The aim of this study was to evaluate the impact of intensive care unit (ICU)-acquired infection on long-term survival and quality of life.
Long-term survival was prospectively evaluated among hospital survivors who had stayed in a mixed, university-level ICU for longer than 48 hours during a 14-month study period during 2002 to 2003. Health-related quality of life was assessed using the five-dimensional EuroQol (EQ-5D) questionnaire in January 2005.
Of the 272 hospital survivors, 83 (30.5%) died after discharge during the follow-up period. The median follow-up time after hospital discharge was 22 months. Among patients without infection on admission, long-term mortality did not differ between patients who developed and those who did not develop an ICU-acquired infection (21.7% versus 26.9%; P = 0.41). Also, among patients with infection on admission, there was no difference in long-term mortality between patients who developed a superimposed (35.1%) and those who did not develop a superimposed (27.6%) ICU-acquired infection (P = 0.40). The EQ-5D response rate was 75 %. The patients who developed an ICU-acquired infection had significantly more problems with self-care (50%) than did those without an ICU-acquired infection (32%; P = 0.004), whereas multivariate analysis did not show ICU-acquired infection to be a significant risk factor for diminished self-care (odds ratio = 1.71, 95% confidence interval = 0.65–4.54; P = 0.28). General health status did not differ between those with and those without an ICU-acquired infection, as measured using the EuroQol visual-analogue scale (mean ± standard deviation EuroQol visual-analogue scale value: 60.2 ± 21 in patients without ICU-acquired infection versus 60.6 ± 22 in those with ICU-acquired infection). The current general level of health compared with status before ICU admission did not differ between the groups either. Only 36% of those employed resumed their previous jobs.
ICU-acquired infection had no impact on long-term survival. The patients with ICU-acquired infection more frequently experienced problems with self-care than did those without ICU infection, but ICU-acquired infection was not a significant risk factor for diminished self-care in multivariate analysis.
Nosocomial infections increase mortality and costs in intensive care units (ICUs) [1-3]. Furthermore, they increase length of stay in the ICU as well as the frequency and duration of organ failures . We previously showed that ICU-acquired infection was an independent risk factor for hospital mortality, even after adjustment for age and Acute Physiology and Chronic Health Evaluation (APACHE) II and Sequential Organ Failure Assessment (SOFA) scores, in a series of 335 ICU patients with ICU stay longer than 48 hours . However, long-term outcome has not been studied in detail in patients acquiring an infection during their ICU stay. Studies have shown that sepsis patients who survive critical illness are at greater risk for post-ICU death than are control individuals; furthermore, survivors have been reported to have poor functional outcomes [6,7]. Reduced quality of life has also been reported among patients with acute respiratory distress syndrome (ARDS) as compared with critically ill control patients, but postdischarge mortality did not appear to be increased in ARDS patients [8,9]. ICU-acquired infections during critical illness impose a major burden on the costs and outcomes of intensive care; we addressed the question of whether these infections also have an impact on long-term mortality and quality of life in a prospectively study conducted in a subgroup of survivors after discharge.
Materials and methods
Study location and population
The study was conducted in Oulu University Hospital, which is a 900-bed tertiary level teaching hospital. All patients admitted into the ICU during the period from May 2002 to June 2003 whose ICU stay was longer than 48 hours were included in the study. The study protocol was approved by the hospital's ethics committee. The distribution of infections on admission and the epidemiology and contribution of ICU-acquired infections to hospital mortality were reported previously [5,10,11]. This substudy concentrated on the situation following hospital discharge.
For all study patients the following information was collected: age, sex, smoking habits, alcohol abuse, presence of chronic underlying diseases (chronic obstructive pulmonary disease, ischaemic heart disease, chronic hepatic disease, chronic renal disease, previous stroke or transient ischaemic attack, diabetes, malignancy or immunosuppressive medication), severity of underlying diseases and organ dysfunctions on admission (assessed using APACHE II  and SOFA ), and diagnostic category on admission.
The presence of infection was recorded using criteria required by the US Centers for Disease Control and Prevention (CDC) [14,15] but with the following modifications. A catheter-related infection was deemed to be present if the same strains of bacteria were isolated in blood cultures and in a semiquantitative catheter tip culture, with no other site of infection. A catheter-related infection was also diagnosed if the patient had a positive semiquantitative catheter tip culture while blood cultures showed no growth or were not done, and there were clinical signs of infection, no other infection site was present and the patient exhibited a favourable response to antimicrobial therapy. Secondary bacteraemia was recorded when the same strains of bacteria were isolated in blood culture and in culture from a site of infection. Ventilator-associated pneumonia was defined according to criteria proposed by an international panel . Pneumonia was diagnosed when a new and persistent infiltrate that was not otherwise explained appeared on chest radiographs, along with the presence of any two of the following: fever (temperature > 38°C) or hypothermia (temperature < 36°C), leucocytosis (> 10 × 103/mm3) or leucopenia (< 4.0 × 103/mm3), and new purulent tracheal aspirate.
Lengths of stay in the ICU and at hospital were recorded. Postdischarge mortality data were obtained from the hospital database, which had been updated with data from Central Statistical Office of Finland on 25 January 2005.
Measurement of health-related quality of life
Health-related quality of life (HRQOL) was measured using the five-dimensional EuroQol (EQ-5D) questionnaire, which has been described in detail elsewhere . It has been recommended and widely used for measuring HRQOL in critical care [18-21]. In short, the questionnaire contains two parts: the EQ-5D self-classifier, a self-reported description of current health problems according to five items (mobility, self-care, usual activities [work, housework, family and leisure activities, and so on], pain/discomfort and anxiety/depression) each with three response alternatives (1 = no problems, 2 = moderate problems, 3 = severe problems). The second part is a visual-analogue scale (EQ-VAS) ranging from 0 (worst possible health state) to 100 (best possible health state), on which the patients rate their current health. A weighted health state index, the EuroQol 5D Index, based on the five dimensions and ranging from -0.11 ('worse than death') to 1 ('perfect health'), was also calculated .
All survivors were mailed the following materials in January 2005: a cover letter explaining the objectives of the study and requesting the patient's or their relatives' collaboration in completing the questionnaire; a copy of the EQ-5D questionnaire; and a form with accessory questions regarding each patient's subjective overall assessment of their health status compared with the situation before ICU treatment and their current employment status. If there was no initial response to the questionnaire, the patients were contacted by phone by a trained ICU study nurse, who repeated the questions on the phone exactly, according to the EQ-5D questionnaire, and asked the patient to answer 'yes' or 'no'.
Data registration and statistical analysis
The data were entered into a SPSS database (SPSS Data Entry, version 2.0; SPSS Inc., Chicago, IL, USA). Summary statistics for continuous or ordinal variables are expressed as the median with the 25th to 75th percentiles or as the mean and standard deviation (SD). The multivariate Cox regression model was used to assess the impact of ICU infection on long-time survival, whereas the other parameters in the final model were selected on statistical grounds (P < 0.05). The log-linearity assumption of the continuous variables was checked by creating a design variable based on quartiles, and the assumption of proportional hazards was evaluated graphically by log-minus-log survival plots. Log-rank test results are presented for Kaplan-Meier survival curves. The impact of ICU-acquired infection on EQ-5D self-care (no problems or some problems) dimension was evaluated by logistic regression analysis. The linearity assumption of the continuous variable for age did not hold, and a dichotomous covariate at age 50 years (< 50 versus ≥ 50) was therefore created. Goodness-of-fit was evaluated using the Hosmer-Lemeshow test. The other variables entered into the multivariate Cox and logistic regression models in addition to ICU-acquired infection were the APACHE II score, chronic underlying disease, infection on ICU admission, sepsis, severe sepsis or septic shock on admission, community or hospital-acquired pneumonia on admission, admission diagnostic category (medical, surgical nontrauma, trauma, neurological), ICU length of stay, SOFA score on admission and on ICU discharge, and the normal face validity parameters of age, sex, smoking habits and alcohol abuse. No significant interactions or collinearities between ICU infection and the other parameters in either multivariate model were found. Further, nested models were compared using the likelihood ratio test to select the best model. Two-tailed P values are reported, and the analyses were performed by the SPSS (version 12.0.1; SPSS Inc.) software.
Characteristics of intensive care unit admissions
The study population is presented in Figure 1. The main demographic data and clinical characteristics of 272 patients discharged from hospital are presented in Table 1. There were significantly more patients with trauma on admission among the patients who developed an ICU infection (P < 0.001) and more medical admissions among the patients who did not develop an ICU infection (P < 0.001). The median APACHE II scores did not differ between the groups, whereas the median SOFA score on admission was higher and the ICU length of stay longer in the group of surviving patients with an ICU-acquired infection. The following ICU-acquired infections were recorded in 55 patients: ventilator-associated pneumonia (17), surgical site infections (14), lower respiratory tract infection (14), intra-abdominal infections (6), sinusitis (6), soft tissue or skin infections (4), primary or catheter-associated bacteraemia (2), secondary bacteraemia (1), urinary tract infection (1) and other infections (2).
Table 1. Main demographic data and clinical characteristics of patients discharged from hospital
Figure 1. Study population. ICU, intensive care unit.
Of the 272 patients discharged from hospital, 85 died after discharge (31.3 %). The median follow-up time after hospital discharge for the whole study population was 22 months (25th to 75th percentile: 16 to 26 months) and that for the survivors was 24 months (21 to 28 months). This was also the median time for completing the EQ-5D questionnaire form. Among the patients with no infection on admission, seven patients with an ICU infection (26.9%) and 10 patients without an infection (21.7%; P = 0.41) died. The corresponding numbers of deaths among the patients with infection on admission were as follows: six patients with an ICU infection (27.6%) and 60 patients without an ICU infection (35.1%; P = 0.40). The long-term survival curves did not differ between the patients with and those without an ICU-acquired infection (Figure 2). Furthermore, based on the multivariate Cox model, ICU-acquired infection did not increase long-term mortality (Table 2). The adjusted hazard ratio for the effect of ICU-acquired infection on posthospital mortality in the multivariable Cox regression model was 0.83 (95% confidence interval [CI] 0.47–1.46).
Table 2. Effect of ICU-acquired infection on posthospital mortality in multivariable Cox regression model
Figure 2. Survival curves of ICU patients after discharge from hospital. The patients who were alive on 25 January 2005 were censored. ICU, intensive care unit.
Health-related quality of life
Quality of life data were obtained from 142 of the 187 survivors (75.9%). The questionnaire was completed by the patient in 121 cases (85.2%), by relatives in 12 cases (8.5%), and by a district or ward nurse in nine cases (6.3%). A total of 113 respondents returned the questionnaire by mail, and 29 were contacted by phone. The nonrespondents were significantly younger than the respondents, but no significant differences were observed in sex, APACHE II scores, admission diagnostic category, presence of chronic underlying disease, or length of stay in the ICU or in hospital (Table 3). Only 36% of those employed before the ICU episode (n = 47) had resumed their previous jobs, whereas 54.4% had quit because of the illness that led to ICU admission.
Table 3. Main demographic data and clinical characteristics in respondents and nonrespondents to the EQ-5D questionairre
The EQ-5D self-classifier data of the study groups with and without an ICU-acquired infection are presented in Table 4. The patients with an ICU infection had more problems on the self-care dimension (P = 0.004), but there were no other differences. However, multivariate analysis did not identify ICU-acquired infection to be a risk factor for diminished self-care (odds ratio = 1.71, 95% CI 0.65 to 4.54; P = 0.28; Table 5). Overall, two-thirds of the patients suffered from moderate or extreme pain.
Table 4. EQ-5D results in long-term survivors
Table 5. Effect of ICU-acquired infection on diminished self-care in multivariate logistic regression analysis
The mean ± SD value on the EQ-VAS was 60.2 ± 21 among the patients without an ICU-acquired infection, and the corresponding figure for those with an ICU-acquired infection was 60.6 ± 22. The difference between the means was -0.41 (95% confidence interval = -9.30 to +8.48; P > 0.9). The mean ± SD EQ-5D weighted health state index (EuroQol 5D Index) was 0.715 ± 0.24 for those without an ICU infection and 0.725 ± 0.23 for those with an ICU infection. The difference between the means was -0.01 (95% CI = -0.11 to +0.09; P = 0.84).
The current general level of health compared with the status before ICU admission did not differ between the groups, because 52 (47.7%) of patients without and 16 (57.1%) of those with an ICU-acquired infection perceived their health status to be worse (P = 0.40). The difference between proportions was -9.4% (95% CI = -28.2% to +10.9%).