The study was conducted in the Complejo Hospitalario de Toledo. This is a tertiary level 800-bed hospital belonging to SESCAM (Castilla La Mancha Health Service) with a referral area of 435 000 inhabitants. The ED has an internal medicine area attended by a staff EP and resident physicians from various medical specialties. During 2008 and 2009, an average of 416 and 430 emergencies/day, respectively, were evaluated and the incidence of CAP in patients ≥18 years of age was 0.92% and 0.98% of the patients seen in the ED (3.21 and 3.56 cases/1000 inhabitants/year, respectively). Patients with CAP may be discharged from the ED (including those who remain under observation for up to 24h) or admitted to the intensive care unit (ICU) or to the SSU (short stay unit) of the pneumology, internal medicine or geriatric wards or, to a lesser extent (<3%) may be admitted to other departments with their own 24-h duty specialists (nephrology and hematology). The EP takes the decision regarding admission and the department to which the patient is initially admitted, except in the case of the ICU, where the duty intensive medicine specialist is consulted.

This was an observational, single-blind study with prospective follow-up of patients in two phases: before and after intervention consisting of implementation of the “Management of CAP in EDs” (SEMES-SEPAR 2008) CPG1,22 along with training sessions for all EPs and residents of the center on the CPG, as described below.

Two independent collaborators unknown to the rest of the EPs throughout the study evaluated the initial eligibility of the subjects who attended the ED between 1 January 2008 and 1 August 2009, until 200 patients were consecutively included in the pre-intervention phase (1 January 2008–30 September 2008) and another 200 in the post-intervention phase (4 October 2008–1 August 2009). To be included, patients had to meet the following criteria: adult patients (≥18 years of age), diagnosed with CAP in the ED by the treating physicians. Immunosuppressed patients (human immunodeficiency virus infection, solid organ transplant, splenectomy, more than 30 days treatment with 10mg or more prednisone per day or equivalent, or treatment with other immunosuppressive drugs) and patients hospitalized during the previous 14 days were excluded. The two collaborators were also responsible for making an evaluation (blinded for the rest of the EPs) of the definitive eligibility, and patients were withdrawn if the final diagnosis of CAP was not maintained by the treating physician after 30 days [codes 481, 482, 483, 485, 486 and 507 of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)], if they had a final diagnosis of tuberculosis or a diagnosis of viral etiology, influenza, mycosis or lung abscess (ICD-9-MC codes 480, 487, 484 and 513). Patients diagnosed with a second episode of CAP during the study period were also excluded. The study met the ethical requirements of our hospital and was approved by the Clinical Research Ethics Committee of the site. All patients were followed up via the computerized medical records of the ED, hospital and primary care center.

Between 1 and 3 October 2008, both the full text and a brochure of the “Management of CAP in EDs” (SEMES-SEPAR 2008) CPG1,22 were distributed to all the EPs and resident physicians of the center, training sessions were given, and the physicians were asked to systematically apply the recommendations from then on. The two independent collaborators unknown to the other EPs were responsible for applying the patient evaluation criteria. To evaluate and analyze the actions of the EPs, “appropriate management” was defined as each case in which the measures and treatments applied by the physicians coincided with the SEMES-SEPAR 20081,22 recommendations, and “inappropriate management” was when they did not. This evaluation was carried out independently for the request for complementary tests in the ED (clinical laboratory, microbiology and biomarkers), the evaluation of prognosis and decision regarding the destination of the patient according to the Pneumonia Severity Index (PSI) and choice and administration of treatment in the ED. When there was discrepancy between the two evaluators, the case was excluded. CAP was defined by the presence of acute symptoms consistent with CAP (cough, dyspnea, fever, pleuritic chest pain, altered level of consciousness, etc.) and radiological findings (presence of new pulmonary infiltrates on X-ray). It was defined as severe and requiring evaluation by the ICU specialist when patients met one major or three minor criteria of the American Thoracic Society-Infectious Diseases Society of America (ATS/IDSA 2007).6

The PSI was used to evaluate the severity and prognosis of patients with CAP and the decision regarding admission, according to the risk categories created by the original authors,23 but with the inclusion of some additional criteria (also explained and communicated in the training sessions), in accordance with the 2008 SEMES and SEPAR recommendations. Accordingly, patients in PSI risk categories IV–V and those in risk categories I–III presenting one or more of the additional risk factors or criteria listed in Table 1 had to be hospitalized.

The patients were divided into two groups: pre-intervention (preG) and post-intervention (postG). All sociodemographic, clinical, exploratory, clinical laboratory and radiological variables included in the PSI,23 the ATS/IDSA ICU admission criteria (2007)6 and the 2001 International Sepsis Definitions Conference24 criteria for the definition of S, SS and SSh were collected. The PSI value and grade, and the Charlson index,25 both original and weighted for age and presence of S, SS, SSh and severe CAP, were evaluated. In addition to these variables, the following were determined: prior antimicrobial treatment (during the 72h prior to consultation in the ED), the profile of the treating EP (resident, ED staff, another specialist), presence of diabetes or chronic obstructive pulmonary disease (COPD), number of comorbidities (including those recorded for the PSI, plus diabetes and COPD), request for laboratory tests as indicated by the CPG (including, per protocol, complete blood count, coagulation, biochemistry and blood gases), request for appropriate microbiological studies according to the CPG (blood cultures, antigens in urine, sputum culture), pulse oximetry (O2 saturation), C-reactive protein (CRP) and procalcitonin (PCT) requests and values, administration and timeliness of starting antibiotics (within the first 4h), appropriate choice of antibiotic regimen and dosing, change in antibiotic regimen in hospitalized patients within the first 48h or in patients treated at home within 72h after discharge, overall and intravenous treatment duration and side effects and complications. Finally, outcome variables included re-admission to the ED and hospital re-admission within 30 days, times to clinical stabilization and lengths of hospital stays, microbiological diagnosis, initial destination of patient (discharge, observation, short stay unit, ward, ICU or death in ED) and its appropriateness, re-evaluation of discharged patients after 24–72h, confirmation of clinical and radiological resolution after 3–6 weeks and mortality in patients who were discharged, admitted to the ward and to the ICU, and overall 30-day mortality.

It was calculated that with 200 patients per study phase, a precision of not less than ±10% would be obtained in the 95% confidence estimate for the percentage difference between both phases. Moreover, the statistical power of a 5% two-tailed comparison of the proportions would be at least 80%, in the case of population differences from 14%. With this sample size, the probability that the 95% confidence interval of the mean difference between the periods would not include the zero value would be equal to or greater than 0.8 for a Cohen effect size of at least 0.2.

Means and standard deviations (SD), ranges, medians and percentages were used as appropriate for describing the demographic, clinical, progress and treatment characteristics of the patients in both phases. In addition, for the estimation of population values, the 95% confidence interval (95% CI) limits were calculated for means and percentages. Comparison of percentages between study phases (pre and post-CPG implementation) was carried out using the Chi-squared test or Fisher's exact test, as appropriate. Comparison of scale and ordinal variables was performed using the Student's t-test and the Mann–Whitney U test, as applicable. Furthermore, in order to control for any confounding bias as far as possible, an unconditional logistic regression model was applied, using in-hospital mortality and 30-day mortality as dependent variables, and as independent variables, the study phase (pre and post-CPG implementation) and the clinical, demographic and treatment characteristics which could influence mortality and be distributed heterogeneously between the phases; in this way, the adjusted odds ratio (OR) of the association between mortality and phase and their 95% CIs were calculated. Similarly to control for confounding factors, a multiple linear regression model was used to investigate whether the possible association between treatment duration and group was maintained after controlling for the patient's destination (discharge, observation, short-stay unit, ward or ICU). Finally, the cumulative proportion of patients who remained hospitalized throughout the period was calculated using the Kaplan–Meier method and compared between phases using the log rank test; this analysis was repeated, stratifying for group according to PSI and destination of the patient. The same procedure was used to analyze time to clinical stabilization. The statistical analysis was carried out using SPSS 11 software for Windows, and significance was considered as a P-value <.05; all comparisons were two-tailed.

The sociodemographic, clinical and severity characteristics, comorbidities (and original and weighted Charlson index) and the distribution of the patients according to the PSI classification of all study patients are presented in Table 2. Differences between groups did not reach the nominal level of significance for any of the variables, and moreover, their magnitudes did not seem to be relevant, except in the case of use of antibiotics within the previous 72h and the criteria for severe sepsis, which were 9% and 7.2% more common, respectively, in the postG group. There were no significant or important differences between the preG and the postG in radiological findings (incidence of multilobar or bilateral CAP and pleural effusion), or in the clinical laboratory or blood gas parameters (Table 3). Although there were significant differences in obtaining the final microbiological diagnosis (12% vs 40.5%, P<.001), the distribution of the pathogens identified (Fig. 1) was proportional and no significant differences were shown (P=.22).

Fig. 1.

Microbiological diagnosis.

(0.31MB).

There were no significant differences between the preG and postG in the profile of the ED treating physician (resident/ED staff/other specialists), with proportions of 52.52%–41.41%–6.06% vs 54.5%–42.5%–3%, respectively (P=.339).

As can be seen in Table 3, there were significant differences between the preG and the postG in requesting CRP and/or PCT in the ED (19.28% vs 90% respectively; P<.001). However (and perhaps as a consequence of greater use in the postG), CRP and PCT levels were lower in the postG (74.82±18.72 vs 58.90±22.01mg/dl, for CRP and 4.89±11.31 vs 2.51±12.88ng/ml, for PCT; P<.001 in both cases). It can also be seen in Table 3 how differences of a greater magnitude appear in the concentration of both markers in the intermediate groups (PSI II-IV). When the mean CRPs of all the study patients were compared by clinical severity subgroups, a significant difference (P<.05) was found between patients without S and patients with S (P=.028), SS (P<.001) and SSh (P=.024), and between patients with S and patients with SS (P=.011), but no statistical difference was found between patients with SSh and patients with S and SS (P>.05). Similarly, when the mean PCT levels were evaluated, differences (P<.05) were found between patients without S and those with SS (P=.042) and SSh (P<.001), between patients with S and those in the SS (P=.017) and SSh (P<.001) subgroups, and between patients with SS and those with SSh (P<.001). In addition, in the patient set an association (P<.05) was shown between requesting biomarkers and early administration (<4h) of antimicrobials in the ED (82.6% vs 48.2%), the appropriate choice of antibiotic (66.5% vs 88.5%), the correct decision regarding the destination of the patient (89.3% vs 69.8%), requesting the appropriate clinical laboratory tests (92.2% vs 63.5%) and the correct request for microbiological testing (86.2% vs 61.2%). Table 4 shows the associations between biomarker levels and obtaining a microbiological diagnosis and in-hospital and overall 30-day mortality.

There is an obvious difference between the preG and the postG (in all cases with P<.001 and differences between 22% and 60%) in performance variables with the antibiotic in the ED that may change or influence the progress, disease course and prognosis of patients with CAP (Table 5). The administration of antibiotics in the ED, early administration during the first 4h, the appropriate choice of regimen and dosing, early administration of antibiotics and intravenous fluids for SS and SSh in the ED were greater in the postG. In contrast, the number of patients in whom the initial antibiotic regimen was changed, either on the ward or in the ICU, or for patients who were discharged, was significantly lower in the postG (see number of cases and percentages in Table 5). The total duration of antibiotic treatment in hospitalized patients was longer in the preG, both overall (11.84±3.81 vs 10.43±3.35 days; P<.001) and in each of the PSI groups (Table 5). This was also the case for the duration of intravenous antibiotic treatment in the overall patient set (6.76±4.50 days in the preG vs 5.76±3.01 days in the postG; P=.042) and in the PSI subgroups (Table 5). After adjusting for patient destination (discharge, observation, short-stay unit, ward, ICU or death) and PSI group, the mean difference continued to be higher in the preG for the total duration of antibiotic treatment (mean=1.6; 95% CI: 1.0–2.3 days) and for the time of intravenous administration (mean=1.4; 95% CI: 0.6–2.2 days).

There were significant and relevant differences (P<.001 and differences greater than 25% in all cases) in favor of the postG in the decisions of the EP on specific actions concerning care (Table 6) with respect to the appropriate request for clinical laboratory and microbiology testing (blood cultures, antigen in urine, sputum culture). Moreover, while a microbiological diagnosis was achieved in 12% of the preG cases, the corresponding figure in the postG was 45.5% (P<.001). There was also a greater number of changes in decisions (regimens or patient destination) after biomarkers were interpreted in postG (10.81% vs 26.11%; P=.052). Finally, it is worth remarking that 12.56% of patients in the preG did not have measurement of O2 saturation, compared to 2% in the postG (P=.022); this difference is most obvious in the PSI I subgroup (27.8% vs 2.6%; P=.010).

Discharge directly from the ED was more common in the preG, while in the postG, the decision was more frequently taken to keep the patient under observation and there was a higher rate of ICU admissions from the ED (values are shown in Table 7). In the preG, 35% of patients had an initially inappropriate destination (inappropriate discharge or admission) compared to 3.6% of the postG (P<.001) (Table 7). The appropriateness of the destination of the patient from the ED according to the PSI recommendations and the additional criteria indicated in Table 1 was also improved in the postG (64.5% vs 96.5%; P<.001). Specifically, “inappropriate discharges” in the PSI IV–V groups fell from 35.5% (preG) to 2% (postG) (P<.001) and inappropriate admissions or discharges in the PSI I–III groups fell from 44% to 5.1% (P<.001). Both the check-up of the patient 24–72h after discharge and the confirmation of clinical and radiological resolution 3–6 weeks later was significantly greater in the postG (P<.01), while the number of re-admissions to the ED after initial discharge was lower in the postG (P=.01).

Duration of hospitalization and the time to achieve clinical stabilization according to the criteria of Halm et al.26 are shown in Table 7. Mean time to clinical stabilization in patients admitted to the short-stay unit and the ward or under observation was significantly higher in the preG (4 days; 95% CI: 3.6–4.4 days) than in the postG (3 days; 95% CI: 2.7–3.3 days). This result was maintained after adjusting by group for PSI or destination (discharge, observation, short-stay unit, ward or ICU). Differences were found between the preG and the postG, with the overall duration of hospital stay being greater in the preG (8.73±5.49 vs 7.59±3.97 days, P=.01). The difference was maintained after adjusting for PSI category and for patient destination (discharge, observation, short-stay unit, ward or ICU) (P=.011). Regarding the hospital stay data (Fig. 2), it is worth pointing out that while in the first phase, the time needed for 50% of the patients to be discharged from the ward or the short-stay unit was 9 days (95% CI: 8–10 days), in the postG the corresponding figure was 7 days (95% CI: 6–8 days); moreover, 8 days (95% CI: 7–9 days) were required in the preG and 11 days in the postG (95% CI: 9–13 days) for 75% of patients to be discharged. After controlling for PSI group and for place of admission, there was no appreciable change in these differences.

Fig. 2.

Distribution of discharges and freeing-up of beds (Kaplan Meier).

(0.09MB).

Raw mortality data are shown in Table 8. No significant differences were found between preG and postG for mortality in discharged patients (1.1% vs 0%, P=1) or in patients admitted to the ICU, although this was higher in the preG cases compared to the postG cases (41.66% vs 25%, P=.306). However, there were differences in in-hospital mortality (20% vs 6.60%, P=.004) and in overall 30-day mortality (15% vs 8.5%, P=.044). The results by PSI group are presented in Table 8. In addition, it was found that selecting the wrong antibiotic regimen in the ED compared to the correct choice was related with both in-hospital mortality (P=.004) and overall 30-day mortality (P=.008). Early administration of antibiotics in the ED was related with lower overall 30-day mortality (31.7% vs 15.3%, P=.068). Finally, when in-hospital mortality adjusted for related covariates (age, number of comorbidities, Charlson index, PSI group and presence of S, SS or SSh criteria) was compared between the preG and the postG, the difference remained significant at P=.002 and OR: 0.191 (95% CI: 0.067–0.545). The situation was similar for overall 30-day mortality adjusted for related covariates (age, number of comorbidities, Charlson index, PSI group and presence of S, SS or SSh), with the differences being maintained (OR: 0.270; 95% CI: 0.123–0.590, P=.001).