- Email: [email protected]
An analysis of two prevalence surveys of nosocomial infection in German intensive care units
journal
of Hospital
Infection
(1997)
35, 97-105
An analysis of two prevalence nosocomial infection in German units P. Gastmeier*,
M. Schumacher+,
surveys intensive
F. Daschnert
of care
and H. Ridden*
*Institute for Hygiene, Environmental Medicine and Occupational Health of the Free University Berlin, Hindenburgdamm 2 7,12 203, Berlin; t Institute for Medical Informatics and Medical Biometry of the Albert-LudwigsUniversity Freiburg; and $ Institute for Environmental Medicine and Hospital Epidemiology of the Albert-LudwigsUniversity Freiburg, Germany Received
11 April
1996;
revised
manuscript
accepted
7 June
1996
Summary:
In 1995, the results of two extensive prevalence studies on hospital-acquired infections were published. Both studies included a prevalence component for German intensive care units (ICUs), but provided very different infection rates. A comparison of the methods used revealed that the data from the ICUs included in the German section of EPIC (European Prevalence of Infection in Intensive Care), reflected the situation in the ICUs of large hospitals. The situation in ICUs with fewer than 600 beds was quite different, and led to the lower overall rate of infection as seen in the NIDEP (Nosocomial Infections in Germany-Surveillance and Prevention) study. Additionally, the NIDEP data permitted the calculation of device-associated, device-day, infection rates for urinary tract infections, pneumonia and bacteraemia. The differences between the ICUs in the two hospital groups were mainly due to a lower use of patient devices with regard to urinary catheters, central venous lines and respiratory ventilators. Keywords:
ICU;
EPIC;
NIDEP;
prevalence
surveys.
Introduction In the last few years, two extensive nosocomial infection prevalence studies in German intensive care units (ICUs) were published. The first was the European Prevalence of Infection in Intensive Care (EPIC) Study in 1992.’ A total of 1417 intensive care units from 17 countries in Western Europe took part in this study, of which 268 were German ICUs. The overall prevalence rate of nosocomial infection in these units was 25.4%. The other study is entitled NIDEP (Nosocomial Infections in Germany-Surveillance and Prevention), in which the first part was a prevalence study in 72 representative German hospitals in 1994 (to be published). One aim of this Correspondence 0195-6701/97/020097+09
to: Dr Petra
Gastmeier
$12.00/O
97
P. Gastmeier
98 Table
I. Methods
Type study
of prevalence
Definitions nosocomial Selection
used for infections of ICUs
Type of investigated intensive care units
Investigators
Validation investigators Exclusion patients
of the criteria
for
Recording of nosocomial infections Recording of diagnostic and therapeutic interventions Recording of microbiology reports
Statistical
analysis
et al.
used in the two prevalence
studies
EPIC
NIDEP
One day point-prevalence study (29.04.92)
One day point prevalence study on different days different hospitals over month period in 1994 CDC
CDC Invitation to participate was sent to all directors of intensive care and heads of microbiology departments in hospitals identified as having an ICLJ Coronary care units, specialized pediatric units, and special care infant units were not included in the survey Untrained hospital staff (of the ICU or infection control staff of the hospital) No
in a 10
Random selection of hospitals from the German Hospital Register according to hospital size, all ICUs belonging to the hospitals were included. Only medical, interdisciplinary included
Four trained two hygiene
surgical units
and were
physicians institutes
from
Two types of validation used No exclusion criteria
were
Patients occupying a bed in the ICU for less than 24 h, patients less than 10 years of age Two types: hospital-acquired and ICU-acquired For up to one week prior to the study day
For up to one week the study day
Whenever available inclusion of the results of bacteriological sampling undertaken on or before the day of study and available before May 6, 1992 Collected centrally, double data entry followed by a series of computer validation tests
Microbiology available prior to prevalence available day Collected data entry series of tests
Only
one
type:
nosocomial prior
to
reports for up to one week the study until the day, no reports after the prevalence centrally, followed computer
double by a validation
latter study was to investigate the prevalence of hospital-acquired infections in the medical, surgical, gynaecological/obstetrical departments and ICUs in each hospital studied. A prevalence rate of 15.3% for nosocomial infection was found in ICUs. This difference is significant and led to the question: ‘what is the real nosocomial infection rate in German ICUs?’ Methods
of the two studies
The basic surveillance methods used for both studies are shown in Table I. There are many similarities, but also some significant differences.
Infection
in ICUs
99
Selection of hospitals In the EPIC study, all hospitals in any country identified as having an ICU were contacted, and were invited to participate. Only 268 hospitals incorporating 2010 patients gave their consent to the study on the prevalence day. The total number of beds in German ICUs has been estimated as 19 705.’ In the NIDEP study, representative hospitals, randomly selected from the German Hospital Register according to size, were contacted and invited to participate. 3 The majority (75%) agreed and for the remaining 25%, substitute hospitals were selected in the same manner. The ICUs were automatically included in the study if the hospital was selected to participate and if its director agreed to be included. Selection of investigators In the EPIC study, nosocomially-infected patients were identified by the hospital staff. No information was available on the type or seniority of the hospital staff. In many hospitals, infection control nurses or hospital epidemiologists may have been involved in identifying these patients. In Germany, the majority of such staff would have only limited experience in the surveillance and identification of nosocomial infections. In other hospitals, the unit staff were exclusively responsible for the documentation of nosocomial infections, again most of them having little or no experience in surveillance. In the EPIC study they received written information on CDC definitions and on how to complete the survey record forms.’ In the NIDEP study, four physicians from the two institutes of hygiene were trained in the use of CDC definitions, and were validated during two validation periods, before starting and after finishing the surveillance period, in a Berlin hospital. Additionally, there was validation by case studies during the surveillance period. The selected hospitals were equally distributed among the four physicians. During the surveillance period, these physicians investigated the patients in each hospital on their prevalance day. According to the NIDEP guidelines, a nosocomial infection was considered prevalent as long as the infection was active or under treatment on the prevalence day. It made no difference whether the infection was acquired inside the ICU or in another hospital unit. In the EPIC study, nosocomial infections were divided into two groups: (1) hospital-acquired infections defined as infections present on admission to the ICU and considered to be related to the present hospital admission; and (2) ICU-acquired infections defined as infections having originated in the ICU and active or under treatment on the day of the study, but not clinically manifest at the time of admission to the ICU. In both studies, important and frequently used devices were recorded for a period of up to one week preceding the study day. In the NIDEP study, the onset of nosocomial infections was additionally recorded if they occurred within this period (seven-day window). Thus, for patients without a nosocomial infection on day six before the prevalence day, it was possible
100 Table
P. Gastmeier II. Demographic
data
of investigated NIDEP
ICUs and study
Number of investigated ICUs ICUs in university hospitals ICUs in university-affiliated hospitals ICUs in community hospitals Unit size O-5 beds 6-10 beds 211 beds Type of unit Medical Surgical Mixed Specialized Number of investigated patients Prevalence of community-acquired infections Prevalence of nosocomial infections (confidence interval) Hospital-acquired ICU-acquired *Calculated
from
the given
et al. results
in
the EPIC
study
EPIC
NIDEP
268 102 (44.7%) 10 (4.4%) 116 (50.9%)
4 &%) 19 (21.3%) 66 (74.2%)
85 (31.7%) 140 (52.2%) 32 (11.9%)
15 (16.9%) 53 (59.5%) 21 (23.6%)
24 (9.6%) 32 (12.8%) 180 (72.3%) 13 (5.2%) 2010 11.2%
21 (23.6%) 16 (18.0%) 52 (58.4%) 515 14.2%
25.4% (24.O;X;%)*
and
the
15.3% (12.518.6%)
17.3;
data.
to calculate device-associated incidence density rates for the window from the data obtained according to the method suggested by the NNIS system.’ For this purpose, the number of newly-developed nosocomial infections (urinary tract infections, lower respiratory tract infections and particularly pneumonia, primary sepsis) during the seven-day observation period was registered for the patient group and served as the numerator. The denominators resulted from the number of urinary catheter days, ventilator days and central line days during the observed window of non-infected patients on day six before the prevalence day. The device-associated deviceday infection rate was defined as a ratio of the number of device-associated infections for a given site and the total number of device days (multiplication factor 1000). Thus it was possible to account for exposure to the major extrinsic risk factors for nosocomial infections in ICU patients. Results
A total of 10 038 ICU patients was investigated in the EPIC study, 2010 (20.0%) were German patients.’ The NIDEP study was a prevalence study of 14 966 patients. Most of them were medical, surgical or gynaecological/ obstetrical patients, with only 515 cases from the ICUs of the 72 representatively selected hospitals. Table II lists the demographic data and results of the ICUs investigated
Infection Table
III. Prevalence of nosocomial infections of NIDEP rates and device-associated device-day infection rates
Rate
Prevalence Urinary tract infections Lower respiratory tract infections Primary sepsis Patients with at least one infection Device-utilization rate Urinary catheter Ventilation Central lines Device-associated device-day infection rate Urinary tract infections (1563 device days) Pneumonia (829 device days) Primary sepsis (1528 device days)
101
in ICUs ICU-patients, according
device-utilization to hospital size
Patients in hospitals 2600 beds n=356
Patients in hospitals >600 beds n=159
All
1.1 5.9 1.7 11.2
5.0 15.7 3.1 23.9
2.3 8.9 2.1 15.3
55.2 28.9 52.7
68.7 42.0 70.5
59.5 33.3 58.6
2.0
5.2
3.2
12.6
14.2
13.3
2.2
1.7
2.0
patients
n=515
in both studies. The high percentage of university and university-affiliated hospitals participating in the EPIC-study is remarkable. There were similarities in the data on community-acquired infections, and significant differences in the data on nosocomial infections. Table III gives a selective description of NIDEP data from ICU patients in hospitals with fewer or more than 600 beds. The rates of infection were more than twice as high in those hospitals with more than 600 beds. These hospitals had a similar infection prevalence rate, as that found in the German EPIC ICUs. The different uses of the most common devices in ICU patients of smaller and larger hospitals were seen with regard to the use of urinary catheters, central lines and ventilators. For the calculation of deviceassociated device-day infection rates, 1563 urinary catheter days, 829 ventilator-days and 1528 central-line days could be considered. Pneumonia and bacteraemia rates were similar, urinary tract infection rates were found to be much higher in larger hospitals. In the total EPIC study population, 63% were mechanically ventilated, 78.3% had some form of intravenous catheterization and 75.2% had an indwelling urinary catheter.’ There were no data available on the occurrence of these risk factors, especially with regard to the German EPIC patients. The same principle applies to the overall use of antibiotics and the availability of microbiology reports in the German EPIC patient group. It
102
P. Gastmeier
et al.
is known that 62.3% of all EPIC patients received antibiotics on the prevalence day, and microbiology reports were available for 85.0% of the nosocomially-infected patients.’ In the NIDEP ICU patient group, 52.8% had antimicrobial treatment on the prevalence day, and a microbiology report was available for 71.0% of the nosocomial infected patients. Discussion
In the last 20 years, several incidence studies of injection have been carried out in German ICUs. Incidence rates between 124-32.8% were found.610 However, most of these studies were performed in university hospitals.6-9 For this reason, there are no representative data on the situation in German ICUs, either in those of smaller community hospials or in medium-sized hospitals. The EPIC study is the largest of its kind in Europe.’ The NIDEP study is the largest prevalence study on nosocomial infections in Germany that focuses on all patients, not only ICU patients. The results of the two prevalence studies are, therefore, useful as an overview over the real nosocomial infection rate in German ICUs. In the EPIC study, a greater number of German ICUs were investigated than in the NIDEP-study, but the data of the latter were based on representatively selected hospitals. Thus, a larger number of community hospitals were involved, most of them being medium-sized hospitals with fewer than 600 beds. On the other hand, only one university hospital with four ICUs was included in the NIDEP study because of the representative selection of hospitals. Most of the university ICUs (102) were included in the group of German EPIC hospitals, but only 50.9% of the participating ICUs were represented by community hospitals. This proportion of participating ICUs was due to selecting the hospitals by invitation. University ICUs generally have more problems with nosocomial infections, and were, therefore, more likely to be interested in participating. A comparison of the prevalence rate of NIDEP study patients in hospitals with more than 600 beds, with the overall prevalence rate of the EPIC study yielded very similar results. The rate obtained in the EPIC study can thus be considered representative for most of the ICUs in large German hospitals. The data from European ICU societies suggested that the sample of the EPIC study represented 40% of all European ICUs. Concerning the United Kingdom and Switzerland, it is known that about 70% of all the eligible units participated.’ It is not clear to what extent the results were influenced by the difference in selecting the investigators. More correct and reliable data seem to be recorded when trained and validated physicians used CDC criteria. Data for the accuracy of diagnosing nosocomial infections are available for these investigators. During two validation periods, they determined nosocomial infections with a sensitivity of 89.0% and a specificity of 99.3% when
Infection
in ICUs
103
comparing them with a gold standard. When infections are recorded by the hospital staff, colonization cases are frequently confused with true infections (lower specificity). Conversely, when hospital staff are uncertain about a case, they may decide against a nosocomial infection (lower sensitivity). This could be of particular importance because of the high percentage of lower respiratory tract infections that occur as a result of nosocomial pneumonias according to CDC definitions.“-” The distribution of nosocomial infections was similar to that in the two studies and to the results of another recent extensive prevalence study in ICUs reported in the literature.” In this study, organized by the Society of Healthcare Epidemiology of America, 1671 intensive care patients in 118 ICUs were investigated. There may be two other points that are important for the differences in the EPIC and NIDEP results: in the EPIC study, patients occupying an ICU bed for less than 24 h were excluded, however this was not the case in the NIDEP study design. A dilution effect could, therefore, result from considering short-stay patients in the denominator of the NIDEP study. There was also a higher percentage of microbiology reports available in the EPIC study. The majority of the hospitals in EPIC would have had their own microbiology laboratories, providing more microbiological data compared with medium-size hospitals. If a nosocomial infection is suspected and no microbiology report is available, the infection could not be confirmed according to the strictly applied CDC criteria. Microbiology reports provide primary evidence when detecting infections such as those of the urinary tract and blood. This was sometimes the case in the NIDEP study. Comparison of the different prevalence rates in the ICUs of hospitals with more and fewer than 600 beds, in terms of different device use, raises the following question: are the higher rates observed in the ICUs of large hospitals only due to the higher device utilization rate or do they practice different device utilization? In the NIDEP study, nosocomial infections and the use of devices were not only recorded for the prevalence day. The same data were recorded for the six days preceding the prevalence day, or for the hospitalization period until the prevalence day, if the latter was less than six days. Thus it was possible to calculate device-associated device-day infection rates as those recommended for interhospital comparison by the NNIS system.” The rates were not different for urinary tract infections, pneumonia or bacteraemia. The device-associated device-day infection rates for the NIDEP study population is in agreement with the data published for the NNIS system.” Here, a much larger number of device days (505 078 urinary catheter-days, 251 229 ventilator-days and 321 810 central-line days) were analysed, and the following mean values were calculated for the combined medical/surgical and surgical ICU patient group (similar combination as the NIDEP ICU patients): 6.1 for catheter-associated urinary tract infections, 12.7 for ventilator-associated pneumonia and 4.9 for central-line-
104
P. Gastmeier
et al.
associated bloodstream infections. These rates are in agreement with those obtained in the NIDEP study. Another explanation for the differences between the ICUs in larger and smaller hospitals could be the higher incidence of serious underlying diseases among patients in university hospitals. To characterize the degree of severity of a patient’s disease, the APACHE II score was used in the EPIC study, but it no longer played a role in the multiple regression model of infection risk factors.’ In the NIDEP study, no risk-score system was used, only a documentation of single risk factors was carried out. It is, therefore, difficult to make a comparison of the predisposing factors in the two patients’ groups in the ICUs of hospitals with fewer of more than 600 beds. It is probably not so important to carry out a separate analysis of the underlying diseases of the patients because the degree of the patients’ illness may also be reflected by the extent of device use. Knowledge about the nosocomial infection rate in a country is essential to gain an overview of the magnitude of the problem to enable planning infection control activities and improving patient care. Basically two methods can be used for calculating infection rates: prevalence and incidence. Prevalence studies have various disadvantages, but they also have the advantage of being timesaving and inexpensive. For this reason, they were preferentially chosen for studies on a national scale, and used in the EPIC and NIDEP study. In most of the extensive prevalence studies, the hospitals participated on a voluntary basis. This made it easier for the organizing team of the study because these hospitals usually keep their own infection records. However, selecting hospitals in this manner very often leads to a non-representive infection rate. The data from the German EPIC ICUs, therefore, reflects mainly large hospitals with more than 600 beds. In order to get an actual overview over the situation, the hospitals must be representatively selected according to a prespecified sampling plan, even if this involves more work during the study period and even if it is more time-consuming and expensive. To compare differences between hospitals or groups of hospitals, determinating device-associated device-day infection rates on the basis of an ongoing surveillance system as recommended by the NNIS system,5’15 is more suitable than overall prevalence rates. Based on the results obtained from this analysis, we concluded that the lower infection rates in the ICUs of hospitals with fewer than 600 beds are mainly due to a low device utilization. There were only slight differences in the device-associated, device-day, infection rates. Hence, we conclude that there is a similar need for infection control activities concerning the device use in the ICUs of smaller and larger hospitals. The NIDEP-study Dr. U. Hartenauer, study.
was supported Miinster, for
by the German providing the
Ministry of Health. data of the German
We thank Professor ICUs of the EPIC
Infection
in ICUs
105
References 1. Vincent J-L, Bihari DJ, Suter PlLl et al. The prevalence of nosocomial infection in intensive care units in Europe-results of the European Prevalence of Infection in Intensive care (EPIC) Study. JAMA 1995; 274: 639-644. 2. Statistisches Bundesamt. Statistisches yahrbuch fiir die Bundevsrepublik Deutschland 1994. Metzler & Poeschel Verlag 1995. 3. Deutsches Krankenhausaerzeichnis. Deutsche Krankenhausverlaesnesellschaft mbH Dtisseldorf, 1992. 4. Garner JS, Emori WR, Horan TC, Hughes JM. CDC definitions for nosocomial infections. Am Y Infect Control 1988: 16: 128-140. 5. NNIS: Nosoco&iacinfection rates fo; interhospital comparison: limitations and possible solutions. Infect Control Hosp Epidemiol 1991; 12: 609-621. 6. Daschner F. Krankenhausinfektionen in einem Universit%tsklinikum. Dtsch med W’schv 1981; 106: 101-105. 7. Daschner F, Frey P, Wolff G, Baumann PC, Suter P. Iiosocomial infections in intensive care wards: a multicenter prospective study. Intensizx Care Med 1982; 8: S-9. 8. Hartenauer U, Thtilig B, Lawin P, Fegeler W. Infection surveillance and selective decontamination of the digestive tract (SDD) in critical ill patients-results of a controlled study. Infection 1990; 18 (Suppl. 1): 22-30. 9. Beck K, Kaufhold \V, Vock D, GroBer J. Zum infektiijsen Hospitalismus in einer Abteilung fiir Neurochirurgie-eine prospektive Studie. Z Klilz Med 1986; 41: 751-754. 10. Gastmeier K. Prospektive Studie zur Erfassung nosocomialer Infektionen auf einer multidisziplin&+en Intensivstation. Anaesthesiol-Reanim 1988; 13(6): 327-332. 11. Scheld WM, lLIande11. Nosocomial pneumonia: Pathogenesis and recent advances in diagnosis. Ret Infect Dis 1991; 13(Suppl. 9): S743-S751. 12. Tablan OF, Anderson LJ, Arden NH, Breiman RF, Butler JC, McNeil MM. The Hospital Infection Control Practices Advisory Committee. Itifect Control Hosp Epidemiol
1994; 15: 587-627. 13. 14.
15.
A’Court C, Garrard CS. Nosocomial pneumonia in the intensive care unit: mechanisms and significance. Thorax 1992; 47: 465473. Brown RB, Colodny SM, Drapkin nlS, File T, Schenfeld L, %Iaki D, Per1 T. Critical Care Committee SHEA. One-day prevalence study of 118 intensive care units. (Abstract) 5th Annual Meeting of SHEA San Diego April 1995. National Nosocomial Infection Surveillance System (NNIS) Semiannual report, May 1995. Am J Infect Control 1995; 23: 377-85.
of Hospital
Infection
(1997)
35, 97-105
An analysis of two prevalence nosocomial infection in German units P. Gastmeier*,
M. Schumacher+,
surveys intensive
F. Daschnert
of care
and H. Ridden*
*Institute for Hygiene, Environmental Medicine and Occupational Health of the Free University Berlin, Hindenburgdamm 2 7,12 203, Berlin; t Institute for Medical Informatics and Medical Biometry of the Albert-LudwigsUniversity Freiburg; and $ Institute for Environmental Medicine and Hospital Epidemiology of the Albert-LudwigsUniversity Freiburg, Germany Received
11 April
1996;
revised
manuscript
accepted
7 June
1996
Summary:
In 1995, the results of two extensive prevalence studies on hospital-acquired infections were published. Both studies included a prevalence component for German intensive care units (ICUs), but provided very different infection rates. A comparison of the methods used revealed that the data from the ICUs included in the German section of EPIC (European Prevalence of Infection in Intensive Care), reflected the situation in the ICUs of large hospitals. The situation in ICUs with fewer than 600 beds was quite different, and led to the lower overall rate of infection as seen in the NIDEP (Nosocomial Infections in Germany-Surveillance and Prevention) study. Additionally, the NIDEP data permitted the calculation of device-associated, device-day, infection rates for urinary tract infections, pneumonia and bacteraemia. The differences between the ICUs in the two hospital groups were mainly due to a lower use of patient devices with regard to urinary catheters, central venous lines and respiratory ventilators. Keywords:
ICU;
EPIC;
NIDEP;
prevalence
surveys.
Introduction In the last few years, two extensive nosocomial infection prevalence studies in German intensive care units (ICUs) were published. The first was the European Prevalence of Infection in Intensive Care (EPIC) Study in 1992.’ A total of 1417 intensive care units from 17 countries in Western Europe took part in this study, of which 268 were German ICUs. The overall prevalence rate of nosocomial infection in these units was 25.4%. The other study is entitled NIDEP (Nosocomial Infections in Germany-Surveillance and Prevention), in which the first part was a prevalence study in 72 representative German hospitals in 1994 (to be published). One aim of this Correspondence 0195-6701/97/020097+09
to: Dr Petra
Gastmeier
$12.00/O
97
P. Gastmeier
98 Table
I. Methods
Type study
of prevalence
Definitions nosocomial Selection
used for infections of ICUs
Type of investigated intensive care units
Investigators
Validation investigators Exclusion patients
of the criteria
for
Recording of nosocomial infections Recording of diagnostic and therapeutic interventions Recording of microbiology reports
Statistical
analysis
et al.
used in the two prevalence
studies
EPIC
NIDEP
One day point-prevalence study (29.04.92)
One day point prevalence study on different days different hospitals over month period in 1994 CDC
CDC Invitation to participate was sent to all directors of intensive care and heads of microbiology departments in hospitals identified as having an ICLJ Coronary care units, specialized pediatric units, and special care infant units were not included in the survey Untrained hospital staff (of the ICU or infection control staff of the hospital) No
in a 10
Random selection of hospitals from the German Hospital Register according to hospital size, all ICUs belonging to the hospitals were included. Only medical, interdisciplinary included
Four trained two hygiene
surgical units
and were
physicians institutes
from
Two types of validation used No exclusion criteria
were
Patients occupying a bed in the ICU for less than 24 h, patients less than 10 years of age Two types: hospital-acquired and ICU-acquired For up to one week prior to the study day
For up to one week the study day
Whenever available inclusion of the results of bacteriological sampling undertaken on or before the day of study and available before May 6, 1992 Collected centrally, double data entry followed by a series of computer validation tests
Microbiology available prior to prevalence available day Collected data entry series of tests
Only
one
type:
nosocomial prior
to
reports for up to one week the study until the day, no reports after the prevalence centrally, followed computer
double by a validation
latter study was to investigate the prevalence of hospital-acquired infections in the medical, surgical, gynaecological/obstetrical departments and ICUs in each hospital studied. A prevalence rate of 15.3% for nosocomial infection was found in ICUs. This difference is significant and led to the question: ‘what is the real nosocomial infection rate in German ICUs?’ Methods
of the two studies
The basic surveillance methods used for both studies are shown in Table I. There are many similarities, but also some significant differences.
Infection
in ICUs
99
Selection of hospitals In the EPIC study, all hospitals in any country identified as having an ICU were contacted, and were invited to participate. Only 268 hospitals incorporating 2010 patients gave their consent to the study on the prevalence day. The total number of beds in German ICUs has been estimated as 19 705.’ In the NIDEP study, representative hospitals, randomly selected from the German Hospital Register according to size, were contacted and invited to participate. 3 The majority (75%) agreed and for the remaining 25%, substitute hospitals were selected in the same manner. The ICUs were automatically included in the study if the hospital was selected to participate and if its director agreed to be included. Selection of investigators In the EPIC study, nosocomially-infected patients were identified by the hospital staff. No information was available on the type or seniority of the hospital staff. In many hospitals, infection control nurses or hospital epidemiologists may have been involved in identifying these patients. In Germany, the majority of such staff would have only limited experience in the surveillance and identification of nosocomial infections. In other hospitals, the unit staff were exclusively responsible for the documentation of nosocomial infections, again most of them having little or no experience in surveillance. In the EPIC study they received written information on CDC definitions and on how to complete the survey record forms.’ In the NIDEP study, four physicians from the two institutes of hygiene were trained in the use of CDC definitions, and were validated during two validation periods, before starting and after finishing the surveillance period, in a Berlin hospital. Additionally, there was validation by case studies during the surveillance period. The selected hospitals were equally distributed among the four physicians. During the surveillance period, these physicians investigated the patients in each hospital on their prevalance day. According to the NIDEP guidelines, a nosocomial infection was considered prevalent as long as the infection was active or under treatment on the prevalence day. It made no difference whether the infection was acquired inside the ICU or in another hospital unit. In the EPIC study, nosocomial infections were divided into two groups: (1) hospital-acquired infections defined as infections present on admission to the ICU and considered to be related to the present hospital admission; and (2) ICU-acquired infections defined as infections having originated in the ICU and active or under treatment on the day of the study, but not clinically manifest at the time of admission to the ICU. In both studies, important and frequently used devices were recorded for a period of up to one week preceding the study day. In the NIDEP study, the onset of nosocomial infections was additionally recorded if they occurred within this period (seven-day window). Thus, for patients without a nosocomial infection on day six before the prevalence day, it was possible
100 Table
P. Gastmeier II. Demographic
data
of investigated NIDEP
ICUs and study
Number of investigated ICUs ICUs in university hospitals ICUs in university-affiliated hospitals ICUs in community hospitals Unit size O-5 beds 6-10 beds 211 beds Type of unit Medical Surgical Mixed Specialized Number of investigated patients Prevalence of community-acquired infections Prevalence of nosocomial infections (confidence interval) Hospital-acquired ICU-acquired *Calculated
from
the given
et al. results
in
the EPIC
study
EPIC
NIDEP
268 102 (44.7%) 10 (4.4%) 116 (50.9%)
4 &%) 19 (21.3%) 66 (74.2%)
85 (31.7%) 140 (52.2%) 32 (11.9%)
15 (16.9%) 53 (59.5%) 21 (23.6%)
24 (9.6%) 32 (12.8%) 180 (72.3%) 13 (5.2%) 2010 11.2%
21 (23.6%) 16 (18.0%) 52 (58.4%) 515 14.2%
25.4% (24.O;X;%)*
and
the
15.3% (12.518.6%)
17.3;
data.
to calculate device-associated incidence density rates for the window from the data obtained according to the method suggested by the NNIS system.’ For this purpose, the number of newly-developed nosocomial infections (urinary tract infections, lower respiratory tract infections and particularly pneumonia, primary sepsis) during the seven-day observation period was registered for the patient group and served as the numerator. The denominators resulted from the number of urinary catheter days, ventilator days and central line days during the observed window of non-infected patients on day six before the prevalence day. The device-associated deviceday infection rate was defined as a ratio of the number of device-associated infections for a given site and the total number of device days (multiplication factor 1000). Thus it was possible to account for exposure to the major extrinsic risk factors for nosocomial infections in ICU patients. Results
A total of 10 038 ICU patients was investigated in the EPIC study, 2010 (20.0%) were German patients.’ The NIDEP study was a prevalence study of 14 966 patients. Most of them were medical, surgical or gynaecological/ obstetrical patients, with only 515 cases from the ICUs of the 72 representatively selected hospitals. Table II lists the demographic data and results of the ICUs investigated
Infection Table
III. Prevalence of nosocomial infections of NIDEP rates and device-associated device-day infection rates
Rate
Prevalence Urinary tract infections Lower respiratory tract infections Primary sepsis Patients with at least one infection Device-utilization rate Urinary catheter Ventilation Central lines Device-associated device-day infection rate Urinary tract infections (1563 device days) Pneumonia (829 device days) Primary sepsis (1528 device days)
101
in ICUs ICU-patients, according
device-utilization to hospital size
Patients in hospitals 2600 beds n=356
Patients in hospitals >600 beds n=159
All
1.1 5.9 1.7 11.2
5.0 15.7 3.1 23.9
2.3 8.9 2.1 15.3
55.2 28.9 52.7
68.7 42.0 70.5
59.5 33.3 58.6
2.0
5.2
3.2
12.6
14.2
13.3
2.2
1.7
2.0
patients
n=515
in both studies. The high percentage of university and university-affiliated hospitals participating in the EPIC-study is remarkable. There were similarities in the data on community-acquired infections, and significant differences in the data on nosocomial infections. Table III gives a selective description of NIDEP data from ICU patients in hospitals with fewer or more than 600 beds. The rates of infection were more than twice as high in those hospitals with more than 600 beds. These hospitals had a similar infection prevalence rate, as that found in the German EPIC ICUs. The different uses of the most common devices in ICU patients of smaller and larger hospitals were seen with regard to the use of urinary catheters, central lines and ventilators. For the calculation of deviceassociated device-day infection rates, 1563 urinary catheter days, 829 ventilator-days and 1528 central-line days could be considered. Pneumonia and bacteraemia rates were similar, urinary tract infection rates were found to be much higher in larger hospitals. In the total EPIC study population, 63% were mechanically ventilated, 78.3% had some form of intravenous catheterization and 75.2% had an indwelling urinary catheter.’ There were no data available on the occurrence of these risk factors, especially with regard to the German EPIC patients. The same principle applies to the overall use of antibiotics and the availability of microbiology reports in the German EPIC patient group. It
102
P. Gastmeier
et al.
is known that 62.3% of all EPIC patients received antibiotics on the prevalence day, and microbiology reports were available for 85.0% of the nosocomially-infected patients.’ In the NIDEP ICU patient group, 52.8% had antimicrobial treatment on the prevalence day, and a microbiology report was available for 71.0% of the nosocomial infected patients. Discussion
In the last 20 years, several incidence studies of injection have been carried out in German ICUs. Incidence rates between 124-32.8% were found.610 However, most of these studies were performed in university hospitals.6-9 For this reason, there are no representative data on the situation in German ICUs, either in those of smaller community hospials or in medium-sized hospitals. The EPIC study is the largest of its kind in Europe.’ The NIDEP study is the largest prevalence study on nosocomial infections in Germany that focuses on all patients, not only ICU patients. The results of the two prevalence studies are, therefore, useful as an overview over the real nosocomial infection rate in German ICUs. In the EPIC study, a greater number of German ICUs were investigated than in the NIDEP-study, but the data of the latter were based on representatively selected hospitals. Thus, a larger number of community hospitals were involved, most of them being medium-sized hospitals with fewer than 600 beds. On the other hand, only one university hospital with four ICUs was included in the NIDEP study because of the representative selection of hospitals. Most of the university ICUs (102) were included in the group of German EPIC hospitals, but only 50.9% of the participating ICUs were represented by community hospitals. This proportion of participating ICUs was due to selecting the hospitals by invitation. University ICUs generally have more problems with nosocomial infections, and were, therefore, more likely to be interested in participating. A comparison of the prevalence rate of NIDEP study patients in hospitals with more than 600 beds, with the overall prevalence rate of the EPIC study yielded very similar results. The rate obtained in the EPIC study can thus be considered representative for most of the ICUs in large German hospitals. The data from European ICU societies suggested that the sample of the EPIC study represented 40% of all European ICUs. Concerning the United Kingdom and Switzerland, it is known that about 70% of all the eligible units participated.’ It is not clear to what extent the results were influenced by the difference in selecting the investigators. More correct and reliable data seem to be recorded when trained and validated physicians used CDC criteria. Data for the accuracy of diagnosing nosocomial infections are available for these investigators. During two validation periods, they determined nosocomial infections with a sensitivity of 89.0% and a specificity of 99.3% when
Infection
in ICUs
103
comparing them with a gold standard. When infections are recorded by the hospital staff, colonization cases are frequently confused with true infections (lower specificity). Conversely, when hospital staff are uncertain about a case, they may decide against a nosocomial infection (lower sensitivity). This could be of particular importance because of the high percentage of lower respiratory tract infections that occur as a result of nosocomial pneumonias according to CDC definitions.“-” The distribution of nosocomial infections was similar to that in the two studies and to the results of another recent extensive prevalence study in ICUs reported in the literature.” In this study, organized by the Society of Healthcare Epidemiology of America, 1671 intensive care patients in 118 ICUs were investigated. There may be two other points that are important for the differences in the EPIC and NIDEP results: in the EPIC study, patients occupying an ICU bed for less than 24 h were excluded, however this was not the case in the NIDEP study design. A dilution effect could, therefore, result from considering short-stay patients in the denominator of the NIDEP study. There was also a higher percentage of microbiology reports available in the EPIC study. The majority of the hospitals in EPIC would have had their own microbiology laboratories, providing more microbiological data compared with medium-size hospitals. If a nosocomial infection is suspected and no microbiology report is available, the infection could not be confirmed according to the strictly applied CDC criteria. Microbiology reports provide primary evidence when detecting infections such as those of the urinary tract and blood. This was sometimes the case in the NIDEP study. Comparison of the different prevalence rates in the ICUs of hospitals with more and fewer than 600 beds, in terms of different device use, raises the following question: are the higher rates observed in the ICUs of large hospitals only due to the higher device utilization rate or do they practice different device utilization? In the NIDEP study, nosocomial infections and the use of devices were not only recorded for the prevalence day. The same data were recorded for the six days preceding the prevalence day, or for the hospitalization period until the prevalence day, if the latter was less than six days. Thus it was possible to calculate device-associated device-day infection rates as those recommended for interhospital comparison by the NNIS system.” The rates were not different for urinary tract infections, pneumonia or bacteraemia. The device-associated device-day infection rates for the NIDEP study population is in agreement with the data published for the NNIS system.” Here, a much larger number of device days (505 078 urinary catheter-days, 251 229 ventilator-days and 321 810 central-line days) were analysed, and the following mean values were calculated for the combined medical/surgical and surgical ICU patient group (similar combination as the NIDEP ICU patients): 6.1 for catheter-associated urinary tract infections, 12.7 for ventilator-associated pneumonia and 4.9 for central-line-
104
P. Gastmeier
et al.
associated bloodstream infections. These rates are in agreement with those obtained in the NIDEP study. Another explanation for the differences between the ICUs in larger and smaller hospitals could be the higher incidence of serious underlying diseases among patients in university hospitals. To characterize the degree of severity of a patient’s disease, the APACHE II score was used in the EPIC study, but it no longer played a role in the multiple regression model of infection risk factors.’ In the NIDEP study, no risk-score system was used, only a documentation of single risk factors was carried out. It is, therefore, difficult to make a comparison of the predisposing factors in the two patients’ groups in the ICUs of hospitals with fewer of more than 600 beds. It is probably not so important to carry out a separate analysis of the underlying diseases of the patients because the degree of the patients’ illness may also be reflected by the extent of device use. Knowledge about the nosocomial infection rate in a country is essential to gain an overview of the magnitude of the problem to enable planning infection control activities and improving patient care. Basically two methods can be used for calculating infection rates: prevalence and incidence. Prevalence studies have various disadvantages, but they also have the advantage of being timesaving and inexpensive. For this reason, they were preferentially chosen for studies on a national scale, and used in the EPIC and NIDEP study. In most of the extensive prevalence studies, the hospitals participated on a voluntary basis. This made it easier for the organizing team of the study because these hospitals usually keep their own infection records. However, selecting hospitals in this manner very often leads to a non-representive infection rate. The data from the German EPIC ICUs, therefore, reflects mainly large hospitals with more than 600 beds. In order to get an actual overview over the situation, the hospitals must be representatively selected according to a prespecified sampling plan, even if this involves more work during the study period and even if it is more time-consuming and expensive. To compare differences between hospitals or groups of hospitals, determinating device-associated device-day infection rates on the basis of an ongoing surveillance system as recommended by the NNIS system,5’15 is more suitable than overall prevalence rates. Based on the results obtained from this analysis, we concluded that the lower infection rates in the ICUs of hospitals with fewer than 600 beds are mainly due to a low device utilization. There were only slight differences in the device-associated, device-day, infection rates. Hence, we conclude that there is a similar need for infection control activities concerning the device use in the ICUs of smaller and larger hospitals. The NIDEP-study Dr. U. Hartenauer, study.
was supported Miinster, for
by the German providing the
Ministry of Health. data of the German
We thank Professor ICUs of the EPIC
Infection
in ICUs
105
References 1. Vincent J-L, Bihari DJ, Suter PlLl et al. The prevalence of nosocomial infection in intensive care units in Europe-results of the European Prevalence of Infection in Intensive care (EPIC) Study. JAMA 1995; 274: 639-644. 2. Statistisches Bundesamt. Statistisches yahrbuch fiir die Bundevsrepublik Deutschland 1994. Metzler & Poeschel Verlag 1995. 3. Deutsches Krankenhausaerzeichnis. Deutsche Krankenhausverlaesnesellschaft mbH Dtisseldorf, 1992. 4. Garner JS, Emori WR, Horan TC, Hughes JM. CDC definitions for nosocomial infections. Am Y Infect Control 1988: 16: 128-140. 5. NNIS: Nosoco&iacinfection rates fo; interhospital comparison: limitations and possible solutions. Infect Control Hosp Epidemiol 1991; 12: 609-621. 6. Daschner F. Krankenhausinfektionen in einem Universit%tsklinikum. Dtsch med W’schv 1981; 106: 101-105. 7. Daschner F, Frey P, Wolff G, Baumann PC, Suter P. Iiosocomial infections in intensive care wards: a multicenter prospective study. Intensizx Care Med 1982; 8: S-9. 8. Hartenauer U, Thtilig B, Lawin P, Fegeler W. Infection surveillance and selective decontamination of the digestive tract (SDD) in critical ill patients-results of a controlled study. Infection 1990; 18 (Suppl. 1): 22-30. 9. Beck K, Kaufhold \V, Vock D, GroBer J. Zum infektiijsen Hospitalismus in einer Abteilung fiir Neurochirurgie-eine prospektive Studie. Z Klilz Med 1986; 41: 751-754. 10. Gastmeier K. Prospektive Studie zur Erfassung nosocomialer Infektionen auf einer multidisziplin&+en Intensivstation. Anaesthesiol-Reanim 1988; 13(6): 327-332. 11. Scheld WM, lLIande11. Nosocomial pneumonia: Pathogenesis and recent advances in diagnosis. Ret Infect Dis 1991; 13(Suppl. 9): S743-S751. 12. Tablan OF, Anderson LJ, Arden NH, Breiman RF, Butler JC, McNeil MM. The Hospital Infection Control Practices Advisory Committee. Itifect Control Hosp Epidemiol
1994; 15: 587-627. 13. 14.
15.
A’Court C, Garrard CS. Nosocomial pneumonia in the intensive care unit: mechanisms and significance. Thorax 1992; 47: 465473. Brown RB, Colodny SM, Drapkin nlS, File T, Schenfeld L, %Iaki D, Per1 T. Critical Care Committee SHEA. One-day prevalence study of 118 intensive care units. (Abstract) 5th Annual Meeting of SHEA San Diego April 1995. National Nosocomial Infection Surveillance System (NNIS) Semiannual report, May 1995. Am J Infect Control 1995; 23: 377-85.