Lesson 6: Investigating an Outbreak

Exercise Answers

Exercise 6.1

Nine cases of cancer in a community represents a cluster — a group of cases in a given area over a particular period of time that seems to be unusual, although we do not actually know the size of the community, the background rate of cancer, and the number of cases that might be expected. Nonetheless, either the health department or the community or both is concerned enough to raise the issue. Under these circumstances, an investigation may be justified for several reasons.

1. Because the number of expected cases is not known (or at least not stated), one reason to investigate is to determine how many cases to expect in the community. In a large community, nine cases of a common cancer (for example, lung, breast, or colon cancer) would not be unusual. If the particular cancer is a rare type, nine cases even in a large community may be unusual. And in a very small community, nine cases of even a common cancer may be unusual.
2. If the number of cancer cases turns out to be high for that community, public health officials might choose to investigate further. They may have a research agenda — perhaps they can identify a new risk factor (workers exposed to a particular chemical) or predisposition (persons with a particular genetic trait) for the cancer.
3. Control and prevention may be the justification for additional investigation. If modifiable risk factors are known or identified, control and prevention measures can be developed. Alternatively, if the cancer is one that can treated successfully if found early, and a screening test is available, then investigation might focus on why these persons died from a treatable disease. If, for example, the nine cases were cancers of the cervix (detectable by Pap smear and generally nonfatal if identified and treated early), a study might identify: a) lack of access to healthcare; b) physicians not following the recommendations to screen women at appropriate intervals; and/or c) laboratory error in reading or reporting the test results. Measures to correct these problems, such as public screening clinics, physician education, and laboratory quality assurance, could then be developed.
4. If new staff need to gain experience in conducting cluster investigations, training might be a justification for investigating these cases. More commonly, cancer clusters generate public concern, which, in turn, often results in political pressure. Perhaps one of the affected persons is a member of the mayor’s family. A health department needs to be responsive to such concerns, and should investigate enough to address the concerns with facts. Finally, legal concerns may prompt an investigation, especially if a particular site (manufacturing plant, houses built on an old dump site, etc. ) is accused of causing the cancers.

Exercise 6.2

First, you should check the dates of onset rather than dates of report. The 12 reports could represent 12 recent cases, but could represent 12 cases scattered in time that were sent in as a batch.

However, assuming that all 12 reports of tuberculosis and the 12 of West Nile virus infection represent recent cases in a single county, both situations could be called clusters (several new cases seen in a particular area during a relatively brief period of time). Classifying the cases as an outbreak depends on whether the 12 cases exceed the usual number of cases reported in August in that county.

Tuberculosis does not have a striking seasonal distribution. The number of cases during August could be compared with: a) the numbers reported during the preceding several months; and b) the numbers reported during August of the preceding few years.

West Nile virus infection is a highly seasonal disease that peaks during August-September-October. As a result, the number of cases in August is expected to be higher than the numbers reported during the preceding several months. To determine whether the number of cases reported in August is greater than expected, the number must be compared with the numbers reported during August of the preceding few years.

Exercise 6.3

This illustrates that a case definition is a method for deciding whether to classify someone as having the disease of interest or not, not whether they actually do or do not have the disease. Patients 1 and 4 may have mild cases, and Patient 7 may have leukemia and eosinophilia-myalgia syndrome, but are classified as non-cases under the revised definition.

Exercise 6.4

A case definition is a set of standard criteria for determining whether an individual should be categorized as having a particular disease or health-related condition. For an outbreak, a case definition consists of clinical criteria and specification of time, place, and person. A case definition can have degrees of certainty, e.g., suspect case (usually based on clinical and sometimes epidemiologic criteria) versus confirmed case (based on laboratory confirmation).

The outbreak appeared to be limited to students (no adults reported illness), but included both tour groups. Some students had severe abdominal pain and diarrhea and stool cultures positive forE. coli O157. Clearly these should be counted as case-patients. Some students had the same symptoms but negative cultures. Should they be counted as case-patients? Still others had the same symptoms but no stool testing. Should they be counted as case-patients? Finally, two students had single bouts of diarrhea, but no abdominal pain and negative cultures.

No one case definition is the absolutely correct case definition. One investigator could decide to include those with symptoms but without testing as suspect or probable cases, while another investigator could exclude them. Similarly, one investigator might put a great deal of faith in the stool culture and exclude those who tested negative, regardless of the presence of compatible symptoms, while another investigator might allow that some stool cultures could be “false negatives” (test negative even though the person actually has the infection) and include them in a suspect or probable or possible category. The two students with single bouts of diarrhea but no abdominal pain and negative cultures seem least likely to have true cases of E. coli infection.

Similarly, the beginning time limit could be set on December 2, the date that Tour A departed, or could be set later, to account for the minimum incubation period.

So, one case definition might be:

PERSON: Any tenth-grade student who went on either tour
PLACE: Limited to students at city high schools
TIME: Onset since December 2? 3? 4?
CLINICAL: Confirmed stool sample positive forE. coli O157:H7, regardless of symptoms
SUSPECT: Self-reported severe abdominal pain and diarrhea >2 episodes/day, with stool culture not done; or self-reported abdominal pain and diarrhea >2 episodes/day and stool culture negative

Exercise 6.5

Exercise 6.6

1. 
   Image Description
2. The date of onset of the earliest case was June 28. Subtracting the minimum incubation period (7 days) from June 28 points to June 21. The median and modal date of onset was June 29. Subtracting the average (say, 12 days) from June 29 points to June 17. So the most likely exposure period was sometime around June 17 through June 21, give or take a day or two on either side. Indeed, the investigators determined that exposure most likely occurred on June 19, when all ill persons either actively participated in or were nearby the sifting of dirt that probably harbored the organism.

Exercise 6.7

Cell phones are quite popular. Noting that most if not all of the 17 patients had used cell phones does not indicate that cell phones are the cause of brain cancer. An epidemiologic study that compares the exposure experience of the case-patients with the exposure experience of persons without brain cancer is necessary. A case-control study is the design of choice, since 17 persons with the disease of interest have already been identified.

As many as possible of the 17 persons with brain cancer should be enrolled in the case-control study as the case group. A group of persons without brain cancer need to be identified and enrolled as the control group. Whom would you enroll as controls? Remember that controls are supposed to represent the general exposure experience in the population from which the case-patients came. Controls could come from the same community (randomly selected telephone numbers, neighbors, friends) or the same healthcare providers (e.g., patients treated by the same neurologist but who do not have brain cancer). Once case-patients and controls are identified and enrolled, each would be questioned about exposure to cell phones. Finally, the exposure experience of case-patients and controls would be compared to determine whether case-patients were more likely to use cell phones, or use particular types of phones, or used them more frequently, or for longer cumulative time, etc.

The alternative to a case-control study is a cohort study. For a cohort study you would have to enroll a group of cell phone users (“exposed group”) and a group of persons who do not use cell phones (“unexposed group”). You would then have to determine how many in each group develop brain cancer. Since brain cancer is a relatively rare event, you would need rather large groups in order to have enough brain cancer cases for the study to be useful. Therefore, a cohort study is less practical than a case-control study in this setting.

Exercise 6.8

1. The appropriate measure of association for a case-control study is the odds ratio.
2. The odds ratio is calculated as the cross-product ratio: ad ⁄ bc.
   Odds ratio = 15 × 23 ⁄ 8 × 7 = 6.16 = 6.2
3. With a chi-square of 9.41 and a 95% confidence interval of 1.6–25.1, this study shows a very strong (odds ratio = 6.2) association between histoplasmosis and working in Building X. This finding is quite statistically significant (chi-square = 9.41 corresponds to a p-value between 0.01 and 0.001). And although the 95% confidence interval indicates that the study is compatible with a seemingly relatively wide range of values (1.6–25.1), most of these values indicate a strong if not stronger association than the one observed.

Exercise 6.9

The first step in answering this question is to compare the attack rates (% ill) among those who ate the meal and those who did not eat the meal. Since the % ill is a measure of risk of illness, you could calculate a risk ratio for each meal.

Clearly, the September 18 lunch has the highest risk ratio. It has a relatively high attack rate (though not the highest) among those who ate the meal, and the lowest attack rate among those who did not eat the meal. Furthermore, almost all of the cases (50 out of 54) could be “accounted for” by that lunch.

In contrast, although the September 18 breakfast has a high attack rate among those who ate that meal, it has a relatively high attack rate among those who did not eat that breakfast, and most importantly, it can only account for one-sixth (9 out of 54) of the cases. Perhaps the September 18 breakfast was a minor contributor, but most of the illness probably resulted from exposure that occurred at the September 18 lunch.