Applied Epidemiology for the Infectious Diseases Physician

The word epidemiology originates from the Greek epi- (meaning “surrounding”), demos (meaning “people”), and logos (meaning “the study of”). Thus epidemiology literally refers to the study of population occurrences. We now define epidemiology as the study of the distribution and determinants of health-related states or events in specified populations, and the application of this study to the control of health problems.¹ Historically, the discipline of epidemiology reaches back in time over 2500 years and includes contributions from thinkers such as Hippocrates, John Graunt, James Lind, and William Farr. However, John Snow, an English anesthesiologist, is widely considered the “father of field epidemiology.” His famous study identified a single water pump on Broad Street as the cause of a cholera epidemic in the Golden Square of London. In 1854, twenty years before the invention of the microscope, Snow mapped the distribution of households with cholera cases during the epidemic and recognized that the cases were clustered around a particular water pump, the Broad Street pump. To confirm his suspicion that contaminated water from this pump was causing the cholera epidemic, Snow removed the handle of the pump, and the outbreak ended.²

The field of epidemiology has expanded from the investigation of infectious disease outbreaks, to the study of injuries and violence, and to the examination of any exposure that influences health outcomes, including molecular and genetic factors.² Epidemiology is used in many ways, such as for evaluating a population's health, informing individual decisions, and establishing causality.² Using rigorous statistical methods, epidemiology can answer complex questions, such as whether risk patterns correlate with exogenous factors or whether a public health program modifies risk.² For infectious diseases clinicians, the epidemiologic factors associated with a case or cluster of cases can be used to great benefit, potentially decreasing the time to diagnosis, limiting the need for extensive testing, and tailoring the use of broad-spectrum antibiotics. Epidemiologic data may also be used to inform policies or guidelines that affect groups or individuals, such as choosing which populations to vaccinate, setting speed limits on a highway, or determining how many times a week to exercise.²

What Can Epidemiology Tell the Infectious Disease Clinician?

Epidemiology asks the clinician to think of the patient's community and exposures, as well as the individual patient. Is my patient's respiratory infection part of an outbreak of influenza in the community and, if so, what can be done to identify and control it? Interaction with the local health department or hospital infection control staff in this instance might have a broader impact than treating a single patient but will also require understanding how epidemiologic questions are formulated and answered.

The Epidemiologic Triad Model: Agent, Host, and Environment

To better inform clinical and public health decisions, epidemiologists (and clinicians) try to find causal links between exposures and health outcomes. If this is an influenza outbreak, how is it being spread, and how severe are the illnesses? The epidemiologic triad, or triangle, is one of the simplest models used to describe disease causation.² In this model disease arises from the interaction between three components: an external “agent,” a susceptible “host,” and an “environment” that supports the transmission of the agent from its source to the host (Fig. 13.1).² By tradition the agent represents an infectious microbe, such as a virus, bacterium, or parasite. However, this component of the model does not sufficiently describe all diseases; some can be chronic processes with multiple contributing causal factors. Host susceptibility can depend on an individual's genetic makeup, immune status, and host defenses, such as skin integrity, gastric acidity, ciliary function, and other factors. Host immunity, acquired immunodeficiency, and principles of immunization are discussed in Chapters 4 to 6, 118, and 316. Environmental factors that promote agent transmission to a susceptible host include the presence of animals and other reservoirs of infectious agents, vector populations, crowding, poor sanitation, climate, and geography.² A detailed understanding of the epidemiology, pathogenesis, and ecology of a disease is required to construct optimal epidemiologic triads (agent, host, and environment).

The Sufficient-Component Cause Model of Epidemiology

A second causation model that considers the multiple factors that contribute to disease is called the “sufficient-component cause model.”³ In this model each disease pathway has “component causes” or factors that contribute to disease. Different combinations of component causes are considered “sufficient causes” of disease when they complete the causal pathway. A component cause that is present in every combination of factors in the causal pathway is considered a “necessary cause” because the pathway cannot be complete without it. For example, human papilloma virus (HPV) infection is a necessary cause of cervical cancer because persons must be infected with this virus to develop HPV-associated cervical cancer. However, HPV infection is not a sufficient cause of cervical cancer because not all individuals infected with HPV develop cervical cancer.

Other Methods to Characterize Infection Transmission

Multiple classifications exist to describe modes of disease transmission or the ways in which an agent leaves its natural environmental reservoir to infect a susceptible host. One classification distinguishes “direct” from “indirect” transmission.² Direct transmission includes direct contact and droplet spread (short-range aerosols produced by sneezing, coughing, or talking), whereas indirect transmission includes airborne, vehicle-borne, and vector-borne (mechanical or biologic) spread.²

Descriptive Epidemiology

Descriptive epidemiology is used to characterize events in terms of person, place, and time and might answer the following questions: What health risks exist for this population? Who is getting influenza this winter? Who are the population members at risk? Is the influenza strain circulating in nursing homes, hospitals, or schools? Which geographic locations are associated with increased risk? How has the level of risk changed over time? Case reports and case series are types of descriptive studies conducted in clinical settings.⁴ Case reports describe the clinical features of a disease, along with demographic and historical details relevant to its presentation. Case series include the presentation, clinical features, and other significant facts pertaining to multiple patients with similar health conditions. Case reports (and series) often represent the first or novel presentations of a condition and can be very valuable from both a clinical and an epidemiologic perspective. For example, ribavirin was recently shown to be effective in a series of patients with chronic hepatitis E, not an original indication.⁵ Descriptive epidemiology may also assist in recognition of an emerging disease or in generating hypotheses about causal factors and factors that increase the risk of disease.²

Analytic Epidemiology

When one would like to more definitely establish a relationship between an exposure and a disease, analytic epidemiology is required. Hypotheses about the relationship are normally formulated, and a study design is selected based on the available data, the research question, ethical considerations, feasibility, and issues of validity and efficiency. Analytic epidemiologic studies fall into two categories: experimental and observational² (Table 13.1). An observational study has no intervention but attempts to link the exposure to the disease passively. In an experimental study the exposure is actively managed and the outcome measured prospectively. Precision, validity, and bias should all be considered when designing an epidemiologic study but are covered elsewhere.²

Experimental Studies

In an experimental study an investigator assigns similar groups of individuals (clinical trial) or communities (community trial) to different exposures or interventions and then follows the groups over time to ascertain the effects of the exposure.² A classic example of a clinical trial is a vaccine trial, where one group of individuals is assigned to receive an immunization to prevent a particular disease and another group of individuals to receive a placebo. There are numerous examples of vaccine trials, including the definitive demonstration of the effectiveness of influenza immunization in children.⁶ After a defined period of time, the investigator monitors each participant to diagnose any occurrence of the disease in question, the investigator compares the rates of infection between the vaccine and placebo groups to determine if the vaccine successfully protected recipients from the disease. In a community trial, groups of individuals undergo randomization to receive an intervention, and investigators ascertain community health outcomes after a period of time relevant to the intervention and health outcome under investigation. Community trials are ideal for evaluating broad-based interventions, such as case management algorithms and health-related policies. An example of a community trial is testing the utility of long-lasting insecticide-treated bed net distribution on malaria incidence in endemic areas.⁷

The most common types of observational study designs are cohort and case-control studies. Cohort studies are conceptually similar to experimental studies. The investigator evaluates the exposure status of participants and then follows them over time to evaluate the incidence of one or more health outcomes in each exposure group. If a health outcome is significantly greater in the exposed group relative to the unexposed group, then the exposure is considered to be associated with the disease. Cohort studies differ from experimental studies in that the investigator does not assign individuals to an exposure group; rather, the exposure group is determined passively. Cohort studies can be performed prospectively or retrospectively. In a prospective study the exposure status is established in advance, and the participant is followed over time to determine the occurrence of the outcome of interest. In a retrospective study both the exposures and the outcomes have already occurred at different points in time. The investigator collects the retrospective data and compares the rates of the outcome in the exposed and unexposed groups. A recent example of a well-designed cohort study described the frequency and distribution of pregnancy outcomes after Zika virus infection.⁸

Determining the Appropriate Epidemiologic Study Design

When determining which study design is most appropriate, it is important to consider the strengths and weaknesses of each (see Table 13.1). Experimental studies are ideal for evaluating an intervention for a particular outcome with a high degree of validity, such as when an intervention is expected to have a small effect or when a highly controlled setting with randomized participants is needed to minimize bias and confounding.¹² Only randomization, preferably blinded, can reduce the effect of unidentified factors or confounders that can potentially bias the outcome. The disadvantages of experimental studies include high cost and ethical considerations related to exposing participants to certain risks. Observational studies can be used to evaluate a wider range of exposures, including preventions, treatments, and causes of disease. The primary limitation of observational studies is the inability to control for confounding factors. Cohort studies are useful for investigating rare exposures, multiple outcomes, and outcomes for which little is known about exposure. Compared with other study designs, case-control studies generally take less time, require fewer resources, and can provide information about multiple exposures when little is known about the etiology of a particular health outcome.¹² In addition, case-control studies are more efficient than cohort studies for studying rare diseases and diseases with long incubation or latency times because they do not require long-term prospective follow-up.¹² Case-control studies are also more efficient when the population of interest is dynamic or when the study requires expensive or hard-to-obtain exposure data.¹² Although caution must be exercised with temporal inference, cross-sectional studies are useful because they can be performed in a short time period and yield highly generalizable results when based on large samples from the general population. Likewise, ecologic studies are rapid, inexpensive, and simple, and they can provide useful information about the context in which populations live.¹³

Basic Biostatistics

A basic understanding of biostatistics is required to interpret medical literature or to plan appropriate epidemiologic studies. This section will cover measures of disease frequency, measures of association, and significance of association only. For more extensive reading on biostatistics, please refer to the classic references 2, 12, and 14. For the use of biostatistics in clinical trials see Chapter 52.

Two-by-Two Tables

To understand how to calculate measures of association, it is important to understand the basic construct of a two-by-two table. In many epidemiologic studies exposure and outcome data are dichotomous (e.g., “yes” or “no”). Thus the relationship between exposure and outcome can be cross-tabulated in a two-by-two table, which has two categories of exposure and two categories of outcome (Fig. 13.2¹⁴). By convention, outcome status is placed in columns, and exposure status is placed in rows. The four cells of the two-by-two table are labeled a, b, c, and d, and each letter refers to the number of persons with the outcome indicated by the column heading and the exposure status indicated by the row heading.

Risk ratio (relative risk)=Riskexposed/Riskunexposed=(a/[a+b])/(c/[c+d])

Risk ratio (RR) describes the excess risk of a health outcome in exposed persons compared with unexposed persons, who represent the background or expected risk. Mathematically, the RR is expressed as the risk in the exposed group divided by the risk in the unexposed group. A risk ratio greater than 1.0 indicates that risk is greater in the exposed group when compared with the unexposed group, whereas RR less than 1.0 indicates that the risk in the exposed group is less than in the unexposed group. This happens when the exposure protects against a health outcome, such as an immunization that protects against an infectious disease.²

Odds ratio (cross-product ratio, relative odds)=ad/bc.

Odds ratios (ORs) are measures of association used in case-control studies to approximate RR. The OR is also called the “cross-product ratio” because it is obtained by calculating the cross products of the cells in a two-by-two table, with “a” × “d” in the numerator and “b” × “c” in the denominator. In case-control studies, epidemiologists enroll a group of “cases” with the outcome and a comparable group of “controls” without the outcome. The number of individuals in each group is determined a priori, and the size of the populations from which the cases and controls arise is often unknown. This lack of population denominator data prevents direct calculation of the RR. However, the OR approximates the RR when the outcome of interest is rare.

Disease Prevention

Prevention includes a wide array of interventions aimed at promoting health by reducing risks that lead to poor health outcomes or disease. Primary, secondary, and tertiary prevention are three terms that describe the points in the disease process targeted by an intervention.² Primary prevention aims to prevent the occurrence of disease. This is done by preventing exposure to hazards that cause poor health outcomes, that is, changing unhealthy or unsafe behaviors that lead to poor health outcomes and increasing resistance to poor health outcomes if exposure occurs. Examples of primary prevention include legislative measures that control health hazards, education about healthy behaviors, and immunization against diseases.² Secondary prevention aims to minimize the impact of a poor health outcome that has already occurred. This can be done by identifying and treating a health condition as soon as possible to prevent its progression, implementing strategies to prevent recurrence, and modifying activities to help persons return to their original state of health. Examples of secondary prevention include screening tests that identify diseases during their earliest stages, such as hepatitis C screening. Tertiary prevention aims to lessen the impact of a poor health outcome with lasting effects, such as chronic disease or injury leading to disability. This can be done by managing long-term ailments to lessen disability, improve quality of life, and lengthen life expectancy. Early antiretroviral therapy for human immunodeficiency virus (HIV) infection is an example of tertiary prevention.

Disease Control: Quarantine and Isolation

There are many strategies to mitigate the spread of infections, and those relevant to hospital infection control are given in Chapter 298. Some strategies targeting public health intervention, namely quarantine and isolation, and a short primer on outbreak investigation are discussed as follows.

Quarantine (separating and restricting movement of individuals exposed to a communicable disease) and isolation (separating and restricting movement of individuals suspected or confirmed to be infected with communicable disease) are public health strategies used to prevent exposure of the public to individuals or populations infected with a communicable disease. In an increasingly connected world, public health officials must interact with multiple entities and many levels of government to balance public health and humanitarian efforts with economic priorities, such as international travel and trade. In the 2014 West African Ebola epidemic, for example, the CDC partnered with WHO, the International Organization for Migration, nongovernmental organizations, and domestic entities, such as the US Department of Homeland Security and state and local health departments, to prevent international transmission of Ebola while minimizing disruption of economic activities.³² Domestically, in addition to communicating with travelers via travel notices and airport messaging, CDC teams provided training and equipped US entry-point personnel to conduct screening using body-heat sensing devices and questionnaires to identify symptomatic and potentially exposed travelers arriving in the United States. If a potentially exposed individual was identified through screening, the domestic quarantine strategy relied on measures such as active self-monitoring, controlled movement, and daily communication with public health officials (for high-risk individuals, such as health care workers) during the 21-day Ebola incubation period, with the aim of applying the least restrictive measures necessary to prevent potential transmission.³²

Outbreak Investigation Primer

Outbreaks of infectious disease can arise anywhere in the world and pose a threat to local and global populations without respect for political borders, geographic separation, or cultural differences. Many current conditions facilitate this potential for rapid spread of infectious diseases, such as the globalization of the food supply, overuse of antibiotics, the growth of megacities with severe crowding, our increasing proximity to animals and vectors of disease, and even climate change.³³

There have been numerous recent outbreaks that have seized the public's attention. The SARS epidemic of 2003 and the recent resurgence of measles in North America and Europe were watershed events and hastened the strengthening of the IHR. These were followed by multiple outbreaks of imported foodborne gastroenteritis caused by Cyclospora and Salmonella, as well as by MERS, Nipah virus, pandemic H1N1 influenza, and, most recently, Ebola virus disease and Zika virus syndrome.^33,34 The scope and impact of these outbreaks underscore the importance of maintaining an informed network of clinicians who understand the dynamics of an outbreak and how to investigate one. This section describes the key concepts involved in an outbreak investigation, discusses basic transmission dynamics, and provides a simple guide for conducting an outbreak investigation.

An outbreak is an increase beyond expectation in the number of cases of a disease or condition, occurring in a specific population in a defined geographic location and period of time.¹⁴ The cases are epidemiologically related, although this linkage is often not initially evident and may only be discovered after thorough investigation. For example, in the 1981 multistate outbreak of Salmonella muenchen, the epidemiologic link was discovered only when case patients reported smoking marijuana more frequently than control patients. This rare Salmonella species was then isolated from marijuana samples found in the homes of patients with the illness.³⁵

Outbreaks very often have an infectious origin, although some can be due to noninfectious agents, such as food intoxication or even hysteria. It is often difficult to define how many cases beyond expectation constitutes an outbreak, but even one case can indicate an outbreak if the disease has been eradicated (smallpox), eliminated (poliomyelitis in Europe), or is novel to humans (highly pathogenic H5N1 avian influenza in the Americas).

In the absence of timely control measures, outbreaks can spread and lead to epidemics or even pandemics. Epidemics are conceptually identical to outbreaks but are more widely disseminated in time and space, such as the cholera epidemic in London during the mid-1800s and SARS in Southeast Asia in 2003. Pandemics spread globally and may persist through months, years, or decades. Examples of pandemics include such historic scourges as bubonic plague in medieval days and influenza in 1918-19, as well as AIDS or the influenza A/H1N1 pandemic of 2009 more recently.

Many infectious diseases are endemic in certain settings and occur routinely. Malaria, dengue and enteropathogens that cause diarrheal disease in the tropics are all examples. Endemnicity does not preclude the occurrence of outbreaks, however, that may occur during point-source foodborne outbreaks when large pools of susceptible individuals are exposed at a single time. Vibrio parahemolyticus and Norovirus, for example, are endemic causes of sporadic gastroenteritis but can also lead to outbreaks when contaminated seafood is eaten raw.^36,37 Often, all that is required for an outbreak to occur is a sufficiently large, naïve population that is exposed to an infectious inoculum of the agent.

The methodology for conducting outbreak investigations and the study designs and methods used have been well described in the literature.^2,14,38–41 Reviews of published outbreaks suggest that foodborne transmission accounts for nearly half of all reported outbreaks. Most efforts to improve the standardized reporting of outbreaks have been focused on either foodborne disease or nosocomial infections.^42,43 The CDC collects and collates data on outbreaks in the United States and maintains a dashboard that includes visualizations and statistics through 2016. Outbreak data can be explored at the National Outbreak Reporting System, which is found at https://wwwn.cdc.gov/norsdashboard/.⁴⁴ A current list of ongoing outbreaks, both domestic and significant international, can be found at https://www.cdc.gov/outbreaks/index.html.

Incubation period is the time elapsed from exposure to a certain infectious agent until the development of symptoms and is often expressed as a range. In an outbreak investigation the incubation period is a key parameter that may help to discern between multiple potential etiologic causes. For example, symptoms of foodborne staphylococcal intoxication often appear 30 minutes to 6 hours after exposure, whereas the incubation period for salmonellosis is 6 to 72 hours. So, abrupt gastroenteric illness in multiple subjects who recently shared a common meal is less likely to be caused by an agent with a longer incubation period like Salmonella. However, it should not be mistakenly assumed that infectious agents with lengthy incubation periods cannot cause outbreaks because transmission of Mycobacterium tuberculosis during air travel has been demonstrated on several occasions.⁴⁵ Cases from outbreaks caused by agents with long incubation periods will typically appear over extended time periods, making their common origin less obvious. The routine application of genetic sequencing of pathogens may assist in defining the relatedness of cases that are separated by time and space.⁴⁶

Transmission Modes

Point source outbreaks can occur after the exposure of a group of people to an agent during a single, short period of time. Cases often present in a single group during a short time period, corresponding to the range of the incubation period, unless there is secondary transmission. The epidemic curve demonstrates an abrupt onset and gradual descent with a single peak. Food and beverages are common point sources, and single-source outbreaks very often present as gastroenteric illness (Fig. 13.3).⁴⁷

Person-to-person, or propagated, outbreaks are characterized by the presence of two or more clusters of cases in time separated by approximately one median of the incubation period, thus suggesting secondary transmission. The distance between clusters of cases usually becomes less clear as the outbreak progresses because the ranges of the incubation periods tend to blend into each other. Respiratory transmission is the most efficient mechanism of person-to-person transmission, the most classic example being the influenza virus. Measles, adenovirus, and pneumonic plague may also be transmitted in this manner. (Fig. 13.4)⁴⁸

In continuous-source outbreaks, exposure to the agent can occur over an extended period. Therefore cases appear over longer periods of time, often substantially beyond the range of the incubation period. Cases sometimes present in several clusters, but the timing between clusters does not necessarily correspond closely to the median incubation period. Vector-borne and zoonotic infections are typical of this transmission pattern, but waterborne, respiratory, and nosocomial outbreaks have occasionally been described as continuous-source outbreaks as well. (Fig. 13.5)⁴⁹

Chains of Transmission

Exploring the connection between initial waves of cases and those occurring later often reveals important interactions that can occur during propagated outbreaks with person-to-person transmission or a continuous common source. Clearly identified transmission waves can be observed, such as those related to cultural practices such as burial traditions in Africa⁴⁹ or the transmission of Nipah virus in a Bangladeshi community (Fig. 13.6).⁴⁸ Transmission chains can also provide hints about the incubation period of the agent, although these may be difficult to tease out of the data.

New molecular diagnostic tools are revolutionizing the study of outbreaks, especially in elucidating chains of transmission where the cases are separated by time and distance. Whole-genome sequencing of pathogens requires less time, money, and equipment than ever before in the United States and in many high-income settings.⁴⁶ Next-generation sequencing platforms are increasingly available in the field, as is the ability to interpret the massive volumes of data inherent to this technology. Genetic sequencing can be used to precisely determine transmission chains, relatedness of pathogens to known standards, and even characteristics such as antimicrobial resistance and virulence, and are extremely valuable tools for outbreak management.⁴⁶

Outbreak Investigation and Response

Conducting an investigation may help us to understand the mode of transmission of the disease, identify the etiologic agent and who may be at risk of infection, and ultimately prevent additional cases and reduce the overall morbidity and/or mortality rates. Conducting an outbreak investigation may also allow us to evaluate the sensitivity and specificity of a surveillance system, evaluate or implement intervention strategies (i.e., vaccination, social distancing, or removal of a point source) and contribute to the epidemiology and scientific knowledge of the disease.

A systematic, step-by-step approach to conducting an outbreak investigation is imperative for identifying the source of the outbreak and for controlling and preventing additional cases. This systematic approach can be divided into 13 distinct steps (Table 13.2), but these are not rigid in their order, and several steps are often accomplished simultaneously.^2,14

1. 1. Preparation for an outbreak should involve prestaging of standardized sample collection materials, personal protective equipment, and most important, the presence of trained personnel. It is also important to think about permissions for conducting the study, whether it is the Institutional Review Board (IRB) of the hospital or CDC, or jurisdictional approval if you are traveling to another state or country.
2. 2. Conducting an outbreak investigation can be resource intensive. Before initiating the investigation, it is imperative to first determine whether or not an outbreak (or pandemic) is actually occurring. Several data sources may be available to help determine if the number of observed cases exceeds that of the expected baseline number (i.e., notifiable disease registries, death registries, hospital discharge summaries, etc.). It is important to recognize that new or improved diagnostic tests, a new or enhanced surveillance system, or simply increased awareness of a disease may artificially indicate that an outbreak is occurring.
3. 3. Verifying the diagnosis often goes hand-in-hand with confirming the existence of an outbreak. It may be necessary to collect additional biologic samples and, if possible, request specialized diagnostic procedures or have the results confirmed at a secondary reference laboratory. However, in the event of an outbreak of a new pathogen, it may be necessary to rely on clinical diagnosis alone, as was the case in the early stages of the SARS epidemic. Furthermore, it is highly recommended to interview patients to gather additional clinical and epidemiologic features.
4. 4. Enumerating the number of cases during an outbreak is only possible once a standard case definition has been established. This can be one of the most difficult and contentious components of an outbreak investigation. The case definition is generally based on clinical features, such as sudden onset of fever higher than 38°C, cough or sore throat, difficulty breathing, and so forth. The case definition is almost always restricted by person (i.e., children younger than years, no history of yellow fever vaccine), place (i.e., patients in a specific wing of a hospital, attendees of a county fair), and time (i.e., persons with illness onset within the previous 24 hours). The initial case definition is often quite broad, to capture all possible cases; however, as the investigation proceeds it generally becomes more refined and divided into subcategories, such as suspect (i.e., fever only), probable (i.e., fever with cough and shortness of breath, death with history of fever and cough), and confirmed (i.e., fever and cough with laboratory confirmation of influenza virus H5N1 infection).
5. 5. Counting and tracking cases is most easily achieved by constructing a line listing of cases. Data in the line listing should include information such as symptom onset date, patient identification data, clinical data, demographic data, laboratory data, and some risk factor or epidemiologic data. (Fig. 13.7)

   
   

6. 6. The next step, and often most revealing, is descriptive analysis of the epidemiologic data. The line listing is converted to visual depiction of the data, referred to as the epidemic curve (as described earlier), which is graphed with time on the x-axis (typically symptom onset data) and number of cases on the y-axis (Figs. 13.3 to 13.5). The epidemic curve can give clues about trend (i.e., person-to-person transmission or point/common source exposure), size of outbreak, and incubation period. The data can also be displayed in map form, revealing potential information such as common source exposure or clustering of cases. Global Positioning System (GPS)/Global Information Systems (GIS) are tools now commonly used in modern epidemiologic outbreak investigations.
7. 7. The next step, albeit a process that likely begins from notification of the first case, is to begin to develop hypotheses about the cause of illness and/or source of infection. In general, conducting the initial descriptive epidemiologic analysis of the data will give us clues to better refine our hypotheses. Furthermore, it is important that the investigator speaks with an adequate representative sample of the initial patients to understand potential links between cases. It is important to remember that the working hypotheses must be testable with statistical methods (i.e., chi-square analysis, logistic regression, etc.).
8. 8. After careful consideration and development of our initial hypotheses, we must evaluate these hypotheses with analytic methods. Two methodologies are typically used for outbreak investigations: cohort studies or case-control studies. These are described earlier under the “Analytic Epidemiology” section.
9. 9. Once the hypotheses have been tested using conventional analytic epidemiologic methods, it may be necessary to refine the hypotheses by conducting additional studies (i.e., laboratory or environmental testing) to further support the conclusions. For example, identifying Cyclospora cayetanensis in the raspberry filling of a wedding cake would greatly increase your epidemiologic evidence that eating cake at the wedding was significantly associated with developing diarrheal illness.⁶⁰ If no conclusive results were initially revealed, a new hypothesis may be proposed.
10. 10. The CDC recommends an explicit step to reconcile epidemiologic findings with the available laboratory and environmental data that have been collected. This is an important step but is often done in parallel with the hypothesis testing.
11. 11. The primary purpose of conducting an outbreak investigation is to not only determine the source of the infection but also simultaneously control the spread of the outbreak and prevent additional cases. For example, during an outbreak of yellow fever it may be necessary to vaccinate any at-risk populations, apply insecticide, and educate the community regarding ways to reduce mosquito breeding sites. Planning for this step is often done in an empirical manner at first and refined as better data are accumulated.
12. 12. As noted earlier, disease surveillance is one of the key components of a public health system and often is responsible for identification of an outbreak. It is important to remember to continue or even enhance surveillance during an outbreak, to assure that cases are not missed and the control measures that have been implemented are effective.
13. 13. Finally, it is imperative to communicate all findings. Unfortunately, this last step is often overlooked. The results of the study should be presented to the stakeholders (i.e., hospital staff, local health authorities, scientific community) so that others can learn from the investigation and recommendations.

Although infectious diseases do not necessarily respect or recognize boundaries and jurisdictions of cities, states, and countries, it is important to remember that local and state health departments, as well as national and international authorities have specific roles and responsibilities. An investigation of a nosocomial outbreak limited to an individual hospital or a diarrheal disease outbreak at a single daycare center would require coordination at the local health department level only. However, if it is determined in the course of the outbreak investigation that the source of the nosocomial outbreak was in multiple states (e.g., from contaminated blood products), the investigation would require coordination with other states and likely national authorities (i.e., CDC) to coordinate the investigation and response across borders. Similarly, if an outbreak was occurring in multiple countries (a pandemic such as SARS or influenza), then coordination would need to occur via WHO and the GOARN.^16,17