Mary L. Andrus
Teresa C. Horan
Robert P. Gaynes
Definition of Surveillance
Surveillance is “the ongoing, systematic collection, analysis, and interpretation of health data essential to the planning, implementation, and evaluation of public health practice, closely integrated with the timely dissemination of these data to those who need to know” [1]. A healthcare-associated infection (HAI) surveillance system may be sentinel event based, population based, or both. A sentinel infection is one that clearly indicates a failure in the hospital's efforts to prevent HAIs and, in theory, requires individual investigation [2,3]. For example, blood transfusion from a source-patient already identified as a hepatitis carrier should always prompt investigation because it clearly indicates a failure of a hospital's safeguards. Denominator data usually are not collected in sentinel event-based surveillance. Sentinel event-based surveillance will identify only the most serious problems and should not be the only surveillance system in the hospital. Population-based surveillance (i.e., surveillance of patients with similar risks) requires both a numerator (i.e., HAI) and denominator (i.e., number of patients or days of exposure to the risk).
Historical Perspective
Since the first edition of this book, there has been a great deal of discussion and debate among professionals over the desirability of continuing routine surveillance, argued by some to be too personnel intensive in an era of constrained hospital budgets. As this discussion continues, an account of the development of concepts and techniques of surveillance should be considered. Many of these techniques were developed to meet emerging problems, and the basic concept of surveillance has been found effective in reducing HAI risk. Knowledge of the historical reasons for these developments may help improve the efficiency and effectiveness of surveillance without discarding well-conceived approaches that remain effective.
The use of surveillance methods to control HAIs dates back at least to the classic work of Dr. Ignaz Semmelweis in Vienna in the 1840s [4]. Although the Semmelweis story is best remembered as the first demonstration of person-to-person spread of puerperal sepsis and of the effectiveness of hand washing with an antiseptic solution, an equally important achievement was Semmelweis' rigorous approach to the collection, analysis, and use of surveillance data. In contrast, the concurrent work of Dr. Oliver Wendell Holmes on the same subject in the United States was based primarily on the traditional anecdotal case-study approach of clinical medicine.
P.74
Semmelweis' investigation constitutes an amazingly contemporary example of the effective use of surveillance in addressing a widespread HAI problem. When he assumed the directorship of the obstetric service at the Vienna Lying-in Hospital in 1847, the apparent risk of maternal mortality had been at high levels for >20 years. The eminent clinicians of the day, in fact, considered the risks to be no more than the expected endemic occurrence that could not be influenced. Semmelweis first undertook a retrospective investigation of maternal mortality and set up a prospective surveillance system to monitor the problem and, later, the effects of control measures. The initial results of his retrospective study of annual hospital mortality showed clearly that the maternal mortality level, which he measured by calculating yearly mortality rates, had increased tenfold following the introduction in the 1820s of the new anatomic school of pathology, which used autopsy as its primary teaching tool. Based on the use of ward-specific mortality, Semmelweis calculated that the risk of death on the ward used for teaching medical students was at least four times higher than that on the ward used for teaching midwifery students. After the septic death of his mentor suggested the presence of a transmissible agent, Semmelweis used the findings from his retrospective surveillance study to implicate the practices of the medical students. After observing their daily routines, he surmised that students might be transferring pathogens from cadavers to the parturient women and that washing hands with a chlorine solution might prevent this transmission. Subsequently, his prospective surveillance data documented a dramatic reduction in maternal mortality immediately following the institution of mandatory hand washing before entering the labor room.
Apparently, due to his abrasive manner, lack of diplomacy, and inability to organize his statistical data into a concise and convincing report, Semmelweis failed to win over his clinical colleagues to his discovery. Within 2 years, he was dismissed from the staff of the hospital, and his successor gradually allowed the strict hand-washing measures to decline. In the absence of continuing surveillance, the epidemic promptly resumed and lasted well into the early part of the 20th century, its severity and means of prevention apparently unappreciated by several more generations of clinicians.
This story illustrates one of the main impediments to infection control today: In the absence of careful epidemiologic data and a diplomatic presentation, clinicians, who are oriented almost entirely toward the treatment of individual patients, often fail to appreciate the severity of the HAI pathogen transmission problem and sometimes resist control measures. It also points out the utility of surveillance in identifying problems and developing and applying control measures. From a methodologic viewpoint, Semmelweis' efforts encompassed almost all aspects of the modern surveillance approach: retrospective collection of data to confirm the presence of a problem; analysis of the data to localize the risks in time, place, and person; controlled comparisons of high- and low-risk groups to identify risk factors; formulation and application of control measures; and prospective surveillance to monitor the problem, evaluate the implemented control measures, and detect future recurrences. The main shortcoming of his approach was in not diplomatically educating his powerful colleagues with a careful report of his findings.
Despite Semmelweis' historical model, the modern era of HAI surveillance grew more from mid-20th century experience. The importance of surveillance for disease control in general arose in the effort to control tropical diseases among troops stationed in the Pacific Theater in World War II. At the end of the war, most of the epidemiologists of the “Malaria Control in War Areas Unit” were transferred to a civilian facility to apply their surveillance and control strategies to the control of malaria in the southern United States. Located in Atlanta, near the endemic areas, the unit was first named the Communicable Diseases Center and later became the Centers for Disease Control and Prevention (CDC). Since the large number of reports of malaria indicated the disease to be widespread, a surveillance system was immediately set up to define the problem. However, as investigators examined each reported case, they found virtually all of the reports had errors in diagnosis. Thus, the mere activity of surveillance “eradicated” the malaria epidemic in the United States [5].
Because of this and similar successes, when the pandemic of staphylococcal infections swept the nation's hospitals in the mid-1950s, CDC staff members were quick to apply the concepts of surveillance to the problem. When asked to assist in investigating a staphylococcal epidemic in a particular hospital, these early investigators often met strong resistance from clinicians and hospital administrators convinced that no unusual infection problems were present in their hospitals. In instances when CDC staff members were able to continue the investigations, the collection and reporting of surveillance data regularly changed those attitudes to strong concern over the documented problems and eagerness to apply control measures. These initial investigations thus confirmed a nationwide staphylococcal epidemic and led the CDC to sponsor several national conferences to discuss the problem [5].
By the early 1980s, some critics were questioning the effectiveness and cost benefit of routine HAI surveillance, although a growing number of hospitals were increasing their surveillance efforts rather than cutting them back [6]. Surveillance was, and remains, a time-consuming activity requiring about 40 to 50% of the time of an infection control professional (ICP) [7]. The inability of some hospitals to establish an adequate number of ICP positions (at least one full-time equivalent per 250 beds) has been a major contributor to disenchantment within the infection control profession.
Several factors have influenced contemporary practices favoring robust surveillance activities in infection control
P.75
programs. First, the results of the Study on the Efficacy of Nosocomial Infection Control (SENIC) project strongly substantiated the importance of surveillance along with control measures to reduce HAI rates and provided the scientific basis for surveillance of HAIs [8,9,10]. The conclusion was that hospitals that are effective in reducing their HAI rates have an organized, routine, hospitalwide surveillance system. Second, the requirements of the Joint Commission on Accreditation of Healthcare Organizations (JCAHO, formerly JCAH) have legitimized the need for personnel to perform surveillance [11,12]. Third, the surveillance practices developed in infection control have begun to influence other aspects of the hospital's quality monitoring and improvement activities. The strategies of targeting surveillance to reduce specific endemic problems and monitoring to assess the intervention's effectiveness were incorporated into JCAHO's 1994 infection control standards for accreditation and were applied to hospital quality-assurance programs to reduce noninfectious complications [13,14]. The increasing pressure to continually improve quality is certain to broaden the use of surveillance to aid in the prevention of HAIs [15].
Goals of Surveillance
A hospital should have clear goals for doing surveillance. These goals must be reviewed and updated frequently to address new HAI risks in changing patient populations, such as the introduction of new high-risk medical interventions, changing pathogens and their resistance to antibiotics, and other emerging problems. The collection and analysis of surveillance data must be performed in conjunction with a prevention strategy. It is vital to identify and state objectives of surveillance before designing a surveillance program and starting it.
Reducing HAI Rates Within a Hospital
The most important goal of surveillance is to reduce the risks of acquiring HAIs. To achieve this goal, specific objectives for surveillance must be defined based on how the data are to be used and the availability of financial and personnel resources for surveillance [7,8]. Objectives for surveillance can be either outcome or process oriented or both. Outcome objectives are aimed at reducing HAI risk and their associated costs. By using comparative HAI rate analysis and providing feedback to patient-care personnel, outcome data are useful in demonstrating where gaps in HAI prevention activities exist. Process monitors, on the other hand, help to identify delivery of care problems that can have an effect on patient outcomes. Examples of process monitors are observing and evaluating patient-care practices, monitoring equipment and the environment, and providing education. Much of the time spent performing surveillance should be devoted to monitoring patient-care process objectives because policy seldom equals practice. However, these activities are of limited value without clearly stating outcome objectives. While HAI surveillance has other legitimate purposes, the ultimate goal is to use this process objective to achieve the outcome objective: decreases in HAI rates, morbidity, mortality, and cost.
Establishing Endemic Rates
Surveillance data should be used to quantify baseline rates of endemic HAIs. This measurement provides hospitals with knowledge of the ongoing HAI risk in hospitalized patients. Most HAIs, perhaps 90 to 95%, are endemic (i.e., not part of recognized outbreaks) [16]. Thus, the main purpose for surveillance activities should be to lower the endemic HAI rate rather than identify outbreaks, and many hospital ICPs report that their presence on the wards may be sufficient to influence HAI rates [17]. However, the mere act of collecting data does not usually influence HAI risk appreciably unless it is linked with a prevention strategy. Otherwise, surveillance is no more than “bean counting,” an expensive exercise without focus that today's hospitals can ill afford and that ICPs will ultimately find dissatisfying.
Identifying Outbreaks
Once endemic HAI rates have been established, ICPs and hospital epidemiologists may be able to recognize deviations from the baseline that sometimes indicate infection outbreaks. This surveillance benefit must be balanced with the relatively time-consuming task of ongoing data collection because only a small proportion of HAIs, perhaps 5 to 10%, occur in outbreaks [14]. Moreover, HAI outbreaks often are brought to the attention of ICPs by astute clinicians or laboratory personnel much more quickly than by the analysis of routine HAI surveillance data. This lack of timeliness often limits the use of routine HAI surveillance in identifying outbreaks in a hospital.
Convincing Medical Personnel
Convincing hospital personnel to adopt recommended preventive practices is one of the most difficult tasks of an infection control program. Familiarity with the scientific literature on hospital epidemiology and infection control is effective in influencing behavior only if the hospital personnel believe the information is relevant to the specific situation in question. Studies in the literature may not address the many varied situations encountered in a particular hospital. Using information on one's own hospital to influence personnel is one of the most effective means to address a problem and apply the recommended techniques to prevent HAIs. If surveillance data are analyzed appropriately and presented routinely in a skillful manner, medical personnel usually come to rely on them for guidance. The feedback of such information often is quite
P.76
effective in influencing behavior of healthcare workers to adopt recommended preventive practices [9]. A team approach with ICPs working with clinicians from a variety of disciplines is particularly effective.
Evaluating Control Measures
After a problem has been identified through surveillance data and control measures have been instituted, continued surveillance is needed to ensure that the problem is under control. By continual monitoring, some control measures that seemed rational can be shown to be ineffective. For example, the use of daily meatal care to prevent nosocomial urinary tract infections (UTIs) seemed appropriate but did not control infection [18]. Even after the initial success of control measures, breakdowns in applying them can occur, requiring a constant vigil including the continued collection of surveillance data.
Satisfying Accrediting and Regulatory Agency Surveyors
Satisfying the requirements of accrediting organizations, such as JCAHO, is a very common use of HAI surveillance data but one of the least justifiable. The collection of surveillance data merely to satisfy a surveyor who visits a hospital once every three years (or occasionally more often) is a largely unproductive use of resources. JCAHO also changed its requirements to avoid this task-oriented process of collecting data as unproductive when it altered its standards in 1990. Hospitals are now required to use HAI surveillance in a directed manner to initiate specific interventions designed to lower the risk of HAIs in patients [11]. JCAHO's Agenda for Change has motivated hospitals to use HAI surveillance for its originally intended purpose: to change the outcome of patient care by reducing HAI risk.
Defending Malpractice Claims Previously
One concern about the collection of HAI surveillance data has been that it would create a record that could be used against the hospital in a malpractice claim related to an HAI. A strong surveillance component in an infection control program will demonstrate, however, that a hospital is attempting to detect problems rather than conceal them (see Chapter 17). Additionally, the records of infection control committees are considered privileged in most states and are not discoverable in civil court proceedings. Therefore, surveillance often is helpful in defending against malpractice claims, and it is rarely, if ever, a hindrance.
Comparing Hospitals' HAI Rates
Traditionally, HAI surveillance has been recommended solely for gaining understanding of and reducing HAI rates within individual hospitals. The idea of comparing HAI rates among hospitals, though often suggested by administrators and quality-assurance supervisors, has generally been discouraged by infection control physicians and practitioners. They argue that the mix of intrinsic HAI risk of the patients in different hospitals renders differences among the hospital rates virtually uninterpretable. Studies performed by CDC, however, have suggested that interhospital comparisons can be useful in reducing HAI risk [19,20] if the rates are specific to a particular site of HAI (e.g., UTI) and control for variations in the distributions of the major risk factor(s) for that type of infection (e.g., duration of indwelling urinary catheterization). Conversely, using a single number to express a hospital's overall HAI rate falls short as a valid measure largely because suitable overall risk adjusters for infections of all types are lacking [20,21,22,23]. Therefore, a hospital's overall HAI rate, as presently derived, should not be used for interhospital comparisons.
Public Reporting of HAI Rates
Many state and national initiatives to mandate or induce healthcare organizations to publicly disclose information regarding institutional and physician performance are underway. Mandatory public reporting of healthcare performance is intended to enable stakeholders, including consumers, to make more informed decisions about healthcare choices and has taken several forms such as report cards and honor rolls. As of this writing, 13 states require hospitals to report publicly their HAI rates [24]. The methods of surveillance and reporting have differed from state to state. In an effort to provide guidance and to establish more uniformity, the CDC's Healthcare Infection Control Practices Advisory Committee (HICPAC) issued a guidance document on public reporting of HAIs in 2005 [25]. Presently the evidence on the merits and limitations of using an HAI public reporting system as a means to reduce HAIs is insufficient. The HICPAC guidance is intended to assist policymakers, program planners, consumer advocacy organizations, and others tasked with designing and implementing HAI public reporting systems. Challenges for meaningful interpretation of publicly reported HAI rates include accuracy in identifying HAIs, risk adjustment to account for varying degrees of risk among the sampled patient population, and the method of expression of the HAI rates that can range from complicated indices to single and perhaps overly simplistic rates intended to measure a facility's experience. HICPAC has made no recommendation for or against mandatory public reporting of HAI rates. Some investigators have recommended that efforts should be directed instead to creating acceptably accurate, objective measures of quality of care and of outcomes, such as process and/or surrogate measures that all healthcare facilities can use [26]. Process measures assess the delivery of care rather than the outcomes. These measures can be useful when their link to beneficial or adverse outcomes is well established.
P.77
For example, appropriate delivery of perioperative antibiotic prophylaxis is a case in point of a process measure that emerged from decades of research [27]. Surrogate measures are objective indicators of readily ascertained events that are sufficiently correlated with HAIs to provide useful information about the actual institutional HAI rate [28]. For example, the surgical site infection (SSI) rates following coronary artery bypass, Cesarean section, or breast surgery appear to correlate closely with the proportion of patients who receive extended courses of inpatient antibiotics and is a useful indicator of a hospital's SSI outcomes for those procedures [29].
Methods of Surveillance
Collecting the Data
Definition of Events to Be Surveyed
Carefully defining those events to be surveyed is important as an initial step in developing a HAI surveillance system. Applying accepted definitions systematically in the data-collection process is another key step. In attempting to understand the relationship between UTI and urinary catheterization, for example, it is necessary first to define or establish criteria to decide what will be called a UTI and what will be considered urinary catheterization. Once the event to be surveyed has been defined as concisely and precisely as possible and the criteria for determining its occurrence have been established, then these definitions and criteria are applied systematically and uniformly henceforth. Ideally, all members of the population at risk for the infection would be systematically monitored for the presence or absence of the criteria elements that define the infection being surveyed.
The CDC has published guidelines for determining the presence and classification of HAIs [30]. These guidelines are not rigorous definitions of disease but serve as practical, operational HAI surveillance definitions for most hospitals regardless of their size or medical sophistication. The exact definitions to be used in surveillance in individual hospitals are not as critical as having the infection control committee obtain the concurrence of key hospital staff members who will be applying them. Such widespread advance agreement is necessary to avoid later having the results of surveillance disqualified by disagreements over the definitions.
Role of the ICP
A variety of methods of collecting HAI data has been described [31,32,33]. In general, the most satisfactory and practical method employs a person (or persons), often called the ICP, whose job description includes collecting and analyzing surveillance data. Details of the qualifications, functions, and responsibilities of the ICP are described elsewhere (see Chapter 5). The ICP reports prevention and surveillance information to the infection control committee and should work directly with the committee chairperson or the hospital epidemiologist. The traditional choice of a nurse to fill this position has been primarily based on the person's professional training and ability to interact primarily with other healthcare personnel in the data-collection process. Experience has shown, however, that persons other than nurses can function well in this position, particularly in the surveillance process. The original studies of surveillance, conducted by CDC in the 1970s, indicated that one full-time-equivalent ICP could conduct surveillance for approximately 250 acute-care hospital beds and have sufficient remaining time for other infection control duties and responsibilities [10]. In recent years, however, as the average length of a patient's hospital stay has decreased and patients' severity of illness has increased, the ICP's job has become more demanding, requiring an increase in infection control resources. ICPs continue to set priorities for the use of their time in ways that will maximize their impact on infection risks.
Minimal Data to Collect About Infections
The precise information collected in conjunction with each HAI depends on the surveillance objective and may vary according to the institution, service, site of infection, or causative agent. Certain essential identifying data, however, can be recommended: the patient's name, age, gender, hospital identification number, ward or location within the hospital, service, and date of admission; the date of onset of the infection; the site of infection; the organism(s) isolated in culture studies; and the antimicrobial susceptibility pattern of the organisms isolated. Additional information should be collected only if it will be analyzed and used by the hospital. Some institutions may wish to include the primary diagnoses of the infected patient, an assessment of the severity of underlying illness(es), the name(s) of the attending physician(s) or other staff who attended the patient, whether the patient was exposed (before the onset of the HAI) to therapies that may predispose to infection (e.g., surgery; antimicrobials, steroids, or immunosuppressive therapy; or instrumentation), what antimicrobial agents were used to treat the infection, and some assessment of mortality related to the HAI. It is important to record the presence or absence of particular risk factors (e.g., the use of invasive devices such as urinary catheters for UTIs, ventilators for pneumonias, and central line catheters for primary bacteremia).
Denominators
The methods for collecting the denominators for HAI rate determination have been controversial and have sometimes been the most labor-intensive aspect of generating HAI rates [29]. Historically, the common denominator was the number of patients admitted to or discharged from the hospital or a particular ward or service [17]. These totals served as a crude estimate of the number of patients at HAI risk. Another denominator used was patient-days rather
P.78
than admissions or discharges. Patient-day denominators are derived by summing the number of days that all patients stayed in the hospital during the surveillance period. Since the patient-day denominator includes length of stay, the resulting HAI rates are at least partially adjusted for differences among patients' lengths of stay. Admissions, discharges, and patient-days can be for the hospital as a whole and for individual patient-care units. These data can be obtained either electronically or in a written monthly report from the hospital medical records department or business office. However, these denominators fail to take into account differences in patient risks for HAIs, such as exposures to certain invasive devices. Calculation of overall rates using these denominators is not recommended [22].
Another method is to collect a record on every patient at HAI risk rather than on only infected patients and to use summary denominator data for rate calculation. As hospitals have become more computerized, this has become easier and more common. In this method, the record contains the risk factors for individual patients (e.g., each patient undergoing a surgical procedure), and then the information about any subsequent HAI is added to the patient's record when detected. Computerization usually is necessary to manage these detailed records. An ICP can use a computerized patient database and computer software to calculate HAI rates, including denominators for whatever time period and patient population selected. While entering data on all patients at risk might seem more time-consuming than other methods, it is preferable for priority-directed surveillance projects, such as generating surgeon-specific SSI rates by risk index category or device-associated infection surveillance by specific nursing care unit. This method offers ICPs more flexibility in the types of rates that can be generated. A full database enables them to calculate rates by any risk factor or combination of risk factors. This alternative is made even more attractive by the ability to “download” denominator records from other hospital software. For example, the records of all operations can be downloaded periodically from the operating room database. As HAI surveillance becomes increasingly automated and directed toward specific prevention objectives, detailed denominators on individual patients at risk are expected to be employed more often (See Chapter 9).
Identifying Sources of Infection Data
To ensure the most complete enumeration of HAIs, ICPs seek a variety of infection information sources from both within and outside the hospital [17]. These active techniques of case finding are used in almost all hospitals and are strongly preferred to passive techniques. The active techniques allow more complete detection of “cases” and provide for the ICP to visit the patient-care areas regularly, interact with and provide consultation to the medical and nursing staffs, and gain firsthand awareness of HAI problems. Passive techniques include asking physicians or staff nurses to fill out infection report forms or relying solely on reviews of computerized microbiology reports. The usefulness of passive techniques is limited by their inaccuracy in routine detection of HAIs. Hospitals relying on passive techniques typically find extremely low HAI rates, but these usually are due to underreporting rather than good patient-care practice [17]. While the use of administrative data (e.g., billing codes) can be a useful screening tool, this approach also has major shortcomings as a single methodology for identifying infections [34]. The actual collection of HAI data is typically accomplished by a combination of manual and electronic surveillance techniques. The ICP often has electronic access to demographic information, microbiology and pharmacy reports, and sometimes radiology dictation summaries. Many of the data elements used to satisfy the criteria for HAIs, however, are found in the patient chart or by consultation with physicians or nursing or respiratory therapy staff.
The Microbiology Laboratory
Of all case-finding methods, one of the most useful is the periodic (usually daily) review of microbiology laboratory reports. Data-mining computer applications also may be used to identify microorganisms of special significance to the patient population requiring specific infection prevention measures. A review of microbiology results may be performed each morning before patient-care unit rounds, so any new or potential HAIs can be identified at that time. Such review requires that the ICP understand the infectious and epidemiologic potential of various microorganisms; such knowledge might be achieved through laboratory training or a basic infection prevention and control course and should be reinforced by periodic continuing education. However, a review of microbiology laboratory reports alone is not sufficient for the identification of HAIs because (1) cultures are not obtained for all infections or may be handled incorrectly, (2) some infectious agents (e.g., viruses) will not be identified in many hospital laboratories, and (3) for some types of infections (e.g., SSIs and pneumonia), the identification of a potential pathogen from a culture specimen does not mean that infection is present and such infections require clinical detection and verification (see Chapter 10).
Patient-Care Area Rounds
Periodic (preferably daily) patient-care rounds should be included as an integral part of an effective surveillance program. The purposes of such rounds are to identify new HAIs and to follow up previously identified HAIs. New HAIs may be identified outright by physicians or nurses working in the area visited and by review of patient records, temperature records, patients having high-risk procedures (e.g., surgery, central line insertion, intubation and mechanical ventilation, indwelling urinary catheters), and patients in isolation or receiving antimicrobial therapy. Visiting the
P.79
patient-care unit also may allow direct assessment of the patient and documentation of visible HAIs. The added value of patient-care rounds is the opportunity for the ICP to interact directly with patient-care staff and initiate measures that could prevent HAIs in the future.
Postdischarge Follow-Up
With the progressive reduction in the average length of patient stay in U.S. hospitals, an increasing percentage of some HAIs, most notably SSIs, become manifest after hospital discharge. The percentage of SSIs that becomes apparent after discharge ranges from 20 to 60% [35,36,37,38,39,40,41], and at least one study has shown the presence of a strong selection bias in rates based only on in-hospital surveillance. Since the average postoperative length of stay of surgical patients influences the probability that SSIs will be recognized while the patient is in the hospital, this variable must be taken into account when analyzing and evaluating a hospital's SSI rate.
Multihospital analyses of SSI rates in the SENIC project used length of stay as a covariate in multiple regression analyses [10]; others have suggested using the incidence density (i.e., patient-days in the denominator of the SSI rates) [42]. Choice of postdischarge surveillance method is controversial and problematic. Many ICPs follow up with postcard inquiries or telephone calls to the surgeon 21 to 30 days after the date of the operation to determine whether an SSI occurred [39,40,41,42]. Though this creates additional work for the surveillance staff, the amount of work may be commensurate with the gain in completeness and accuracy of the SSI rates. Patients themselves are rarely contacted and questioned about signs and symptoms of SSI following selected operative procedures, but the reliability of this information is questionable. Postdischarge antibiotic exposure also may be a resource-efficient adjunct for surveillance of SSIs after discharge [43].
Other Sources
Additional infection information may be obtained through a periodic review of radiology, pharmacy, and laboratory reports; records of personnel health clinic visits; and autopsy reports. The exclusive use of other methods of infection data collection, such as reviewing patients' medical records after discharge or the use of infection report forms filled out by attending physicians or floor nurses, is less satisfactory from the standpoint of infection control. Reviewing patients' medical records after discharge results in a failure to apply indicated infection control precautions and may result in excessive morbidity or mortality among patients or hospital personnel. The use of infection report forms filled out by attending physicians or floor nurses has been used in a number of hospitals, but it suffers from the lack of systematic application of standard definitions and criteria for detecting HAIs and from variation in the thorough reporting of HAIs.
Consolidating and Tabulating Data
Consolidating the HAI data in ways that make them more understandable helps users identify potentially important relationships or patterns of infection that may not be apparent from the raw data on data collection forms. The most effective analyses of surveillance data may take many forms depending on the objective being addressed by the surveillance. For routine, total hospital surveillance, ICPs often simply analyze HAI data in single-variable frequency tables (e.g., number of HAIs by hospital location, body site, or pathogen) and two-way cross-tabulations (e.g., a line listing with the number of HAIs by body site and by pathogen for each patient-care unit), and tally antimicrobial susceptibility patterns for each pathogen by HAI site. In recent years, however, more imaginative analyses have been done, including three-way rate tables (e.g., pathogen by site by patient-care unit), four-way rate tables (e.g., susceptibility patterns by pathogen by site by care unit), and more complex cross-tabulations. Usually these tabulations are performed for subsets of the patients, such as those on a particular service or the patients of a specific surgeon. To do these adequately requires a computer to analyze the patient databases.
Although beginning surveillance personnel should first master the standard frequency and cross-tabulation routines mentioned previously, more experienced ICPs can try more creative ways of organizing the infection data. The basic purpose of tabulating the HAIs is to gain a new understanding of when, where, and in whom the HAIs are occurring. One of the most frequent mistakes made when the raw data are organized is to make initial tabulations hastily and proceed with calculating rates without pausing to examine the data. It often is useful to read over the original listings and the simple tabulations for an initial synthesis of the data that can suggest the need for additional tabulations, graphs, and listings. For example, finding an increased rate of bacteremia in surgical patients might call for a tabulation of bacteremia for each surgical subspecialty or for each surgical ward and surgical intensive care unit and for comparisons with similar rates from previous months or for the same months the previous year. This process of exploration of the data has no rigid rules; what is right is that which works!
Calculating Rates
Definition of Rate
After the initial tabulations of the HAIs have been completed, the infection control staff should have a strong indication of where HAI problems might be occurring. Since these initial analyses are based solely on examination of the overall number of HAIs (numerator data), further analysis involving the calculation of rates is necessary to develop stronger evidence.
A rate is an expression of the probability of occurrence of some particular event, and it has the form k(x/y), where
P.80
x, the numerator, equals the number of times an event has occurred during a specific time interval; y, the denominator, equals a population from which those experiencing the event were derived during the same time interval; and k equals a round number (100, 1,000, 10,000, 100,000, and so on) called a base. The base used depends on the magnitude of x/y, and it is selected to permit the rate to be expressed as a convenient whole number. For example, if five HAIs were found among 100 patients in a given month, the value of x/y would be 0.05 HAIs per patient per month; to express the rate as a convenient whole number, x/y would be multiplied by the base number 100, giving five HAIs per 100 patients per month. If 50 HAIs were found among 10,000 patients in a month, the base number 1,000 would be used to express the rate as 5 HAIs per 1,000 patients per month. It is important to emphasize that in determining a rate, both the time interval and the population must be specified, and these must apply to both the numerator (x) and the denominator (y) of the rate expression. A practical way of generating a rate is to enter the denominator data, obtained in the data collection stage, below the appropriate numerator figures in the frequency and cross-tables of infections tabulated earlier. From these numerators (indicating the numbers of HAI) and their denominators (indicating the numbers of patients, or patient-days, at HAI risk), HAI rates can be calculated. Alternatively, currently available computer software can display the numerator and denominator and calculate the rate automatically from a database containing all patients at risk and the associated HAIs that occurred.
|
Figure 6-1 Incidence of infection is three during either time A or B (three new cases were added during each time period); prevalence of infection during time A is 40% and during B, 60% (four cases and six cases respectively, occur in each period of time; and point prevalence of infection at time C is 30% (at that point in time, three cases exist). |
Types of Rates
Three specific kinds of rates—prevalence, incidence, and attack rate—are fundamental tools of epidemiology and, as such, should be familiar to infection control personnel.Prevalence is the number of episodes of the disease found to be active within a defined population either during a specified period (period prevalence) or at a specified point in time (point prevalence). These concepts are discussed further in a later section of this chapter. Incidence is the number of new episodes of disease that occur in a defined population during a specified time period. The incidence rate is obtained by dividing the number of new episodes by the number of people in the population at risk during the specified time period. In Figure 6-1, which portrays the infection status of 10 hospitalized patients, the incidence of infection during either time period A or B, for example, would be three, since three new infections began among the 10 patients in each time period. Assuming that period A was 1 month and period B was 3 months, the incidence rates would be three infections per 10 patients at risk (30%) per month in period A and 10% per month in period B (i.e., exactly equivalent to 30% per 3 months).
An attack rate is a special kind of incidence rate. It is usually expressed as a percentage (i.e., k = 100 in the rate expression), and it is used almost exclusively for describing outbreaks where particular populations are exposed for limited periods of time (e.g., in common-source outbreaks). Since the duration of an outbreak is reasonably short, the period of time to which the rate refers is not stated explicitly but is assumed. This is what distinguishes an attack rate from an incidence rate when the period of time is always stated. If 100 infants in a newborn nursery, for example, were exposed to a contaminated lot of infant formula over a 3-week period, and if 14 of the infants developed a characteristic illness believed to be caused by the contaminated formula, the attack rate for those infants exposed to the formula would be 14%. Note that the incidence rate would be 14 cases per 100 infants per 3 weeks, preferably expressed as 4.67 cases per 100 infants per week.
Choice of Numerator and Denominator
Basically two types of incidence rates can be calculated: The infection ratio is the number of infections divided by the number of patients at risk during the specified period, and
P.81
the infection proportion is the number of patients with ≥1 infections divided by the number of patients at risk during the period. Since approximately 18% of patients with HAIs have >1 infection, the infection ratio is usually nearly 1.27 times larger than the infection proportion [44]. In practice, most ICPs have found the infection ratio to be far easier to obtain than and equally as useful as the infection proportion; consequently, the commonly used term infection rate has come to refer specifically to the infection ratio. Nevertheless, when presenting results, it is important to specify which method has been used for calculating the rate.
Another approach is to use the number of patient-days at risk during the period of surveillance (i.e., the sum of all days spent by all patients in the specified area during the time period covered) for the denominator in a rate calculation. This rate is referred to as the incidence density. To get an idea of how the two types of rates compare, rates based on the number of patient-days (R) usually are smaller than those based on admissions or discharges (r) by a factor approximately equal to the average length of stay of the patients (k), that is:
For example, if the HAI rate were five HAIs per 100 admissions per month and the patients' average length of stay was 10 days, one would expect to find a rate of approximately five HAIs per 1,000 patient-days. The incidence density is useful primarily in two situations: (1) when the infection rate is a linear function of the length of time a patient is exposed to a risk factor (e.g., indwelling urinary or intravenous catheter) and (2) when the duration of follow-up will influence the measured infection rate (e.g., SSI rates when no postdischarge surveillance is done). The relative merits of the alternative ways of controlling for differences in length of stay, including using the incidence density, performing postdischarge surveillance of surgical patients, and using multivariate analytic methods to control for length of stay, are yet to be clearly defined.
The denominator must reflect the appropriate population at risk as precisely as possible. In determining the attack rate of SSIs among patients on the urology service, for example, only those urology patients who actually undergo a surgical procedure that results in a wound capable of being infected should be included in the denominator. Practical difficulties in obtaining such refined denominators, however, often dictate the use of a less precise, summary denominator (e.g., the total number of admissions or discharges from the urology service during the appropriate period).
In the CDC's National Nosocomial Infections Surveillance (NNIS) system, hospitals that performed surveillance in intensive care units (ICUs) used device-days (e.g., ventilator-days or central line [CL]-days) as the denominator data [45]. The use of device-days may seem like a subtle change from patients, but the choice of denominator was critical for purposes of interhospital comparison. This is more fully illustrated by the distribution of several rates for hospital ICUs (Figure 6-2). The top histogram of this figure shows the number of central-line-associated bloodstream infections (CLABSIs) per 100 patients, the middle histogram shows the number of central line-days divided by patient-days (i.e., central line-utilization), and the bottom histogram shows CLABSIs per 1,000 central line-days. For hospital unit A, the rate on the top histogram, which uses the number of patients as the denominator, was nearly five times higher than the median. However, the middle histogram shows that hospital unit A had the highest central line-utilization rate; that is, >80% of patient-days were also central line-days. Using central line-days as the denominator of the rate helps to take into account this high utilization of central lines. Hospital unit A's CLABSIs rate was slightly lower than the median (bottom histogram). Although hospital unit A was no longer an outlier, its high central line-use may need to be reviewed for appropriateness. On the other hand, for hospital unit B, the CLABSIs rate (top histogram) was near the median, and its central line-use (middle histogram) was low. When its rate was calculated using central line-days as the denominator, it was quite high, suggesting the need to review central line-placement and maintenance practices.
Analyzing the Results
Comparing Patient Groups
Analysis implies careful examination of the body of tabulated data in an attempt to determine the nature and relationship of its component parts. This includes comparing current HAI rates to determine whether significant differences exist among different groups of patients. Suppose, for instance, that both the gynecology and general surgery services had 8 catheter-associated UTIs during a given month; however, during the same month, 20 patients who had indwelling catheters were discharged from the gynecology service and 100 other patients were discharged from the general surgery service. Thus, the rates for gynecology and general surgery patients are 40% and 8%, respectively. Determining whether the difference observed between these infection rates is significant (i.e., higher than what we would expect by random or chance occurrence alone, if indeed no real difference exists) requires the use of a statistical process known as significance testing.
Several tests of significance (e.g., the chi-square test, Fisher's exact test for cross-tables, or Student's t-test for comparison of sample means) should be familiar to epidemiologists and ICPs (see Chapter 8). Currently available software packages for computers make even the most sophisticated statistical testing procedures very accessible to all infection control departments [26,42]. In the preceding example, the difference between the observed infection rates (40% versus 8%) is highly significant at p < 0.001, according to the Fisher's exact test. This means
P.82
that a difference as large as or larger than that observed would be expected to occur by chance alone less than 1 time in 1,000. Thus, it is very likely that there is a real difference between the infection rates on the two services, and further investigation is indicated to explain why such a difference exists. If the ICP wants to compare the hospital HAI rates to that of other hospitals, such comparisons can be made only when comparison data that are risk stratified and risk adjusted using the same definitions and protocols are available [43].
|
Figure 6-2 Comparison of the distribution of bloodstream infection (BSI) rates (patient based and central line based) and central line utilization in combined coronary and medical intensive care units, National Nosocomial Infections Surveillance (NNIS) System, October 1986–December 1990. A and B represent individual hospital ICU rates. Arrows indicate the median [20]. |
Comparing Rates Over Time
Another type of analysis involves the comparison of current HAI rates with previous rates within the same patient care area or population to determine whether significant changes have occurred over time. Current rates and those of preceding periods can be visually inspected in tabular form, or the rates can be plotted on a graph to detect changes of potential importance. Potentially important deviations from baseline rates then should be tested for statistical significance (see Chapter 8), and further investigation should be undertaken if indicated (see Chapter 7). Although it is convenient to compare rates each month, caution is needed when the denominator of a rate is small. This may be particularly true when examining SSI rates. Tests of significance must often be performed when the estimate of a surgeon's SSI rate is unstable due to a small number of procedures performed.
Identifying Clusters
Screening to identify clusters of similar strains or specific patterns of antimicrobial susceptibility is another potentially valuable analytic tool for detecting outbreaks, especially when it is applied to particular pathogens on
P.83
specific wards or in particular geographic areas. Pulsed-field gel electrophoresis or other molecular techniques applied to microorganisms can have additional value in identifying clusters (See Chapter 10).
Analyzing Surveillance Data
Analysis of surveillance data must not be limited to HAI rates. Most ICPs are familiar with analyzing the numerator data, specifically the numbers and types of HAIs and their associated pathogens, but an analysis of the denominator data alone can be extremely revealing. For example, if an ICP determines that a hospital's utilization of central lines in ICUs is high compared with that in another hospital ICUs (see hospital unit A in the middle panel of Figure 6-2), a review of appropriateness of device use may be needed. Also, after SSI rates have been assessed, useful information can be obtained by further exploring the distribution of risk factors in each of the risk index categories. For example, while a surgeon with more than the expected number of herniorrhaphy operations in the higher SSI index categories may be operating on more high-risk patients, he or she also may be consistently exceeding the 75th percentile for the duration of surgery for the herniorrhaphy procedure, thus increasing the patients' risk of SSI. The question must be asked, “Are patients unnecessarily being placed at risk of an HAI?” Because examination of appropriateness of medical care and device use is of major interest to performance improvement personnel [21,46,48,49], ICPs may find areas for collaboration with their performance improvement colleagues.
Interpretating the Results
Many consider interpretation of the data to be the final step in analysis; it is the intellectual process by which meaning is ascribed to the tabulated and analyzed body of information. The interpretation may vary from no significant change in the HAI rates to the detection of a serious endemic problem or outbreak in the hospital. Often, however, more information, particularly that obtained through further investigation directed at problem areas identified by the analysis of the surveillance data, will be necessary for the final interpretation of the data. Additional uses of other information collected through surveillance, such as the time of onset of HAI, are described in Chapters 5 and 7.
Reporting the Data
The tabulated data, or at least their analyses and interpretations, should come to the attention of those people in the hospital who can take appropriate actions. A periodic report containing the tabulated data and the analytic results and their interpretations should routinely be submitted to the infection control committee and maintained on record in the hospital. Weekly or even daily reports may be necessary during outbreak or unusual situations. It is inefficient to include a line listing of HAIs in this report. When the analysis yields tables that contain insufficient data to justify inclusion in the report, the tables should be retained and a summary table released whenever sufficient data have accumulated.
The data should be displayed in graphic form to provide clinicians and/or administrators with visual evidence of the existence of a problem and the need to take preventive action. Simple, creative graphics are particularly effective. The ICP should not assume that those individuals have the time or epidemiologic expertise to interpret the data without clearly presented graphics and narratives. The availability of computer software for infection control allows graphic analyses to be performed efficiently and accurately.
In the reporting phase and throughout the surveillance process, measures should be taken to ensure the privacy of the information collected on patients and hospital staff members. For example, the ICP should keep all surveillance forms that list patients by name under very tight security, including a locked filing cabinet or other secure storage. Reports should not mention patients or staff members by name unless there is a good reason for doing so, and the distribution of the reports should be limited to those who need to know. The infection control committee should establish a policy on information privacy, including specific procedures for handling records or reports that identify patients or staff members (e.g., surgeon-specific SSI rates or laboratory data implicating an employee as a human disseminator of an outbreak organism).
Prevalence
Definition
Prevalence is a count of the number of all episodes of active disease (existing and new) during a specific period. The prevalence rate is the number of active disease episodes divided by the number of patients at risk for disease during the period. When the period used for the calculation is relatively long, such as a month or more, the measurement is called period prevalence. In Figure 6-1, the period prevalence rate of infection during Time A would be 4/10, or 40%, and during Time B, 6/10 or 60%. When the period of time used for the calculation is relatively short, such as ≤1 day, the measurement is called point prevalence (i.e., the number of all currently active episodes of disease, old and new, at a given instant in time). The difference between point and period prevalence is arbitrary; an interval that is considered a point on one time scale may become a period on a different time scale.
The Prevalence Survey
The prevalence survey, as applied to HAIs, consists of a systematic study of a defined population for evidence of
P.84
HAI at a given point in time; such a survey derives the point prevalence rate for the population. Typically, after a short period of training to standardize definitions and methods, a team of surveyors visits every patient in the hospital on a single day and detects all active HAIs by studying the medical records and examining patients or discussing patients with the clinical staff when necessary. When there are more patients than the survey team can visit in one day, the prevalence study is conducted over several days with care taken to ensure that each patient is visited only once.
If the object is to obtain the prevalence rate of infections (the usual procedure), then those infections that have resolved before the day of the visit are not counted. The point prevalence rate then is calculated by dividing the number of infections active on the day of the visit by the number of beds visited and multiplying by an appropriate base number (usually 100). If, on the other hand, the object is to obtain the point prevalence rate of infected patients, then all patients with currently active or resolved infections contracted during the current hospitalization are counted. The point prevalence rate is calculated by dividing the number of infected patients by the number of beds visited and multiplying by the appropriate base number. To be consistent with the usual way of defining the incidence rate (i.e., the infection ratio), prevalence surveys should be designed to measure the point prevalence rate of infections.
The relative magnitudes of these various measures are complicated but can be deduced from the fundamental relationship between prevalence and incidence [50]:
From this general relationship, it is apparent that prevalence rates (both point and period) are always higher than the comparable incidence rates, and the longer the duration of the infections, the greater the difference will become. As explained in the earlier discussion of the two types of incidence rates, the infection ratio is always higher than the infection proportion measured on the same population as long as some patients develop >1 infection. In contrast, the prevalence rate of infections is always lower than the prevalence rate of infected patients because the duration of an infected patient's hospitalization is almost always longer than the duration of the active infection. Interestingly, point and period prevalence rates measured on the same population are usually approximately the same although, due to the larger number of patients studied in a period prevalence survey, estimates of period prevalence are usually more precise than those of point prevalence.
In general, the most useful measure to derive from surveillance is the incidence rate because it provides an estimate of the risk of HAI uncluttered by differences in the durations of various HAIs. The main reason that prevalence rates have been used is simply that a point prevalence survey requires much less effort and can be completed much more rapidly than the ongoing, daily surveillance needed to obtain incidence rates. Point prevalence surveys have two main disadvantages: First, due to the influence of the duration of infections, the prevalence rate overestimates patients' risk of acquiring infection; second, except in the largest hospitals, the number of patients included in a point prevalence survey (i.e., the number of beds) is usually too small to obtain precise enough estimates of rates to detect important differences (e.g., a difference between the BSI rates on medicine and surgery wards) with statistical significance. Because of these limitations, prevalence surveys are generally useful primarily when a “quick and dirty” estimate is needed and time or resources to obtain a more useful measure of the incidence are insufficient [51,52].
Uses of Prevalence Surveys
Secular Trends
Repeated prevalence surveys in the same institution have been used to document secular trends in the epidemiology of HAIs and to demonstrate effectiveness of infection control measures [51,54]. Prevalence studies in large hospitals have shown shifts in the predominant pathogens associated with HAIs and in the patterns of antimicrobial use for hospitalized patients [55]. However, limitations in the numbers of patients studied and variations in the types of prevalence rates determined in the various surveys have complicated the interpretation of these results. In general, incidence rates derived from ongoing surveillance, although more time-consuming, are much better for detecting and examining secular trends. Prevalence rates should not be compared with incidence rates.
Estimation of Surveillance Accuracy
Prevalence surveys have been used to evaluate a hospital's ongoing surveillance system [56,57]. Typically, a survey team, using the same standard definitions used for routine surveillance by the ICP, visits all patients in the hospital to detect all active HAIs. By comparing the HAIs identified during the prevalence survey with those detected by the routine surveillance system and under the assumption that the survey team correctly detected all active HAIs, an estimate of the percentage of true HAIs detected by the routine surveillance system is derived. Although this percentage approximates the sensitivity of routine surveillance (i.e., the probability that the routine surveillance system will detect a true HAI), the statistic has been referred to as an efficiency factor since the difficulty of reliably determining whether HAIs are active at the time of the prevalence survey introduces some error into the assessment [57]. Because the efficiency factor approximates the sensitivity, it can be used to correct the monthly routine estimates of the incidence rates for the degree of under ascertainment.
One of the weaknesses of this application in past studies has been the failure to estimate the specificity of routine HAI surveillance in addition to its sensitivity (efficiency). Specificity is the probability of correctly classifying a patient
P.85
as uninfected, and it reflects how often the ICP records an HAI when one was not really present. Unfortunately, the specificity of case finding by ICPs in estimating incidence rates in routine surveillance has not been thoroughly studied [58,59,60]. The process of correcting incidence rates with the efficiency factor assumes that specificity is a perfect 1.0. Corrections for lower levels of specificity would give lower estimates of the incidence rates.
Estimation of Incidence from Prevalence
Modified prevalence surveys can be used to derive a crude approximation of the incidence rates that would have been obtained from continuous surveillance [61]. In this use, surveys are performed at regular intervals (e.g., weekly) that must be considerably less than the average duration of the stay of infected patients. Only HAIs that began since the preceding survey are tabulated. The denominator of the rates should include the number of admissions during the interval plus the census at the start of the period. Obviously, the shorter the interval between prevalence surveys, the more closely the final estimate approximates the true incidence rate. For modified prevelance surveys, some loss in completeness, accuracy, and timeliness in detecting HAIs, as compared to the results of surveillance studies, must be weighed against the potential benefit from a smaller time commitment to case finding.
Other Uses
Perhaps the best uses of prevalence studies are to make valuable estimates of antimicrobial usage patterns, to evaluate the adherence to proper isolation practices, and to monitor practices related to high-risk procedures, such as use of intravenous or urinary catheters [55,62,63]. In one study, investigators used sequential prevalence surveys to estimate the impact of their infection control program on the HAI risk [51].
Finally, a single institution has pooled data across prevalence surveys conducted over multiple years to understand the epidemiology of HAIs in a subset of their patient population for the purpose of designing specific prevention interventions [64].
Priority-Directed Surveillance
Since the mid-1980s, the trend has been away from continuously monitoring all patients for all HAIs in all parts of a healthcare facility (“facilitywide” or “comprehensive” surveillance) in favor of an approach that targets specific patient-care areas, infection sites, infections with certain organisms, or patient-care processes (“priority-directed” surveillance). When the level of effort is matched with the seriousness of the HAI problem, the method is called “surveillance by objective” [65]. Although such targeting was initially motivated by the need to reduce the amount of personnel time devoted to surveillance in hospitals with inadequately staffed programs, this approach has proven beneficial in reducing HAI rates in certain high-risk patients [66,67,68].
Unit-Directed Surveillance
Many hospitals have sought to maximize the use of personnel time by directing surveillance toward HAIs in patient-care areas with the highest HAI risk (e.g., ICUs, oncology units). For example, in one hospital, the HAI rate in the ICU was three times higher than in general medical-surgical patients [69]. ICUs tend to house patients who are most susceptible to HAIs—that is, the patients most likely to have suppressed immune systems, to be undergoing invasive diagnostic or therapeutic procedures, or to be receiving intensive nursing and medical care with the attendant risk of person-to-person pathogen spread. Focusing scarce resources on a few relatively small units has the advantages of greatly simplifying the surveillance effort and of preventing HAIs in the patients with the highest risks and greatest likelihood of sustaining severe and life-threatening HAIs.
Another patient-care area that should be considered for unit-directed surveillance is the high-risk nursery (HRN). Neonates in the HRN are highly susceptible to HAIs. Host and environmental factors unique to patients in this unit contribute to the high HAI risk (e.g., low birth weight). BSIs are among the most common HAIs in all birth weight groups, this frequency differing dramatically from that in adult patients with HAIs and should be a major focus for prevention [70]. The NNIS system used the unit-directed surveillance method in its ICU and HRN Surveillance Components [70,71,72].
Site-Directed Surveillance
Site-directed surveillance focuses on detecting one or more specific types of HAI (e.g., BSI) occurring among either all hospitalized patients or some subset of patients, such as those with certain devices (e.g., central lines), who have undergone certain procedures (e.g., hemodialysis), and/or are treated in certain patient-care areas (e.g., ICU). The latter example represents a combination with the unit-directed approach. The CDC's National Healthcare Safety Network (NHSN) uses this combined approach in its device-associated and procedure-associated modules [73]. In those modules, the facility chooses each month which device-associated HAI in which patient-care area and/or which procedure-associated HAI (SSI or postprocedure pneumonia) to monitor following which procedure.
Surgical Site Infections
Surveillance of SSIs is an important part of a prevention program. Unit-directed surveillance is ineffective for SSIs because these HAIs may not manifest clinically while the patient is on a particular unit. For priority-directed
P.86
SSI surveillance, all patients undergoing operations are enrolled in a surveillance registry at the time of the operation, and information on several key risk factors are recorded at that time. The risk factors most likely to be useful in the analysis are the wound classification, type and duration of the operation, and a measure of the severity of the patient's underlying disease, such as the number of underlying diagnoses or the physical status classification of the American Society of Anesthesiologists (ASA) [74,75,76]. Patients are visited regularly during hospitalization by the ICP to detect SSIs. Attempts should be made to follow up all, or a subset, of the patients for SSIs occurring after discharge [35,36].
The tabulation process should be conducted monthly and should involve three steps. First, each patient is assigned to an intrinsic infection risk category by using a multivariate risk-factor scale. Second, the SSI rate is calculated for the patients in each category of the intrinsic infection risk category—not just for clean wounds or low-risk categories. Third, two reports are compiled, one displaying each surgeon's category-specific monthly rates over time (e.g., over 1 or 2 years), and a second comparing the category-specific rates of each surgeon with those of his or her colleagues, for the service overall, and possibly an SSI rate aggregated from other hospitals [20]. Finally, the reports are given to the chief of the surgical service to distribute among and discuss with the practicing surgeons. With this plan, the surveillance time devoted to SSIs would be directed toward the surveillance effort shown to be the most effective in reducing the problem [10,35,77].
Pneumonia
A surveillance plan targeted outside ICUs and toward specific prevention objectives for healthcare-associated pneumonia (HAP) should be used. The development of HAP in ventilated patients has been the subject of much investigation and controversy. Although clinical criteria together with cultures of sputum or tracheal specimens may be sensitive for bacterial pathogens, they are nonspecific in patients with mechanically assisted ventilation [78,79]. Consensus recommendations for standardization methods to diagnose pneumonia in clinical research studies of ventilator-associated pneumonia have been proposed [80,81,82]. However, these approaches are generally not applicable to the surveillance setting. Further studies are needed to help in the diagnosis of this common clinical entity in the nonresearch setting.
For HAP in inpatients who are not on ventilators, the first consideration is to differentiate pneumonia among surgical and medical patients. The vast majority of preventable episodes of pneumonia among surgical patients are postoperative pulmonary infections representing progression of the usual atelectasis syndrome most commonly following operations of the chest and upper abdomen. In contrast, most preventable pneumonias among adult medical patients are hypostatic infections related to the failure to frequently turn patients with diminished levels of consciousness.
On pediatric and newborn services, the most serious preventable pneumonias follow person-to-person spread of HAIs with viruses (e.g., respiratory syncytial virus. Postoperative pneumonias can be detected with little additional effort if surgical patients are already being monitored for SSIs. Analysis of these rates can be performed for each type of operative procedure or by surgeon. Reports of these analyses then should be discussed just as the SSI rates are (see “Surgical Site Infections”).
The ICP also should identify patient-care areas where patients with strokes, drug overdoses, and other high risks for hypostatic pneumonia are congregated. In such areas, all patients at high risk should be followed regularly to detect all episodes of pneumonia. The pneumonia rates among these patients should be regularly reported to the charge nurses of the specified nursing units, and continuing in-service education on the importance of frequently turning these patients and providing pulmonary assistance to prevent pneumonia should be provided.
The ICP should regularly visit units caring for infants or children at high risk for pneumonia from respiratory syncytial virus and similar agents to monitor informally the frequency of upper respiratory infections among patients and employees, especially in the fall and winter months [83]. When these infections become evident, virology studies should be done immediately to detect the presence of viruses. When such pathogens are found, the staff members should be warned of the imminent danger and instructed in meticulous contact precautions to use for infected patients and employees [84]. Approaches to other types of HAP (e.g., Legionnaires' disease or influenza) are described in the revised CDC Guideline for the Prevention of Nosocomial Pneumonia [84].
Bloodstream Infections
While BSIs comprise only about 10 to 15% of all HAIs, their morbidity, mortality, and increasing frequency demand a high degree of surveillance effort [8,85]. However, one must control for a single, major risk factor before rate calculation becomes meaningful, that is, exposure to an intravascular device. Any device-associated BSI should be examined for correctable errors in patient management (e.g., failing to use all protective barriers when inserting the line, failing to change the site of intravenous catheters frequently enough, or improperly sterilizing arterial pressure-monitoring devices). Such errors must be corrected. Calculation of a CVC-BSI rate using central line-days is useful for monitoring endemic problems that may be corrected by changes in policies, practices, or selection of equipment [20].
Urinary Tract Infections
Many ICPs invest little time in the surveillance of UTIs since they are generally of less consequence to the
P.87
patient than are other HAIs. This relative inattention remains controversial because overall, UTIs remain the most frequent HAIs [86]. Surveillance efforts should be directed to identifying hospital areas where patient-care personnel are not properly managing urinary catheters or other urinary instrumentations. Then catheter-associated UTI rates (stratified or adjusted by categories of the duration of urinary catheterization) on different patient-care areas could be calculated. If rates are higher than desired, an assessment of the indications for inserting and discontinuing catheters and the techniques of aseptically caring for them should be done. By periodically reporting how the UTI rates of the different areas compared, practices could be improved in those areas with consistently higher rates. Alternatively, the infection control staff could annually conduct a 1-month prevalence study of UTIs and catheter care practices. This amount of effort would be commensurate with the magnitude of the problem and would reinforce prevention at least on a yearly basis.
Other Infections
Since HAIs at body sites other than those mentioned are comparatively rare, little time should be spent on their surveillance [20]. Instead, the ICP should depend on other hospital departments to recognize the rare outbreaks of unusual infections and to recognize common factors that might tie the patients together, such as a single unusual organism, spatial clustering, or a relationship with some diagnostic or therapeutic device. For example, the employee health service should maintain surveillance to detect problems of Mycobacterium tuberculosis transmission to employees; the director of the newborn nursery should notify the ICP of clusters of staphylococcal pyoderma; and staff who work in the microbiology laboratory should be alert to clusters of unusual pathogens. Only when a suspicious cluster or relationship is found would a more detailed investigation involving the calculation of rates be undertaken. The level of effort is commensurate with the magnitude of the problems, but the efforts are likely to detect problems if they occur. ICPs must remain alert to such problems regardless of which priority-directed surveillance approach is employed.
Surveillance by Objective
While unit-directed surveillance is efficient and effective for high-risk areas (e.g., ICUs), this approach is less practical when dealing with other patient-care areas of the hospital. An alternative approach, referred to as surveillance by objective, matches the level of surveillance effort to the seriousness of the HAI problem [8,9]. Instead of focusing on geographic areas, this priority-directed approach focuses on the types of HAIs to be prevented and assigns levels of effort commensurate to the relative seriousness of the problems.
The first prerequisite of this approach is to establish priority rankings based on the relative seriousness of the different types of HAIs. Two possible parameters for setting these priorities are compared in Table 6-1. In the past, the main parameter had been simply the relative frequency of the different types of HAIs. If this measure were used, UTI would be given the highest priority followed by SSI with pneumonia; primary BSI would receive a lower priority just ahead of a large number of relatively rare infections at other sites [86].
|
TABLE 6-1 |
|||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||
An alternative parameter that appears to be a better measure of the relative seriousness of the various types of HAIs is the total hospital cost attributable to each of the HAIs [87]. This measure reflects both the relative frequency of the HAIs and the relative degree of morbidity as expressed by the costs of the extra days and extra ancillary services necessary to treat the HAIs. By this criterion, BSI
P.88
would constitute the most serious problem followed by SSI with ventilator-associated pneumonia and UTI in third and fourth places, respectively (see Table 6-1).
HAIs have a substantial impact on patient outcomes in the hospital setting. With trends in healthcare toward shorter hospital stays, increasing the use of invasive devices, increasing the rates of antibiotic resistance, and rising public interest in decreasing HAI-related morbidity and mortality, surveillance activities will require more time and expertise than in the past. Resources for infection surveillance and prevention have not increased proportionally with these demands. Maintaining successful features of traditional surveillance systems, adopting novel surveillance strategies such as using surrogate measures for HAI rates, and employing new approaches to collecting and using information with new technologies such as the Internet will make healthcare surveillance an even better tool for prevention.
References
P.89
P.90