Five Cases Within 0.3 Miles: How Software Caught a Salmonella Outbreak Before the Lab Did

On May 21, 2019, before anyone had interviewed a single patient or a laboratory had typed a single isolate, New York City's health department knew it had a problem. Its software had flagged five salmonellosis patients living within a 0.3-mile (0.48-km) radius, all falling ill during May 11–17. A cluster that size should turn up about once every 2.3 years. The next day, interviewers found that three of the five had eaten prepared food from the same grocery store. The alarm came 9 days before serotyping results were available.

How the software works

Every weekday, the NYC Department of Health and Mental Hygiene runs the free SaTScan software's prospective space-time permutation scan statistic over salmonellosis reports, using each patient's geocoded home address and "event date" (illness onset, or specimen collection if onset is unknown), to look for recent, local increases. The city set this up in February 2019, so it could spot clusters without waiting for lab subtyping — ahead of CDC PulseNet's switch, on July 15, 2019, from pulsed-field gel electrophoresis (PFGE) to whole genome sequencing (WGS). WGS characterizes pathogens in far more detail, but takes longer, and some outbreaks can slip through when culture-independent tests leave no isolate or labs can't sequence every one.

The investigation

CDC FoodCORE-funded graduate student interns try to interview every salmonellosis patient quickly, usually within about 2 days of the report. The cluster alert told them what to listen for.

15 NYC residents met the case definition — lab-confirmed Salmonella after eating food from the store: 14 customers and one food handler, who fell ill after 10 of the customers and so was not the source. They ate there during May 8–20 and fell ill May 14–21; none shared a household; the median age was 42. The most common food was chicken — rotisserie chicken, chicken salad or chicken soup — eaten by 10.

The 17 cases in the SaTScan cluster, by date reported to the health department, New York City, May–June 2019. CDC.

All 11 patient isolates available were Salmonella Blockley, differing by a median of zero alleles by WGS; all nine tested by PFGE were indistinguishable. The same strain came from the leftover chicken.

What inspectors found

• A walk-in refrigerator at 51°F (10.6°C) instead of 40°F (4.4°C) or below.
• Raw chicken prepared inside the walk-in cooler, and gloved hands used to open its doors mid-preparation.
• A poorly calibrated thermometer, improper hot and cold holding, and inadequate cooling of cooked food.
• Eight environmental swabs were negative for Salmonella, but two of four ready-to-eat samples had high fecal coliform counts, and three held Bacillus cereus (70–670 CFU/gram, below the 10⁵–10⁶ CFU/gram that causes illness) — consistent with food held at the wrong temperatures.

Faster than before

The early start made the rest possible: testing food handlers, inspecting the store, deciding which isolates to sequence first, and collecting leftovers before they were thrown away. The local findings later fed a multistate investigation of S. Blockley linked to chicken. In a city with about 1,000 salmonellosis cases a year, the authors write, automated cluster detection — built on fast electronic lab reporting, geocoding and well-staffed interview teams — can catch local, community outbreaks the lab alone would find late.

Sources

Based on Latash J, Greene SK, Stavinsky F, et al., "Salmonellosis Outbreak Detected by Automated Spatiotemporal Analysis — New York City, May–June 2019," MMWR Morbidity and Mortality Weekly Report volume 69, number 26, Centers for Disease Control and Prevention; a work of the United States government in the public domain.