Expected and observed mortality in critically ill patients receiving initial antibiotic therapy
- First Online:
- Cite this article as:
- Janisch, T., Wendt, J., Hoffmann, R. et al. Wien Klin Wochenschr (2012) 124: 775. doi:10.1007/s00508-012-0276-0
- 136 Views
To evaluate the predictors of mortality in critically ill patients receiving initial antibiotic therapy (IAT; < 48 h after admission).
Six hundred thirty-one consecutive patients admitted to an intermediate care (IMC) unit were included. IAT was initiated in 227 patients. Laboratory markers, interventions, medications, systemic inflammatory response syndrome (SIRS) and sepsis criteria, length of stay, and hospital mortality as well as expected mortality, based on the SAPSII-expanded score, were assessed retrospectively. Failure of IAT was defined as a rise in C-reactive protein (CRP) or leukocyte count on day 3 compared with the values on admission.
Patients with IAT were significantly older (67 ± 14 vs. 64 ± 14 years; p = 0.006) and had a higher prevalence of chronic renal failure (33 vs. 23 %; p = 0.015), chronic obstructive pulmonary disease (COPD; 27 vs. 16 %; p = 0.002), malignoma (17 vs. 9 %; p = 0.007), acute renal failure (11 vs. 4 %; p = 0.001), respiratory failure (22 vs. 7 %; p < 0.001), and a shock index < 1.0 (21 vs. 8 %; p < 0.001). Although patients with IAT did not have significantly different expected mortality compared with patients without IAT (19.2 vs. 14.5 %; p = 0.144), they did have a significantly higher observed mortality (16.7 vs. 3.7 %; p < 0.0001). Based on the number of SIRS criteria (0, 1, 2, or 3–4) or sepsis criteria (no sepsis, sepsis, or severe sepsis) fulfilled, expected mortality (16.4, 18.2, 20.6, or 21.0 %, respectively; p = 0.955/17.5, 18.3, or 23.4 %, respectively; p = 0.689) did not differ in IAT patients. In contrast, observed mortality differed significantly (4.8, 10.6, 20.6, or 29.4 %, respectively; p = 0.029/8.3, 19.7, or 29.3 %, respectively; p = 0.013). Patients who responded to IAT did not differ regarding comorbidities, SIRS or sepsis criteria, but they had a lower observed mortality (11.9 vs. 26.3 %; p = 0.008) than patients who failed to respond to IAT. Central venous lines were more frequently present in patients with failure to IAT when compared with those with response (51 vs. 22 %; p = 0.009). In the subgroup of patients with acute myocardial infarction (AMI), those with IAT (n = 41) were treated less frequently according to the current cardiac guidelines than those without (n = 124)
Patients with IAT have a high morbidity burden and higher observed than expected mortality. The SAPSII-expanded score does not seem to precisely estimate the risk of in-hospital mortality in these patients. Failure of response to IAT was associated with an even higher mortality. Whether central venous lines and nonadherence to cardiac care guidelines influence the mortality of patients with IAT should be investigated in further studies.
KeywordsAntibioticSepsisMortalitySAPSII expandedCritically ill
= Initial (< 48 h after admission) antibiotic therapy
= Simplified Acute Physiology Score II (Intensive care score to predict mortality)
= C-reactive protein (inflammatory marker)
= Systemic inflammatory response syndrome
= Chronic obstructive pulmonary disease
= Coronary artery disease
= Acute myocardial infarction
= Angiotensin-converting enzyme
= Troponin T
= Noninvasive ventilation
= Gastrointestinal bleeding
= Intermediate care
= Standard operating procedure
Erwartete und beobachtete Mortalität kritisch kranker Patienten mit initialer antiinfektiver Therapie
Evaluation von Mortalitätsprediktoren bei kritisch kranken Patienten mit initialer antiinfektiver Therapie (IAT) (< 48 h nach stationärer Aufnahme).
Sechshunderteinunddreißig Patienten, die aufeinander folgend auf eine Intermediate Care Station aufgenommen wurden, wurden in die Studie eingeschlossen.
Eine initiale antiinfektive Therapie (IAT) wurde bei 227 Patienten begonnen. Laborwerte, Interventionen, Medikationen, SIRS und Sepiskriterien, Dauer des Aufenthaltes und die Mortalität während des Krankenhausaufenthaltes ebenso wie die nach SAPS II Score erwartete Mortalität, wurden retrospektiv erfasst.
Bei einem Anstieg von C-reaktiven Protein (CRP) oder Leukozyten am 3. Tag nach Aufnahme, verglichen mit den Werten vom Aufnahmetag, wurde ein Versagen der initialen antiinfektiven Therapie angenommen.
Patienten, welche eine IAT erhielten, waren signifikant älter (67 ± 14 vs. 64 ± 14 Jahre; p = 0,006) und hatten eine höhere Prävalenz von chronischer Niereninsuffizienz (33 vs. 23 %; p = 0,015), COPD (27 vs. 16 %; p = 0,002), Malignomen (17 vs. 9 %; p = 0,007), akutem Nierenversagen (11 vs. 4 %; p = 0,001), respiratorischer Insuffizienz (22 vs. 7 %; p = 0,001) und hämodynamischen Schock (21 vs. 8 %; p = 0,001). Obwohl Patienten mit IAT keine signifikant unterschiedliche erwartete Mortalität (SAPS-II-EXPANDED) im Vergleich zu Patienten ohne IAT hatten (19,2 vs. 14,5 %; p = 0,144), hatten die Patienten mit IAT eine signifikant erhöhte beobachtete Mortalität (16,7 vs. 3,7 %;p = 0,0001). Die beobachtete Mortalität divergierte signifikant nach Anzahl der erfülltem SIRS-Kriterien (0, 1, 2 oder 3–4) (4,8, 10,6, 20,6 bzw. 29.4 %; p = 0.029) bzw. Sepsis-Kategorien (keine Sepsis, Sepsis, schwere Sepsis) (8,3, 19,7 bzw. 29,3 %; p = 0.013), während sich die erwartete Mortalität in diesen Gruppen nach SAPS-II-Expanded nicht wesentlich unterschied (16,4, 18,2, 20,6 bzw. 21,0 %, p = 0,955/ 17,5, 18,3 bzw. 23,4 %; p = 0,689). Patienten, die auf die initiale antiinfektive Therapie ansprachen, unterschieden sich nicht in Bezug auf Komorbititäten, SIRS oder Sepsiskriterien, hatten aber eine niedrigere beobachtete Mortalität (11,9 vs. 26,3 %; p = 0,008) als die Patienten, bei denen die IAT versagte. Patienten, bei denen die IAT versagte, hatte häufiger Zentrale Venenkatheter als Patienten, welche auf die IAT ansprachen (51 vs. 22 %; 0,009). In der Subgruppe der Patienten mit akuten Myokardinfarkt wurden die Patienten, welche eine IAT erhielten (n = 41) weniger häufig nach den aktuellen kardiologischen Leitlinien behandelt, als die Patienten, welche keine IAT erhielten (n = 124).
Patienten mit initialer antiinfektiver Therapie haben eine höhere Morbidität und die beobachtete Mortalität ist höher als die Erwartete. Die Krankenhausmortalität dieser Patienten wird vom SAPS-II-expanded-Score nicht präzise vorhergesagt, während einfache Sepsis-Parameter eine gewisse Vorhersagekraft haben. Ein Versagen der IAT war mit einer nochmals erhöhten Mortalität assoziiert. Ob Zentrale Venenkatheter oder die nicht-Einhaltung der kardiologischen Leitlinien Einfluß auf die Mortalität hat, sollte in weiteren Studien untersucht werden.
SchlüsselwörterAntiinfektive TherapieSepsisMortalitätSAPS II expandedKritisch krank
Sepsis is an important cause of mortality in critically ill patients . It is accepted that early initial antibiotic treatment (IAT) is fundamental and mandatory in patients with suspected sepsis . However, despite poor public awareness of sepsis , physicians caring for patients with sepsis recognize the difficulty of defining and diagnosing sepsis and are aware that they frequently miss the diagnosis . Despite evidence-based treatment guidelines , the outcome of patients with sepsis is still poor . There is an overwhelming evidence that the adequacy of IAT is associated with reduced mortality, whereas a failure of response to IAT is associated with high mortality [6–10]. The definition of IAT adequacy is frequently based on whether the prescribed antibiotic matches the in vitro susceptibility of a detected pathogen. However, the pathogen is often unknown at the time of antibiotic administration. In addition, it seems that antibiotics are often prescribed in circumstances where their benefit is not proven . A lack of sepsis detection, limited knowledge of pathogens, and overuse of antibiotics are a few of the challenges physicians face while caring for critically ill patients.
This study was conducted to compare the predicted (SAPSII) and the observed mortality of patients receiving IAT.
Material and methods
Study population of critically ill medical patients
The charts of all 947 consecutive patients admitted to a step-up intermediate care (IMC) unit between December 2004 and December 2005 were retrospectively analyzed. These patients were transferred from a standard care ward, emergency room, or other hospitals.
The one hundred thirty-two surgical patients admitted prior to an elective (n = 81) or emergent (n = 51) cardiac surgery were excluded. Thirty-five patients were excluded because no monitoring was performed due to clear “do-not-resuscitate” orders and strict palliative care that started on the day of admission; these patients were admitted because they had a high nursing requirement, they needed analgesic titration, or there was a lack of capacity in the department of palliative care. Seven hundred eighty patients formed the primary study population. Thirt-one patients were excluded because of incomplete clinical data. Of the remaining 749 patients, 118 were transferred to regular wards at external hospitals.
The remaining 631 medical patients, all of whom were admitted to our hospital and were discharged home or died in our hospital, formed the study population.
Patients requiring invasive mechanical ventilation who developed multiorgan failure or needed more than one vasopressor were not admitted to the IMC ward but rather to an intensive care unit (ICU) ward.
Two hundred twenty-seven patients with suspected infections received initial antibiotic therapy (IAT) on the day of admission.
The decision to start antibiotic treatment was based on clinical signs such as the SIRS/sepsis criteria or a rise in laboratory markers (C-reactive protein (CRP) and leukocytes). Empiric treatment was up to the attending physician and according to current guidelines. There was no local standard operating procedure (SOP) for recommendation of different antimicrobial agents.
The admission diagnosis, vital signs, comorbidities, and SAPSII-expanded data were assessed and validated by two of the authors (Johannes Wendt and Thorsten Janisch). Initial medication within the first 48 h was recorded.
Laboratory data were defined as the primary laboratory test on the day of admission for all patients. The test included the following: hemoglobin, hematocrit, leukocytes, platelets, creatinine, bilirubin, CRP, potassium, and sodium bicarbonate. Inflammatory markers such as CRP and leukocyte count were assessed again on day 3.
Medication use in the first 48 h were defined as either “yes” or “no” for aspirin, ACE inhibitors, statins, beta blockers, calcium-channel blockers, diuretics, sedative medication (neuroleptics or benzodiazepine), and antibiotics.
Response or failure to IAT
Failure of IAT was defined as a rise in CRP or leukocytes on day 3 after admission compared with the initial levels on the day of admission. Response to IAT was defined as a decrease in CRP and leukocytes on day 3 after admission compared with the initial levels on the day of admission.
SAPSII score and predicted death rate
Diagnostic tests (ECG, sonography, echocardiography, X-ray, and CT-scan) and interventions within the first 48 h (monitoring, arterial lines, central venous catheter, left heart catheterization, emergency endoscopy, and noninvasive ventilation (NIV)) were assessed.
Definition of acute renal failure, respiratory failure, sepsis
Acute renal failure was defined as urine output below 500 ml per day or an increase in serum creatinine of more than 100 % over 24 h, given an initial value above 1.2 mg/dl. Respiratory failure requiring NIV was defined as acute respiratory distress in patients for whom nasal oxygen or oxygen given by mask was not sufficient to maintain an O2 saturation above 90 %; NIV was initiated in these patients. SIRS and sepsis were defined according to current guidelines .
For the statistical analysis, the commercially available software package SPSS for Windows (Version 16.0) was used. For continuous variables, ANOVA was used. For comparisons of categorical variables (observed vs. expected mortality), a chi-squared analysis or a cross tabulation analysis was performed. A probability value < 0.05 was considered statistically significant. Continuous data are expressed as means ± standard deviations, and qualitative data are expressed as frequencies.
Clinical characteristics in the study population
Patients with IAT were significantly older than patients without IAT. Average age of the study population was 65 ± 14 (with IAT 67 ± 14, without IAT 64 ± 14) and had a higher burden of chronic (COPD, chronic renal failure, and malignoma) and acute (acute renal failure, respiratory failure, and shock index > 1) comorbidities; 66 % were male (68 % with and 62 % without IAT). The average SAPSII score was 31 ± 9.
Admission diagnoses were acute myocardial infarction (AMI; n = 165), arrhythmia (n = 127), chest pain/dyspnoea (n = 119), heart failure (n = 66), intoxication (n = 28), respiratory failure due to COPD/pneumonia (n = 49), metabolic disorders/electrolyte imbalance/pancreatitis/anaphylaxis/pancytopenia (n = 28), bleeding (n = 23), pulmonary embolism (n = 16), acute renal failure (n = 10), myo/endo/pericarditis (n = 6), seizure/apoplexy (n = 3), vascular disease (n = 14), aortic dissection (n = 2), or sepis/infection (n = 22).
Initial antibiotic therapy
IAT was initiated in 227 patients. Narrow/moderate-spectrum penicillins were given to 4 patients (Flucloxacillin n = 2, Amoxicillin n = 1, and Ampicillin n = 1). Broad-spectrum penicillins were given to 62 patients (Piperacillin/Sulbactam n = 43, Ampicillin/Sulbactam n = 14, Mezlocillin n = 3, and Amoxicillin/clavulanic acid n = 2). First-generation cephalosporines were given to 2 patients (Cefazolin n = 2), second-generation cephalosporines were given to 87 patients (Cefuroxime n = 87), and third-generation cephalosporines were given to 5 patients (Ceftriaxone n = 5).
Quinolones were given to 32 patients (Moxifloxacin n = 16, Ciprofloxacin n = 8, and Levofloxacin n = 8). Carbapenems were given to 23 patients (Imipenem n = 20 and Meropenem n = 3). Various antibiotics were given to 12 patients (Vancomycin n = 3, Clindamycin n = 3, Metronidazole n = 3, Gentamicin n = 1, Cotrimoxazole n = 1, and Fosfomycin n = 1).
Outcome in the study population
Table 1 shows the outcome of patients with IAT vs. those without IAT. Patients receiving IAT had a significantly higher mortality and longer length of stay in the hospital.
Outcome of patients with vs. without IAT
Patients without IAT
Patients with IAT
p-value (with vs. without IAT)
Length of stay
17 ± 28
13 ± 22
24 ± 36
Predicted mortality (SAPSII expanded)
Association of SIRS and sepsis criteria with expected mortality based on the SAPSII-expanded score vs. observed mortality in patients with IAT
As shown in Table 2, the predicted mortality (SAPSII expanded) did not differ significantly, but the observed mortality incrementally increased as the severity of SIRS/sepsis increased. The SIRS/sepsis classification revealed a stronger diversification of outcomes than the SAPSII score.
Association of SIRS and sepsis criteria on admission with expected and observed mortality in 227 patients with IAT
Expected mortality (SAPSII expanded; (%))
Observed mortality (%)
SIRS criteriaa present according to published guideline (14)
No SIRS criteria
1 SIRS criteria
2 SIRS criteria
3–4 SIRS criteria
21.0 (p = 0.955)
29.4 (p = 0.029)
Sepsis definition according to the published guidelines (14)
23.4 (p = 0.689)
29.3 (p = 0.013)
Clinical factors and outcomes in patients who did or did not respond to IAT
In patients who did not respond to IAT, there was no significant difference in regard of acute or chronic morbidities. In addition, neither antibiotic class nor the SIRS or sepsis criteria were associated with response or failure. However, patients with no response to IAT had more often received a central venous line than patients who did respond (51 vs. 33 %; p = 0.009). However, the predicted mortality did not differ (20.5 vs. 18.6 %; p = 0.146). Patients who did not respond to IAT had significantly higher observed mortality (26.3 vs. 11.9 %; p = 0.008; Table 3). This finding is illustrated by a Kaplan–Meier plot in Fig. 1. The different outcomes stratified by failure of IAT and presence of sepsis are shown in Fig. 2.
Outcome in patients with IAT stratified by response to IAT (response to IAT was defined as fall of CRP and leukocytes on day 3 compared with admission values)
All patients with IAT
Patients with laboratory response on day 3
Patients without laboratory response on day 3
67 ± 15
66 ± 15
69 ± 12
415 (66 %)
274 (68 %)
141 (62 %)
Length of stay
24 ± 36
25 ± 42
20 ± 17
Predicted mortality (SAPSII expanded)
Subgroup analysis in patients with acute myocardial infarction
Table 4 summarizes the differences between patients with AMI who did and those who did not receive IAT. Patients with AMI and IAT were 7 years older and had more frequent renal failure (12 vs. 1 %). However, guideline-based therapy, including ACE inhibitors, left heart catheterization, and statin therapy, were performed less often in patients with AMI and IAT. The observed mortality in patients with AMI and IAT was significantly higher than in patients with AMI without IAT (17.1 vs. 3.2 %; p = 0.006).
Presentation, treatment, and outcome in patients with myocardial infarction stratified by IAT
Patients without IAT
Patients with IAT
p-value (without vs. with IAT)
Patients with main diagnosis AMI
68.8 ± 10.6
67.0 ± 11.0
74.1 ± 7.1
Observed in-hospital mortality
Multiple logistic regression analysis
As shown in Table 5, the main significant factors associated with mortality in our study population were sepsis and failure of IAT response. Renal failure, respiratory failure, and malignoma were also associated with observed mortality, whereas IAT itself was not. Interestingly, the presence of a central venous line and arterial catheter was also significantly associated with the observed mortality.
Association of different variables on observed mortality analyzed by multiple logistic regression analysis in the study population (n = 631)
Failure of IAT
Central venous line
Acute renal failure
Left heart catheter
Chronic renal failure
Shock index > 1
IAT is fundamental and mandatory in patients with suspected sepsis . However, antibiotics are also given to patients with suspected infections without sepsis. In this study, we endeavored to compare the predicted (SAPSII) and the observed mortality in these patients.
One major result is that patients with IAT have an immensely higher burden of acute (acute renal failure, respiratory failure, and shock index > 1.0) and chronic (chronic renal failure, COPD, and malignoma) comorbidities. They more frequently received oxygen therapy, central venous lines, arterial lines, and emergency endoscopy. This may reflect a greater need for intensive therapy in patients with IAT. Patients with IAT had higher hospital mortality compared with those without IAT.
Patients receiving IAT are a highly selected group of patients with a high burden of critical illness and an unfavorable prognosis. However, based on our multivariate analysis, the main factors contributing to high mortality were sepsis and failure of IAT response, not IAT itself. This finding is in accordance with recent studies [15, 16] in which failure of IAT was also associated with higher mortality. Therefore, we conclude that IAT itself does not put patients at risk for higher mortality, but rather, the high burden of comorbidity frequently leads to death if their infections cannot be controlled by antibiotic therapy. We did not investigate the additional sepsis therapy  the patient did or did not receive; thus, we cannot be certain that this might have an influence to our results.
It is of concern that the placement of central venous lines and arterial catheters is associated with higher mortality. It is known that central venous lines are associated with infections, which are partially preventable . Still, in this study, the association between central venous lines and the failure of IAT response might be secondary to the fact that the need for central venous lines is surrogate markers of hemodynamic or respiratory instability. As there is a lack of evidence regarding whether central venous lines influence the prognosis of patients with infections, this factor should be investigated in further studies.
We also discovered that patients with AMI and IAT received less guideline-based therapy and had a higher mortality rate. However, this finding was not statistically significant. Nonadherence to guideline-based cardiac therapy, especially fewer catheter interventions, should be investigated in a larger cohort to determine its effects on mortality.
The second major result is that the SAPSII-expanded score did not precisely predict in-hospital mortality and that the simple SIRS and sepsis criteria better estimated this outcome. This is in accordance with a very small study from Wunder et al. . While larger studies found an association of SAPSII with sepsis prognosis, the accuracy of SAPSII for this special population was not perfect [19, 20].
Our study has several limitations. There was no SOP for initiation IAT. We have no valid data about the suspected or proved infection site. Blood cultures were not obtained for every patient receiving IAT; thus, the data regarding positive or negative blood cultures cannot be used for statistical analysis. As it is an observational and retrospective study and not a randomized controlled trial, no firm therapeutic conclusions can be drawn. However, we highlight the excessive mortality in patients who received IAT, specifically when inflammatory markers increased during the first 48 h after admission. Thus, strategies are necessary to predict early IAT failure and to escalate the aggressiveness of IAT.
IMC patients with IAT have a higher mortality than expected by SAPSII-score calculations. We therefore conclude that SAPSII should be used with caution to predict outcome in these patients with infections and IAT. In contrast, SIRS and sepsis criteria are valuable tools for risk prediction in patients receiving IAT. Whether central venous lines and nonadherence to cardiac care guidelines influence mortality in patients with IAT should be investigated in further studies.
Conflict of interest
The authors declare that there is no actual or potential conflict of interest in relation to this article.