A Cluster of Miscarriages Near a Landfill: What the Data Can and Cannot Say
A county health department receives a complaint from residents of a neighborhood adjacent to a closed municipal landfill. Over 18 months, 7 women in a 400-household subdivision report miscarriage; the expected number based on county rates is 4.5. Residents point to a 2019 groundwater report showing low-level trichloroethylene (TCE) at 3 µg/L in one monitoring well, below the 5 µg/L MCL, and to odors. Local media have run the story. The health department must decide whether to conduct a full investigation, and how to communicate. The state health department has limited staff and asks the county to triage.
1. Assess whether this constitutes a cluster warranting full epidemiologic investigation, and justify your reasoning with the numbers.
Read the worked answer
Seven observed versus 4.5 expected gives a ratio of about 1.56. With small numbers, the 95% confidence interval around 7 is wide (roughly 2.8-14.4 by exact Poisson), so the observed count is entirely compatible with chance. A single 18-month window in a 400-household area, with self-referred cases, is highly susceptible to ascertainment bias: women who are worried are more likely to report, and miscarriages that would otherwise be unrecognized (early pregnancy loss) may be over-represented in a concerned community. That does not mean 'do nothing,' but it means a full case-control study is unlikely to be informative or feasible now. The appropriate triage is: (1) verify case ascertainment (confirm pregnancies and outcomes, define the time window and geographic boundary a priori); (2) review the environmental data critically, noting that 3 µg/L TCE is below the MCL and that a single well does not establish residential exposure; (3) check for a plausible exposure pathway (is the well a drinking-water source? is there vapor intrusion?); and (4) conduct enhanced surveillance over a defined period rather than a one-off investigation. The judgment point is that a ratio above 1 is not evidence of causation, and small-area clusters are usually statistical noise unless the excess is large, sustained, and pathway-plausible.
2. Write the key messages the health department should deliver to residents, and identify what it should not say.
Read the worked answer
Say: (1) what is known - the number of reported cases, the expected number, and that the difference is within the range expected by chance; (2) what is being done - case verification, review of the groundwater and vapor data, and a defined period of enhanced surveillance with a report-back date; (3) what the exposure data mean - the detected TCE is below the drinking-water standard, and the department will assess whether the well is a residential source and whether vapor intrusion is possible; (4) practical steps - if residents use private wells, testing is available; report symptoms to their clinicians; (5) a named contact and a commitment to share results. Do not say: 'there is no risk' (overclaims certainty), 'the landfill caused this' (unsupported), or 'it's just statistics' (dismissive). Avoid releasing individual case information. The judgment point is risk communication that is honest about uncertainty while giving people actionable steps and a timeline.