New science and a second federal audit
Published 13 October 2026
The scientific account of the outbreak advanced in 2018. A PNAS paper by Sammy Zahran and colleagues examined the relationship between water-source changes, disinfectant levels and Legionnaires' disease. The article page gives an early February publication date and the later journal issue, so it belongs before the July federal audit rather than in the year of the outbreak itself. Its authors reported that disease risk rose after the switch and returned towards earlier levels after the return to purchased water. Their statistical findings supported a system-wide explanation for the outbreak. The paper did not merely repeat the state release's unresolved position from January 2016; it used data across places and time to evaluate a causal hypothesis. The results were the authors' scientific estimates, with the uncertainty and study design that accompany such estimates. F017, pnas.org, Assessment of the Legionnaires' disease outbreak in Flint, Michigan, 2018-02-20 F015, michigan.gov, Increased cases of Legionnaires Disease Investigated in Genesee County, 2016-01-13
The researchers used free chlorine as an indicator of conditions that could permit bacterial growth. They described chlorine demand as complex, affected by organic matter, metals, ammonia, temperature, corrosion and other factors. Limited historical data prevented them from identifying all the constituents responsible for that demand. The paper explicitly declined to treat low residual chlorine as the sole physical cause of disease. That limit belongs beside the result, because a strong statistical relationship is not a complete reconstruction of every chemical and biological step. What the team could observe was variation in a routinely measured disinfectant residual and in reported illness. Their argument was that this measure offered a useful signal of disease risk within an operating municipal system, not that one measurement fully explained every infection. F017, pnas.org, Assessment of the Legionnaires' disease outbreak in Flint, Michigan, 2018-02-20
Before the source change, the eight monitoring locations had relatively similar chlorine patterns. After it, both differences between sites and variation over time grew. The paper described one location with a sustained low residual, another with persistently higher values and a third with considerable fluctuation. After the switch back, the extreme variation largely subsided, with an exception at one location. Treatment at a plant and conditions at a distant monitoring point are related but different observations. The authors compared Legionnaires' disease presence by census tract and week, and assigned water quality monitors to parcels using occupancy, water age and the pipe network. The spatial pattern was part of the evidence for a distribution-system problem. F017, pnas.org, Assessment of the Legionnaires' disease outbreak in Flint, Michigan, 2018-02-20
The study compared Flint with other areas and divided the timeline into four exposure phases: before the change, after the change before boil advisories, after advisories and after the return. It used a difference-in-differences approach, comparing changes over time in the affected area with changes in comparison areas. The authors treated the non-Flint pattern as an estimate of what Flint might have experienced without the source switch. That counterfactual is a modelling assumption, supported in the paper by similar earlier risk patterns, rather than an observation of an alternative Flint that actually existed. The design allowed the researchers to separate a water-regime effect from some other changes common to places and seasons. It still required careful interpretation of the units in each result: people, tracts and weeks were not interchangeable denominators. F017, pnas.org, Assessment of the Legionnaires' disease outbreak in Flint, Michigan, 2018-02-20
In the abstract, the authors estimated a 6.3-fold rise in the odds of a Flint resident presenting with Legionnaires' disease after the switch, with a ninety-five percent confidence interval from 2.5 to 14.0. They also reported higher odds at low weekly neighbourhood chlorine levels, with separate estimates for thresholds below 0.5 and 0.2 milligrams per litre. Those values are odds ratios in the paper's models, not a statement that 6.3 times as many residents were all infected. The interval indicates substantial uncertainty around the estimate. The paper associated a decline in risk after advisories with possible avoidance of water, but its abstract described that mechanism as likely, not as behaviour documented for every patient. F017, pnas.org, Assessment of the Legionnaires' disease outbreak in Flint, Michigan, 2018-02-20
On July 19, 2018, the EPA inspector general issued the fuller audit, Management Weaknesses Delayed Response to Flint Water Crisis. It followed the 2016 management alert and examined additional failures in oversight and response. The audit attributed implementation lapses to the city and state, while identifying federal management weaknesses that delayed intervention. As of May 2018, it said some emergency-order actions were complete and others remained underway. The report's date distinguishes a review of earlier failure from an assessment of the response then in progress. Its conclusion was not that every agency had been inactive or that nothing had improved. It was that unclear roles, weak risk assessment, ineffective communication and inadequate proactive oversight had kept available evidence from producing timely federal protection. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
The audit treated Flint's missing service-line inventory as a regulatory failure as well as an obstacle to repair. High-risk sampling sites could not be properly selected without knowing where lead connections were. It said the inventory submitted after the January 2016 emergency order was incomplete and inaccurate, and that the city did not update it until August 2016. This connects two parts of the story often considered separately. A city's records of its buried lead service lines bear on how monitoring sites are chosen. If high-risk homes are missing from the list, apparently compliant results may not answer the intended question. The inventory weakness was also described as a problem beyond Flint, making the case relevant to national oversight without claiming that every water system had the same conditions. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
On treatment, the audit said the state directed another six-month monitoring round instead of continuous corrosion control. It recorded a rise in the city's reported ninetieth-percentile result from six to eleven parts per billion, while explaining that neither result excused the treatment requirement at issue. The action level of fifteen parts per billion and the five-part threshold discussed for treatment were different regulatory measures. A result below fifteen was not proof that omitting corrosion control was permissible. This was the core of the audit's criticism of the state interpretation. The later return brought treated water with corrosion control, but the network had already been damaged. The audit said lead did not fall below the federal action level until late 2016, and described supplementary treatment added in December 2015. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
The audit recommended annual controls to verify state monitoring of all Lead and Copper Rule requirements, including sampling-site identification and continuous treatment. It also recommended revising monitoring and corrosion-control protocols. EPA agreed in principle with the annual-oversight recommendation, but the inspector general did not accept its proposed approach or pilot as an adequate substitute for actual annual controls. Recommendation 1 therefore remained unresolved at the report's release. Agreement over the value of oversight was not the same as agreement on a corrective action sufficient to close the finding. The distinction is visible in the audit itself: eight other recommendations had accepted planned actions and dates, but they were still pending completion. An accepted plan is a procedural status, not proof that the planned work had already occurred. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
The report described the federal-state relationship as overly dependent on partnership. Region 5 sought collaboration, but the inspector general said that preference limited effective oversight when the state did not meet requirements. It pointed back to a 2010 contractor review of state programme reductions. Region 5's reply to that 2010 review described the reductions as temporary and limited to activities unrelated to public health, and the reductions continued despite the contractor's recommendation. The audit said reductions continued and grew through 2015. This was a documented warning about programme capacity predating the river switch, examined by the audit in 2018. The report's criticism was not of cooperation itself. It was of cooperation without clearly understood limits and without an effective response when a partner failed to carry out the minimum programme. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
Examples of those reductions show why administrative rules could matter to health. The audit listed failure to issue violations for late monitoring reports, missing lead-rule forms and deficient consumer reports. Late or incomplete records could delay recognition of contamination, obscure whether the required sampling pool had been used or leave customers without complete information. The inspector general described potential effects, rather than proving that each listed omission caused a particular Flint illness. Even so, the examples challenge a simple divide between paperwork and protection. Sampling locations, result deadlines and public notices were among the requirements the table links to reduced state staffing. Calling them administrative does not make them irrelevant to whether a dangerous change is recognised while there is still time to intervene. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
Accounts of communication differed. Federal personnel told the auditors that their concerns had been clear and consistent, while a state manager disputed having received an instruction to begin corrosion control at the point claimed. The audit also traced the attempt to obtain a headquarters interpretation of the rule. Discussions began before the formal request, but that request did not come until September 30, 2015. The resulting November memo addressed large systems generally and was of limited use to the immediate Flint situation. The report therefore identifies more than a message simply not being sent. It describes disputed meaning, delayed escalation and a general clarification arriving after months of exposure. A clear record of who asked what, when, and with what urgency was part of the management capacity the agency lacked. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
Risk assessment was another weakness. The auditors argued that individual signals should have been considered together: a source change, complaints, bacterial violations, high household lead and missing treatment. They acknowledged that bacterial violations alone would not indicate lead contamination. The finding concerned the combined picture of a system under strain. The report said the regional administrator had not received the problem in a way that conveyed its gravity and that no widespread public-health issue appeared apparent to that office at the time. This is the administrator's account reported by the auditors, not proof that the underlying risk did not exist. It explains a failure in how information was assembled and presented. The agency had pieces of evidence without a process that reliably joined them into an escalating warning. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
Regional staff told the auditors that aesthetic complaints were common and often unconnected with lead. The inspector general accepted neither that point as a complete answer nor a claim that every bad-smelling tap must contain lead. Its argument was that complaints, particularly their volume and persistence alongside other signals, could be an important indicator. The report contrasted the regional response with EPA's later drinking-water action plan, which called for indicators to identify troubled systems and direct assistance and oversight. In this account, residents' observations belonged in risk assessment, even when they were not diagnostic tests. Referring callers back to the same local or state authorities could keep information circulating without changing protection. The audit urged an agency response that used such reports as evidence requiring assessment. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018
The audit also distinguished oversight tools from emergency powers. Region 5 had used some approaches, but did not turn to other authorities when those approaches failed to produce swift protection. The report discussed grant conditions and enforcement powers among the options not used. It said managers either did not recognise those powers or did not consider them appropriate. This widened the 2016 alert's focus on an emergency order into an account of ongoing programme management. Federal oversight could involve conditions, reviews, information and formal enforcement, rather than a choice only between persuasion and a dramatic emergency intervention. The report's lesson was that the agency needed to know its available tools, set responsibilities and recognise when one approach was no longer enough. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018 F093, epa.gov, EPA Office of Inspector General, Management Alert on Drinking Water Contamination in Flint, October 2016
The response itself still required adjustment. The audit records a November 2016 amendment to the emergency order after EPA saw limited progress on some actions despite good-faith effort. The amendment set new goals and allowed achievable milestones. By the audit's later status date, treatment oversight, sampling, outreach and technical assistance had all been part of federal work. Yet the report said not all required actions were finished in May 2018, and officials continued recommending filters for drinking and cooking. Compliance improvement and unfinished response thus appear in the same source. The story cannot end solely at the source return, the issuance of an order or a result below the action level. Each is a different milestone, with duties that may remain after it. F091, epa.gov, EPA Office of Inspector General, Management Weaknesses Delayed Response to Flint Water Crisis, Report 18-P-0221, July 19, 2018