Belinda Ostrowsky
Introduction and Importance/Expanding Role of Epidemiology in the Healthcare Facilities
The term epidemiology is derived from the Greek epi (on or upon), demos (people or population), and logos (word or reason). Literally, it means “the study of things that happen to people”; historically, it has involved the study of epidemics [1,2]. The Harvard School of Public Health defines epidemiology as “the study of the frequency, distribution and determinants of disease in humans. … Epidemiologists use many approaches, but the ultimate aim of epidemiologic research is the prevention or effective control of human disease.” [3] For many years, the population for discussion in this text would have been solely/predominately hospitalized patients and the terms hospital-acquired or nosocomial infection would have been used. Since the spectrum of healthcare and the interaction of different types of healthcare facilities (including hospital, long-term care, rehabilitation, or ambulatory care facilities) have expanded in recent years, a more appropriate term healthcare-associated infections (HAI) will be used in this chapter (still used interchangeably with nosocomial as appropriate)[4].
In hospitals alone, HAIs account for an estimated 2 million infections, 90,000 deaths, and $4.5 billion in excess health care costs annually [4,5]. There has been a shift in the patient population that healthcare facilities care for, especially in hospitals, to more complicated patients, including those who are more severely ill (multiple comorbidities and the need for intensive care unit [ICU] level of care) and an increasing number of patients who are severely immunocompromised. More devices and procedures are used in patients and for longer durations of time. In recent cost-conscious times, there are been staffing shortages (decreasing staff to patient ratios). In addition, over several decades, antimicrobial resistant pathogens (ARPs) and emerging infectious diseases have emerged. All of these have added to the challenge of preventing and controlling HAIs [6].
In addition, in the last several years, broadened concerns for medically adverse events by reports such as the Institutes of Medicine's (IOM) To Err is Human: Building a Safer Health System [7], including medical errors, have illustrated the use of basic epidemiological methods and the expansion of the role of those in infection control and healthcare epidemiology. Recently, there also has been a campaign by the Consumer's Union for the public disclosure of HAI rates [8]. This organization states that the goal of its campaign is to help consumers find the best quality of care by promoting the public disclosure of HAI rates. If hospitals disclose this key information, it says, consumers and employers can select the safest hospitals, and competition among hospitals will quickly force the worst to improve. In several states, legislation that
P.4
obligates hospitals to report specific infection information has been passed. Professional infection control and healthcare epidemiology organizations and healthcare epidemiologists throughout the United States are assisting in framing model legislation and helping hospitals how to comply with these legislations [8,9]. Concerns are that state reporting systems should be based on reliable data, adhere to recommended practices that have been shown to reduce the risk of HAIs and improve patient care, protect patient confidentiality, and reflect the fact that some institutions treat more seriously ill patients.
All of these challenges illustrate the important and evolving role that those in infection control and healthcare epidemiology have in healthcare facilities. Although the circumstances may change, it is the working knowledge of the principals of epidemiology and especially of the subtleties that apply to HAI/adverse events that is essential and sets us apart from other healthcare workers (HCWs). The healthcare epidemiologist will be the one person who has the skills to analytically review occurrences and design studies to evaluate risk factors and interventions (i.e., wielding the power of epidemiology to impact prevention and control of HAIs).
It is with this in mind that the focus of this chapter is to review the basic principles of epidemiology with special emphasis on the relationship of these to HAIs. This first chapter in previous editions of this text, Hospital Infections [1] had been relatively stable and written by one of the fathers of infectious diseases epidemiology and nosocomial infections. Although much of the basic format of this chapter is unchanged, the authors hope to update this classic chapter including new nomenclature, advanced epidemiological methods and updated examples, and references for the basic epidemiological principals related to HAI.
Definitions
Infection and Colonization
Although terms such as infection, infectious disease, subclinical infection, and colonization are used frequently, the subtleties of these terms often are confusing. The term infection implies the successful multiplication of a microbe on or within a host. The term infectious disease applies when signs and symptoms result from infection and its associated damage or altered physiology [1,10].
If the infection provokes an immune response only, without overt clinical disease, it is a subclinical or inapparent infection. Colonization implies the presence of a microorganism in or on a host with growth and multiplication of the microorganism but without any overt clinical expression or detected immune reaction at the time it is isolated [1]. Subclinical or inapparent infection refers to a relation between the host and microorganism in which the microorganism is present; there is no overt expression of the presence of the microorganism, but there is interaction between the host and microorganism that results in a detectable immune response, such as a serologic reaction, a skin test conversion, or a proliferative response of white blood cells to antigens from infecting organisms [1]. Therefore, special tests to detect immune responses may be needed to differentiate colonization from subclinical infection. In many instances, there is an absence of such data and the situation is considered colonization.
A carrier (or colonized person) is an individual colonized with a specific microorganism and from whom the organism can be recovered (i.e., cultured) but who shows no overt expression of the presence of the microorganism at the time it is isolated [1,11]; a carrier may have a history of previous disease due to that organism, such as typhoid.
Colonization is a natural process in the development of natural flora. In the neonate, this process occurs within days to weeks of delivery after which the neonate's normal flora is similar to that seen in adults [11]. Whether colonization occurs long or immediately before infection, it can play a major role in the development of HAIs. In many instances, colonization is a necessary precedent to infection. It is worth discussing colonization in more detail since there is heated discussion in the infection control/healthcare epidemiology community about screening for colonization with ARPs and the role and extent of isolation policies/practices. Those who advocate screening and aggressive infection control practices/isolation contend that colonized patients represent a large predominately unrecognized population who can serve as an unchecked reservoir for infection and that once the proportion of colonization patients reaches a threshold, this may lead to high burden of infection with these ARPs for which antimicrobial treatments are limited [12,13]. Those who oppose these screening programs point to limited resources/cost, competing emerging issues and the concerns about the strength of studies/data to support these efforts and practices [14].
Dissemination and Related Concepts
Dissemination, or shedding of microorganisms, refers to the movement of organisms from a person carrying them into the immediate environment [1]. This could be illustrated by culture samples of air or surfaces or other inanimate objects onto which microorganisms from the carrier may have been deposited. Shedding studies may be conducted in specially constructed chambers designed to quantitate dissemination. While shedding studies occasionally have been useful to document unusual dissemination [15], they have generally not been useful in identifying carriers whose dissemination has resulted in infection in other persons. In the hospital setting, dissemination is most effectively identified by means of surveillance in which the occurrence of infection among contacts is noted.
P.5
In some hospitals, culture surveys of all or selected asymptomatic staff may be conducted in an attempt to identify carriers of certain organisms. Even in outbreak settings, such surveys lack practical relevance, can be costly, and can actually be misleading. This practice identifies only those who are culture positive and does not in itself reliably separate colonized persons into disseminators versus nondisseminators. The practice could erroneously identify a HCW as the source and have serious ramifications for his or her future. Instead, culture surveys should be directed by sound surveillance and epidemiological investigation to identify the potential source [16]. Additional laboratory studies to confirm the presence of a suspected HCW or patient disseminator may then be undertaken.
In some instances, dissemination from a carrier has been reported to be influenced by the occurrence of an unrelated disease such as a second infection [17]. One report, for example, suggested that infants carrying staphylococci in their nares disseminate staphylococci only after the onset of a viral respiratory infection. Such infants are called cloud babies. In another instance, a physician disseminated staphylococci from his skin because of a reactivation of chronic dermatitis. Desquamation of his skin led to the transmission of staphylococci (probably by skin squames) to patients with whom he had contact. Dissemination of tetracycline-resistant Staphylococcus aureus from individuals carrying this organism who were treated with tetracycline has been reported. The risk of dissemination is generally greater from individuals with disease caused by that organism than from individuals with subclinical infection or colonization with the organism [17].
Contamination refers to microorganisms that are transiently present on body surface (e.g., hands) without tissue invasion or physiologic reaction. Contamination also refers to the presence of microorganisms on or in an inanimate object.
Healthcare-Associated/Nosocomial Infections
In previous editions of this text, the terms hospital-acquired or nosocomial infections would have been used. With the definition for hospital-acquired/nosocomial infections as infections that develop within a hospital or are produced by microorganisms acquired during hospitalization. As discussed previously, the delivery and scope of healthcare and healthcare epidemiology is expanding. The Centers for Disease Control and Prevention (CDC) defines HAIs as infections that patients acquire during the course of receiving treatment for other conditions or that HCWs acquire while performing their duties within a healthcare setting [4]. Even the branch of CDC that was formally the Hospital Infections Program broadened its name to the Division of Healthcare Quality Promotion to reflect this sentiment [4].
The CDC's National Nosocomial Infections Surveillance (NNIS) system was developed in the early 1970s to monitor the incidence of HAIs and their associated risk factors and pathogens. NNIS is the only national system for tracking HAIs [18]. The NNIS system currently is undergoing a major redesign as a Web-based knowledge management and adverse events reporting system. Once implemented, the redesigned system (to be called the National Healthcare Safety Network [NHSN]) will cover new areas of patient safety monitoring and evaluation. Although the NNIS system is for surveillance purposes and NNIS definitions are not necessarily to be held as a gold standard for clinical/therapeutic decisions, some important points about HAIs can be illustrated by NNIS experience and definitions [19]. The first is that a major factor that specifically separates epidemiology and HAIs from other infectious diseases is that there is a quest to step away from a single patient and to standardize definitions to consistently identify trends in HAIs. In beginning to identify a cluster or outbreak, one of the early and consistent steps is to try to find a definition as to what constitutes a case of the HAI in question. Use of uniform definitions is critical if data collected are to be used for inter- or intrafacility comparisons, or to data from aggregated systems (e.g., NNIS) [16,19].
The NNIS/NHSN system defines an HAI as a localized or systemic condition that results from adverse reaction to the presence of an infectious agent (s) or its toxin(s), which was not present or incubating at the time of admission to the hospital/facility. For bacterial HAIs, this means that the infection usually becomes evident ≥48 hours (i.e., the typical incubation period) after admission. Since incubation periods vary with type of pathogen and patient's underlying disease, each infection must be assessed individually.
Two special situations that are usually HAIs are infection in a neonate that results from passage through the birth canal and infection that is acquired in the hospital but does not become evident until after hospital discharge. The majority of HAIs becomes clinically apparent while the patients are still in the facility; however, different studies have given widely varying estimates that between 12%–84% of surgical site infections (SSIs) are detected after discharge from the hospital [20]. Since the length of postoperative stay continues to decrease, many SSIs may not be detected for several weeks after discharge and may not require readmission to the hospital where the operation occurred. In these instances, the patient became colonized/infected while in the hospital, but the incubation period was longer than the patient's hospital stay. This sequence also is seen in some infections of newborns and in most breast abscesses of new mothers (since the length of postpartum stay also is brief).
Two special situations that usually are not considered HAIs are the complication or extension of infection(s) already present on admission unless a change in pathogen or symptoms strongly suggests the acquisition of new infection and the infection in an infant that is known or proven to have been acquired transplacentally (e.g., toxoplasmosis, syphilis) and becomes evident ≤48 hours
P.6
after birth. Infections incubating at the time of the patient's admission to the facility are not HAIs; they are community acquired unless they result from a previous healthcare exposure. However, community-acquired infections can serve as a ready source of infection for other patients or HCWs and thus must be considered in the total scope of hospital-related infections.
Two important principles of HAIs relate to infections that are preventable versus those that are nonpreventable. The term preventable HAI implies that some event related to the infection could have been altered and that such alteration would have prevented the infection from occurring. A HCW who does not perform hygiene on his or her hands between contacts with the urinary collection equipment of two patients, for example, may transmit pathogens from the first patient to the second, which may result in a urinary tract infection. Hand hygiene might have prevented this infection from occurring. The identification of such an event in retrospect, however, is difficult; it is necessary to distinguish this situation from circumstances in which both patients developed infections from their own endogenous flora (e.g., from Escherichia coli). It often is impossible to identify the precise mode of acquisition of individual HAIs. More than one mode of transmission may contribute to the development of the same infection, and not all modes may be preventable.
A nonpreventable infection is one that will occur despite all possible precautions, for example, infection in an immunosuppressed patient due to endogenous flora. It has been estimated that approximately 30% of all reported HAIs are preventable [1]; however, more recent studies documenting the near elimination of catheter-related bloodstream infections (BSIs) in ICU patients suggest that an even higher percentage of HAIs may be preventable. Outbreaks, especially those caused by a common vehicle, potentially are preventable; however, outbreaks/clusters account for only a small number of HAIs (5–10% of all HAIs) [21,22]. Prompt investigation and the institution of rational control measures should reduce the extent of the epidemic. Endemic infections account for the majority of HAIs, and the consistent application of recognized, effective control and prevention measures for endemic infections probably is the single most important factor in reducing the overall level of HAIs.
Source: Endogenous (Autogenous) or Exogenous
Two terms endogenous (autogenous) and exogenous, are helpful in understanding HAIs. Endogenous infections are caused by the patient's own flora; exogenous infections result from transmission of organisms from a source other than the patient. For endogenous infections, that patient either was admitted to the facility colonized with these microorganisms or became colonized at some point during his or her stay at the hospital/facility after admission. It may not always be possible to determine whether a particular organism isolated from a patient with an HAI caused by that organism is exogenous or endogenous, and the term autogenous should be used in this situation. Autogenous infection indicates that the infection was derived from the flora of the patient, whether or not the infecting organism became part of the patient's flora subsequent to admission [1]. Information about current infectious diseases/microorganism problems in the community or in hospital contacts may be useful in differentiating the two sources. For example, in the past, if a patient had an infection with methicillin-resistant S. aureus (MRSA), it probably would have been assumed that this infection was related to acquisition in the healthcare facility. In the last several years, however, episodes of community-acquired MRSA [23,24], have increased and it may be helpful to know the local occurrence of these community-acquired isolates in addition to the patient's recent healthcare-related and antimicrobial exposures. Microbiologic characteristics of the organism such as antibiograms, biochemical testing, and molecular typing (staphylococcal cassette or genotype) may provide additional evidence to support nosocomial versus community origin of these strains/infections.
Spectrum of Occurrence of Cases
Often possible clusters of HAIs are detected through clinical microbiology or infection control surveillance data or by an astute laboratorian or clinician [16,25]. Once a cluster is detected, one must evaluate whether this represents a problem, such as an outbreak. As discussed previously, exploring cases and arriving at a case definition is essential to identify as many cases as possible. Comparing the rate of the event during the cluster to a period before the cluster can establish whether an outbreak is occurring. A few definitions are helpful to characterize disease frequency, including sporadic, endemic (hyperendemic), outbreak, and epidemic.
Sporadic means that episodes occur occasionally and irregularly without any specific pattern. Endemic means that the disease occurs with ongoing frequency in a specific geographic area in a finite population and over a defined time period. Hyperendemic refers to what appears to be a gradual increase in the occurrence of a disease in a defined area beyond the expected number; however, it may not be certain whether the disease will occur at epidemic proportions. An epidemic is a definite increase in the incidence of a disease above its expected endemic occurrence. Outbreak often is used interchangeably with epidemic; however, many use outbreak to mean an increased rate of occurrence but not at levels as serious as an epidemic [1].
An occasional gas gangrene infection among postoperative patients is an example of a sporadic infection. An endemic HAI is represented by the regular occurrence of infections either in a particular site or at different sites that are due to the same organism, occur at a nearly constant rate, and are generally considered by the hospital staff to be
P.7
within expected and acceptable limits. SSIs due to a single organism that follow operations classified as “contaminated surgery,” for example, could represent the endemic level of SSIs.
Plotting a histogram of the distribution of “cases” by time in an “epidemic curve” may aid in confirming the existence of an outbreak (versus sporadic or endemic infections) and developing hypotheses about the mode of transmission [16,26,27]. This can be simply executed on graph paper or by using a variety of software packages, such as Microsoft Excel or PowerPoint. Details on the construction of an epidemic curve are described in the descriptive epidemiology section (also Table 1-2 and Figure 1-1).
|
Figure 1-1 Examples of epidemic curves. (A) Point-source: This epidemic curve represents a point-source exposure. The patients are all exposed to the same source and the curve rapidly rises to a peak and then resolves when the source is removed. (B) Continuous Common source: This epidemic curve represents a continuous common source outbreak. Exposure to the source is prolonged and thus the curve is less peaked than the point-source curve. Here the downward slope of the curve quickly decreases with removal of the exposure. (C) Propagated/Progressive source: This epidemic curve represents a propagated/progressive source. Each case is a source of infection for the subsequent case. There are usually several peaks caused by person-to-person transmission. |
Measures of Disease Frequency-Incidence and Prevalence and Related Measures
To identify that a problem with HAIs exists, it is important to be able to quantify the frequency of disease/event occurrence. The two most commonly used measures of disease frequency are prevalence and incidence. Some unique issues that may occur with HAIs related to these measures of frequency will be reviewed, and some additional measures of incidence (incidence density and cumulative incidence) and prevalence are discussed. Each of these measures has uses in healthcare epidemiology and advantages and disadvantages (Table 1-1).
Incidence is the number of new cases in a specific population in a defined time period [28]. Prevalence is a measure of status rather than newly occurring disease and of people who have the disease at a specific time [28].
Incidence can be described in several ways. Incidence density (also known as the incidence rate) is the number of new events (disease onset) in a specified amount of person-time (hospital or healthcare facility days) in a population at risk [28,29]. Incidence density usually is restricted to the first event (first HAI, i.e., BSI), since second events are not statistically independent events in the same individual
P.8
(i.e., once the patient has had one HAI, he or she is more likely to have a second one). The population at risk includes all patients who have not yet had the first event. Once a patient acquires the first HAI, he or she would not still be a part of the population at risk and would be withdrawn. Patients who never have an HAI would contribute all their hospital-/facility-days to the pool of days at risk for a first event, but patients who became infected would contribute only those hospital-days before the onset of the infection.
|
TABLE 1-1 |
||||||||||||||||||||||||||||||
|
Since the first event is just a number, incidence density has the units of (1/time). In practical use in healthcare epidemiology, HAI rates usually are expressed as number of first events in 1,000 hospital-/facility-days (usually gives a single or double-digit number of events per 1,000 hospital days) [29,30]. The advantage of using incidence density is that it allows a way to correct for time and separat out the duration of exposure from the effect of daily risk. Examples where this is particularly useful in healthcare epidemiology are in comparisons of those with short versus long hospital stays and for peripheral intravenous lines versus central venous catheters [29,30]. In each of these instances, the time at risk is substantially longer for the second group versus the first.
A question that arises with the incidence density is what to do with a second or additional event (e.g., second HAI; i.e., second healthcare-associated BSI) since multiple studies have shown that the subsequent events are not independent. The first guidance is that for quantitative analysis of HAIs, it would be overly simplistic and misleading to sum these nonindependent events and put them over the denominator [29]. The first and each subsequent event actually would be a risk factor for the next infection, which is why it is best to restrict analysis to the first event.
There are more complex methods to include first and multiple/subsequent events in a study using different stratum [31]. For example, one approach is to define the population at risk differently for each occurrence [28]: The population at risk for the first event would consist of individuals who have not experienced the disease before; the population at risk for the second event or first recurrence would be limited to those who have experienced the event (infection) once and once only, and so on. An individual should contribute time to the denominator of the incidence rate for first events only until the time that the disease first occurs. At that point, the individual should cease contributing time to the denominator of that rate and should begin contributing time to the denominator of the rate measuring the second occurrence. If and when there is a second event, the individual should stop contributing time to the rate measuring the second occurrence and begin contributing to the denominator of the rate measuring the third occurrence, and so forth.
P.9
Cumulative incidence is the proportion of all those at risk who ultimately suffer a first event [28,29,30]. In traditional infectious disease epidemiology, this has been termed the attack rate [29]. This is actually not a rate but a proportion. Cumulative incidence is derived from the incidence density and in simple terms could be thought of as the sum of all incidence densities for first events over all of the person-time at risk for the first event. This is a simple proportion and thus has no units. For overall HAIs, the time implied is the course of hospitalization (duration in the facility) until the first event or discharge without first event. There are a few limitations to the cumulative incidence. First, there should be follow-up for all at risk to determine whether they have the first event. However, patients do not all have the same length of hospital stay or remain at risk for the same amount of time. Also, HAIs are time related, and comparing HAI rates among patients with differing lengths of stay can be misleading. The cumulative incidence could be of particular use with an HAI considered to be from a point-source, such as a contaminated fluid or SSI (operation as the point-source) [29,30].
In the past, HAIs were reported as a cumulative incidence of number of infections per 100 discharges. One disadvantage of this aggregation and presentation is that there was no distinction between separate first infection from multiple infections in the same patients (thus, 10 infections per 100 discharges could be 10 infections from one very complicated patient, 10 from 10 different healthy patients, or some description between these extremes; the extremes illustrate how this summary term could be quite different in its clinical and epidemiological relevance and in what intervention methods might be necessary). Another disadvantage is that since one patient could be counted multiple times, this would not take into account lack of statistical independence, thus making comparisons difficult [29].
Unlike incidence measures, which focus on events, prevalence focuses on disease status. Prevalence is defined as the proportion of a population that has disease at a specific point in time [30]. Several terms, such as point prevalence, prevalence proportion, and prevalence rate, aften are used interchangeably. Prevalence depends on the incidence and the duration of the disease. As either of these increases, the prevalence increases. The main useful measure for healthcare epidemiology would be point prevalence for studies such as a cross-sectional study [30] (i.e., a point prevalence survey on a day using cultures to detect colonization/infections with an ARP, such as vancomycin-resistant enterococcus (VRE) or MRSA). This could give an idea of the burden of a problem at a particular point in time to assist in defining that a problem exists, guide decisions to pursue additional studies, and allow for allocation of resources. Of note, populations are dynamic and since individuals are entering and leaving the population, the prevalence can vary based on when it is measured.
Epidemiological Methods
Generally, three techniques are used in epidemiologic studies: descriptive, analytic, and experimental; all may be used in investigating HAIs. Descriptive epidemiology is the foundation for evaluation of HAIs and is used in both surveillance and most investigations of potential problems/outbreaks. Once the initial problem has been defined by descriptive epidemiology, additional studies using analytic and/or experimental methods can be conducted to develop more information about the problem, confirm initial impressions, prove/disprove hypotheses (including identifying risk factors/potential associations/sources or causes), and evaluate the effectiveness of control measures and/or prevention measures.
The presentation of descriptive epidemiology can include case report/case series. A case report is the clinical description of a single patient. A case series is a report of >1 patient. These types of studies/publications are easy to prepare and can serve as examples to other healthcare epidemiologists. These studies also can serve as a resource to generate hypotheses and ideas for additional studies. The disadvantage of this type of study is that patient numbers are small, and the findings may not be generalizable to other populations. In addition, no comparison to other groups has been made.
The analytic study section will discuss case-control and cohort studies, which are the comparative studies frequently used in healthcare epidemiology, especially to explore outbreaks and healthcare epidemiology problems to identify risk factors and potential associations. Additional analytic-type studies that can be used for healthcare epidemiological studies are ecological or cross-sectional studies. Experimental methods of studies include randomized control trials (rarely used in healthcare epidemiology) and quasi-experimental studies (to evaluate an intervention without randomization).
Descriptive Epidemiology
Descriptive epidemiologic studies evaluate the occurrence of disease in terms of time, place, and person [1,26]; each “case” of a disease is first characterized by describing these three attributes. When data from the individual cases are combined and analyzed, the parameters of the outbreak or disease problem should be characterized. Issues that arise regarding time, place, and person in general descriptive epidemiology and specifics to healthcare epidemiology are discussed next.
Time
There are four time trends to consider: secular, periodic, seasonal, and acute [1]. Secular trends are long-term trends in the occurrence of a disease—that is, variations that occur
P.10
over a period of years. An example in HAIs would be the gradual increase in fungal BSIs, including those that are nonalbicans Candida and azol resistant [32]. Periodic trends are temporal interruptions of the secular trend and usually reflect changes in the overall susceptibility to the disease in the population. The upsurge in Influenza A activity every 2 to 3 years, for instance, reflects the periodic trend of this disease and generally is the result of antigenic drift of the Influenza A virus. Seasonal trends are the annual variations in disease incidence related in part to seasons. In general, the occurrence of a particular communicable disease increases when the circumstances that influence its transmission are favorable. The seasonal pattern of both community-acquired and healthcare-associated respiratory disease, for example, is high incidence in the fall and winter months when transmission is enhanced because people are together in rooms with closed windows and are breathing unfiltered, recirculating air. Thus, they have more contact with one another and with droplets/droplet nuclei. There also may be agent and host factors that influence the seasonal trends. Another example is that healthcare-associated Acinetobacter spp infections have a seasonal trend, increasing in the summer and fall [33]. The fourth type of time variation is the acute or epidemic/outbreak occurrence of a disease with its characteristic upsurge in incidence.
As described previously, a graphic representation of the “cases” can assist in confirming the existence of an outbreak, its source, its transmission, the point of an outbreak you are in, and evaluating interventions. The overall shape of the epidemic curve depends on the interaction of many factors: characteristics of the agent (i.e., pathogenicity, concentration, and incubation period), the method of transmission, host factors (i.e., susceptibility and concentration of susceptible individuals), and environmental factors (i.e., temperature, humidity, movement of air, and general housekeeping). A step-by-step guide to creating an epidemic curve is included in Table 1-2 [16,26,27].
The following are a few points to keep in mind while attempting to create/interpret an epidemic curve. The time scale will vary according to the incubation or latency period, ranging from minutes, as in an outbreak of disease following exposure to a toxin or a chemical, to months, as in an epidemic of hepatitis B. The time scale (abcissa, hortizontal scale,x axis) should be selected with three facts in mind: (l) The unit time interval should be less than the average incubation period (commonly one-fourth to one-third of the probable incubation period) so that the true nature of the epidemic curve will be apparent (i.e., all the cases will not be bunched together); (2) the scale should be extended far enough in time to allow all cases to be plotted; and (3) any cases that occurred before the epidemic should be plotted to give a basis for comparison with the epidemic/outbreak experience [1,26,27]
If the epidemic curve starts with the index case (i.e., the first case in the outbreak), the time between the index case and onset of the next case reflects the incubation period if transmission was from the index case directly to the next case—that is, from person to person. The upslope in the curve is determined by the incubation period, the number and concentration of exposed susceptible persons, the number of infected sources, and the ease of transmission. The height of the peak of the curve is influenced by the total number of exposed susceptible individuals and the time interval over which they occur. The downslope of the curve is usually more gradual than the upslope; its gradual change reflects cases with longer incubation periods and the decreasing number of susceptible individuals. The initiation of control measures may contribute to the gradual decline or to a sudden decrease in the appearance of new cases [1].
When interpreting an epidemic curve, it is useful to look at the overall shape of the curve to assist in determining how the outbreak spread throughout the population and, potentially, if the disease is unknown, the initial diagnosis of the disease. For simplicity, there are three main patterns the epidemic curve can take (Figure 1-1) [27]. In a point-source epidemic, persons are exposed to the same exposure over a limited, defined period of time, usually within one incubation period. The shape of this curve commonly rises rapidly and contains a definite peak at the top followed by a gradual decline. Sometimes cases also may appear as a wave that follows a point-source by one incubation period or time interval. This is called a point-source with secondary transmission.
In a continuous common-source epidemic, exposure to the source is prolonged over an extended period of time and may occur over >1 incubation period. The downward slope of the curve may be very sharp if the common source is removed or gradual if the outbreak is allowed to exhaust itself (i.e., affect all susceptible persons).
A propagated (progressive source) epidemic occurs when a case of disease serves as a source of infection for subsequent cases, and those subsequent cases, in turn, serve as sources for later cases. The shape of the curve usually contains a series of successively larger peaks, reflective of the increasing number of cases caused by person-to-person contact until the pool of susceptibles is exhausted or control measures are implemented.
In reality, mixed modes of transmission may occur, and the epidemic curve could include both point-source and propagated cases.
Place
Although outbreaks of HAI occur infrequently in some settings, such as ICUs, they can account for a substantial percentage of the HAIs. In an investigation of HAIs, three different places may be involved. The first is where the patient is when the disease is diagnosed, and the second is where contact occurred between the patient and the agent. If a vehicle of infection is involved, the third place is where
P.11
P.12
the vehicle became contaminated. To implement the most appropriate control and preventive measures, it is necessary to distinguish between these three geographic areas; certain actions may control additional spread from a specific focus but may not prevent new cases from occurring if the source continues to contaminate/infect new vehicles.
|
TABLE 1-2 |
|||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||
An example will help to emphasize the importance of carefully describing the place or places involved in disease outbreaks. In an outbreak of nosocomial salmonellosis, the patients were located on various wards throughout the hospital at the time they developed disease. Individual control measures were directed at each patient on the various wards; however, the place of infection was the radiology department, where barium used for gastrointestinal tract roentgenographic examinations was contaminated with salmonella. The barium had been contaminated in the radiology department, and thus preventive measures directed there terminated the outbreak.
Because transfer of patients between hospital wards or units is common, it may be difficult to attribute an outbreak to a particular geographic area. Infection rates can be calculated for other areas of the hospital and compared to the area(s) with the cluster to aid in identifying the location of the outbreak [16]. In addition, a review of the geographic location of cases using a spot map of the hospital or ICU may suggest the location or pattern of transmission [16,26].
Person
The third major component of descriptive epidemiology is person. Careful evaluation of host factors related to the individual person includes consideration of age, gender, race, immunization status, immunocompetence, and presence of underlying disease that may influence susceptibility (acute or chronic), therapeutic or diagnostic procedures, medications, and nutritional status. In essence, any host factor that can influence the development of disease must be considered and described. Those factors that increase the patient's chance of developing disease are known as risk factors.
Age also can be an important clue to the source of an outbreak of disease. If, in an apparent common-source outbreak, for example, all ages are involved, the source of the outbreak must have been exposed patients scattered through at least several wards. On the other hand, if all the patients involved in an epidemic are women of child-bearing age, in attempting to identify the place of the exposure, the investigation can be narrowed to the obstetric or, possibly, the gynecology ward.
Consideration of therapeutic procedures may be of similar importance. If all patients who developed BSIs due to the same organism have received intravenous fluid therapy, a common source of intravenous fluids could be suspected as the cause of the outbreak.
In addition, knowledge of intrinsic host risk factors is useful because separate risk-specific rates can be calculated, which allows for the comparison of HAIs among patients with similar risk. Severity of illness is a strong confounding variable in outbreaks in healthcare settings. The Acute Physiology and Chronic Health Evaluation (APACHE II) and Diagnostic Related Groups (DRGs) are well-known indices used to assess and control the severity of illness [35,36]. These indices are used to predict the risk of death among ICU patients and for staff resource utilization. In pediatrics, severity of illness scores including the modified
P.13
abbreviated injury severity score (MISS) and Score for Neonatal Acute Physiology (SNAP) have been used to assess neonatal/pediatric populations [36,37].
Analytic Epidemiology
After descriptive epidemiologic review has been performed and hypotheses have been generated, one may need additional studies to identify the source of a problem/outbreak. Settings in which additional efforts should be considered are when resources are available, when the problem/outbreak is associated with high mortality or severe disease, when new or unusual pathogens or methods of transmission are identified, or when the problem/outbreak continues despite implementing control measures. Additionally, the principles involved in these methods have application to surveillance; surveillance data commonly are analyzed by the descriptive method, and such analysis may suggest the need for analytic studies to identify certain features of a disease. The choice of analytical/comparative study depends on resources, time, and size of the problem/outbreak.
A few basic epidemiological principles should be reviewed with an emphasis on how these concepts vary or apply to HAIs. These are study designs/methods (emphasizing case-control or cohort studies but also including ecologic and cross-sectional studies), measures of association, strength of association, and bias/confounding. Although this chapter is not meant to be a primer on general epidemiology, some key points will be reviewed.
Study Designs/Methods
Two frequently used analytic methods include case-control or cohort studies. In both instances, associations that may identify causes and effects are sought. The case-control method starts with the effect (cases) and searches for causative host and exposure factors, and the cohort method starts with potential causative factors and evaluates the effect. The case-control and cohort methods also have been referred to as retrospective or prospective studies, respectively; both methods, however, can be either retrospective or prospective. These terms indicate the temporal frame of reference for the collection of specific data: in a retrospective study, data are collected after the event has occurred; in a prospective study, the data are collected as the event occurs.
Case-Control Study
In a case-control study, case-patients are compared with (a set ratio usually of 1:2 or 1:3) control-patients who do not have the adverse outcome or infection but have had the opportunity for the exposure. The case-control approach has the advantages of being inexpensive, relatively quick, and easily reproducible. It is used most often in acute disease investigations, since the epidemiologist usually arrives after a problem is recognized and often after the peak of the epidemic has passed. It also allows evaluation of many potential associated exposures and for outbreaks/problems that may have persisted for lengthy periods of time.
One of the main controversies with this type of study is the appropriate choice of control-patients [38]. In one investigation of transmission of VRE in 32 hospitals and long-term care facilities, we wanted to explore the characteristics and exposures of those patients with VRE-colonization. The facilities varied in size and provided a range of intensity of medical care. VRE-colonized patients were identified from several different facilities. Thus, for each case-patient, we chose a control-patient from the same facility, and the analysis was matched for the facility [39].
In ICU outbreaks, since the patients are very different (e.g., more severely ill, have more devices, experience more procedures, take more medications) from non-ICU patients, other ICU patients may be the most appropriate controls [25,40,41]. In some studies, controls may be matched to cases on certain factor(s), such as age, gender, or other factors known to predispose to the outcome. In general, random selection of controls is preferred [42,43]. Two concerns about matching are that special statistics are needed for matched analyses and that when matching is performed, no comparisons can be made between case- and control-patients on the factors on which matching was done [42].
Review of the case-patients' medical records should identify several potential sources/risk factors. In the case-control study, a comparision of the presence or absence of these factors in the case-patients and controls is performed to see whether any of these exposures is more likely to be present in cases, suggesting that this may be associated with the outbreak. Use of standardized data collection forms facilitates the systematic review of exposures.
A pitfall in some healthcare outbreaks is that many exposures can be collected. Only biologically plausible exposures should be evaluated. A rule of thumb is that if an exposure is not present in at least 30%–40% of the cases, even if it is more common in cases than controls, it will not account for enough cases to be the source of the outbreak (attributable risk—the amount or proportion of disease incidence/risk that can be attributed to a specific exposure) [40,44,45].
Two important statistical principles should be reviewed at this point relating to errors. Type I error (α error) relates to concluding that a statistical relationship exists when it does not. This may occur in a case-control study when many factors/exposures are evaluated. With multiple tests, a relationship may be found to be statistically significant but represents a false positive. Often these false positive relationships have only borderline significance (p values near 0.05) with weaker magnitudes of association and lack of biological plausibility. The take-home message is
P.14
not to examine factors that are not clinically relevant or biologically plausible since a relationship may be identified purely by performing multiple tests looking at multiple variables [46]. For other studies not related to outbreaks, it best to make an a priori plan in which variables are to be collected and analyzed to prevent this problem.
Type II error (β error) relates to concluding that a factor is not significantly associated with becoming a case when it in fact is related. This error is related to the concept of power. Power is 1 - β error. These concepts are greatly affected by the sample size in a study. In a planned research protocol, set numbers of cases and controls can be enlisted. In an outbreak situation, the number of cases is obviously limited. The main points are that HAI outbreaks/studies may be of a smaller scale and that certain associations may not reach statistical significance; however, trends may still have clinical significance [46].
Another area of controversy is how long variables of interest should be collected in case-and control-patients. This needs to be clear for those reviewing charts/medical records. In one ICU outbreak we investigated, exposure data were collected for the case-patients from SICU admission until the day of diagnosis of their S. marcescens BSI and for controls from the date of SICU admission to the median time that the case-patients developed their S. marcescens BSI (7 days) or discharge date from the SICU if the controls' SICU length of stay was < 7 days. The exposure period for case-patients and controls should be similar [40]. For case-patients only, exposure until the onset of illness should be collected. For example, antimicrobial exposures for acquisition of VRE probably should be collected for the proceeding days or weeks rather than months before onset of colonization. Exposure months before onset of a disease may be present but have no relationship to disease acquisition. Similarly, exposure data should be collected for a preceding biologically plausible period of time. Control exposures to case-patients should be collected for a similar period, not for the entire hospitalization. Often, this pro-cess will lead to a difference with case-patients that really represents the fact that the exposure period in controls (admission to discharge) is longer than in case-patients (admission until onset of disease).
Case-control studies establish only that case-patients were more likely to have been exposed to potential risk factors than were controls. In case-control studies, one can calculate an odds ratio (abbreviated OR), which estimates the relative risk (RR) and measures the strength of the association between the condition and the exposure/risk factor (Figure 1-2) [42,47].
|
Figure 1-2 Example of a two by two table and definitions for measures of effect or association, relative risk and odds ratio. |
Cohort Study
In a cohort study, one assesses the entire population (e.g., all ICU patients from June 2000–December 2001) and evaluates what exposures are more common among those who develop disease/infection than those who do not. As mentioned earlier, a cohort study may be either prospective or retrospective. This distinction depends on when the study is conducted with regard to when the outcome of interest occurs. If patients are identified as exposed and unexposed and then followed forward in time to determine whether they develop the disease, this is a prospective cohort study. If the study is conducted after the time of outcome has already occurred, this is a retrospective cohort study. In either study, subjects are selected based on their exposure to the variable of interest, and these groups are compared based on the outcome.
The main advantage of the cohort study (versus the case-control study) is that if significantly more exposed patients than unexposed patients develop the outcome, this factor may be not only associated but also causally related to the outbreak. In a cohort study, one can quantify the extent to which the exposure increases the risk of developing the condition in a summary term RR (Figure 1-2) [47]. Several cohorts may be followed, each representing a different level of exposure to a factor, thus allowing for the determination of a dose response. Unlike the OR, the RR not only describes that the exposure is associated with the outcome but also denotes causation.
Disadvantages of the cohort study are that they may be costly and time-consuming since patients must be followed in time until a sufficient number develop the outcome of interest (which could be a lengthy period of time if the disease course is slow or the disease is rare). This could lead to loss of follow-up of some in the cohort. Some of
P.15
these limitations are lessened by performing a cohort study retrospectively because the outcome has already happened.
A variant on the traditional case-control and cohort design can be helpful in the evaluation of HAIs. A case-control study within an identified cohort is sometimes termed a “nested case-control” study.
Other Analytics Epidemiology Methods
Ecologic Studies
The studies described so far share the characteristic that the observations made pertain to individuals. Ecologic or aggregate studies conduct research in which the unit of observation is a group of people rather than an individual. The requirement is that information on the population studies be available to measure the exposure and disease distributions in each group. Because the data in ecologic studies are measurements averaged over individuals, the degree of association between exposure and disease need not reflect individual associations. These data may be more easily obtainable than individual-level patient data but as discussed cannot be extrapolated to an individual patient and, in fact, if done could lead to an ecologic fallacy, an error in the interpretation of statistical data, in which inferences about the nature of individuals are based solely on aggregate statistics collected for the group to which those individuals belong. This fallacy assumes that all members of a group exhibit characteristics of the group at large.
In healthcare epidemiology, some examples of ecologic studies are the use of data aggregated for other purposes, such as drug dispending data from hospital pharmacies, and the antimicrobial susceptibility data from the hospital clinical microbiology laboratory. A good example would be hospital data that could show an increase in use of the antimicrobial vancomycin and VRE in enterococcal isolates, but it will not be possible to know whether the patients who received vancomycin were the patients who acquired VRE. This type of data, however, may serve as exploratory data on which to base additional studies.
An interesting use of ecologic data is the CDC's NNIS Intensive Care Antimicrobial Resistance Epidemiology (ICARE) project [48]. During this 4-year study, a subset of NNIS hospitals (50 ICUs at 20 U.S. hospitals) monitored antimicrobial use and ARPs. Participating hospitals reported the grams of select antimicrobial agents administered to patients and the antimicrobial susceptibility results of isolates recovered from clinical specimens from hospitalized patients each month. Microbiologic data were aggregated for each ICU separately, all non-ICU inpatient wards combined, and all outpatient areas combined. Pharmacy data were reported for the same hospital strata except for outpatient areas for which pharmacy data were not available. Amounts of antimicrobial agents reported were standardized by conversion to defined daily doses; for parenteral vancomycin, one daily dose was defined as 2 grams.
The study found that after data were adjusted for changes in MRSA prevalence, changes in specific prescriber practice at ICUs were associated with significant decreases in vancomycin use (mean decrease -48 defined daily doses per 1,000 patient days, p < 0.001). These ICUs also reported significant decreases in VRE prevalence compared with those not using unit-specific changes in practice (mean decrease of 7.5% compared with mean increase of 5.7%, p < 0.001). In this study, practice changes that focused on specific ICUs were associated with decreases in ICU vancomycin use and VRE prevalence. This example illustrates how HAI data may be aggregated for ecologic study, such as the development of defined daily doses and which hospital units' data were compared.
Cross-Sectional Studies
A cross-sectional study is a survey or sampling of a population in which the status of the exposure and the outcome are ascertained at the same time. In healthcare epidemiology, this study type is frequently used to assess the prevalence of a specific disease, such as amount of antimicrobial resistance. A disadvantage to this type of study is that it does not give an idea about transition of status over time. Depending on the populations, a cross-sectional study may be analyzed as a cohort or case-control study.
Measures of Association and Related Concepts
A measure of association provides an index of how strongly two factors under study vary in concert [49]. The more tightly they are linked, the more evidence exists that they are causally related to each other (though not necessarily that one causes the other, since they might both be caused by a third factor). Although this term and “measure of effect” have frequently been used interchangeably, Rothman and Greenland draw the following distinction: associations involve comparisons between groups or populations; effects involve comparisons of the same population (hypothetically) observed in two different conditions; measures of association are typically used to estimate measures of effect [49].
In Figure 1-2, a two-by-two table is constructed and the calculation for relative risk and odds ratio [43,47], the two main measures of association that healthcare epidemiologists deal with on a regular basis, are illustrated. The calculations are not the difficult part; it is understanding what these measures elude about the study and the two populations being compared (and what they do not mean).
Depending on the study performed, either an RR or OR usually will be calculated. First, it is worth distinguishing between risk versus odds. Risk refers to probability and has a numerator with the event/occurrence of interest and a denominator with all possible outcomes including the event of interest. The odds has the same numerator,
P.16
but the denominator includes all possibilities minus the event/outcome of interest.
The RR is the ratio of two probabilities of the outcome in the exposed over the probability of the outcome in the unexposed [49]. The RR can be calculated in cohort studies (and randomized control trials). If there is not a difference in the risk of the exposed versus the unexposed, then the RR = 1. RR > 1.0 implies that the exposed group is more likely to have the outcome than those without the exposure (no effect). An RR < 1 implies that the exposed group was less likely to have the outcome than the nonexposed group (protective) [49].
A study of pyrogenic reactions associated with single daily dosing of intravenous gentamicin illustrates the use of RR in a cohort study [50]. The authors conducted cohort studies in an inpatient service of a large community hospital in Los Angeles, California, following patients for the occurrence of pyrogenic reactions (chills, rigors, or shaking chills) within 3 hours after the initiation of gentamicin. During the epidemic period, 22/152 (15%) patients developed documented pyrogenic reactions following receipt of gentamicin. Pyrogenic reactions were more likely among patients receiving single daily dosing than multiple daily dosing of gentamicin (20/73 [27%] versus 2/79 [3%]; relative risk was 10.8%). Thus, in simple terms, those receiving single daily dosing of gentamicin had a risk of developing a pyrogenic reaction that was 10.8 times higher than the risk of those who received multiple daily dosing.
The OR is less intuitive in its interpretation but is the measure that will be available from a case-control study [49]. In this type of study, the subjects are enrolled based on the outcome of interest (comparing a group with the outcome to a group without the outcome) to determine what proportion in each group has an exposure/risk of interest. An RR cannot be directly calculated, because how common the outcomes/exposures are in the entire population cannot be measured. Only an OR can be calculated. An OR reflects the odds of exposure with the outcome divided by the odds of exposure in study subjects without the outcome. An OR = 1.0 implies no effect.
An RR cannot be calculated from a case-control study in usual situations; however, if a disease is rare, the OR can closely approximate the RR that would have been derived from a cohort study [30,49]. The calculations to support this are in Figure 1-2.
An outbreak of Serratia marcescens BSI traced to an infused narcotic by using a case-control study illustrates the interpretation of OR [51]. To identify risk factors for the BSIs, patients with S. marcescens BSIs were compared to randomly selected controls. Patients with S. marcescens BSIs were more likely to have received fentanyl in the surgical ICU (odds ratio, 31; p < 0.001) and were more likely to have been exposed to two particular respiratory therapists (odds ratios, 13.1 and 5.1; p < 0.001 for both comparisons). One respiratory therapist had been reported for tampering with fentanyl, and his hair sample tested positive for it. Cultures of fentanyl infusions from two case-patients yielded S. marcescens and E. cloacae The isolates from the case-patients and from the fentanyl infusions had similar patterns on pulsed-field gel electrophoresis. After removal of the implicated respiratory therapist, no further cases occurred. To translate some of the ORs into understandable statements, for the odds ratio of 31, 13.1, and 5.1 above, cases (patients with Serratia BSI) were 31 times more likely than controls to have received fentanyl in the surgical ICU, cases were 13.1 times more likely than controls to have been exposed to/received care from respiratory therapist X, and 5.1 times more likely to have been exposed to/received care from respiratory therapist Y.
Strength of Association and Confidence Intervals
Analysis should begin with simple univariate frequencies followed by two-by-two tables for binary outcomes with bivariate analysis (Fisher's exact or chi-square tests) or appropriate tests for continuous variables (parametric t tests or nonparametric tests); (Figure 1-2) [43]. A software package, Epi-Info, is available from the CDC at no cost. This software package is very useful for acquiring, organizing, and interpreting epidemiologic data from questionnaire to final analysis (http://www.cdc.gov/epiinfo/).
It is not the intent of this chapter to discuss the background or derivation of these tests for significance. However, interpreting the results for everyday use in healthcare epidemiology is helpful. The idea of these tests is to assess whether a difference seen between groups compared in the studies is real or could be based on chance alone and to assign a probability that the difference is real. By convention, a p value of < 0.05 is usually considered statistically significant [29,30]. This suggests that there is a ≤5% chance that the difference between the groups is due to chance alone. The 0.05 is the convention but somewhat arbitrary, and there may be cases where a less stringent cut off of 0.1 is used. The pvalue can be affected by sample size; with a large enough sample, even small differences may be statistically significant but may not have clinical relevance (a problem in large database analysis, that is, pooled data such as a nationwide database). On the other hand, a larger difference may not reach statistical significance if the sample size is small (a problem in some HAI outbreaks if small numbers of cases are seen).
In the Serratia outbreak linked to fentanyl contamination described earlier [51], patients with S. marcescens BSI were more likely to have received fentanyl in the surgical ICU (odds ratio, 31; p < 0.001). The p value was actually much smaller <0.000001, thus, there was a less than 1 in 1,000,000 chance that the results seen (that cases were more likely to have received fentanyl in the surgical ICU than controls) is by chance alone.
Due to the limitations of the p value, a 95% confidence interval for the measures of association (ORS and RRs,
P.17
depending on the study performed) provides a range within which the true magnitude of the association lies with a certain degree of assurance. If the range includes 1.0, the p value often is nonsignificant or close to 0.05. Sample size also affected these confidence intervals. Especially since HAI outbreaks may be small, studies often suffer from wide confidence intervals [30]. For example, in the study described earlier, pyrogenic reactions were more likely among patients receiving single daily dosing than there receiving multiple daily dosing gentamicin (20/73 [27%] versus 2/79 [3%]; RR = 10.8, p < 0.01 with a 95% confidence interval of 2.6 to 44.7) [50]. Thus, those receiving single daily dosing had a risk to develop a pyrogenic reaction that was higher than the risk of those who received multiple daily dosing of gentamicin, but it is not clear whether it was 2.6 times higher, 44.7 times higher, or somewhere in between.
Bias and Confounding
Bias is defined as any systematic error in an epidemiologic study that results in an incorrect estimate of the association between exposure and risk of disease [52]. Evaluating the role of bias as an alternative explanation for an observed association is necessary in interpreting any study result. Unlike chance and confounding, which can be evaluated quantitatively, the effects of bias are far more difficult to evaluate and may even be impossible to consider in the analysis. There are two general classes of bias. Selection bias refers to any error that arises in the process of identifying the study populations (discussed in the case-control section regarding appropriate controls). The second general category, observation or information bias, includes any systemic error in the measurement of information on exposure or outcome (discussed in the case-control section regarding how long to collect exposure information for cases and controls). The prevention and control of potential biases must be accomplished through careful study design and meticulous conduct of the study. Once a potential source of bias is introduced, it usually is extremely difficult to correct for its effects analytically. However, it is necessary to estimate both the direction and magnitude that the bias would have on the effect, and investigators should discuss all of these issues fully in published reports to provide readers the maximum opportunity to judge for themselves whether the bias accounts for the observed findings.
Confounding can be thought of as a mix of the effect of the exposure under study on the disease with that of a third factor [53]. This third factor must be associated with the exposure and be independent of that exposure, that is, be a risk factor for the disease. Confounding can lead to an overestimate or underestimate of the true association between exposure and disease and can even change the direction of the observed effect. A number of methods are available to control confounding in the design or analysis of any study. These include restriction, matching, or randomization (in clinical trials) in the design and stratification and multivariate techniques in the analysis. No single method can be considered optimal in every situation. Each has strengths and limitations, which must be carefully considered at the beginning of the study. In most situations, a combination of strategies will provide better insight into the nature of the data and more efficiently control for confounding than a single approach [53]. Common examples of confounders in HAIs are length of stay and severity of illness.
Experimental Epidemiology
Randomized Control Trials
The third method of epidemiologic investigation is the experimental method, which is a definitive method of proving or disproving a hypothesis. The experimental method assumes that risk or protective factors are followed by effects on outcomes and that a deliberate manipulation of these factors is predictably followed by an alteration in the outcomes that could rarely be explained by chance. The two groups selected for study are ideally similar in all respects except for the presence of the study factor in one group. Either the case-control or the cohort method is used to evaluate the interaction between the cause and the effect.
An example of the experimental method is the evaluation of a new drug as treatment for a disease: A group of patients with the disease is randomly divided into two subgroups that are equal in all respects except that one of the subgroups is treated with the experimental drug and the other subgroup (the control group) is given a placebo or another agent known to be effective in treating or preventing the disease. If there is no other variation between the two groups, any differences in the course of the disease may be ascribed to the use of the drug.
The experimental method has less direct use in the investigation of HAI outbreaks today than the other analytic methods such as case-control or cohort studies. The experimental method, however, is useful in assessing general patient care practices and in evaluating new methods to control and prevent disease. Placebo-controlled trials have less use in therapeutic studies because of the needs for informed consent and for preventing the placement of the patient at an unjustified or greater risk in attempting to conduct a specific study.
Thus, while the healthcare epidemiologist may not perform these studies, it is important to have a working knowledge of these types of studies as a comparison to the descriptive and analytic methods described previously so that the strengths and weaknesses of these other designs can be acknowledged (since experimental trials such as random control trials provide the best support for causality). In addition, healthcare epidemiologists play multiple roles at
P.18
facilities and may be asked to aid in the assessment of new products for which experimental trials may have been performed and should be able to read the literature and interpret studies for others at the facility.
Quasi-Experimental Studies (Pre- and Postinterventions)
Quasi experimental studies, however, are used in infection control, particularly when a nonrandomized intervention is put into place, assessments of a baseline are taken before the intervention, and similar data are collected before and after the intervention [30,54]. The advantages of these methods include allowance for study of an intervention when a randomized control trial is not feasible for a number of reasons including ethics (in an outbreak setting, the first priority is to protect patients and control the outbreak, withholding treatment/control measures) may be unethical, logistics (it may be impossible to randomize changes to different patients/units), cost, and acceptability.
One disadvantage of quasi experimental studies is that it may be difficult to control for potential confounding variables that may have changed over the same/similar time as the intervention that could account for the change in outcome in part or in full rather than the intervention. Another disadvantage may be a natural change/range in outcomes that may have happened even without the intervention; thus, it may be difficult to attribute the change to the intervention.
An example of this type of study is described in relationship to preventing transmission of the multidrug-resistant Mycobacteria tuberculosis to patients and HCWs [55]. It was a retrospective cohort study measuring the proportion of case-patients with nosocomial acquired M. tuberculosis and the rate of tuberculin skin test conversion among HCWs before and after implantation of control measures (from the 1990 CDC guidelines: prompt isolation and treatment of patients, rapid diagnostic techniques, negative pressure isolation, and molded surgical masks for HCWs). The study found that the proportion of patients with multidrug resistant strains of M. tuberculosis decreased after the intervention (10/70 [14%] compared to 30/95 [32%] patients before the intervention; RR = 0.5, 95% CI, 0.2 to 0.9).
Chain of Infection
General Aspects
Infection results from the interaction between an infectious agent and a susceptible host. This interaction—called transmission—occurs by means of contact between the agent and the host. Three interrelated factors—the agent, transmission, and host—represent the chain of infection.
The links interrelate with and are affected by the environment; this relation is referred to as the ecology of infection, that is, the relation of microorganisms to disease as affected by the factors of their environment. In attempting to control and/or prevent HAIs, an attack on the chain of infection at its weakest link is generally the most effective procedure. With definition of the links in the chain for each HAI, future trends of the disease should be predictable, and it should be possible to develop effective control and prevention techniques. Defining the chain of infection leads to specific action in contrast to the incorporation of nonspecific actions in an attempt to control a HAI problem.
Disease causation is multifactorial; that is, disease results from the interaction of many factors related to the agent, transmission, and host. The development of disease reflects the interaction of these factors as they affect a person. Thus, some people exposed to an infectious agent develop disease and others do not.
Agent
Agent characteristics
The first link in the chain of infection is the microbial agent, which may be a bacterium, virus, fungus, or parasite. The majority of HAIs is caused by bacteria and viruses; fungi are assuming a greater role and parasites a rare cause. A number of factors help to characterize the agent, including infectiousness, pathogenicity (including virulence and invasiveness), dose, specificity, infectivity, and other agent factors (including antimicrobial resistance) [1,10].
The determination of the number of the susceptible individuals who become infected with an organism to which they are exposed is a measure of the infectiousness of that organism. Host factors can influence the infectiousness of an organism.
The measure of the ability of microorganisms to induce disease is referred to as pathogenicity, and it may be assessed by disease-colonization ratios. An organism with low pathogenicity is alpha-hemolytic streptococcus; it commonly colonizes humans but only rarely causes clinical disease. The pathogenicity of an organism is additionally described by characterizing the organism's virulence and invasiveness.
Virulence is the measure of the severity of the disease. In epidemiologic studies, virulence is defined more specifically by assessing morbidity and mortality rates and the degree of communicability. The virulence of organisms ranges from slightly to highly virulent. Although some organisms are described as avirulent, it appears that any organism can cause disease under certain circumstances. Some naturally occurring organisms have been considered avirulent or of low virulence; however, under certain conditions, such as high doses, host immunodeficiency, or both, disease has resulted from contact with these organisms. For years, Serratia marcescens, for example, was considered to be an avirulent organism; because of this and the easily recognizable red pigment produced by certain strains,
P.19
these organisms were used for environmental studies in hospitals. However, as hospitalized patients became more susceptible to developing infections due to advancing age, comorbid conditions, immunosuppression, and the effects of new diagnostic and therapeutic measures, HAIs due to S. marcescens organisms subsequently became recognized and reported. Invasiveness describes the ability of microorganisms to invade tissues. Some organisms can penetrate the intact integument whereas other microorganisms can enter only through a break in the skin or mucous membranes.
Another important agent factor is dose, that is, the number of organisms available to cause infection. The infective dose of an agent is that quantity of it necessary to cause infection. The number of organisms necessary to cause infection varies from organism to organism and from host to host and is influenced by the mode of transmission.
Microorganisms may be specific with respect to their range of hosts. Brucella abortus is highly communicable in cattle but not in humans. Some Salmonella spp. are common to both animals and humans, but others have a narrow range of specificity; for example, S. typhosa is known to infect only humans.
Infectivity refers to the ability of an organism to spread from a source to a host [1,10]. An infected human may be infective during the incubation period (e.g., hepatitis A), the clinical disease state (e.g., Influenza A), convalescence (e.g., salmonellosis, shigellosis), or some combination of the three. Additionally, an asymptomatic carrier (or colonized person) who does not show evidence of clinical disease may be infective. In some diseases, such as typhoid fever or hepatitis B, a chronic carrier state may develop in which the individual may be infective for a long time, possibly years, while showing no symptoms of illness. However, the microorganisms that most commonly cause HAIs, such as E. coli, Klebsiella, Enterobacter, and Pseudomonas spp., do not demonstrate the same patterns of infectivity or evoke the protective immune responses that typhoid fever or hepatitis B does.
Asymptomatic or subclinical carriers may be the more important source of infection than the clinically infected individual. The staphylococcus carrier provides a classic example of the asymptomatic dissemination of infectious organisms; in this instance, the site of dissemination may be the anterior nares or, at times, the skin. Similarly, the site of asymptomatic streptococcal carriage may be in the pharynx, perianal area, or vagina.
The source of an infection may be an atypical case of a specific disease whose clinical course has been modified by therapy, vaccine (as in measles), or prophylaxis (such as the use of immune serum globulin in hepatitis A). Animals also may provide a source of infection, although this is of less concern in the healthcare settings.
Additional characteristics of the agent that may affect their ability to produce disease are the production of virulence factors/enzymes, antigenic shift and drift (such as seen by Influenza A), and development/acquisition of antimicrobial resistance (via plasmid or gene mutation).
The increase in antimicrobial resistance has had a dramatic affect on HAIs. Changes in antimicrobial sensitivity may make therapy difficult; it can result in an increasing prevalence of the resistant strain, reduce the necessary infecting or colonizing doses of the organism in those receiving drugs to which these strains are resistant, increase the numbers of organisms disseminated from persons colonized with these strains, and potentially increase the frequency of HAIs due to this more resistant strain [1].
Reservoir, Source, and Portal of Exit
All organisms have a reservoir and a source; these may be the same or different, and it is important to distinguish between these potentially different sites if control and/or prevention measures are to be directed at this aspect of the chain of infection. The reservoir is the place where the organism maintains its presence, metabolizes, and replicates. Viruses generally survive better in human reservoirs; the reservoir of gram-positive bacteria is usually a human, whereas gram-negative bacteria may have either a human or an animal reservoir (e.g., Salmonella) or an inanimate reservoir (e.g., Pseudomonas in water).
The source is the place from which the infectious agent passes to the host, either by direct or indirect contact, droplet, airborne, common vehicle, or a vector as the means of transmission. Sources also may be animate or inanimate. The source may become contaminated from the reservoir. For example, a reservoir for Pseudomonas spp. may be the tap water in a hospital; however, the source from which it is transmitted to the patient may be a humidifier that has been filled with contaminated tap water.
The portal of exit for organisms from humans usually is single, although it may be multiple. In general, the major portals of exit are the respiratory and gastrointestinal tracts and the skin and wounds. Blood also may be the portal of exit, as in hepatitis B or human immunodeficiency virus (HIV) infections. However, depending on the organism, any bodily secretion or excretion can be infectious.
Transmission
Transmission, the second link in the chain of infection, describes the movement of organisms from the source to the host. Spread may occur through one or more of five different routes: contact (either direct or indirect), droplet, airborne, common vehicle, and vectorborne (described later based on CDC Guidelines for isolation precautions in hospitals) [56]. An organism may have a single route of transmission, or it may be transmissible by two or more routes. M. tuberculosis, for example, is almost always transmitted by the airborne route; measles is primarily a contact-spread disease but may also be transmitted through the air; salmonellae may be transmitted by contact or by the
P.20
common-vehicle, airborne, or vector-borne routes. Thus, in defining the route of transmission, although one route may be the obvious one involved in an HAI problem, another route also may be operative. Knowledge regarding the route of transmission for a specific pathogen can be very helpful in the investigation of an HAI problem. Such information can point to the source and may allow control measures to be introduced more rapidly.
Contact Transmission
Contact transmission is the most important and frequent mode of transmission of HAI pathogens. Contact transmission can be divided into two subgroups, direct-contact transmission and indirect-contact transmission [56]. Direct-contact transmission involves a direct body surface-to-body surface contact and physical transfer of microorganisms between a susceptible host and an infected or colonized person as occurs when a person turns a patient, bathes a patient, or performs other patient-care activities that require direct personal contact. Direct-contact transmission also can occur between two patients with one serving as the source of the infectious microorganisms and the other as a susceptible host.
Indirect-contact transmission involves contact of a susceptible host with a contaminated intermediate object, usually inanimate, such as contaminated instruments, needles, or dressings, or contaminated hands that are not washed and gloves that are not changed between patients. The intermediate object may become contaminated from an animate or inanimate source. An example is the transfer to susceptible hosts of enteric organisms on an endoscope that initially became contaminated when brought in contact with an infected patient (the index patient). Examples of organisms that can be transmitted via contact are VRE and MRSA.
Droplet Transmission
Droplet transmission, theoretically, is a form of contact transmission. However, the mechanism of transfer of the pathogen to the host is quite distinct from either direct- or indirect-contact transmission. Thus, in the 1996 Guidelines for Isolation Precautions in Hospitals, droplet transmission was considered a separate route of transmission [56]. Droplets are generated from the source person primarily during coughing, sneezing, talking, and performing certain procedures such as suctioning and bronchoscopy. Transmission occurs when droplets (large-particle droplets, >5 µm in size) containing micro-organisms generated from the infected person are propelled a short distance through the air and deposited on the host's conjunctivae, nasal mucosa, or mouth. Transmission via large-particle droplets requires close contact between source and recipient persons, because droplets do not remain suspended in the air and generally travel only short distances, usually 3 feet or less, through the air. Because droplets do not remain suspended in the air, special air handling and ventilation are not required to prevent droplet transmission (as opposed to airborne transmission). Examples of pathogens transmitted by the droplet route are Bordetella pertussisand Neisseria meningitides
Airborne Transmission
Airborne transmission occurs by dissemination of either airborne droplet nuclei (small-particle residue [≤5 µm in size] of evaporated droplets containing microorganisms that remain suspended in the air for long periods of time—hours or possibly days) or dust particles containing the infectious agent. Microorganisms carried in this manner can be dispersed widely by air currents and may be inhaled by a susceptible host within the same room or over a longer distance from the source-patient, depending on environmental factors; therefore, special air handling and ventilation are required to prevent airborne transmission. Microorganisms transmitted by airborne transmission include M. tuberculosis, rubeola, and varicella viruses (including disseminated zoster) [56]. In the last several years, there has been controversy as to whether other emerging pathogens/diseases (SARS, smallpox) might be transmitted via airborne route (which would have implications for isolation and personal protective equipment) [57,58,59]. The airborne route of transmission is more frequently assumed to be the route of an infection than is the case [1]. Creation of an infectious aerosol is more difficult than is usually recognized.
Common-Vehicle Transmission
In common-vehicle-spread infection, a contaminated inanimate vehicle, such as food, water, medications, devices, and equipment, serves as a vector for transmission of the agent to multiple persons [56]. The susceptible host becomes infected after contact with the common vehicle. This transmission may be active if the organisms replicate while in the vehicle, such as salmonellae in food, or passive if the organisms are passively carried by the vehicle, such as hepatitis A in food. Other types of common vehicles include blood and blood products (hepatitis B and HIV), intravenous fluids (gram-negative septicemia), and medications (gram-negative septicemia, fungal infections) in which units or batches of a product become contaminated from a common source and serve as a common vehicle for multiple infections.
Vector-Borne Spread
vector-borne transmission occurs when vectors such as mosquitoes, flies, rats, and other vermin transmit microorganisms [56]; this route of transmission is of less significance in U.S. hospitals than in other regions of the world.
Host
The third link in the chain of infection is the host. Disease does not always follow the transmission of infectious agents to a host. As previously discussed, various agent
P.21
factors play a part; similarly, a variety of host factors must also be surmounted before infection occurs and disease develops. Host factors that influence the development of infections are the site of deposition of the agent and the host's defense mechanisms, referred to as immunity, either specific or nonspecific.
Portal of Entrance
Sites of deposition include the skin, mucous membranes, and respiratory, gastrointestinal, and urinary tracts. Staphylococci spp need a minute breach in the integrity of the skin to gain entrance to the body. Mechanical transmission may occur through the normal skin, as with hepatitis B or HIV viruses on a contaminated needle or in contaminated blood. Abnormal skin, such as a preexisting wound, may be the site of deposition of organisms such as Pseudomonas aeruginosa Mucous membranes may be the site of entrance, as the conjunctiva is for adenovirus.
Another site of deposition is the respiratory tract. The exact area of deposition depends on the size of the airborne particle and the aerodynamics at the time of transmission. Generally, particles ≥5/lm in diameter will be deposited in the upper respiratory tract, whereas those <5/lm in diameter will be deposited in the lower respiratory tract.
Infectious agents may gain entrance to the body through the intestinal tract by means of ingestion of contaminated foods or liquids, contaminated supplemental feedings, and contaminated medications or through contaminated equipment, such as endoscopes inserted into the intestinal tract. The urinary tract may become infected from contaminated foreign objects such as catheters or cystoscopes inserted into the urethra or by the retrograde movement of organisms on the external surface of a catheter inserted into the bladder.
Organisms may gain entrance into the host via the placenta, as occurs in rubella and toxoplasmosis. Transplantation is another method by which microorganisms enter the host; infection may follow renal transplantation if the donated kidney is infected with cytomegalovirus.
An organism may colonize one site and cause no disease, but the same organism at another site may result in clinical disease. E. coli, for example, routinely colonizes the gastrointestinal tract and under normal circumstances does not cause disease; however, the same organism in the urinary tract may cause infection. S. aureus may colonize the external nares without any evidence of disease, but when the same organism colonizes a fresh surgical wound, an SSI may develop.
Nonspecific and Specific Defense Mechanisms
Humans have extensive nonspecific and specific defense mechanisms to protect against infection [1]. Nonspecific defense mechanisms include the skin, mucous membranes, and certain bodily secretions (tears, mucus, enzymes). The local inflammatory response provides another nonspecific host defense mechanism. Other nonspecific protective mechanisms include genetic, hormonal, nutritional, behavioral, and personal hygiene factors. Age, as influenced by these nonspecific factors, is associated with decreased resistance at either end of the spectrum; the very young and the very old frequently are more susceptible to infection. Surgery and the presence of chronic diseases, such as diabetes, blood disorders, certain lymphomas, and collagen diseases, alter host resistance.
Specific immunity results from either natural or artificially induced (i.e., vaccines, immunoglobulin) events. Over the last several decades, medical therapies for conditions such as cancers, solid organ/bone marrow transplantation, and HIV have increased the population of hosts with significant immunosuppression. These hosts have added to the challenges of HAIs.
Host Response
The spectrum of the host's response to a microorganism may range from a subclinical (or inapparent) infection to a clinically apparent illness, the extreme being death. The host may become a carrier of the organism. The clinical spectrum of disease varies from mild to a typical course (although a disease may typically be mild) to severe disease and possible death. The degree of host response is determined by both agent and host factors and includes the dose of the infecting organism, its organ specificity, the pathogenicity of the infecting organism, its virulence and invasiveness, and its portal of entry. Host factors include the quantitative and qualitative level of the specific and nonspecific immunologic factors previously discussed.
The same organism infecting different hosts can result in a clinical spectrum of disease that is the same, similar, or different in various individuals. For example, many cases of what appears to be the outbreak disease may meet the clinical case definition, whereas other cases that epidemiologically are related to the same epidemic may not meet the same case definition. They may, in fact, be cases of the outbreak disease but with a different clinical spectrum (as can occasionally be demonstrated by serologic tests). They also may be cases of other diseases occurring concurrently with the outbreak.
Environment
Patients at healthcare facilities often are confined to a hospital bed and surrounded by multiple medical devices, equipment, and environmental surfaces. Thus, there is concern that the facility environment may play a role in the chain of infection. At times, too much emphasis is placed on the role of the environment; for example, it is inappropriate to take environmental cultures routinely throughout a hospital. However, in investigating HAIs, it may be appropriate to obtain environmental cultures as suggested by the circumstances of the specific problem under investigation [1,16,60]. A compromise and a healthy
P.22
respect for the environment are needed with maintenance that does not deliberately promote the transmission of disease-causing agents to hosts but without excessive control measures that impose unnecessary and ineffective actions on the hospital staff and a consequent loss of efficiency and effectiveness, and a wasting of resources such as personnel time and money [1,16].
Some environmental factors can influence all of the links in the chain of infection, whereas others are more limited in their range of action. Humidity, for example, can influence a multiplicity of factors; it can affect the persistence of an agent at its source, its transmission through the air, and the effectiveness of a host's mucous membranes in resisting infection.
Replication of the agent at its reservoir may depend on certain substances in the environment. The agent's survival is influenced by the temperature, humidity, pH, and radiation at its reservoir or source.
The transmission of agents will be affected by environmental factors such as temperature and humidity as mentioned earlier. Airborne transmission is influenced by air velocity and the direction of its movement. The stability and concentration of an aerosol are directly related to environmental factors. In winter, people tend to be indoors with closed windows and reduced air circulation, and this increases the risk of airborne disease; in summer, room air is air conditioned or diluted with outside air. In outbreaks associated with common vehicle transmission, the temperature of the environment will influence the level of contamination in the vehicle.
The host's resistance mechanisms are affected by environmental factors; for example, in an excessively dry atmosphere, mucous membranes become dry and are less able to protect against microbial invasion. Also, the host's behavioral patterns are influenced by temperature.
Ultimately, the reason to discuss the links in the chain of infection and their modulators, for example, the environment, in such detail is that once these factors have been examined, the most appropriate methods of control and prevention can be determined.
Conclusions
The intent of this chapter was to review some of the basic epidemiological concepts and methods with an emphasis on HAIs. Many of the concepts are expanded upon in later chapters. However, the reader may wish to seek additional resources and references on specific general epidemiology concepts.
References
P.23