INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The relationship between nosocomial infections and in-hospital mortality remains unclear. The report from the Study on the Efficacy of Nosocomial Infection Control (SENIC) project (1) estimated that at least 2.1 million nosocomial infections occurred annually among 37.7 million admissions in United States hospital, and considered 77,000 deaths to be associated with nosocomial infections (2). The highest rates of nosocomial infections are observed in intensive care units (ICUs), which are also the units in which the most severely ill patients are treated and in which the highest mortality rates are observed (3). The latter results from the acute severity of illness of ICU patients and their frequent exposure to therapeutic procedures. Thus, ICU patients are both at risk for acquiring nosocomial infections and at risk for dying. Establishing a relationship between severity of illness, therapeutic activity, occurrence of nosocomial infections, and outcome requires separate analyses of illness severity and therapeutic activity as causes of nosocomial infections, and of nosocomial infections as causes of excess illness severity and extra therapeutic activity. Various systems for scoring severity of illness and extent of therapeutic activity have been developed and applied to ICU populations (4). To date, only illness severity evaluated on admission, as measured with the Acute Physiology and Chronic Health Evaluation (APACHE) score, Therapeutic Intervention Scoring System (TISS), or Pediatric Risk of Mortality Score (PRISM), has been associated with the occurrence of nosocomial infection (7). Similarly, the influence of time on development of nosocomial infection has previously been approached only through the duration of exposure to risk, defined as the stay in the ICU and/or the duration of use of invasive procedures such as mechanical ventilation or catheterization, which reflects therapeutic activity. On the other hand, the specific consequences of nosocomial infections in terms of excess morbidity and excess mortality have recently been investigated through case-control studies (10, 11) that have precisely identified the direct contribution of infection by pairing infected and noninfected patients according to the severity of their underlying disease. By contrast, no study has examined the daily variations of illness severity and intensity of therapeutic activity during patients' ICU stays with regard to nosocomial infections and subsequent deaths.

In the present study, we attempted to determine the roles of evolution of severity of illness and therapeutic activity in the onset of nosocomial infection. To do so, we undertook a case-control study in which the case and control groups were carefully matched on admission for severity of illness and ICU length of stay. Furthermore, we also analyzed the overall effects relative to the development of nosocomial infection of mortality, excess length of stay, evolution of the patient's disease severity, and therapeutic activity after infection.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design

We performed a pairwise, retrospective case-control study with 1:1 matching. The study was conducted in the Service de Réanimation Médicale of Broussais University Hospital in Parisa 10-bed ICU that receives patients from all departments in the hospital and from ICUs of other hospitals. The study period ran from January 1, 1992, to December 31, 1992, during which time 281 patients were admitted in the unit.

Case Identification

Patients who developed pneumonia, urinary tract infection (UTI), primary bacteremia, central venous catheter-related infection, sinusitis, or surgical wound infection according to the definitions given subsequently and at least 48 h after ICU admission were classified as cases. Patients with new and persistent lung infiltrates on chest roentgenograms and with macroscopically purulent tracheal secretions were suspected of having nosocomial pneumonia. These patients underwent fiberoptic bronchoscopy, with the diagnosis of nosocomial pneumonia ascertained by: (1) positivity of a quantitative protected-specimen-brush (PSB) culture, defined as at least one microorganism recovered at a significant concentration ( 10³ cfu/ml) (12); and/or (2) positivity of bronchoalveolar lavage fluid (BALF) upon direct examination, defined as more than 5% of cells containing intracellular bacteria (13). In patients clinically suspected of having an infection, diagnosis of central-venous-catheter-related infection was confirmed by a positive quantitative tip culture with a significant threshold of 10³ cfu/ml (14). Diagnosis of primary bacteremia was confirmed by at least one positive blood culture (two or more blood cultures when coagulase-negative staphylococci were isolated) without another site simultaneously infected with the same microorganism. A UTI was defined by association of the two following criteria: pyuria ( 10 white blood cells [WBC]/mm³) and a urine culture of 10⁵ cfu/ml in patients with clinical signs of infection (fever > 38° C, leukocytosis, abnormal macroscopic appearance of urine, and presence of urinary nitrites). Sinusitis was suspected in patients with fever and/or purulent nasal secretions who had a radiologic opacification of the maxillary sinuses, and was confirmed by a sinus aspirate containing more than five altered polymorphonuclear leukocytes (PMNs) per oil-immersion field, and by a positive microbiologic culture with a quantitative threshold of 10⁴ cfu/ml (15). The diagnosis of surgical-wound infections was based on clinical examination and confirmed by microbiologic analysis of specimens.

Matching and Selection of Controls

A control had to have had no evidence of nosocomial infection at any time during hospitalization in the ICU. A computer-generated list of potential controls was obtained from a data base including 1,169 patients that was constituted over a 4-yr period (from 1991 to 1994). Controls were selected according to the following matching criteria: age (± 5 yr) and APACHE II score calculated on the first day of ICU admission (± 5 points). In addition, the length of stay for controls had to be at least equal to the interval, for cases, from admission to the occurrence of first infection, to ensure the same duration of exposure to risk. The list of potential controls was reviewed for the best possible match, giving highest priority to duration of exposure to risk, APACHE II score, and age. In the case of multiple acceptable controls, the one with the date of ICU admission closest to that of the patient was chosen.

Collection of Data

The following variables were recorded: age; sex; dates of admission and discharge from the ICU; location before ICU admission; severity of underlying medical conditions at admission, stratified according to the criteria of McCabe and Jackson (16) as fatal, ultimately fatal, and nonfatal; first-day and third-day Organ Dysfunction and/or Infection (ODIN) scores (17), based on the presence or absence of cardiac, respiratory, renal, hepatic, neurologic, and/or hematologic dysfunctions and/or infection; Glasgow coma scale (GCS) score (18) at admission; weight; and serum protein level at admission. The severity of illness was evaluated with the APACHE II score (4) and the Simplified Acute Physiology Score (SAPS) (6) calculated every day from ICU admission to discharge. Therapeutic activity was evaluated daily, using TISS (5) and the Omega score (19) before and after the onset of nosocomial infection. The Omega score is composed of therapeutic items accorded 1 to 10 points, and is divided into three categories as follows: Category 1, items entered only at the time of their first application; Category 2, items entered at each application; and Category 3, items entered every day of application. The total score, which covers the entire length of stay, is calculated by adding the points obtained in the three categories (Table 1).

Detailed information on invasive procedures and treatment before and after the onset of nosocomial infection was also recorded: the durations of mechanical ventilation, central venous catheterization by insertion sites, urinary catheterization, and/or use of a nasogastric tube; the number of fibroscopies of the digestive tract or respiratory tract; surgical intervention; and dialysis.

Variables concerning antibiotic and sedative treatments were also noted, including antibiotics received within 15 d before ICU admission, antibiotic regimens given before and after the onset of nosocomial infection, and sedatives administered before infection.

The responsible pathogens, dates of sampling, and sites of each nosocomial infection were also recorded.

Statistical Analysis

The characteristics of patients in both groups were compared through use of the Mann-Whitney nonparametric test for continuous variables and the chi-square test for categorical variables. Wilcoxon's test was used to compare two continuous variables within one group. To identify risk factors independently associated with nosocomial infection, variables found to be significantly different between cases and controls in the univariate analysis were entered into a forward stepwise logistic-regression model (Statistica 4.5; Statsoft, Inc., Tulsa, OK). When patients developed multiple nosocomial infections during their hospitalization, only the first episode was used in the risk factor analysis. A value of p < 0.05 constituted a significant difference.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Effectiveness of Matching

Forty-one of the 281 studied patients developed at least one nosocomial infection, producing a global incidence rate of 14.6 infections per 100 admissions. The results of matching for the criteria listed earlier were as follows: all except one control patient (40 of 41 controls, 97.6%) were matched to cases for duration of exposure to risk. The average duration of exposure to risk of infection was similar for cases and controls (8.7 ± 6.8 d). Forty of the 41 case-control pairs (97.6%) were matched for APACHE II score and 38 of the 41 (92.7%) for age. Overall, matching was successful for 118 of 123 (95.9%) variables used. Eleven other variables were also compared in the two patient groups on admission, and no significant difference was observed between cases and controls (Table 2).

Description of Nosocomial Infections

During their ICU stay, the 41 case-patients developed 98 nosocomial infections (2.4 episodes per patient): 33 urinary-tract infections, 35 bacteremias, 15 pneumonias, 12 central-venous-catheter-related infections, two sinusitides, and one surgical wound infection. Of the 35 episodes of bacteremia, only four were primary; the other 31 complicated the following nosocomial infections: 14 urinary tract infections, eight catheter-related infections, eight instances of pneumonia, and one surgical-site infection. In all, 25 cases (61%) were polyinfected: two nosocomial infections developed in 12 of these cases and three or more nosocomial infections developed in 13 cases. The first episodes of infection were distributed as follows: 26 UTIs, eight pneumonias, four bacteremias, and three central-venous-catheter-related infections.

Risk Factors

Univariate analysis identified eight variables that differed significantly in the cases versus the controls (Table 3). None of the other variables recorded differed significantly. Multivariate analysis, which incorporated these eight variables in the stepwise logistic-regression model, revealed that neurologic failure on the third day after ICU admission was the only variable independently associated with the occurrence of nosocomial infection (p = 0.007).

As shown in Table 4, no significant difference between cases and controls was found for the duration of mechanical ventilation in 38 cases and 29 controls, for central-venous catheterization in 29 cases and 25 controls, for nasogastric intubation in 38 cases and 27 controls, or for urinary catheterization in 41 cases and 33 controls before the onset of nosocomial infection. Moreover, the numbers of digestive or pulmonary fibroscopies and the numbers of surgical interventions before the onset of nosocomial infection were similar in both groups (data not shown).

The time-dependent analysis of APACHE II score, SAPS, and TISS before the onset of nosocomial infection was conducted for the 40 case-control pairs correctly matched for the duration of exposure to risk (Figure 1). Although the three scores did not differ statistically for cases versus controls on admission, as assessed through the matching process, all three scores were significantly higher in cases than in controls at 3 d after ICU admission and at 1 d before the onset of nosocomial infection. Within each scoring system, comparison of the scores measured on admission with those measured on the day preceding (theoretical in control patients) nosocomial infection in each group evidenced significant improvement only in controls, as follows: APACHE II score, 23.7 ± 8.4 on admission and 14.6 ± 7.5 on the day before theoretical infection (p < 0.0001); SAPS, 14.3 ± 5.2 and 9.2 ± 4.1, respectively (p < 0.0001); and TISS, 24.3 ± 11.4 and 16.5 ± 8.4, respectively (p < 0.0001). The same comparison for cases gave: APACHE II score, 24.0 ± 8.0 on admission and 18.5 ± 5.9 on the day before infection (p < 0.05); SAPS, 13.8 ± 4.6 and 12.2 ± 3.3, respectively (p = NS); and TISS, 26.1 ± 9.4 and 24.6 ± 9.0, respectively (p = NS).

View larger version (25K): [in this window] [in a new window]   Figure 1.   Comparison of scores between cases (solid triangles) and controls (open circles) on the first 3 d (D1, D2, D3) in the ICU, 7 d before infection (7 to 1), the day of nosocomial infection (NI), and 5 d after infection (+1 to +5). *p < 0.05, cases versus controls.

Outcome

According to the crude mortality rates observed for cases (24 of 41, 58.5%) and for controls (six of 41, 14.6%), the mortality attributable to nosocomial infections was 44% (p < 0.001) with a relative risk of death of 4.0.

As shown in Table 4, all durations of use of invasive devices were significantly longer in cases than in controls after the onset of nosocomial infection.

The time-dependent analysis of illness severity and therapeutic activity scores after nosocomial infection (Figure 1) showed that the APACHE II score, SAPS, and TISS were significantly higher for cases than for controls, respectively, on the day of nosocomial infection; the APACHE II score was 18.8 ± 7.5 points for cases versus 14.2 ± 6.9 points for controls (p = 0.01); the SAPS was 12.6 ± 5.4 versus 8.8 ± 4.4, respectively (p = 0.01); and the TISS was 25.4 ± 10.0 versus 15.7 ± 8.7, respectively (p = 0.01). On the third day after the occurrence of nosocomial infection, the APACHE II score was 19.1 ± 4.9 points for cases and 13.3 ± 8.5 points for controls (p  0.02); the SAPS was 12.8 ± 4.7 versus 9.0 ± 5.4, respectively (p  0.02); and the TISS was 26.1 ± 8.1 versus 14.7 ± 9.6, respectively (p  0.02).

The ICU stay was significantly prolonged for cases; their antibiotic consumption was also significantly greater, and their therapeutic activity, as evaluated from the TISS and Omega score, was significantly greater than for controls (Table 5). When the survivor pairs were considered alone, the excess length of stay attributable to nosocomial infection was 22 d.

The outcome variables attributable to nosocomial infection were calculated on the basis of three categories of case-patients; those who developed one, two, or three or more infections, by comparison with their respective controls (Figure 2). Differences between patients with one nosocomial infection and their respective controls were not significant, regardless of the method of calculation used. This category included 11 UTIs, two pneumonias, two primary bacteremias, and one central-venous-catheter-related infection. Conversely, case-patients who developed two infections or three or more infections during their ICU stay cost significantly more than did their controls. Crude mortality rates stratified by category were: seven of 16 (43.7%) for cases with one nosocomial infection and none of 16 (0%) for their respective controls (p = 0.003); seven of 12 (58.3%) for cases with two infections and three of 12 (25%) for their controls (p = NS); and nine of 12 (75%) for cases with three or more infections versus three of 12 (25%) for their controls (p = 0.04).

View larger version (12K): [in this window] [in a new window]   Figure 2.   Comparison of lengths of stay, durations of antibiotic treatment, TISS, and Omega scores between nosocomially infected (NI) cases with 1, 2, or  3 infections (solid circles), and their controls (open circles). *p < 0.05, cases versus controls.

Crude mortality (52% in UTI patients versus 82% in patients with other infections; p = NS), duration of ICU stay, and resource use estimated by TISS were not significantly different for patients with UTI and patients with other infections. Conversely, bacteremias, which occurred in 35 patients, were significantly associated with increased mortality as compared with other infections (85% versus 35%, p = 0.005), as were also increased length of stay (26 ± 32 d versus 15 ± 21 d, p = 0.04) and increased therapeutic activity (604 ± 586 versus 313 ± 437 TISS points, p = 0.02). Attributable costs calculated by three different methods based on therapeutic activity, length of ICU stay, and antibiotic consumption measured after the onset of nosocomial infection were significantly higher in case-patients than in controls (data not shown).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We performed a case-control study, intending to evaluate the influences of severity of illness and evolution of therapeutic activity on the development of nosocomial infections, and to estimate the attributable consequences of these infections on ICU patients. Neurologic failure on the third day following admission was identified as the major component favoring the occurrence of nosocomial infection. Patients with no signs of clinical improvement and requiring a sustained high level of therapeutic intervention were also predisposed to acquire an infection during their stay. Regarding the outcome attributable to nosocomial infections, we demonstrated excess mortality, prolonged ICU stay, higher antibiotic consumption, and increased therapeutic activity, which led to considerable cost overruns. We matched cases and controls on admission by using variables known to be strongly correlated with the risk of nosocomial infection (e.g., duration of exposure to risk and severity of illness) and the risk of death (e.g., severity of illness and age). The purpose of such a matching process was to control for initial selection bias and, therefore, to evaluate the role of daily variations of disease severity in the onset of nosocomial infection independently of the patient's initial status. In addition, this matching technique discriminated between the contributions of nosocomial infections and underlying medical conditions on subsequent outcome (i.e., to assess the attributable outcome). Matching was 96% successful for 123 possible variables. To verify independently the adequacy of the matching for the severity of underlying illness, we compared cases and controls with respect to other potentially confounding variables, including the McCabe score, TISS, GCS score, and ODIN score on admission. No significant differences in these indices were observed between cases and controls. We chose to pair cases and controls at admission for daily evaluation of variations in illness severity and therapeutic activity between admission and the day of the first nosocomial infection.

The criteria used to diagnose nosocomial infections in this study were those routinely applied in ICUs, with the application of strict criteria to define bacteriologically proven ventilator-associated pneumonia (VAP) (13). Close monitoring of nosocomial infections through a continuous surveillance system introduced in the unit in 1991 made it most unlikely that an episode of infection would be missed or the number of cases of infection would be overestimated.

The analysis of risk factors was based on disease severity and therapeutic activity scores usually measured in ICUs. Disease severity scores, such as the APACHE II score and the SAPS, are used to evaluate physiologic disturbances (4, 6), whereas organ-failure scoring systems, such as the ODIN model (17), assess major comorbidities. Therapeutic activity scores, such as the TISS and Omega score, reflect the global burden of care applied during the entire ICU stay by adding the number of medical and nursing interventions (5). Many studies (7, 20) have shown that the severity of underlying disease measured within the first 24 h of intensive care is a risk factor for nosocomial infection. By contrast, and to the best of our knowledge for the first time, our study was based on the daily comparison of severity of illness and therapeutic activity in cases versus controls and within the same group during the entire ICU stay. We demonstrated that only neurologic failure on the third day in the ICU remained an independent predictor for the acquisition of nosocomial infection, thereby indicating the importance of persistent impaired consciousness. This organ dysfunction has often been associated with acquisition of nosocomial infection, especially in trauma or neurosurgical patients. In a multivariate analysis of risk factors in 1,325 ICU patients, in which the urinary tract represented the most frequently identified site of infection (13%), Craven and colleagues (21) found that coma, head trauma, and neurologic disease present on admission were univariately associated with nosocomial infection. With a logistic regression model, they also identified monitoring of intracranial pressure as an independent predictor of infection. Similarly, in several studies using multivariate analysis to define independent risk factors, a depressed level of consciousness was also evidenced as significantly predisposing to the development of pneumonia (22); however, the exact relationship between coma and colonization and/or infection of the lower airways remains to be clarified. Besides favoring respiratory-tract colonization, neurologic dysfunction may be a marker of impairment of host defenses or increased use of invasive devices (26).

Studies that previously identified therapeutic activity as a potential risk factor for nosocomial infection were based either on an analysis of invasive-device use (e.g., venous catheterization, intubation, urinary or nasogastric tubes); the utilization of drugs such as sedatives, antacids and H₂-blockers; or the influence of surgical procedures (21, 27). Adjusting for initial disease severity, we found that the duration of invasive- device use did not differ in cases versus controls before the onset of nosocomial infection and, consequently, was probably not a major risk factor for the acquisition of nosocomial infection. On the other hand, our analysis identified two major determinants of the development of infection: (1) a sustained level of therapeutic activity as evaluated with daily TISS measurement, with a significantly greater workload in the days preceding the onset of nosocomial infection; and (2) a constant increase in differences in severity of illness scores between cases and controls in the days preceding the diagnosis of infection. This difference in scores probably reflects both the evolution of underlying disease and/or the early aggravating role of infection itself. Furthermore, it emphasizes the difficulty in clearly separating causes from consequences of nosocomial infections in critically ill patients hospitalized in ICUs, even with the use of daily monitoring of disease severity and/ or therapeutic interventions.

The adverse effects of nosocomial infections have previously been estimated in terms of mortality, morbidity, and other consequences such as economic impact (28). However, few studies, limited to VAP and bacteremia (10, 11), have applied an accurate method to evaluate the roles of length of stay in the ICU prior to the onset of infection and initial disease severity in the additional consequences of nosocomial infections in ICU patients.

We found that mortality attributable to nosocomial infection exceeded 40%, a value that corresponds to a relative risk of death equal to 4.0, thereby confirming the results of previous case-control studies (10, 11). The mean length of ICU stay for the cases was 14 d longer than for the controls, and the difference increased to 22 d when only the survivor-matched pairs were considered. Similarly, the overall level of therapeutic activity applied to patients with nosocomial infections after the onset of nosocomial infection was at least five-fold greater than that for patients without infection. The durations of antibiotic treatment were also longer for the cases, which more often received combination antibiotic therapy with a median number of two antibiotics for a median duration of 13 d, whereas paired controls received one antibiotic for 2 d, corresponding to a fivefold excess in antibiotic-day use. As previously documented (29, 30), the development of multiple infections rather than a single infection during ICU stay significantly increased length of stay and the use of therapeutic resources. Moreover, mortality was significantly greater among case-patients acquiring more than one nosocomial infection than in paired controls. The same results have been reported by Coello and coworkers (30), who found a significant increase in length of stay; microbiology, chemistry, and radiology requests; and number of antibiotic-days for surgical patients with multiple infections than for controls. It must be noted that in our study, the first episode of nosocomial infection was a UTI in 63% of casesan infection usually considered the least severe type of nosocomial infection (32). However, our findings agree with those reported by Haley and colleagues (29) who found an eightfold, statistically significant difference in terms of prolongation of stay for symptomatic UTI occurring as a patient's first nosocomial infection versus UTI occurring after a previous nosocomial infection.

In conclusion, our study suggests that a persistent high level of therapeutic activity and persistently depressed consciousness on the third day after ICU admission are associated with the acquisition of nosocomial infection by critically ill patients hospitalized in a medical ICU. Such nosocomial infections are responsible for excess mortality, prolonged stay, and increased therapeutic activity independently of the initial severity of illness. Thus, nosocomial infections exact a heavy toll on all concernedthe patient, the medical staff, and economic resourcesespecially in cases of multiple infections.