This is a multicenter observational study based on the analysis of data obtained from a prospective registry of hospitalized patients with CAP. Three teaching hospitals from Spain participated in the study: Hospital Clinic of Barcelona (1996–2019) [n=1635]; Hospital Universitario Cruces (2005–2019) [n=627] and Hospital La Fe of Valencia (2012–2019) [n=309]). We prospectively included all hospitalized patients aged ≥80 years with CAP and compared those admitted to the ward and ICU within 24h of hospitalization. Exclusion criteria were: (1) age <80 years; (2) severe immunosuppression (neutropenia after chemotherapy or hematopoietic transplant; drug-induced immune suppression in solid organ transplant or cytotoxic therapy; oral corticosteroid treatment with at least 20mg prednisone or equivalent per day for at least two weeks; HIV infection); (3) active tuberculosis; and (4) patients with do-not-resuscitate (DNR) orders at the time of ICU admission. Pneumonia was defined as a new pulmonary infiltrate on a chest X-ray at hospital admission, with symptoms and signs of lower respiratory tract infection.11 More details about data collection and microbiological diagnosis are described in the Supplementary Material.

For publication purposes, the study was approved by the IRB of our institutions (Register: HCP2009/5451). Written informed consent was waived due to the non-interventional design of the study.

A prior episode of pneumonia was defined as that occurring within the prior 12 months. Severe CAP was defined when at least one major (septic shock or a need for invasive mechanical ventilation) or 3 minor criteria of the 2007 IDSA/ATS guidelines were present (Supplementary Table S1). IDSA/ATS criteria for severe CAP requiring ICU admission11 used the working definition of “direct ICU admission”, i.e., admission directly from the emergence room in the first 24h after consultation to the hospital. Prior antibiotic treatment was considered when patients had taken antibiotics in the previous month. Polymicrobial pneumonia was defined as more than one pathogen being identified as the causative agent. Septic shock was defined as a systolic blood pressure of <90mmHg and peripheral hypoperfusion with the need for vasopressors for >4h after fluid resuscitation. Appropriateness of empirical antibiotic treatment was defined In all patients based on multidisciplinary guidelines for CAP management.11,12

The primary outcome was ICU admission. We divided the study population into two groups: patients hospitalized in the ward and patients admitted to the ICU, including intermediate care units. The secondary outcome was 30-day mortality.

We reported the number and percentage of patients as categorical variables, the median (Q1; Q3) as continuous variables with non-normal distributions, and the mean (standard deviation [SD]) as continuous variables with normal distributions. Categorical variables were compared using the chi-squared test or Fisher's exact test, whereas continuous variables were compared using the t-test or non-parametric Mann–Whitney U test.

We also performed univariable and multivariable logistic regression analyses13 to identify variables predictive of ICU admission. Factors showing an association in the univariable analyses (p<0.10) were incorporated into the multivariable regression model. Final variable selection was performed using the backward stepwise selection method (likelihood ratio) (pin<0.05, pout>0.10). Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated. Single collinearity was evaluated using the Pearson correlation (r) and multicollinearity was examined by means of the variance inflation factor (VIF). Discrimination and calibration were assessed by area under the receiver operating characteristic curve (AUC) and the Hosmer–Lemeshow test, respectively. For the assessment of internal validity, we used ordinary nonparametric bootstrapping with 1000 bootstrap samples and bias-corrected, accelerated 95% CIs.14 We used the multiple imputation method15 for missing data in the analyses to identify variables predictive of ICU admission (Supplementary Table S2). To determine the ability of IDSA/ATS severe CAP criteria to predict ICU admission, we determined sensitivity, specificity and positive and negative likelihood ratio,16 along with 95% CIs. The coincidence between the predictive rule and the clinical decision for ICU admission was assessed with the kappa coefficient of agreement.17 The univariate association of the predictive rule and the severity criteria with ICU admission is also expressed as the relative risk and the 95% CI. The level of significance was set at 0.05 (two-tailed), and all analyses were performed using IBM SPSS Version 26.0 (IBM Corp., Armonk, NY, USA).

During the study period, 10,742 patients with CAP visited the emergency room and 2571 (24%) were hospitalized VEP without a DNR order, of whom 2006 (78%) patients, with a median (Q1; Q3) age of 85 (82; 88) were included in the analysis. 1802 (90%) were admitted to the general ward while 204 (10%) to the ICU (Fig. 1).

Fig. 1.

Flow chart of the study population.

(0.19MB).

The IDSA/ATS criteria identified 519 (26%) of this series of patients as having severe CAP: 58 (11%) had major severity criteria only; 84 (16%) had major and at least one minor severity criteria; and, in the absence of major criteria, 377 (73%) had three or more minor severity criteria. Of those 519 patients with severe CAP, 154 (30%) were admitted to the ICU. Only 50 (3%) of the 1487 patients who were defined as not having severe CAP received ICU admission.

The main clinical characteristics of the study population are shown in Table 1. Patients admitted to the ICU were younger, more often males, and more frequently smoked and abused alcohol. Conversely, when compared to those patients who were not admitted to ICU, those admitted to the ICU had received the influenza vaccine less frequently; were less likely to have chronic neurological diseases; and presented higher rates of diabetes mellitus, altered mental status, higher levels of creatinine, C-reactive protein and serum glucose, and a lower platelet count and PaO2/FiO2. Additionally, when compared to non-ICU-admitted patients, patients admitted to the ICU had higher PSI score; presented severe CAP criteria more frequently; and experienced more pulmonary and extra-pulmonary complications, longer hospital stay and higher 30-day mortality.

ICU-admitted patients had a higher rate of etiological diagnosis than non-ICU patients (Supplementary Table S3). However, we did not find any significant differences regarding the relative proportion of causative pathogens between ICU and non-ICU patients.

Logistic regression analysis indicated that male sex, absence of chronic neurological disease, respiratory rate ≥30breath/min, systolic blood pressure <90mmHg, serum creatinine ≥1.5mg/dL, serum glucose ≥200mg/dL, PaO2/FiO2<250, pleural effusion, multilobar infiltration and septic shock were independently associated with ICU admission (Table 2). The AUC for the ICU admission prediction model was 0.86 (95% CI 0.84–0.89). Internal validation of the final model demonstrated robust results (Supplementary Table S4).

Concordance between IDSA/ATS criteria for severe CAP and ICU admission occurred in 1591 (79%) cases; the kappa coefficient was 0.33 (p<0.001). The predictive rule overestimated the number of patients who would be admitted to the ICU, i.e., 365 (20%) patients who were not admitted to the ICU had severe CAP criteria, whereas only 50 (25%) ICU-admitted patients had non-severe CAP.

Sensitivity and specificity of IDSA/ATS criteria for severe CAP in identifying patients admitted to ICUs were 75% and 80%, respectively. Severe CAP was 3.73 times more frequent among ICU-admitted patients compared to those not admitted to the ICU. Likewise, the probability that ICU-admitted patients had non-severe CAP was only 0.31, when compared to non-ICU patients. Compared with patients with non-severe CAP, patients with severe CAP had an 8.82-fold higher relative risk of ICU admission.

Thirty-day mortality was higher among ICU-admitted patients than in those not admitted to the ICU (p<0.001; Table 1). Differences in mortality between groups remained unchanged when analyzing patients with severe and non-severe CAP separately (Table 3). Table 5 shows the association between the predictive rule and mortality. Severe CAP and the presence of either of the two major severity criteria had the strongest association with mortality. All minor severity criteria were also significantly associated with 30-day mortality. Sensitivity, specificity, positive and negative likelihood ratios, and the risk ratio of severe CAP criteria in predicting 30-day mortality were similar to those of patients admitted to the ICU.

Major severity criteria had the highest association with ICU admission, particularly invasive mechanical ventilation. However, 45 (44%) patients with septic shock who were initially managed and stabilized in the emergency department did not require subsequent ICU admission. In patients with septic shock, 72 (71%) did not receive invasive mechanical ventilation, and these had similar median PSI scores irrespective of their likelihood of ICU admission. However, 30-day mortality was lower in patients with septic shock admitted to the ICU (8 [30%] vs. 27 [60%], p=0.013).

In contrast, the 377 patients with severe CAP defined by the presence of ≥3 minor criteria in the absence of major criteria, less often were admitted to the ICU (Table 4). In this subset of patients, while the PSI score was lower in ICU-admitted patients (138 [123; 154] vs. 145 [130; 160], p=0.045), 30-day mortality was similar (12 [21%] vs. 53 [17%], p=0.37).