Kiln, Cough and a Deadline: A Health Survey in Nadia
A student joins an NGO survey of 120 brick-kiln workers beside a wetland in Nadia. The kiln runs November to May; the survey is in March. Workers are migrants from Jharkhand and Odisha, many on advance payments, and some fear that refusing answers will cost them work. The questionnaire asks about cough, chest pain and smoking; about 40% smoke bidis and most cook with biomass. There is no spirometer, no comparison village, and no dust measurement. The NGO wants a report in six weeks, before a local election, and its coordinator says the report should 'prove' that the kiln is causing disease. A local doctor offers to sign a list of 'respiratory cases' without examining the workers.
1. Given the design, what conclusions can the student legitimately draw, and what would be needed to support a causal claim?
Read the worked answer
The survey is cross-sectional and based on self-report, so it can describe but not prove causation.
Legitimate conclusions:
- the prevalence of reported cough and chest pain in the 120 workers;
- differences in reported symptoms by task, years at the kiln, sex or age, stated as associations;
- a clear statement that smoking, biomass cooking, season and previous exposure are plausible alternative explanations.
Not legitimate: 'the kiln causes disease', or a disease rate for the district, since there is no comparison group and no clinical examination.
To support a causal claim, the student would need:
- a comparison group (non-kiln workers in the same area, or the same workers before and after the season);
- an exposure measure such as task, distance from the kiln, dust sampling or years of work;
- spirometry or clinical examination, not only symptoms;
- control for smoking, age, sex and cooking fuel in the analysis.
The report should state these limits prominently, not in a footnote.
2. How should the student sample workers and obtain consent given the advance payments and the fear of losing work?
Read the worked answer
The debt relationship makes consent fragile, so the procedure matters as much as the form.
- Sampling: obtain the worker roster, then use a systematic random sample stratified by task (moulding, loading, firing) and by sex. Include women, who are often omitted. Record refusals and reasons.
- Setting: interview away from supervisors and the owner, at a time and place the worker chooses.
- Consent: explain in the worker's language that participation is voluntary, that refusal will not affect wages or work, and that no name will be written. Use verbal consent with a witness if literacy is low.
- Debt questions: do not record names, amounts or contractor details that could identify a worker. Report advances only as aggregates.
- No promises: do not offer compensation or imply that the survey will settle debts.
If a worker seems to be answering under pressure, stop and note it. A high response rate obtained through fear is not good data.
3. How should the student respond to the NGO's demand to 'prove' causation before the election and to the doctor's offer to sign a case list?
Read the worked answer
Both demands push the student beyond the evidence.
On 'prove':
- Tell the coordinator that the design cannot prove causation, and that a report claiming proof would be easily attacked and could harm the workers' case.
- Offer a preliminary descriptive report before the deadline, with prevalence estimates, associations, methods and limitations.
- Separate findings from advocacy: the NGO can campaign on the findings, but the report must not overstate them.
On the doctor's offer:
- Decline. A list signed without examination is not a diagnosis and could mislead workers about their health.
- Propose instead a referral pathway: workers with symptoms are told where to go, with their consent, and the project records only that a referral was made.
On timing: releasing a report just before an election can politicise it. Publish with full methods and no personal data, and let the NGO handle its own advocacy separately.