was used
The high R0, estimated early and shared with Chicago partners, underlined the need for active case-finding. With cases appearing in residents who had had one MMR dose, it also supported continued vaccination — including a second-dose campaign 28 days after the first — and asking shelter staff and essential visitors to show measles immunity. City leaders, government agencies and health care partners used the forecasts to plan. The results echo the Operation Allies Welcome model, where a 7-day vaccination delay would have raised the median outbreak size by 50%.
Limitations
Parameters were uncertain and partly chosen from the literature and expert opinion, since the outbreak was too small to estimate them all. The model assumed an exponentially distributed infectious period, overstating its variability. It did not account for the 22 families moved to a quarantine hotel on March 11–12, so it may have overestimated the outbreak. And the residents' true immunity was unknown.
Vaccination coverage in the shelter was below the 95% needed to stop measles spreading. The model code is public, for use by health departments and researchers.
Sources
- Nina B. Masters, Inga Holmdahl, Paige B. Miller and colleagues, "Real-Time Use of a Dynamic Model To Measure the Impact of Public Health Interventions on Measles Outbreak Size and Duration — Chicago, Illinois, 2024," MMWR, May 2024. https://www.cdc.gov/mmwr/volumes/73/wr/mm7319a2.htm
- Model code: https://github.com/CDCgov/measles-model-chicago-2024
- The outbreak itself: "Measles Outbreak Associated with a Migrant Shelter — Chicago, Illinois, February–May 2024," MMWR, May 2024. https://www.cdc.gov/mmwr/volumes/73/wr/mm7319a1.htm
Licence: CC0 1.0 (public domain) · Adapted from www.cdc.gov
1
0
0
0
Comments
