William R. Jarvis
Although healthcare-facility infection control (IC) programs have been shown to be effective in reducing the healthcare-associated infection (HAI) rate, endemic and/or epidemic infections associated with the delivery of healthcare continue to occur [1]. It has been estimated that only approximately one-third of all HAIs are preventable [2]. However, IC programs at U.S. hospitals prevent only ~6% of these HAIs because of incomplete implementation of recommended control measures [2,3]. The goals of any healthcare facility IC program should be to educate healthcare workers (HCWs) on the recommended measures to prevent and control HAIs; conduct active, prospective HAI surveillance; analyze HAI surveillance data to identify endemic or epidemic infection problems warranting further investigation; conduct epidemiologic investigations to identify the source of these problems; and implement control measures and assess their efficacy in preventing and controlling these problems. The purpose of this chapter is to describe the epidemiology of endemic or epidemic HAIs, discuss criteria for determining whether to investigate endemic or epidemic HAIs, and outline the systematic approach to such investigations.
Defining Endemic or Epidemic HAIs
Endemic infections are defined as sporadic infections that constitute the background rate of infection at the healthcare facility; the rate of such infections usually fluctuates from month to month, but overall is not statistically significantly different from the background rate of these infections (see Chapters 6, 30). Of all HAIs, endemic infections account for the majority of infections and are the focus of most IC activities. The predominant pathogens and sites of endemic infection are somewhat similar at different types of healthcare facilities but do differ based on the mix of patients (including underlying diseases and severity of illness) and types of procedures performed and devices used (see Chapter 30) (Tables 7-1 and 7-2) [4,5]. Endemic infections can change in type (pathogen, site, or both) and/or rate as the result of a variety of factors, including opening or moving to a new facility, introducing a new or expanding existing clinical services or specialties (e.g., bone marrow or organ transplant, neonatal intensive care unit [ICU], surgical or medical subspecialty), introducing new diagnostic methods (e.g., laboratory or radiology), and so on.
Most endemic HAIs result from breaks in aseptic technique, most commonly from person-to-person transmission via transient HCW hand carriage of the colonizing or infecting pathogen. Numerous studies have documented that HCWs often fail to wash their hands before and between contacts with patients [6]. Nevertheless, investigating endemic problems transmitted in this way focuses attention on the importance of general IC recommendations (including identifying and isolating infectious patients, HCW hand hygiene, environmental cleaning and disinfection,
P.92
and existing guidelines to prevent these infections) and can be associated with a reduction (often transient) in the rate of these infections. Because many endemic infections are preventable, investigation of such problems by the ICP or hospital epidemiologist may be warranted if the rate of endemic infection is gradually increasing at the institution or the rate of these infections is higher than expected, than reported in the literature, or than reported from other similar institutions.
|
TABLE 7-1 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Epidemic infections are defined as the occurrence of infection at a rate statistically significantly higher than the background rate of such infections; recognized infection clusters often are unexpected and involve either an unusual organism or an organism with an unusual antimicrobial susceptibility pattern (see Chapters 9 and 15). Recognition of clusters of common organisms with common antimicrobial susceptibility patterns may be difficult because they merge with the existing endemic infections. Often IC personnel attempt to determine whether a cluster represents an outbreak based on numerator data only. In such a situation, it is difficult if not impossible to determine whether a detected cluster represents an outbreak unless it involves either a very rare organism (e.g., Vibrio cholera diarrhea) or a common organism with an unusual antimicrobial susceptibility pattern (e.g., vancomycin-resistant Staphylococcus aureus).
|
TABLE 7-2 |
|||||||||||||||||||||
|
|||||||||||||||||||||
Determination of an outbreak should not be based on numerator data alone. Although one episode of nosocomial malaria or cholera in a U.S. hospital represents an epidemic, one cannot determine whether a cluster of nosocomial S. aureus bloodstream infections (BSIs) represent an outbreak unless one can calculate and compare the S. aureus BSI rates in the time periods during and before the cluster. Because of the abrupt nature of epidemics and the feeling that most such outbreaks are preventable, in most situations epidemics warrant investigation.
Recognizing Endemic or Epidemic HAIs
Surveillance is the cornerstone for rapid recognition of endemic or epidemic HAIs (see Chapter 5). In order to detect either endemic or epidemic HAIs, one must have a surveillance method in place to detect such infections. Because the majority of HAIs arise in severely ill patients exposed to invasive devices (e.g., patients in intensive care units) or those undergoing surgical procedures, surveillance for infections in these populations is most important. If no systematic surveillance system is in place, many clusters
P.93
may not be recognized and if they are recognized, it is impossible to determine whether the cluster identified is endemic or epidemic unless appropriate denominator data are obtained and rates are calculated and compared.
If active prospective surveillance using standard definitions and methodologies is not conducted, it may be necessary to carry out a specific retrospective study to determine the background rate of infection before one can determine whether the infection problem detected or recognized is endemic or epidemic [7,8,9,10,11,12,13,14,15,16,17]. In institutions in which active prospective surveillance is not being conducted in the area, when a cluster is detected, a retrospective study must be undertaken to attempt to reconstruct the current and past rates of such infection in the area. Only then can one differentiate between an endemic and epidemic problem.
The early detection of infection clusters and the determination of whether the cluster identified is endemic or epidemic are most readily done if active surveillance for such infections and calculation of infection rates in the area are part of an ongoing, validated process. Once the cluster is identified, the investigator next determines whether the problem is endemic or epidemic in nature.
Differentiating endemic from epidemic HAIs requires review of how the numerator and denominator are collected, validation of the accuracy of these data, and assessment of whether factors exist that might influence either the numerator or denominator data (i.e., surveillance artifact) (Table 7-3). Care must be taken to ensure that surveillance artifact is not leading to an erroneous conclusion that either the number of HAIs or the HAI rate is increasing. A wide variety of factors that can inflate or deflate the numerator or denominator data can lead to surveillance artifact; they include changes in the definitions used to ascertain HAIs, HAI detection methods—including surveillance or laboratory methods—changes in the populations of patients served, devices used, or procedures performed.
Because surveillance artifact influencing either the numerator or the denominator data can prejudice HAI rate comparisons, it is important that the accuracy of these data be determined before rate comparisons are made. Although it is tempting to quickly begin comparative epidemiologic studies to ascertain the source and risk factors for infection, it is imperative that sufficient time be taken before embarking on such analyses to ensure the accuracy of the numerator and denominator data used to demonstrate that the HAI rate is increasing. Otherwise, one will be misled and devote precious personnel resources to conducting an investigation of a less urgent, nonemerging or nonimportant problem.
|
TABLE 7-3 |
|
|
Numerator data often are the first finding that leads to a belief that there is an epidemic. Thus, confirmation of the validity and accuracy of the numerator data is an important first step. To be consistent, the numerator data should be obtained using the same definition over the time periods being compared. For instance, if the ICP has changed the HAI definitions used, it may appear that the HAI rate also changed when it did not; case ascertainment has altered, and reclassification of the HAIs using the “old” surveillance definitions may show no change in the number of HAIs or the HAI rate. Even if the same surveillance definitions are used, for some sites of infection, it may still be necessary to validate the accuracy of the surveillance data.
For instance, if BSIs are being classified by the ICP as primary or secondary BSIs based on microbiology data, but data concerning the presence of a catheter in the patients with BSIs are not being collected, there may be misclassification of BSIs. If a review of the surveillance definitions used for case ascertainment suggests that the sensitivity and/or specificity of the definitions used is low or allows for considerable subjectivity, it may be necessary to validate the accuracy of the case-ascertainment method by having several members of the IC team review each “case” independently and ensure its accuracy.
Because the existence of an outbreak usually cannot be determined by evaluation of the numerator data alone, one needs to be equally precise in ensuring the validity and accuracy of the denominator data that will be used to calculate the HAI rates in the epidemic and pre-epidemic periods. Selection of the denominator to use in calculating HAI rates is critically important. Previous studies have shown that for ICU patients, the duration of a patient's ICU stay, device (e.g., central venous or urinary catheters or mechanical ventilation) exposure and the duration of that exposure, and severity of illness all contribute to the patient's HAI risk [10,11,12,13].
Thus, these important confounding variables must be controlled for by using ICU-specific, device-specific denominator data (e.g., the number of urinary catheter days for the surgical ICU) if valid HAI rate comparisons are to be made [8]. Similarly, use of a surgical patient risk index that attempts to control for some of the most important factors determining a surgical patient's infection risk, such as the type and duration of the surgical procedure, severity of illness, and wound class, is essential if valid HAI rate comparisons are to be made in surgical patients [14,15].
P.94
In most instances, one can determine whether a perceived problem is an endemic or epidemic problem only by comparison of HAI rates during the suspected epidemic and pre-epidemic periods. Care should be taken when comparing the HAI rate at one's own institution with that reported either in the literature or from other institutions. Unless the surveillance methods used (including HAI definitions and case-ascertainment methods), the populations served, the invasive device used, and types and number of procedures performed are similar and the type of denominator data used for rate calculations are the same, it may be misleading to compare one institution's rate with that of another institution. Thus, because of the large number of confounding variables that can influence the numerator, denominator, or HAI rate, one often is safest comparing rates over time at one's own institution. However, even then, one must be cautious in making such comparisons, ensuring that the comparisons include both similar numerator and similar denominator data.
For example, comparison of nosocomial BSI rates in an ICU to which only medical patients are admitted with rates in another unit at that same institution to which both medical and surgical patients are admitted or comparison of nosocomial BSI, UTI, SSI, or pneumonia rates among patients in a medical ICU to rates in patients in surgical ICUs may be misleading. Because both the mix of patients and the devices used can influence HAI rates, controlling for these factors by making comparisons with similar populations and controlling for the duration of patients' stays and devices used is necessary if valid rate comparisons are to be made. These factors are most easily controlled for by making HAI rate comparisons in a particular unit or population within an institution over time.
Differentiating Endemic from Epidemic Infections
No one definition can be used in all situations to differentiate endemic from epidemic infections or define epidemic infections. Depending on the severity of the problem and the administrative and other pressures present, one may be able to precisely and accurately define the numerator and denominator data or perform a “quick and dirty” analysis to determine whether an epidemic is present and an investigation must be initiated. Once one is confident of the validity and accuracy of these data, one is in a position to compare the rate of the infections or other adverse events recognized in the cluster to the background rate of such events and determine whether the cluster is an epidemic or endemic occurrence. If one can confirm the accuracy of the numerator data but not the denominator data, one may be forced to use a less accurate denominator (e.g., the number of patients or the number of patient-days) to calculate the “rate.”
It must be realized that depending on the variation in the possible confounding variables, the findings of the comparison of the “epidemic” to “pre-epidemic” rates (i.e., statistically significantly higher, lower, or unchanged) may be misleading. Furthermore, it may sometimes be difficult, even given the appropriate numerator and denominator data, to determine whether a cluster of infections is endemic or epidemic. For instance, if an endemic rate of infection continues to rise, at some point it may be considered epidemic in nature if recent experience is compared with suitable remote baseline data. In other situations, the lack of background rate data (e.g., no surveillance conducted, new patient population or new diagnostic test introduced at the facility) may preclude the possibility of making rate comparisons. However, in most situations, it is possible to either determine or estimate the background rate and to decide whether the detected cluster of HAIs represents an epidemic or endemic problem. The recent development of a variety of computer statistical software packages has simplified the calculation and comparison of HAI rates; however, the ease with which it is possible to make such comparisons has further highlighted the importance of ensuring the accuracy and validity of the numerator and denominator data used.
Some researchers have suggested conducting prospective surveillance and using threshold programs to determine when the HAI rate increases above a certain level that warrants further investigation. Such programs were used in the 1970s in the National Nosocomial Infections Surveillance system and were found to be unreliable because of normal variations in the HAI rate at most institutions. Although it may be easy to establish a high HAI rate above which further investigation is indicated, the sensitivity of such a system would be low. To date, it has been impossible to design a sensitive and specific threshold program that would identify all epidemics when they occur yet not also highlight other nonoutbreaks as clusters requiring investigation. It has been equally difficult to develop a sufficiently sensitive and specific threshold program for the detection of clusters requiring further investigation.
More recently, data have been published from national surveillance and other studies providing HAI rate distributions for a variety of groups of patients [8,9,10,11,12,13,14,15]. These data can be used to compare the HAI rate at one's own institution and assist in determining whether the endemic HAI rate is too high and should be investigated or whether the HAI rate documented at the institution is significantly higher than at one's own institution in the past and significantly higher than national averages. These national benchmark data are most useful for evaluating one's endemic HAI rate rather than for determining whether an epidemic is in progress. Despite these data, there are many populations for which there are no published benchmark HAI rates. Furthermore, some of the published benchmark rates have not carefully controlled for intrinsic or extrinsic risk factors. Thus, one often is forced to decide whether to further investigate an epidemic or endemic problem based on the limited data one has at the institutional level.
P.95
Deciding when to Conduct an Investigation
Making the decision to conduct an epidemiologic investigation and determining the extent of that investigation depend on a number of factors. These factors differ at the institutional, local health departmental, state health departmental, federal governmental (Centers for Disease Control and Prevention [CDC]), and private consultant levels. In all situations, factors that influence the decision to initiate the investigation include whether the cluster is endemic or epidemic in nature, the morbidity and mortality associated with the infection cluster, and staff availability and expertise.
If a decision is made within the hospital that an investigation is warranted, the necessary staff should be mobilized and the investigation initiated. If the staff's availability or expertise is insufficient to conduct the investigation, hospital personnel can call in a private consultant, the local or state health departments, or the CDC for assistance. Regardless of the outside source, the exact nature and scope of the investigation desired by hospital personnel and what the outside consultant or organization is offering to do should be well defined and outlined before a request for assistance is extended. In addition, because of the changing epidemiology of HAIs and the possible complexity of an investigation, combined epidemiologic and laboratory investigations may be indicated and the capability to perform these investigations should be documented before an invitation is initiated or as soon as the need for such expertise is realized.
At the federal level (Division of Healthcare Quality Promotion, CDC), the decision to initiate an epidemiologic investigation at a healthcare facility is based on the public health importance of the problem, whether the cluster may represent a nationwide problem (e.g., intrinsic product contamination or a serious problem related to a newly introduced medical device), the morbidity and mortality associated with the cluster, the extent to which the investigation may advance knowledge of healthcare epidemiology and infection control, and staff availability [16,17]. Moreover, because combined epidemiologic and laboratory investigations are frequently most useful, the ability to obtain and genetically type the colonizing or infecting isolates associated with the cluster may influence the decision to collaborate in the investigation. Because the CDC is a nonregulatory agency, its collaboration in the investigation requires the approval of and invitation from both the healthcare facility's IC department and administration and the local and/or state health department.
Clusters of infection caused by very unusual organisms, those associated with great morbidity or mortality, and those of epidemiologic importance (introduction of a new multidrug-resistant pathogen) may warrant initiating an investigation without carefully comparing HAI rates in the epidemic and pre-epidemic periods. These clusters include infections caused by very unusual or never previously reported pathogens (i.e., vancomycin-resistant S. aureus, Rhodococcus or Nocardia sp infections in surgery patients), ≥2 infections caused by organisms usually indicative of a carrier (e.g., Group A streptococcus), isolated infections caused by a multidrug-resistant HAI pathogen (e.g., vancomycin-resistant enterococcus, multidrug-resistant Mycobacterium tuberculosis) that, if not controlled, could become endemic or dissemination of a clonal organism, which often suggests an eradicable common source [16,17,18,19,20,21,22,23,24,25,26].
Once a cluster is identified and the decision is made to conduct an investigation, the extent of the investigation must be determined. If the cluster is associated with little morbidity and mortality and involves only colonization or infection of a small number of patients, a brief investigation may be performed. In such circumstances, the possible case-patients' medical records are reviewed, a hypothetical mode of transmission is identified, and interventions are implemented. If a full-scale investigation is to be initiated, a systematic approach should be taken (discussed later). In general, it is recommended that cultures of HCWs, products, solutions, or environmental sources be directed by the epidemiologic data.
Documentation of an epidemiologic association between the outbreak and a product, device, or HCW, with subsequent culture confirmation is preferred to widespread culturing of HCWs, solutions, or equipment with the hope that the source will be identified. Widespread culturing of animate or inanimate objects with the object of serendipitously identifying the source is wasteful of the time of both IC and laboratory personnel (see Chapters 10 and 20). On the other hand, if one cannot preserve the area in which the outbreak is taking place and the inanimate environment is going to be cleaned or the product reprocessed, selected cultures should be obtained before the scene of the outbreak is altered. In addition, it may be important to immediately interview HCWs in the area where the outbreak has occurred because practices may change or recall bias may be introduced if these persons are interviewed days or weeks later.
Last, careful consideration should be given to whether the unit/ward/area should be closed or surgery or other procedures discontinued. If a large number of deaths are associated with the outbreak in one area, closing the area may be warranted. However, if such a decision is made, IC personnel should realize the seriousness of the message being sent. Before closing the area, there should be discussion, and a consensus should be developed concerning what criteria must be met before reopening the area. The decisions to close and reopen an area can be made only at the institutional level. Regardless of whether the unit is closed, IC and administrative personnel should ensure that all potentially relevant materials (devices, medications, solutions, etc.) are saved and quarantined for future evaluation.
P.96
If there is serious morbidity or a large number of deaths or if many patients are affected, serious consideration should be given to identifying a spokesperson, best coordinated through IC and public relations staff, who will regularly update both appropriate institutional (e.g., staff and patients) and external (e.g., regulatory, governmental, or media) personnel. The spokesperson should present enough information to assure others that an appropriate institutional investigation is being conducted but should not prematurely reveal preliminary information. All public inquiries should be directed to the identified spokesperson to ensure that one voice is being heard and conflicting stories do not emerge. Open and honest communication with the local or national media is preferable to disseminating mis-information or refusing to speak to them.
It should be remembered that state and federal laws require that healthcare facilities notify public health authorities of selected adverse events (see Chapter 18). Because laws differ in each state, IC personnel should consult the law in the state where their healthcare facility is located. Any outbreaks that may have a public health impact at the county, state, or national level should be reported to local, state, or federal health officials. Any outbreaks associated with intrinsically contaminated or defective products (including solutions, blood or blood products, or devices) should be reported to the Food and Drug Administration (FDA) through the Med-Watch Program (1-800-FDA-1088). If such an outbreak involves many healthcare facilities in more than one state, federal public health agencies are responsible for the investigations.
Conducting an Epidemiologic Investigation
Saving Critical Materials
The first step in any investigation is to ensure that critically important isolates and/or materials that may be associated with the outbreak are saved. At the first sign of an outbreak, the IC staff should get in touch with the director of the microbiology laboratory and request that any of the infecting organisms from current and past possible “cases” be saved. Laboratory personnel also should be alerted to keep any subsequent isolates of the outbreak strain that may be recovered during the investigation. If this is not done, laboratory confirmation of the subsequent epidemiologic findings in the form of isolate typing will be impossible. In addition, if there is the possibility that an intrinsically or extrinsically contaminated product device may be the source of the outbreak, such solution or devices should be immediately quarantined.
For example, if there is an unusual cluster of BSIs with the same organism(s) in one unit over a short period of time (1 to 5 days), a contaminated product should be seriously considered; it may be prudent to collect all medications, multidose vials, or solutions from the involved area and immediately move them to an area where they can be preserved and protected. If extrinsic contamination of a product or solution is suspected, it may be useful to have personnel save products or solutions used in the area as the epidemiologic investigation is being conducted so that these solutions/products can be cultured or studied if epidemiologic studies establish an association between them and the adverse outcome.
The Source, the Pathogen, the Host, and the Mode of Transmission
Epidemiologic investigations have similarities and differences. The approach is similar, although the risk factors assessed may be different. For this reason, standardized forms usually are not used in such investigations; instead, the data collected during each epidemiologic investigation is individualized depending on the data available on the pathogen, host, and known or suspected modes of transmission. Although the approach may differ slightly for each investigation initiated, in general, epidemiologists use a fairly standard system (Table7-4). Four major areas assessed in any epidemiologic investigation include the source(s), the pathogen(s), the host, and the mode of transmission. Factors in these four areas contribute to the outbreak, and modification of ≥1 of these elements usually can terminate the outbreak. For infection to occur, sufficient organisms must be present, the host must be susceptible to the organism, and opportunity must exist for the host to have contact with the organism. The goal of the epidemiologic investigation is to determine which of these factors is the most important in causing the outbreak and which can most easily be modified to interrupt transmission.
It is important to understand the pathogen and the ecologic niche the organism prefers. For instance, Sternotrophomonas maltophilia or Burkholderia cepacia are increasingly common HAI pathogens that often can be traced to water sources [22,23,24]. Malassezia sp are lipid-loving organisms that usually infect patients receiving intralipid. A nosocomialMalassezia pachydermitis outbreak among neonates in a neonatal ICU was traced via colonization of the hands of HCWs via colonization of the ears of their dogs [25,26]. Aspergillussp usually infect immunocompromised patients and can be found in soil and air [27,28]. Acinetobacter or Serratia spp. emerge in situations of high antimicrobial pressure, and the latter organisms have been traced to intrinsic or extrinsic contamination of antiseptics [29,30]. Legionella sp are typically associated with water sources and primarily infect immunocompromised hosts [31]. Some salmonella
P.97
and most Group A streptococcus infections are traced to personnel carriers [32,33].
|
TABLE 7-4 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
In general, molecular typing of the infecting pathogen can provide valuable information [34] (see Chapter 10). If the infecting organisms are identical (i.e., clonal), the chances are great that there is a common source (e.g., a solution, device, or HCW) and that an epidemiologic investigation can identify the source, the source can be removed, and the outbreak can be terminated. On the other hand, if molecular typing of the infecting organisms shows that they are not identical (i.e., nonclonal), most likely the organism is being introduced from multiple sources and/or transmitted from person to person via HCW hands. An epidemiologic investigation may pinpoint factors increasing the risk of infection (failure to rapidly identify colonized/infected patients and isolate them, failure of HCWs to perform hand hygiene, etc.), but the empiric reinforcement of such practices as isolation of patients and hand hygiene may terminate the outbreak without a more complete investigation. Thus, reviewing the microbiology of the infecting pathogen can provide valuable clues to the likely source of an infection and facilitate generation of hypotheses about that source.
Next, host factors need to be evaluated. For the host to become infected, there must be sufficient numbers of organisms, and the patient must be susceptible to infection. In some instances, host susceptibility is related to age or immunosuppression that is condition specific (e.g., low birthweight neonates), disease specific (e.g., human immunodeficiency virus–infected patients) or condition or medication induced (patients with hematologic malignancy or bone marrow or organ transplant patients). In other instances, the host has become susceptible to colonization or infection because of exposure to medical devices or other surgical or invasive procedures.
The next factor to be considered is the mode of transmission. HAI pathogens may be transmitted by contact (direct, indirect, droplet) or a common source, or it may be airborne (droplet nuclei, skin squamae) or vectorborne (see Chapter 1). Although HAI pathogens can be transmitted by the airborne route (e.g., Aspergillus spp, Influenza, M. tuberculosis,measles, or varicella-zoster viruses), the majority are passed by contact, usually necessitating transfer of the organisms from an infected or colonized patient to a susceptible patient via transient HCW hand colonization [18,21,26,29,35] or by droplet (<3 feet) spread (e.g., respiratory syncytial viruses or adenoviruses). With most organisms transmitted by contact, the infected or colonized patient is the source; for some organisms, such as VRE, multidrug-resistant S. aureus, and respiratory syncytial viruses and rotavirus, the environment may play a role in infection transmission by fomites.
The predominant site of HAI also can help the epidemiologist focus the investigation on the most likely
P.98
route of transmission. For instance, a cluster of BSIs is most likely related to an intrinsically contaminated common vehicle or extrinsically contaminated solution or device (including transducers) or to lapses in aseptic technique on the part of HCWs during intravascular catheter device manipulation [36]. Nosocomial pneumonia often is traced to contamination of respiratory therapy equipment or person-to-person transmission of infecting pathogens via HCWs hands [37]. Nosocomial UTIs are typically attributed to contamination during urinary tract manipulation, open urinary drainage systems, or the insertion or maintenance of a urinary catheter [38]. SSIs often are traced to sources in the operating room [16,18,32,39,40,42] or, more rarely, to the postoperative ICU [41] or the preoperative ward.
When initiating an investigation, a review of the source, pathogen, host, and mode of transmission can help guide the direction of the epidemiologic investigation. The more that is known about these areas, the more focused the investigation can be. In contrast, the less that is known about these areas (for instance, the first nosocomial outbreak of Rhodococcus bronchialis among surgery patients), the broader the scope of the investigation must be at the beginning [18].
Initial “Case” Review
Regardless of whether a brief or detailed epidemiologic study is being carried out, one of the first steps in the investigation of a possible epidemic or outbreak is to review some or all of the medical records of possible case-patients. The purpose of this review is to characterize the population at risk by time, place, and person so that a “case” definition can be developed. If possible, this review should encompass all known possible case-patients. If the number of possible case-patients is large, one may select a sample of them; if either a random or convenience sample is chosen, the reviewer should be aware that bias may thus be introduced and lead to erroneous conclusions.
For infections that have a long incubation period, meaning that the majority of patients experience onset of symptoms after hospital discharge, review of the medical records of only those who are still hospitalized may lead to an underestimation of the extent of the outbreak and may not describe the affected population accurately. Examples of such outbreaks include many insidious postoperative SSIs (caused by, e.g., nontuberculous mycobacteria, Nocardia spp, Rhodococcus spp), S. aureus infections among newborns, and infections with long incubation periods [18,42,43,44]. In outbreaks with long incubation periods, after the case definition is developed, extensive case-ascertainment is necessary.
Line-Listing
During the initial case review, a detailed line-listing should be developed for each patient (Table 7-5). Data collected should include demographic and clinical data, the date of hospital admission, wards/units admitted to and the dates of admission to and discharge from each ward/unit, underlying diseases, date of onset of infection and/or colonization, and, if the outbreak involves infection, whether colonization with the infecting pathogen preceded the onset of infection and at what site infection was noted. Additional information is collected specific to the site of the infection/colonization. For example, with SSIs, information is collected on the pre-, intra-, and postoperative courses of the case-patients; these data should include whether a patient was admitted to an ICU before the surgery, the date and type of each surgical procedure, the HCWs who took part in the surgery, and exposures specific to the surgical procedure, including the type and timing of administration of prophylactic antimicrobials.
For BSI outbreaks, the types of catheters, their dates of insertion, and the duration of intravascular catheterization and the types of intravenous fluids, medications, and monitors should be documented. The epidemiologist or ICPs should collect adequate information to characterize the population at risk sufficiently so that the type of outbreak (i.e., colonization or disease), population (i.e., patients affected), place (i.e., specific wards, ICUs, or surgical areas), and estimated outbreak period can be defined. Sufficient detail should be collected to establish a case definition, taking care to avoid generating an enormous amount of useless information. The primary purpose of this review is to try to identify common characteristics among the patients so that one can describe the time period of the outbreak, the group of patients at greatest risk, and the affected area(s) of the healthcare facility. In the course of this review, investigators should ensure that the HAIs are real and that neither surveillance nor diagnostic artifact has resulted in the “appearance” of an outbreak.
Case Definition
The next step in the investigation of an outbreak is the development of a case definition. Initially, unless the disease is known or the time or place of exposure is well defined, the case definition should be broad. This tentative case definition can be further refined as more information is obtained during the conduct of the investigation. Each case definition should specify the estimated time period of the outbreak, the place or places in the healthcare facility where the outbreak is taking place, and the types of persons who are becoming case-patients. In outbreaks of a pathogen-specific disease, development of the case definition may be relatively simple, for instance, all neonatal ICU patients from June 2 to August 30, 2007, with blood cultures positive for M. pachydermitis [26]. For syndromes for which a direct association with a specific infectious agent cannot be established from the information available (e.g., initial investigations of toxic shock, Legionnaires' disease, or toxic
P.99
P.100
exposures), development of the case definition can be more challenging. In these instances, the case definition should include all of the signs and symptoms common to all of the “tentative cases” reviewed.
|
TABLE 7-5 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
In addition, the case definition may include combinations of signs and/or symptoms common to the majority of the cases reviewed. One such case definition might be any patient in the surgical ICU from January 5 to March 19, 2007, who had a temperature >102°F and a >20 mm drop in systolic blood pressure together with any one or more of the following signs: a white blood cell count >25,000 cells/mm3, a platelet count <20,000/mm3, a bilirubin level of >5 mg/dL, or splenomegaly. When formulating a case definition, a balance must be drawn between sensitivity and specificity. It may be preferable to have a case definition that is very specific rather than overly sensitive; in this way, one can be more certain that each “case” identified is a true case. Furthermore, in subsequent analytic epidemiologic studies, missed “case-patients” may be included in the control or noncase group and create bias toward not finding differences between the case and control or noncase-patients; thus, any significant differences identified are probably even more significant. Alternatively, one may wish to identify “definite,” “probable,” or “possible” case-patients; this approach helps ensure that controls do not include possible or probable case-patients and that case-patients are only those patients most likely to be cases for case-control comparisons.
Case Ascertainment
Having established a case definition, the next step is to conduct extensive case ascertainment. In this process, the IC staff try to identify all of the cases that may have arisen. All potential sources of information should be examined for possible case-identifying information. If the case definition is microorganism based, usually a careful review of the existing microbiology records is all that is needed to identify case patients. Reviewing microbiology methods may be necessary to exclude the possibility that the organism might not be identified or that specimens are referred to an outside laboratory.
Thus, case ascertainment in organism-based outbreaks of BSIs, UTIs, or most SSIs is simplified. Nevertheless, one must be aware of the possibility of changes in culturing frequency upon case detection [44]. However, depending on the site of infection, outbreaks where culturing bias (i.e., clinicians are less likely to document the infection by culture) may exist; for example, nosocomial pneumonia (for which a review of radiology reports in addition to microbiology records may be necessary) or SSIs caused by multiple pathogens (when it may be necessary to review the records of all patients undergoing the procedure to identify those with signs or symptoms of SSIs from whom cultures were not taken) may require more extensive and rigorous case ascertainment. During the case-ascertainment process, all computerized and noncomputerized data sources (including microbiology, radiology, infection control, pharmacy, operating room, surgery, hemodialysis, other invasive procedures, nursing, or patients' medical records) that may facilitate case ascertainment should be considered.
At the same time that case ascertainment is being conducted, one should review an arbitrarily defined preoutbreak period (usually 6 months to 2 years) for the occurrence of the outbreak adverse event. In this way, one can define the background number of such events. This information will be used to calculate the pre-epidemic rate of the adverse event that will be compared to the epidemic period rate to determine whether an outbreak has occurred or is occurring.
The Epidemic Curve (Time)
Using the information obtained from case definition and case ascertainment, an epidemiologic curve (epi curve) should be drawn. This curve depicts the number of case-patients on the vertical axis, or y-axis, and time on the horizontal axis, or x-axis (Figure 7-1). The line-listing information and epi curve can provide data to facilitate generation of hypotheses about the mode of transmission. The time scale used should be shorter than the presumed incubation period of the adverse event because otherwise person-to-person transmission may appear to be common-source transmission. By plotting the time course of the adverse event during the epidemic and pre-epidemic periods, one can compare the occurrence of the adverse event during the epidemic and background periods, identify clusters of the event, and, based on the shape of the epi curve, generate hypotheses about the mode of transmission.
The shape of the epi curve can suggest the possible mode of transmission. If there is an abrupt increase in the number of adverse events over a short time period, the curve suggests a single exposure to a point source of contamination, such as a contaminated product (Figure 7-1a). In contrast, when the epi curve depicts patients over a more prolonged time course, it is most suggestive of person-to-person transmission (Figure 7-1b). In some situations, there may be more than one mode of transmission operating either sequentially or simultaneously (Figure 7-1c). Additional epi curves depicting the dates of culture or possible exposure also may be informative—particularly when the incubation period is long. It also may be useful to draw epi curves with different time intervals (i.e., day, week, month) on the horizontal axis.
Geographic Assessment (Place)
It is useful to closely examine the place of the outbreak to determine whether there is geographic clustering. Use of spot maps may facilitate recognition of clustering of case-patients that is not otherwise obvious. Do all of the case-patients come from one ward or unit? If all the case-patients
P.101
are from one ward/unit, one can plot graphically the dates of admission and discharge from that unit (or multiple units) for each case-patient and ask whether there is continuous overlap among the patients. If so, this finding would suggest person-to-person transmission from one colonized or infected patient to the next. If all case-patients have an infection that can be transmitted via the airborne route, is plotting the patients by room location consistent with the ventilation system airflow diagrams or direction of airflow demonstrated by smoke tube? If the case-patients are located in several different wards or units, have they all had common exposure to medications, solutions, devices, procedures, or rooms where a procedure may have been performed? If the outbreak is at one surgical site, have all procedures been performed in one operating room or outpatient surgical suite? Geographically plotting the case-patients by a variety of dates, including those of the time of culture, onset of illness, possible exposures, and so forth, may all be useful.
|
Figure 7-1 Example of epidemic curves. a: Distribution of patients with bloodstream infections associated with a contaminated common source medication. b: Distribution of patients with S. aureus skin infections associated with person-to-person transmissions. c: Distribution of patients with gastroenteritis associated with a common contaminated food source followed by patient-to-patient transmission. |
Host Factors (Person)
Further review of the characteristics of the case-patients then is undertaken to try to define the most likely risk factors of infection. What role are specific underlying host factors playing in the outbreak? For instance, are all neonatal ICU patients at risk or only those <1,500 g? Are all surgical patients at risk or only those undergoing cardiac surgery? Are all hematology-oncology patients at risk or only those who have prolonged neutropenia? Each of the case-patients' host characteristics should be reviewed to determine whether there are common features among these patients.
Factors to assess include those intrinsic to the host (i.e., age, gender, race, underlying diseases, and nutritional status) and those extrinsic to the host (e.g., receipt of medications or solutions or environmental exposures—admission to an ICU or procedure room; HCW exposures; exposure to therapeutic measures that might alter host susceptibility, e.g., receipt of antimicrobials; or exposure to invasive
P.102
procedures or to invasive devices, e.g., central venous or urinary catheters, mechanical ventilation, or arterial pressure monitoring).
Refinement of the Case Definition and Reassessment of the Need for Further Investigation
At this point, the basic elements of the simple epidemiologic investigation are complete. IC should refine the case definition based on the data accumulated. Only the essential elements reflecting the time, place, and persons who have experienced the adverse event should be included. For example, if the line-listing shows that all (or nearly all) of the case-patients are <1,500 g and have stayed in the neonatal ICU for a median of 5 days, the case definition might be modified from that of all neonatal ICU patients to those <1,500 g who stayed in the unit for ≥5 days. Careful addition of patients or deletion of those initially thought to be case-patients who do not fit the case definition should take place. It is imperative that the refined case definition be as precise and accurate as possible because the results of subsequent epidemiologic studies are contingent on this case definition.
If the case definition is in error, it may lead to erroneous conclusions. As mentioned previously, an overly restrictive case definition will exclude true case-patients and diminish the likelihood of finding an association between an exposure and the case-patients. In contrast, an overly inclusive case definition will result in the inclusion of noncase-patients in the case-patient group and may obscure possible associations. In most circumstances, it is preferable to begin the investigation by being overly inclusive and become more restrictive as the investigation proceeds. In some investigations, it may not be possible to divide all patients into case-patient and noncase-patient or control populations. If it has not already been done, one may at this point need to define possible and/or probable case-patients and then perform further analyses including and then excluding these patients in the case-patient group.
Occasionally, the review of the case-patients' medical and microbiology records suggests that there is a cluster of positive cultures but that the patients do not have clinical illness [16]. Such findings should alert one to the possibility of a pseudoepidemic, that is, an apparent cluster of positive cultures that may be false positives or contaminants. In the 1960s and 1970s, such outbreaks most often were associated with cross-contamination in the laboratory during manual processing of cultures. In the 1980s and 1990s, clusters of positive cultures increasingly have been associated with cross-contamination via automated systems [16,17]. Pseudoepidemics have been traced to a variety of sources, including intrinsically or extrinsically contaminated antiseptics or culture media, cross-contamination of cultures during manual or automated processing, or intrinsically or extrinsically contaminated blood collection tubes. Whenever there is an unusual cluster of positive cultures, particularly if it is an uncommon or environmental organism and the case-patients do not have clinical signs or symptoms indicating HAI, evaluation of the specimen collection and processing methods to exclude a pseudoepidemic should be initiated.
At this point in the investigation, IC personnel may have arrived at the most likely source and mode of transmission of the outbreak. Because of personnel, time, or other constraints, it may be decided to introduce a number of IC interventions thought likely to reduce or interrupt transmission and then discontinue the investigation. If the outbreak is similar to others previously reported, it may be possible simply to implement the previously documented successful IC measures. If this is done, it is imperative that surveillance be maintained for the adverse event so that one can document the effectiveness of the implemented interventions. If the interventions do not reduce (to acceptable levels) or interrupt transmission, it may be necessary to initiate a more comprehensive comparative investigation.
Is Outside Assistance Needed?
If it is decided that a more comprehensive epidemiologic investigation is warranted because the outbreak (a) has not been terminated by the introduced control measures; (b) is unusual, complex, or associated with substantial morbidity or mortality; (c) is of major public health importance; or (d) represents an opportunity to advance our knowledge and/or understanding of healthcare epidemiology, one should re-evaluate whether the personnel, time, and/or expertise are sufficient to conduct the required comprehensive studies. Such investigations may demand a substantial investment in IC, laboratory, and statistical/computer resources. The comprehensive investigation may involve exhaustive evaluation of potential risk factors; a series of several case-control and/or cohort studies; complex analytic techniques, including multivariate analyses and modeling to control for confounding variables; HCW questionnaires, interviews, or observational studies; epidemiologically directed environmental, product, device, and/or HCW cultures; and implementation of interventions and evaluation of the effectiveness of these interventions.
If the adverse event is serious enough to consider a comprehensive investigation, sufficient resources should be dedicated to the investigation to permit the rapid initiation of the study, identification of the source, and introduction of control measures. If healthcare facility IC personnel add the conduct of such an investigation to their ongoing responsibilities, it may mean that the comparative studies will be conducted over several months. By the time the probable source of the outbreak is identified, it is possible either that the device, solution, or source will have been reprocessed or discarded or that the epidemiologically implicated HCW will no longer be colonized with the infecting strain.
P.103
If outside advice or assistance is desired, a variety of experts is available including personnel at the local or state health departments, colleagues at other healthcare facilities in the same area, experts at nearby academic centers with an interest in healthcare epidemiology, private consultant epidemiologists, or the CDC. State and local health departments may be able to assist in arranging epidemiologic and/or laboratory support. Healthcare facility administrative and IC personnel or personnel at the state health department can request CDC assistance. In CDC-conducted on-site investigations, both epidemiologic and laboratory support is provided at no cost to the healthcare facility.
Regardless of the outside source of assistance, the investigation should be considered and conducted as a collaborative effort with the healthcare facility personnel. Without the assistance of healthcare facility personnel, a comprehensive investigation cannot be efficiently conducted, and it will be these personnel who will need to implement the recommended control measures. Furthermore, local personnel are likely to be much more knowledgeable about available data sources, local IC and other policies and practices, HCW changes, and the particular control measures that can realistically be successfully implemented at their institution.
The Comprehensive Investigation
The first step in conducting the comprehensive investigation is to review the basic investigation's line-listing, case definition, and case-ascertainment methods to ensure that no shortcuts were taken that may influence the comparative studies. Was the case definition based on a review of all possible cases or just a sample? If sampling was used, was it random sampling, which may be representative of all case-patients, or convenience sampling (i.e., those medical records that were readily available)? Could the case review sampling have biased the case definition? Except in very large outbreaks, it is preferable to review the medical records of all potential case-patients.
Occasionally, it is the small number (often one or two) of patients who are otherwise very similar to the case-patients but who did not have the adverse condition who provide insight into the possible cause of the outbreak. Was the case ascertainment complete? If not, it should be extended and repeated. Does the case ascertainment need to be broadened to a review (and/or contacting) of patients who have been discharged from the hospital or to communicating with IC personnel at nearby hospitals or home care? Was the line-listing extensive enough, or is further review of the case-patient medical records necessary to obtain information on additional exposures or clinical findings? Was the denominator used to calculate the rate of the adverse event and to compare the rates of these events in the epidemic and pre-epidemic periods the correct denominator? If the denominator was not correct, is there a way to obtain the correct denominator, or must one estimate the denominator?
For example, if the outbreak involves BSIs in the medical ICU but in the initial investigation one had to use either the number of medical ICU patients or the number of patient-days as the denominator, can one obtain the more appropriate denominator, that is, the number of days medical ICU patients spent on central venous catheters? If these data are not available or have not been collected, it may be unrealistic to try to derive them by reviewing the medical records of all medical ICU patients (assuming that the central venous catheter insertion, presence, and removal dates are well documented in the medical records).
However, it would be possible to estimate these denominator data. One could randomly sample time periods during the pre-epidemic and epidemic time periods and, for several days or a week in the pre-epidemic and epidemic periods, determine for all patients in the medical ICU how many had central venous catheters and what proportion of the ICU days those patients were on such catheters. Then one can determine a “central venous catheter use factor” (i.e., the proportion of patient-days spent on central venous catheters times the number of patients in the unit) for each time period. In this way, one can estimate the number of central venous catheter days in the unit during the epidemic and pre-epidemic periods. Such estimation may be necessary when either the necessary data are not available or review of records containing the needed data would be excessively laborious and/or time-consuming.
Comparative Epidemiologic Studies
Once one has defined the case, performed comprehensive case ascertainment, developed an extensive line-listing, and drafted epidemic curves, one is ready to develop and test hypothesized risk factors and modes of transmission. This can be done through either a case-control or a cohort study (see Chapters 1 and 8). In a case-control study, all or some of the case-patients are compared with a group of patients who did not experience the adverse event (controls) for exposure to the potential risk factors. In contrast, in a cohort study, such exposures are compared among all the patients in the involved area during the specified period.
Numerous factors influence the decision to conduct a case-control or cohort study including statistical and practical considerations. In a cohort study, one can calculate the relative risk or a quantitative measure of the strength of the association between the exposure and the risk of developing the adverse condition (See Chapter 8). In contrast, in a case-control study, one can estimate the strength of the association between the exposure and the adverse condition only by calculating an odds ratio; one can determine only that case-patients were more likely to be exposed to the factor than were controls. Adverse conditions with a high attack rate lend themselves to cohort
P.104
studies whereas conditions with low attack rates are better evaluated through a case-control study. If the duration of the outbreak is short and the number of patients in the unit during the outbreak is small, it may be feasible to conduct a cohort study. On the other hand, if the duration of the outbreak is long (months or years) and the number of patients in the involved unit(s) is large (e.g., >100), it may be impractical to conduct a cohort study, and a case-control study may be more realistic. Depending on the duration of the outbreak, the size of the population at risk, and the extent of the data being obtained from the medical records, a case-control study may take days to weeks whereas a cohort study may require weeks to months.
Cohort Study
In a cohort study, all patients in the unit or units of interest or undergoing the particular procedure of interest are evaluated for exposures to the variable(s) of interest. Such a study can be conducted either prospectively or retrospectively. Because case-control studies are easier to perform, take less time, and tend to be more powerful, retrospective cohort studies are less frequently undertaken to evaluate nosocomial endemic or epidemic problems. However, if the case-control study data are insufficient to determine the cause of the outbreak and/or the case-control study has narrowed the population at risk, carrying out a prospective cohort study may be useful.
Case-Control Study
In contrast, in a case-control study, the case-patients are compared with a similar population without the adverse condition. In a case-control study, selection of the controls is critical. Unless it cannot be avoided, historical controls (i.e., patients present in the unit but in the time period before the outbreak) should not be used. Such patients may not have had the same opportunity for exposure and may not be similar to the population affected. Random selection of patients in the same unit at the same time as the case-patients is preferred. If one is concerned that there could be differences in ≥1 factors influencing the types of exposures of case-patients and controls, a small, focused case-control study focusing on these factors can be done before the comprehensive case-control study. For instance, if by examination of the line-listing, one becomes concerned that birthweight (e.g., <1,500 g) and duration of neonatal ICU stay (i.e., >5 days) influenced the types of exposure of the patients, one can first evaluate these two factors. If they are found to be significantly different between case- and control-patients, selection of controls could be limited only to those neonates who weighed <1,500 g and who stayed in the neonatal ICU >5 days.
There are several other methods to control for such confounding variables (matching, stratification, or multivariate analyses). In matching, one selects controls by matching ≥1 factors (e.g., birthweight, duration of stay, severity of illness, underlying disease) with the case-patients. It must be realized that such matching factors cannot be evaluated as potential risk factors. Furthermore, the more factors one wishes to match between case- and control-patients, the more difficult it is to identify controls without reviewing all of their records looking for these factors. It is preferable to randomly select the controls and evaluate the factors under consideration for their strength of association with the adverse condition. Then, if they are found to be statistically significant, they can be controlled for through stratified or multivariate analyses. In contrast, in matched analyses, it is assumed that these factors are not significant without assessing whether the assumption is true.
Several important decisions must be made when conducting a case-control study. The first is whether to include all or a sample of the case-patients. In general, it is best to include all case-patients to increase the power of the study and to avoid introduction of bias. Second, one must decide how to select the controls. To enhance the likelihood of determining the source of the outbreak, controls should have similar opportunity for exposure to the potential risk factors as the case-patients. In other words, they should be selected from the same population (same ward or unit or undergoing the same procedure) as the case-patients and be hospitalized during the same period as the case-patients. Third, once the population of potential controls is identified, one must decide how many controls to select for each case-patient.
The number of controls selected is based on statistical power and practical considerations. If the attack rate is high and the number of case-patients is large, one control per case-patient may be sufficient. If the number of case-patients is small (e.g., <20), one should select 2 to 3 controls per case-patient. The proportional increase in power associated with the selection of >3 controls per case declines markedly, and the proportional increase in power by selecting >4–5 controls usually does not justify the increased volume of work. Last, one must decide how the controls are going to be selected. There are at least two selection methods: random and stratified or proportional sampling.
The random selection of controls is the easiest and most widely used method. With this method, all potential controls are listed and then selected using a random numbers table. Random numbers tables can be found in most statistics books and may be generated by numerous statistical software packages. If one has 500 potential controls and wishes to select 50, one just lists the controls in numerical order from 0 to 500 and then generates 50 random numbers between 1 and 500. Alternative random selection methods include choosing every “nth” control from the list of controls or selecting the patient who came before and/or after the admission or surgery of the case-patient. The random numbers method is preferred because the other methods may introduce bias—all patients are not admitted to a unit randomly and in outbreaks associated
P.105
with surgery, patients at risk may not undergo surgery at night or on weekends.
If one is concerned about the distribution of some factor among the case-patients, such as admission to one of several ICUs or an SSI that affects patients who have undergone a variety of surgical procedures, one may wish to select controls using the proportional or stratified method. The more factors one is concerned about, the more difficult and tedious it will be to match these factors. To identify a match, one needs to review the characteristics in the control-patients' medical records; this process requires review and discarding of large numbers of controls until the desired factors are identified. In contrast, using the stratified or proportional method, controls are placed into categories based on the variables of concern (e.g., type of surgery, admission into ICU), and then the proportion of controls desired is randomly selected from each category. For example, if in an SSI outbreak 10% of case-patients had cardiac surgery, 20% had neurosurgery, 30% had orthopedic surgery, and 40% had general surgery, the controls would be distributed into categories by the type of surgery they had undergone. Then they would be selected from each surgical category in the same proportion as the case-patients.
Once the case- and control-patients are selected, one can compare the exposures of interest. These exposures should be compared statistically using one of the available statistical software packages. If only one risk factor is statistically significantly different between case- and control-patients, no further epidemiologic analytic studies are necessary. If there are significant differences in 2–3 risk factors, further analysis using stratification may demonstrate which of these factors is the most important. In addition, if case- and control-patients differ significantly in ≥2 factors, multivariate analyses can be performed to identify the independent importance of these factors. For many HAI outbreak investigations, numerous risk factors are identified as significant. Because some of these variables are confounding factors, they are highly correlated with the causative risk factors. Multivariate techniques can be used to try to identify the independent importance of the factors found to have significance in univariate analyses. Multivariate techniques require the close cooperation of the epidemiologist and the statistician for appropriate conduct and interpretation.
Observational Studies
Often outbreaks associated with healthcare facilities are the result of failure by HCWs to fully comply with current recommendations or policies. Therefore, IC personnel should observe HCWs performing the procedures in question to document the adequacy of their understanding and compliance with IC recommendations. Observational studies can help generate hypotheses about the cause of the outbreak, or they can confirm the findings of the epidemiologic studies. For instance, if the comparative analyses have linked transmission of the etiologic agent to a particular product and HCW, observation of this person preparing or manipulating the product may identify how the product was contaminated.
IC personnel should review all written policies associated with the practices in question, interview supervisory staff, and identify any changes in the procedures in question before, during, and after the outbreak. IC personnel should observe the implicated procedures and/or HCWs and directly question the personnel who are performing the implicated procedures. It may be useful to observe HCWs on all shifts and to distribute a questionnaire asking the HCWs about their particular practices. For instance, in questionnaires distributed to them, HCWs often claim to perform hand hygiene before and between contacts with patients, but observational studies usually document that such hand hygiene occurs ~30%–40% of the time [6].
Culture Surveys
Cultures of the environment and/or HCWs should not be performed before the comparative epidemiologic studies are completed. Cultures of the environment may identify the causative agent, but this may represent secondary contamination rather than the source of contamination. Many common HAI pathogens, such as gram-negative organisms and fungi, can be isolated from the environment even in the absence of an outbreak. Similarly, a positive culture from an HCW can represent the source of the outbreak but also may represent secondary contamination from the environment, colonized or infected patients, transient carriage, or carriage of the organism independent of the outbreak. Because identification of an HCW carrier may require removal from work, conclusive epidemiologic data identifying a person as associated with transmission in the outbreak is needed to warrant such action. Random culture surveys of products, the environment, or personnel are costly both in materials and time.
Once the comparative epidemiologic studies have been completed, cultures should be obtained from the epidemiologically implicated sources (products or personnel). At the same time, several other cultures should be carried out on HCWs or products representing controls to avoid the risk of identifying the product or HCW before the investigation is completed and appropriate recommendations can be made. The performance of cultures of only those products or HCWs who are epidemiologically implicated reduces the number of cultures to be obtained, the personnel resources needed to conduct the cultures (they should not be performed by the implicated HCW[s]), and the burden on the laboratory. Cultures of nonsterile areas (e.g., floors, sinks, walls) and other animate or inanimate objects that do not have plausible connections to the outbreak are a waste of valuable resources and may generate uninterpretable data.
P.106
If the epidemiologic data identify a source yet the cultures of the person or product do not disclose the causative agent, one should not abandon the hypothesis. The epidemiologic data should always supercede the laboratory data. When low-level contamination of a product occurs, it may require cultures of large numbers of the product to pinpoint the contamination. Furthermore, the cultures of the implicated HCW may give negative results if the colonized site is not cultured or if the HCW shows positive results only intermittently or has treated himself or herself and is no longer culture positive. IC personnel should never vindicate a source that has been identified by a well-designed epidemiologic study even in the absence of culture confirmation.
Because the appropriate method of culturing inanimate and animate objects varies, IC personnel should consult with the microbiologist before such culturing is initiated. For cultures of HCWs' hands, the broth hand-wash method is preferred [45]. For water or dialysate used in hemodialysis, the pour-plate or milepore filter method is preferred [46]. For cultures of the environment when low numbers of organisms are anticipated, the use of premoistened swabs and inoculation into nutrient broth may facilitate recovery of the organism. When investigating outbreaks associated with possible airborne transmission, the use of settle plates or air samplers may expedite recovery of the implicated pathogen [19,20,28,47,48,49,50,51,52].
Typing of the Outbreak Strain
When the outbreak strains are available from the case-patients and the environment or implicated HCW(s) or products, it is essential that they be saved for potential future studies. Although outbreaks can be clonal (caused by one strain) or nonclonal (caused by multiple strains of one species); typing the outbreak strains can be a vital addition to the epidemiologic study. A wide variety of nonmolecular and molecular typing methods are available for many of the pathogens associated with outbreaks in healthcare facilities (SeeChapters 10 and 16) [34]. The finding of clonality among isolates increases the likelihood that the outbreak is caused by a contaminated product or colonized HCW, whereas the finding of a nonclonal outbreak increases the likelihood that the pathogen is being transmitted via HCW hands from one source (patient or environment) to another patient. For this reason, it may be useful to type some outbreak strains, particularly those that are common to the environment, water, or HCW. Such typing may influence the direction of the comparative studies.
Typing methods include biotyping, antimicrobial susceptibility testing, multilocus enzyme electrophoresis, serotyping, phage typing, and a variety of molecular methods, such as plasmid analysis, restriction endonuclease analysis, chromosomal analysis, ribotyping, restriction fragment length polymorphism, or pulsed-field gel electrophoresis. In general, the molecular methods—in particular, pulsed-field gel electrophoresis—have become the most useful for epidemiologic typing [34]. However, the field of molecular typing is evolving rapidly, and polymerase chain reaction–based pulsed-field gel electrophoresis typing and other methods are proving useful for organisms for which previous nonmolecular or molecular typing methods have been deemed inadequate.
Making Recommendations and Evaluating the Efficacy of the Recommendations
Once the epidemiologic and laboratory studies have been completed, one can assess the results and make appropriate recommendations to terminate the outbreak and prevent further transmission. Often these recommendations are based on existing guidelines. Several outbreak investigations have documented that often transmission stems from failure to fully implement IC guideline recommendations, not that the guideline recommendations are inadequate [50,51,52]. Occasionally, the outbreak investigation identifies a practice needing revision or a contaminated product that should be removed. More often, the investigation finds lapses in aseptic technique that require further HCW education. Thus, it is imperative that once the recommendations are made, follow-up studies be initiated to ensure that there is compliance with the recommendations and that transmission is terminated.
Furthermore, during the investigation, one has the opportunity to improve general IC measures throughout the hospital (even if the outbreak is taking place in only one unit) and to improve methods of documentation of information that would have made the investigation easier. For instance, if, during an investigation of BSIs in ICU patients, it is impossible to determine the dates of central venous catheter insertion and removal, the prospective collection of such data in a standardized manner by ICU staff should be included in the recommendations.
To ensure that the outbreak investigation is conducted efficiently and that hospital HCWs fully implement the recommendations, hospital administrative personnel should be actively involved in the investigation and follow-up. IC personnel should alert hospital administrative staff as soon as the outbreak is detected. IC personnel should update administrative staff during the course of the investigation and then review with them the investigation findings and recommendations. Follow-up meetings with administrative staff should review the adequacy of implementation of the recommendations and determine whether the outbreak has been terminated.
For complex outbreaks that involve a variety of services or are causing significant morbidity or mortality, a task force, which should include administrative staff, should be formed at the beginning of the investigation, and this group should be responsible for monitoring the adequacy of implementation of the recommendations and evaluation of their efficacy. The successful conclusion of
P.107
any investigation of an endemic or epidemic problem is the documentation through follow-up data that the endemic or epidemic problem has been controlled or terminated by the implemented control measures. Any successful investigation requires the close cooperation of a wide variety of people from many hospital departments. It becomes the responsibility of these people to translate the recommendations into action.
If the investigation is conducted in close collaboration with staff members of other departments, they will in turn help with the implementation and monitoring of the recommendations in their departments. To educate and inform those associated with the outbreak or the investigation, it is important that a report be written at the conclusion of the investigation summarizing the methods used, the results, and the recommendations. This report should be circulated to all of those assisting with the study, all service chiefs or heads of departments where the outbreak took place, and the administrative, risk management, and hospital public relations departments. In this way, all necessary personnel are informed about the findings and recommendations of the investigative team. Feedback of follow-up data to these departmental staff members will help them improve conditions and practices in their departments. Investigations of endemic or epidemic problems should be viewed as collaborative efforts between the IC personnel and other departments. With the cooperation of other departments, a successful outcome is enhanced; without it, success is unlikely.
References
P.108