Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Feb 20;115(8):E1730-E1739.
doi: 10.1073/pnas.1718679115. Epub 2018 Feb 5.

Assessment of the Legionnaires' Disease Outbreak in Flint, Michigan

Affiliations
Free PMC article

Assessment of the Legionnaires' Disease Outbreak in Flint, Michigan

Sammy Zahran et al. Proc Natl Acad Sci U S A. .
Free PMC article

Erratum in

Abstract

The 2014-2015 Legionnaires' disease (LD) outbreak in Genesee County, MI, and the outbreak resolution in 2016 coincided with changes in the source of drinking water to Flint's municipal water system. Following the switch in water supply from Detroit to Flint River water, the odds of a Flint resident presenting with LD increased 6.3-fold (95% CI: 2.5, 14.0). This risk subsided following boil water advisories, likely due to residents avoiding water, and returned to historically normal levels with the switch back in water supply. During the crisis, as the concentration of free chlorine in water delivered to Flint residents decreased, their risk of acquiring LD increased. When the average weekly chlorine level in a census tract was <0.5 mg/L or <0.2 mg/L, the odds of an LD case presenting from a Flint neighborhood increased by a factor of 2.9 (95% CI: 1.4, 6.3) or 3.9 (95% CI: 1.8, 8.7), respectively. During the switch, the risk of a Flint neighborhood having a case of LD increased by 80% per 1 mg/L decrease in free chlorine, as calculated from the extensive variation in chlorine observed. In communities adjacent to Flint, the probability of LD occurring increased with the flow of commuters into Flint. Together, the results support the hypothesis that a system-wide proliferation of legionellae was responsible for the LD outbreak in Genesee County, MI.

Keywords: Legionella pneumophila; chlorine residual; drinking water.

Conflict of interest statement

The authors declare no conflict of interest.

Figures

Fig. 1.
Fig. 1.
Spike in LD cases coincident with switch in water supply and increased variation observed in the Flint water distribution system. (A) Quarterly LD incidence in Genesee County, MI, 2010 through 2016. The count of LD cases in Genesee County as compiled in the Michigan Disease and Surveillance System at the quarterly time step. Bars in gray correspond to the preswitch period, bars in maroon correspond to the postswitch period, and bars in navy correspond to the switch back period. (B) Free chlorine at eight monitoring locations in Flint’s water distribution system, 2013–2016. Free chlorine (mg/L as Cl2) was reported weekly during the three water regime phases defined above (vertical lines) and the periods and dates (year/week) shown at eight locations in Flint.
Fig. 2.
Fig. 2.
Probability of observing a case of LD in Genesee County during four phases of the Flint water crisis. (A) The four phases of water exposure risk are defined. Phase A is the period before switch with water supplied by the DWSD from Lake Huron. Phase B is the period after switch to Flint River water treated by the City of Flint and before water boil advisories. Phase C is the period after switch to treated Flint River water and after boil advisories. Phase D is the period after switch back to water derived from Lake Huron. Start and end dates for each phase are indicated. (B) The probability of observing a case of LD in Flint and non-Flint census tracts by phases in the Flint water crisis. The estimated probability (with 95% confidence intervals) of observing a case of LD in a census tract in each of the four phases of water regime exposure risk in Oakland and Wayne census tracts (control group, non-Flint tracts, navy) and in Flint census tracts (treatment group, Flint, maroon) are shown. Estimated probabilities are derived with all other model covariates (i.e., meteorological and demographic) fixed at sample means.
Fig. 3.
Fig. 3.
The probability of an LD case being observed in a given week as a function of free chlorine residual. The estimated probability is calculated for each census tract within the Flint water distribution system for a given week as a function of free chlorine (mg/L as Cl2) before, during, and after the change in water supply. The probabilities are estimated with other observed model covariates (i.e., meteorological and demographic) fixed at sample means. Bars indicate 95% confidence intervals.
Fig. 4.
Fig. 4.
LD incident risk in non-Flint census tracts by commuter flow to Flint. LD incident risk is shown as a function of the number of commuters from Genesee County locations other than Flint either before (navy) or after (maroon) the switch to the Flint River as the Flint municipal water source. Probabilities are estimated with all other model covariates fixed at their sample means, and bars indicate 95% confidence intervals.

Similar articles

See all similar articles

Cited by 9 articles

See all "Cited by" articles

Publication types

Feedback