PHLT 8522 Week 7 Analytic Comparison Example

Three Candidate Sources, One Identical Test: A Case-Control Comparison of Time Downwind of a Riverfront Demolition, a Park Roost Trimming and an Athletic Field Regrading Among Composite Histoplasmosis Cases

Student Name

Walden University

PHLT 8522: Epidemiological Investigations

Instructor Name

Month Day, Year

Three Candidate Sources, One Identical Test: A Case-Control Comparison of Time Downwind of a Riverfront Demolition, a Park Roost Trimming and an Athletic Field Regrading Among Composite Histoplasmosis Cases

Question and Design

Last week's memo named the Front Street warehouse demolition as the working hypothesis and gave two rival sources, tree trimming at a starling roost in Riverside Park and regrading of an athletic field to the north, their own predictions. This paper reports a case-control comparison designed to test all three with the same instrument, the analytic step that Goodman et al. (1990) place after description in a field investigation. Cases were confirmed or probable community cases under the outbreak definition. The three cases employed by the demolition company are analyzed separately, because their exposure was occupational and would dominate any comparison. All figures are composite.

Cases and Controls

Of 15 community cases, 14 completed the interview; one declined, and her answers to the brief initial form, which placed her home in the northeast quarter, were not used in the analysis. Controls were drawn from the same part of the city as the cases, using address-based sampling within the three census tracts where cases lived or worked, and screened to exclude anyone with pneumonia or fungal testing since September. Three controls were enrolled per case, frequency matched on ten-year age group, for 42 controls in all. Wacholder et al. (1992) emphasize that controls should represent the same source population as the cases, so that a control who fell ill would have been found and counted; address-based sampling in the same tracts approximates that base. Controls answered the same questionnaire, by the same interviewers, in the same week as the cases they were matched to.

Exposures, Defined Before Interviews

Each exposure was defined before the first interview, so that no definition could be adjusted to fit the answers.

The interview used a calendar of September with local events and weather marked on it, so that respondents could anchor their recall, and a map on which they traced their usual routes.

Table 1

Exposure Definitions for the Three Candidate Sources

Note. Downwind zone fixed on October 14, before interviews began on October 16.

Results

Time downwind of the warehouse was strongly associated with illness: ten of 14 cases and 12 of 42 controls spent two or more hours outdoors in the zone during the demolition period, an odds ratio of 6.25. Neither rival showed an association. Park visits and field visits were about as common among controls as among cases.

Table 2

Exposure Among Cases and Controls, by Candidate Source (Composite Figures)

Note. Composite figures; Woolf confidence intervals.

A Dose Gradient

Odds rise with hours spent in the zone, from 3.33 for under five hours to 15.0 for five hours or more, although cells this small make the upper estimate unstable. A gradient does not prove causation on its own, but it is harder to explain by a simple recall difference than a single yes-or-no comparison, since recall bias would need to grow with the reported amount of time.

Table 3

Hours Outdoors in the Downwind Zone, September 8-19 (Composite Figures)

Note. Composite figures; the zero-hours row includes respondents with under two hours.

The Occupational Cases

All three company employees with histoplasmosis worked on the Front Street demolition crew during September 8 to 19. None worked only at the company's other sites. Work records, provided by the employer with the employees' consent, show they operated excavators and hand-cleared debris in the upper floors before the structure was brought down, without respirators. That pattern matches the third falsifying test set out in the hypothesis memo, which the warehouse passed.

Limits of the Comparison

The study is small, and the confidence interval for the warehouse runs from 1.6 to nearly 24, so the size of the association is uncertain even if its presence is not. Recall is imperfect at four to six weeks, and cases, having been ill, may have reconstructed their September more carefully than controls; the calendar and route map were used to reduce that difference. Interviewers also asked about the park and the field with the same care as the warehouse zone, and the null results for those two sources suggest that cases were not simply reporting more outdoor time of every kind.

Living or working in the zone could stand in for other features of the northeast quarter, though no other candidate source lies there. Matching by age group controls only for age, and controls were not matched on outdoor habits, so cases who simply spend more time outside could appear more exposed to every source; the flat results for the park and field argue against that explanation here. Rothman and Greenland (2005) caution that any single study is one piece of evidence, and the weight of this result rests on its agreement with the descriptive pattern, the timing of onsets and the occupational cases.

Conclusion

Of three candidate sources tested on identical terms, only the warehouse demolition shows an association with illness, with a dose gradient and support from the occupational cases. The rival sources show none. Laboratory evidence from the site is still pending, and the decision brief that follows considers what to do before it arrives.

References

Goodman, R. A., Buehler, J. W., & Koplan, J. P. (1990). The epidemiologic field investigation: Science and judgment in public health practice. American Journal of Epidemiology, 132(1), 9-16. https://doi.org/10.1093/oxfordjournals.aje.a115647

Rothman, K. J., & Greenland, S. (2005). Causation and causal inference in epidemiology. American Journal of Public Health, 95(S1), S144-S150. https://doi.org/10.2105/AJPH.2004.059204

Wacholder, S., McLaughlin, J. K., Silverman, D. T., & Mandel, J. S. (1992). Selection of controls in case-control studies: I. Principles. American Journal of Epidemiology, 135(9), 1019-1028. https://doi.org/10.1093/oxfordjournals.aje.a116396