Cluster Investigations of Non-Infectious Health

Events

Goals

Describe cluster investigations of non-infectious health events

Discuss key factors which should be considered before carrying out a cluster investigation

Outline the basic steps of a cluster investigation

Cluster investigations of non-infectious diseases

Critical public health function May link specific exposures to

diseases Example: Limb deformities in infants

related to maternal use of thalidomide in Europe in 1960s (1)

Led to U.S. legislation requiring rigorous testing process for approval of new pharmaceutical products

Notable cluster investigations

Non-infectious disease cluster investigations may be difficult Hard to confirm apparent geographic or

temporal excess in case numbers Supposed clusters may represent normal

disease patterns Confounding factors such as age Different pathogenic processes may result

in diseases that look alike but are not linked Example—primary brain cancer vs. brain

metastases spread from cancer in another organ

Non-infectious disease cluster investigations may be difficult Often impossible to establish a

definitive cause-and-effect relationship Small case numbers Problems isolating single potential exposure Difficulty in reconstructing exposure

histories (2)

Large-scale epidemiologic studies may be required Difficult to carry out

Ensuring a successful investigation Standardized step-wise process for

receiving/evaluating cluster reports Centralized tracking system, data collection

tools, clear lines of communication Well-trained staff and adequate resources

Experienced investigators, access to laboratories

Perceived problems must be addressed responsibly and sympathetically Effective, credible communication with public

and other agencies

To investigate or not? Investigating a link between exposure and

disease may be impossible but it is important to respond to threats perceived by the public

Keep in mind: Value of using a step-wise process with clear

decision points Share policy of using step-wise process with medical

community, general public, media Deliberate and transparent approach when carrying

out any investigation Recognize local concern but stay within stated

investigation process Develop effective methods of communication

Basic steps in investigating non-infectious disease clusters

Each step in a cluster investigation requires: Collecting and

analyzing data Decision to take

immediate action (if needed)

Decision to proceed to next step or not (3)

Figure 1. Flowchart of cluster investigation

Step 1: Initial ascertainment of cluster

Begin by collecting data: Identifying information from person

reporting the cluster Demographic information for cluster

cases Clinical information on cluster cases Identifying information for cluster

cases

Step 1—continued Enter information collected into a

tracking system Example: EpiInfo, Microsoft Access or

Excel Notify health department staff, local

health officers and appropriate agencies

Begin seeking information on disease causes and compare this information with the reported cluster

First decision point Based on initial information decide

whether to continue the investigation Criteria for continuing include:

Clinically similar health events without a plausible alternative etiology

Apparent excess occurrence of such health events

Plausible temporal association with the possible exposure(s)

A disease present in an demographic group where it is not usually found

One or more cases of a very rare disease

If investigation ends

Create a brief summary report and share with person reporting cluster and health department supervisors

If investigation is halted, explain why to person reporting. Example: variety in diagnoses (e.g.,

different types of cancers) argues against a common origin

Step 2: Assessment of excess occurrence Estimating excess occurrence

Confirm whether the number of cluster cases is greater than expected

Estimate an occurrence rateNumber of people with the health event Total population at risk

Population at risk = all people in the geographic area where the exposure occurred over a designated time period

To estimate an occurrence rate Select an appropriate geographic area

and time period Geographic area should include all persons

at risk for the health event but not large enough to include those not at risk

Designated time period should be consistent with time period during which supposed exposure took place

Defining the geographic area and time period too narrowly or too broadly may over- or under-estimate problem

How size of geographic area affects occurrence rate

Figure 2. Finding the occurrence rate in the population at risk

Occurrence rate of 20% (left) vs. 8% (right)

Determining cases and finding a reference population Determine which cases from the reported

cluster to include in a preliminary analysis Find a reference population comparable to

the population in which the cluster appeared Example—residents from a similar geographic

area Estimate an expected occurrence rate for

the reference population from existing surveillance data

Compare occurrence rates Compare observed occurrence rate

based on the cluster with the expected rate from the reference population

Use appropriate statistical tests to compare rates 5 or more cases and appropriate

denominator—Chi-square tests or Poisson regression

Small case numbers—group cases across geographic areas or time periods

Case Verification Case definition should include clinical criteria and

restrictions on time, place, and person Sensitive case definition

Broad criteria, may include several related diseases or health events, captures more true cases but includes false positives

Specific case definition Narrow criteria, focuses on one health event, uses

confirmatory testing, excludes true cases (false negatives)

Example—cluster of cancer cases linked to benzene exposure

Sensitive case definition = diagnosis of any form of blood cancer

Specific case definition = diagnosis of leukemia

Using multiple case definitions Example – investigation of childhood

cancer cases in Dover Township and Toms River, NJ, 1995 (4)

Industrial pollutants released into Toms River contaminated Dover’s municipal well

Investigation of all childhood cancers and subgroups of selected cancers

Childhood cancer cases in New Jersey, 1995 (4)

Observed and expected occurrence rates compared by calculating standardized incidence ratios and 95% confidence intervals

SIR = observed cases (or rate) expected cases (or rate) where = 1 no excess occurrence

> 1 possible excess occurrence< 1 observed is less than

expected

Childhood cancer cases in New Jersey, 1995 Table 2. Childhood cancer incidence in Toms

River census tracts, 1979-1995, children 0-4 years

Case-Verification Examine case-patients’ medical records Refer to relevant health registries Obtain copies of relevant laboratory,

pathology, or other reports Obtain clinical/laboratory re-evaluations

(e.g. retest biopsy or other specimens) May need to do additional case-finding

Case-Finding In an expanded assessment:

Reconsider initial case definition Reassess geographic/time boundaries Ascertain all potential cases within

geographic and time boundaries Identify appropriate database sources Perform literature review Assess likelihood that clustered events are

related to supposed exposure(s)

Case-Finding Review additional data sources or medical

records Formal surveys of the community reserved for

later stages in the investigation If excess occurrence of disease confirmed

with evidence of association with supposed exposure, consider etiologic study

If excess occurrence not confirmed or confirmed with no plausible relationship to supposed exposure, conclude investigation

Step 3: Determining the feasibility of an etiologic study

First, determine epidemiologic and logistical feasibility of an etiologic study

Construct a testable hypothesis Clearly state hypothesis Include the target population, health

event(s) and exposure(s) of interest

Determining feasibility Pros and cons of different study

designs Potential challenges and ways to

address them Potential for finding additional cases,

expanding the case definition and changing the time/geographic periods

Collecting additional data and associated costs

Etiologic study—measuring exposure

Do clinical or environmental tests for the exposure exist? How sensitive are they? Given the lapse of time since

exposure will the test be useful? Is the reported exposure history a

good predictor of true exposure?

Determining study benefits May be difficult to determine whether

an etiologic study will justify the effort Etiologic studies may not be successful

unless disease is rare or frequency has suddenly increased

Etiologic agent must be measurable and leave a physiologic response

Appropriate unexposed control group is needed—levels of exposure must vary within population to carry out study (6)

Assess study implications Consider epidemiologic and policy

implications Consider community reactions If etiologic study is feasible and likely

benefits justify the effort, carry out study If etiologic study is logistically impossible,

too expensive or will not affect policies or programs, end investigation

Step 4: Conducting an etiologic investigation Etiologic study should generate

knowledge about broader epidemiologic and public health issues raised

Begin by writing a formal study protocol Lay out steps in data collection,

processing, quality assurance and data analysis

Further study design decisions will be unique to the particular study

Conclusion Cluster investigations allow public

health officials to interact with the community and be responsive to public needs

May provide information about previously unsuspected exposure-disease relationships

Can be an unproductive drain on public health resources

References1. Lenz W. Kindliche mißbildungen nach medikament-einnahme

während der gravidat [Malformations in children after a drug taken during pregnancy]. Dtsch Med Wochenschr. 1961;86:2555–2556.

2. Cartwright RA. Cluster investigations: Are they worth it? Med J Aust. 1999;171(4):172. http://www.mja.com.au/public/issues/ 171_4_160899/cartwright/cartwright.html. Accessed August 13, 2008.

3. CDC. Guidelines for investigating clusters of health events. MMWR Morb Mortal Wkly Rep. 1990;39(RR-11):1-16. http://www.cdc.gov/mmwr/preview/mmwrhtml/00001797.htm. Accessed August 13, 2008.

4. New Jersey Department of Health and Senior Services and ATSDR. Childhood Cancer Incidence Health Consultation: A Review and Analysis of Cancer Registry Data, 1979-1995, for Dover Township (Ocean County), New Jersey. 1997. http://www.state.nj.us/health/eoh/hhazweb/cansumm.pdf. Accessed August 13, 2008.

References5. Bender AP, Williams AN, Johnson RA, Jagger HG.

Appropriate public health responses to clusters: The art of being responsibly responsive. Am J Epidemiol. 1990;132:S48-S52.

6. Rothman KJ. A sobering start for the cluster busters’ conference. Am J Epidemiol 1990;132:S6-S13.

7. Fischoff B, Lichtenstein S, Slovic P, et al. Acceptable Risk. Cambridge, UK: Cambridge Univ Press; 1981.

8. Greenberg MR, Wartenberg D. Understanding mass media coverage of disease clusters. Am J Epidemiol. 1990;132:S192-5.

9. Covello VT, Allen F. Seven Cardinal Rules of Risk Communication. Washington, DC: US Environmental Protection Agency, Office of Policy Analysis; 1988. OPA publication 87-020.