Abstract
The purpose of the study was to validate the criteria used in the guidelines of the American Thoracic Society (ATS) for severe community-acquired pneumonia (CAP). Severe pneumonia was defined as admission to the intensive care unit (ICU). Overall 331 nonsevere (84%) and 64 severe cases (16%) of CAP were prospectively studied. Mortality was 19 of 395 (5%) and 19 of 64 (30%), respectively. Single severity criteria as well as the ATS definition of severe pneumonia were assessed calculating the operative indices. A modified prediction rule including minor (baseline) and major (baseline or evolutionary) criteria was derived. Single minor criteria at admission had a low sensitivity and positive predictive value. Defining severe pneumonia according to the ATS guidelines had a high sensitivity (98%). However, specificity and positive predictive value were low (32% and 24%, respectively). A modified prediction rule (presence of two or three minor criteria [systolic blood pressure < 90 mm Hg, multilobar involvement, PaO2 /Fi O2 < 250] or one of two major criteria [requirement of mechanical ventilation, presence of septic shock]) had a sensitivity of 78%, a specificity of 94%, a positive predictive value of 75%, and a negative predictive value of 95%. The ATS definition of severe pneumonia was highly sensitive but insufficiently specific and had a low positive predictive value. Our suggested modified rule had a more balanced performance and, if validated in an independent population, may represent a more accurate definition of severe CAP.
Severe community-acquired pneumonia (CAP) represents a frequent and potentially life-threatening condition. About 10% of all hospitalized patients with CAP require admission to the intensive care unit (ICU) (1), and the mortality of these patients reaches 20 to 50% (2-8). Although the epidemiology and prognosis of severe CAP have repeatedly been investigated in the last decade (2-11) there is no universally accepted definition for this condition. A corresponding definition, however, may be helpful in clinical assessment of patients and may contribute to a more homogeneous approach in clinical research of this condition.
The guidelines for the initial management of adults with CAP published by the American Thoracic Society (ATS) in 1993 have included 10 criteria in order to provide a tentative definition of severe illness, and the presence of any one of them was used to determine a pneumonia case as severe (12). With the exception of the ratio of arterial oxygen tension to fraction of inspired oxygen (PaO2 /Fi O2 ) < 250, all proposed severity criteria have been shown to be associated with death by others (4, 7, 8, 11-17). Up to now, however, there has been only one preliminary communication about validation of these severity criteria (18). In this analysis, using the MediQual (MQ) clinical database, risk factors as defined by the ATS guidelines for severe pneumonia correlated well with clinical and economic outcomes. However, the predictive potential of any of the single criteria, as well as of the definition of severe pneumonia provided by the guidelines, have not been validated in an individual hospital setting.
Therefore, the aim of the present study was to validate the set of criteria for severe CAP as well as its definition as proposed in the ATS guidelines. In addition, we made an attempt to optimize the severity criteria according to a new prediction rule (19).
Between October 1996 and December 1997 we prospectively studied 422 consecutive and nonselected, not severely immunosuppressed patients hospitalized with CAP at our 1,000 bed teaching hospital. All patients had symptoms suggestive of a lower respiratory tract infection, a new infiltrate on chest radiograph, and no alternative diagnosis emerging during follow-up. Severe immunosuppression was defined as presence of solid organ or bone marrow transplantation, human immunodeficiency virus (HIV) infection, and neutropenia < 1 × 109/L. In addition, treatment with steroids was considered as severely immunosuppressive in daily doses > 20 mg prednisolone-equivalent for more than at least 2 wk, and in any dose as part of an immunosuppressive combination regimen with azathioprine, cyclosporin, and/or cyclophosphamide.
Overall 64 patients were admitted to the ICU according to clinical judgment of the physician in charge. These patients were defined as having severe CAP. Twenty-seven patients died from pneumonia as an expected terminal event of a chronic disabling illness, were not admitted to the ICU, and therefore were excluded from the analysis. Thus, the remaining 331 patients represented the group hospitalized with nonsevere CAP.
The following parameters were recorded at admission: age, gender, smoking and alcohol habits, comorbidity, clinical symptoms (cough, dyspnea, chest pain, body temperature, confusion [i.e., disorientation with regard to person, place, or time that is not known to be chronic, stupor, or coma], respiratory rate, heart rate, arterial systolic and diastolic blood pressure, blood gas analysis [Po 2, Pco 2, PaO2 /Fi O2 ], chest radiograph pattern [alveolar, interstitial or mixed infiltrate, number of lobes affected, uni- versus bilateral affection, pleural effusion]), and serum creatinine. Acute Physiology and Chronic Health Evaluation (APACHE) II scores were calculated in patients admitted to the ICU. At the clinical endpoints of hospital discharge or death, the following parameters were additionally retrieved: definite microbial etiology, admission to the ICU, complications (respiratory failure requiring mechanical ventilation, septic shock, renal failure, progression of radiographic infiltrates by ⩾ 50% in the first control chest radiograph), treatment regimens, and in-hospital outcome. In patients transferred from another hospital to the ICU, severity criteria were recorded from the time of admission to the transferring hospital.
The ATS guidelines for the initial management of adult CAP provide 10 criteria for severe pneumonia. These include three parameters reflecting respiratory failure (respiratory rate > 30/min at admission, PaO2 /Fi O2 < 250, requirement of mechanical ventilation); one set of three radiographic parameters (bilateral involvement or multilobar involvement or increase in the size of opacities on chest radiograph by ⩾ 50% at 48 h); and four criteria reflecting circulatory compromise (systolic blood pressure < 90 or diastolic blood pressure < 60 mm Hg, requirement of vasopressors for > 4 h, and total urine output lower than 80 ml in 4 h or acute renal failure requiring dialysis). The presence of any single criterion was defined as severe pneumonia and admission to the ICU was strongly recommended.
The application of these criteria is hampered by the following reasons. First, five criteria were not exactly defined with regard to the time during clinical course at which they should be applied, i.e., PaO2 / Fi O2 < 250, chest radiograph showing bilateral involvement or multilobar involvement, and systolic blood pressure < 90 mm Hg or diastolic blood pressure < 60 mm Hg. Second, multilobar involvement was not exactly defined. Moreover, the significance of progressive infiltrates on chest radiograph was not linked to the clinical condition. Third, the urine output is usually not exactly assessed in patients with nonsevere pneumonia. Therefore, the criteria had to be interpreted and simplified. From our viewpoint, a reasonable interpretation would be that the first group of parameters refers to baseline criteria available and assessed at admission: (1) respiratory rate > 30/min; (2) severe respiratory failure as defined by a PaO2 /Fi O2 < 250; (3) chest radiograph showing bilateral involvement; (4) involvement of > 2 lobes (in the following referred to as “multilobar involvement”); (5) systolic blood pressure < 90 mm Hg; (6) diastolic blood pressure < 60 mm Hg. The second group of parameters is assessed either at admission or during clinical course: (1) requirement for mechanical ventilation; (2) increase in the size of opacities on chest radiograph by ⩾ 50% compared with the first chest radiograph, in the presence of nonresponse to treatment (no defervescence < 38.3 grades C within 72 h of in- hospital antimicrobial treatment and/or persistent clinical symptoms) or deterioration (in the following referred to as “progressive infiltrates”); (3) requirement of vasopressors > 4 h; (4) total urine output lower than 80 ml in 4 h or acute renal failure requiring dialysis. The third criterion obviously was intended to include patients with septic shock. Accordingly, all patients meeting the definition of septic shock as suggested by Bone and coworkers (20) (i.e., presence of sepsis, hypotension, perfusion abnormalities [e.g., confusion, renal failure, multiorgan failure], requirement of fluid resuscitation and vasopressors) required vasopressors > 4 h. Thus, the term “requirement of vasopressors for > 4 h” was substituted with “septic shock”. The fourth criterion was called “renal failure” and assessed as (1) serum creatinine ⩾ 2 mg/dl at any time during hospitalization in a patient without known previous impairment of renal function, or (2) increase in serum creatinine of ⩾ 2 mg/dl from known baseline values in patients with previous impairment of renal function, or (3) acute renal failure requiring dialysis. Because the second group of parameters clearly implies more severe illness, we referred to these criteria as “major” criteria, whereas the first group of parameters were called “minor” criteria (Table 1).
| Baseline (“minor”) criteria assessed at admission |
|---|
| 1. Respiratory rate > 30/min |
| 2. Severe respiratory failure (PaO2 /Fi O2 < 250) |
| 3. Bilateral involvement in chest radiograph |
| 4. Involvement of > 2 lobes in chest radiograph (multilobar involvement) |
| 5. Systolic blood pressure < 90 mm Hg |
| 6. Diastolic blood pressure < 60 mm Hg |
| “Major” criteria assessed at admission or during clinical course |
| 1. Requirement for mechanical ventilation |
| 2. Increase in the size of infiltrates by ⩾ 50% in the presence of clinical |
| nonresponse to treatment or deterioration (progressive infiltrates) |
| 3. Requirement of vasopressors > 4 h (septic shock) |
| 4. Serum creatinine ⩾ 2 mg/dl or increase of ⩾ 2 mg/dl in a patient with |
| previous renal disease or acute renal failure requiring dialysis (renal |
| failure) |
The 10 severity criteria were tested for their ability to correctly classify severe pneumonia in our population. Admission to the ICU was used as reference for the definition of severe pneumonia. The following operative indices of each criterion were assessed in 2 × 2 tables: (1) sensitivity (ratio of predicted severe to truly severe cases); (2) specificity (ratio of predicted nonsevere to truly nonsevere cases); (3) positive predictive value (ratio of predicted severe cases to number of cases meeting the criterion [= incidence of criterion]); (4) negative predictive value (ratio of predicted nonsevere cases to number of cases not meeting the criterion). The prognostic implications of each severity criterion were assessed by chi-square tests with in-hospital mortality within 30 d as the outcome variable and by multivariate logistic regression. Operative indices were also determined for the ATS definition of severe pneumonia (presence of at least one of 10 criteria).
In order to assess which criteria available on admission independently were associated with ICU admission, the six baseline parameters (minor criteria) were included in multivariate logistic regression models. Those baseline parameters with an independent association to severity were included in prediction rules, assigning one point for the presence of each variable. The rules were tested by grouping cases with different total scores (ranging from one up to the maximum of independent parameters available) as predicting severe pneumonia versus cases with less than these scores as predicting nonsevere pneumonia. A multivariate logistic regression analysis including these variables was performed to determine its association with death. Similarly, the four major criteria were included in a multivariate analysis in order to determine those independently associated with severity and mortality. Finally, the performance of two combined rules including minor and major parameters was determined, which assigned as severe pneumonia any case in the presence of (1) two of three independent minor criteria or (2) one of two (or three) independent major criteria. Again, the prognostic implications of this rule were assessed by a multivariate logistic regression analysis including all minor and major criteria of the rule.
Results are expressed as mean ± SD. The Student's t test for continuous variables and the chi-square test to compare proportions were used. Multivariate logistic regression analysis was always performed by stepwise forward selection. All p values are two-sided and the level of significance was set at 5%.
Of the 395 patients (260 male, 135 female, mean age 68 ± 18 yr) 331 (84%) had nonsevere pneumonia and 64 (16%) severe pneumonia. The main baseline clinical characteristics of both groups are compared in Table 2. The 27 patients excluded from analysis had a mean age of 82 ± 10 yr, and 41 comorbidities (15 had one, 10 had two, and two had three comorbidities), with neurologic illness as the most frequent condition (n = 17, 42%). According to the general performance and prognosis of the patients, eventual ICU admission was denied.
| Parameter | Nonsevere Pneumonia | Severe Pneumonia | ||
|---|---|---|---|---|
| Age, yr† | 69 ± 18 | 63 ± 16 | ||
| Sex, male/female | 213/118 | 47/17 | ||
| Comorbidity present, n (%) | 250/331 (76) | 51/64 (80) | ||
| Type of comorbidity, n (%) | ||||
| Cardiac | 63 (19) | 7 (11) | ||
| Pulmonary | 145 (44) | 33 (52) | ||
| Renal | 19 (6) | 6 (9) | ||
| Hepatic† | 20 (6) | 10 (16) | ||
| Central nervous system | 41 (12) | 3 (5) | ||
| Diabetes mellitus | 54 (16) | 14 (22) | ||
| Neoplastic‡ | 36 (11) | 1 (2) | ||
| Miscellaneous | 6 (2) | — | ||
| Body temperature ⩾ 38.3° C, n (%)† | 182/331 (55) | 21/64 (33) | ||
| Cough, n (%) | 261/331 (79) | 45/64 (70) | ||
| Dyspnea, n (%)§ | 212/331 (64) | 60/64 (94) | ||
| Chest pain, n (%) | 112/289 (39) | 16/47 (34) | ||
| Confusion, n (%)§ | 61/331 (18) | 35/64 (55) | ||
| Respiratory rate > 30/min, n (%)† | 138/317 (44) | 38/59 (64) | ||
| PaO2 /Fi O2 < 250, n (%)§ | 102/289 (35) | 40/63 (64) | ||
| Pco 2 > 44 mm Hg, n (%)‖ | 43/289 (15) | 21/63 (33) | ||
| Systolic blood pressure < 90 mm Hg, n (%)§ | 2/326 (1) | 7/61 (12) | ||
| Diastolic blood pressure < 60 mm Hg n (%)‡ | 15/326 (5) | 9/61 (15) | ||
| Heart rate > 125 beats/min, n (%)‡ | 29/325 (9) | 11/61 (18) | ||
| Type of infiltrate, n (%)‡ | ||||
| Alveolar | 261/331 (79) | 41/64 (64) | ||
| Interstitial | 4/331 (1) | 3/64 (5) | ||
| Mixed | 66/331 (20) | 20/64 (31) | ||
| Bilateral involvement, n (%)§ | 48 (15) | 26 (41) | ||
| Multilobar involvement, n (%)§ | 37 (11) | 33 (52) | ||
| Pleural effusion, n (%)† | 44/331 (13) | 17/64 (27) | ||
| Requirement for mechanical ventilation, n (%)§ | 0 | 37/64 (58) | ||
| Progressive infiltrates, n (%)§ | 27/331 (8) | 18/64 (28) | ||
| Septic shock, n (%)§ | 0 | 24/64 (38) | ||
| Renal failure, n (%)§ | 12/331 (4) | 19/64 (30) |
The microbial etiology could be determined in 182 of 395 (46%) cases. Streptococcus pneumoniae accounted for 65 of 227 (29%) of pathogens, atypical bacterial and viral pathogens for 97 of 227 (43%), and nonpneumococcal nonatypical microorganisms for 65 of 227 (29%).
The severity of pneumonia was reflected by a significantly higher frequency of afebrile status, dyspnea, confusion, respiratory rate > 30/min, PaO2 /Fi O2 < 250, Pco 2 > 44 mm Hg, systolic blood pressure < 90 mm Hg, diastolic blood pressure < 60 mm Hg, heart rate > 125 beats/min, interstitial and mixed alveolar/interstitial infiltrates, bilateral and multilobar involvement, and pleural effusion (Table 2). The incidences of these parameters were comparable with those reported in previous studies. Mortality was 19 of 395 (5%) in the whole population of 19 of 64 (30%) in those with severe pneumonia.
Fifty-one of 64 patients with severe pneumonia (80%) were admitted to the ICU within the first 24 h of hospital admission, six (10%) within 48 h, three (5%) within 72 h. The remaining four patients (6%) were admitted on Days 4, 10, 11, and 12 after hospital admission. Three of 63 patients (5%) were transferred from another hospital to our ICU.
Both parameters reflecting arterial hypoxemia (respiratory rate > 30/min and PaO2 /Fi O2 < 250 at admission) were neither sensitive nor specific (sensitivity 64% and 57%, specificity 64% and 65%, respectively) and had low positive predictive values (22% and 28%, respectively) for ICU admission. The requirement of mechanical ventilation was 100% specific by definition, but only 58% sensitive. The presence of septic shock was also 100% specific, but had an even lower sensitivity of 38%. All three radiographic criteria had a high specificity with 86%, 89%, and 92%, respectively, but the positive predictive values remained low (35%, 47%, and 40%, respectively). Systolic blood pressure < 90 mm Hg and diastolic blood pressure < 60 mm Hg were highly specific (99% and 95%, respectively) but at the cost of a very low sensitivity (12% and 15%, respectively). Renal failure had a sensitivity of 30% and a specificity of 96% (Table 3).
| Criteria | Sensitivity n (%) | Specificity n (%) | PPV n (%) | NPV n (%) | ||||
|---|---|---|---|---|---|---|---|---|
| Respiratory rate > 30/min | 38/59 (64) | 179/317 (57) | 38/176 (22) | 179/200 (90) | ||||
| PaO2 /Fi O2 > 250 | 40/63 (64) | 187/289 (65) | 40/142 (28) | 187/210 (89) | ||||
| Bilateral involvement in chest radiograph | 26/64 (41) | 283/331 (86) | 26/74 (35) | 283/321 (88) | ||||
| Multilobar involvement | ||||||||
| in chest radiograph | 33/64 (52) | 294/331 (89) | 33/70 (47) | 294/325 (91) | ||||
| Systolic blood pressure < 90 mm Hg | 7/61 (12) | 324/326 (99) | 7/9 (78) | 324/378 (86) | ||||
| Diastolic blood pressure < 60 mm Hg | 9/61 (15) | 311/326 (95) | 9/24 (38) | 311/363 (86) | ||||
| Requirement for mechanical ventilation | 37/64 (58) | 331/331 (100) | 37/37 (100) | 331/358 (92) | ||||
| Progressive infiltrates | 18/64 (28) | 304/331 (92) | 18/45 (40) | 304/350 (87) | ||||
| Septic shock | 24/64 (38) | 331/331 (100) | 24/24 (100) | 331/371 (89) | ||||
| Renal failure | 19/64 (30) | 319/331 (96) | 19/31 (61) | 319/364 (88) |
Mechanical ventilation was required in 18 of 37 (47%) patients within the first 4 h of initial management in the emergency department. Septic shock occurred in 17 of 24 (71%) and renal failure in 13 of 19 (68%) within this time limit. Thus, these major criteria were often evolutionary, implying clinical deterioration during hospital stay in 32 of 80 (40%) occasions.
All severity criteria, except PaO2 /Fi O2 < 250 at admission, were significantly associated with death. The prognostic implications of each criterion are listed in Table 4. In the multivariate analysis including all 10 criteria, only requirement of mechanical ventilation (relative risk [RR] 44.3, 95% confidence interval [CI] 7.5 to 260) and septic shock (RR 8.4, 95% CI 1.8 to 36.6) remained independently associated with death.
| Criteria | RR | 95% CI | p Value | PPV (%) | ||||
|---|---|---|---|---|---|---|---|---|
| Respiratory rate > 30/min | 2.7 | 1.01–7.6 | < 0.05 | 7 | ||||
| PaO2 /Fi O2 < 250 | 1.3 | 0.6–3.2 | 0.52 | 6 | ||||
| Bilateral involvement in chest radiograph | 2.5 | 1.03–6.2 | < 0.05 | 10 | ||||
| Multilobar involvement | ||||||||
| in chest radiograph | 5.1 | 2.2–12.2 | < 0.0001 | 14 | ||||
| Systolic blood pressure < 90 mm Hg | 12.0 | 4.9–29.3 | < 0.001 | 44 | ||||
| Diastolic blood pressure < 60 mm Hg | 7.6 | 3.1–18.4 | < 0.001 | 25 | ||||
| Requirement for mechanical ventilation | 82.2 | 19.8–342.2 | < 0.0001 | 46 | ||||
| Progressive infiltrates | 4.5 | 1.9–10.9 | < 0.001 | 16 | ||||
| Septic shock | 33.5 | 13.9–80.3 | < 0.0001 | 54 | ||||
| Renal failure | 5.4 | 2.2–13.3 | < 0.01 | 19 |
The definition of severe pneumonia as the presence of at least one severity criterion as originally proposed in the ATS guidelines in 332 patients (84%) with complete data sets had a high sensitivity of 98% (57 of 58). However, specificity remained low (32%, 88 of 274). The positive predictive value was 24% (57 of 243) and the negative predictive value 99% (88 of 89). When only the proportion of patients with severe pneumonia admitted to the ICU within the first 24 h of hospitalization was included (summing up to n = 319 patients), thereby eliminating the potential confounder of clinical deterioration not predictable at admission, the performance remained virtually unchanged (sensitivity 98% [44 of 45], specificity 32% [88 of 274], positive predictive value 19% [44 of 230], and negative predictive value 99% [88 of 89]).
Prediction of ICU admission by baseline (minor) criteria. The inclusion of the baseline (minor) criteria in multivariate analysis revealed that only three parameters remained independently associated with need for ICU admission: systolic blood pressure < 90 mm Hg (RR 13.2, 95% CI 2.2 to 78.7), multilobar involvement (RR 6.7, 95% CI 3.5 to 13.1), and PaO2 /Fi O2 < 250 (RR 2.7, 95% CI 1.4 to 5.2). Of these baseline criteria, 28 of 60 (47%) patients needing ICU admission and having complete data sets met one, 20 patients (33%) two, and two patients (3%) three criteria; on the other hand, 10 patients (17%) met none. The operative characteristics of discriminant rules including these three criteria are listed in Table 5. The rule consisting of two out of three criteria achieved a sensitivity of 33%, a specificity of 94%, and a positive predictive value of 56%. Again, performances were very similar if only the proportion of patients with severe pneumonia admitted to the ICU within the first 24 h after hospitalization was included in the prediction analysis (sensitivity 33% [15 of 46], specificity 94% [268 of 284], positive predictive value 52% [15 of 31], negative predictive value 90% [268 of 299]). Multivariate regression analysis including these three variables revealed that only systolic blood pressure < 90 mm Hg (RR 14.4, 3.1 to 66) and multilobar involvement (RR 4.5, 1.6 to 12.4) were independent predictors of death.
| Rule | Sensitivity n (%) | Specificity n (%) | PPV n (%) | NPV n (%) | ||||
|---|---|---|---|---|---|---|---|---|
| One of three | 28/60 (47) | 182/284 (64) | 28/130 (22) | 182/214 (85) | ||||
| Two of three | 20/60 (33) | 268/284 (94) | 20/36 (56) | 268/308 (87) | ||||
| Three of three | 2/60 (3) | 284/284 (100) | 2/2 (100) | 268/308 (83) |
Prediction of ICU admission by major criteria. Among major criteria, only mechanical ventilation (RR undefined), septic shock (RR undefined), and renal failure (RR 5.6, 95% CI 1.6 to 18.8) remained independent predictors of severe pneumonia. Considering only the criteria of mechanical ventilation and septic shock, 20 patients (33%) had one severity criterion, 18 patients (30%) had two, and 22 (37%) none; when renal failure was considered additionally, 18 patients (30%) had one severity criterion, 14 patients (23%) two, and nine (15%) three; 19 patients (32%) had none.
Prediction of ICU admission by combined rules of minor (baseline) and major criteria. If the rule obtained with minor criteria was combined with major criteria by assigning each case as severe pneumonia if at least two of three baseline (minor) criteria or one of two major criteria (requirement of mechanical ventilation or septic shock) were met, the operative indices in 344 patients (87%) with complete data sets resulted in a sensitivity of 78%, a specificity of 94%, a positive predictive value of 75%, and a negative predictive value of 95% (and 77% [36/47], 94% [268/284], 69% [36/52], and 96% [268/279], respectively, if only the proportion of patients with severe pneumonia admitted to the ICU within the first 24 h after hospitalization was included) (Table 6).
| Rule | Sensitivity n (%) | Specificity n (%) | PPV n (%) | NPV n (%) | ||||
|---|---|---|---|---|---|---|---|---|
| First rule | 47/60 (78) | 268/284 (94) | 47/63 (75) | 268/281 (95) | ||||
| Second rule | 49/60 (82) | 258/284 (91) | 49/75 (65) | 258/269 (96) |
Overall, 20 patients (33%) met two minor criteria, another two patients (3%) three minor criteria, 20 patients (33%) met one major criterion, and 18 (30%) two criteria. Nine patients (15%) were classified as having severe pneumonia exclusively because of minor criteria, 25 (42%) because of major criteria, and 13 (22%) patients met both minor and major criteria. All four patients with severe pneumonia excluded because of incomplete data sets met the severity criteria (Table 7).
| Criteria | n | % | ||
|---|---|---|---|---|
| Patients meeting the rule | 47 | 78 | ||
| Baseline (minor) | 9 | 15 | ||
| Two criteria | 9 | 15 | ||
| Baseline (minor) and major | 13 | 22 | ||
| Two minor criteria + one major criterion | 5 | 8 | ||
| Three minor criteria + one major criterion | 1 | 2 | ||
| Two minor and two major criteria | 6 | 10 | ||
| Three minor and two major criteria | 1 | 2 | ||
| Major | 25 | 42 | ||
| One criterion (+ one minor criterion) | 6 8 | 10 13 | ||
| Two criteria (+ one minor criterion) | 1 10 | 2 17 | ||
| Patients not meeting the rule | 13 | 22 | ||
| One minor criterion | 10 | 17 | ||
| None | 3 | 5 | ||
| Patients with incomplete data sets | 4 | |||
| Two minor criteria | 1 | |||
| Two minor criteria and one major criterion | 1 | |||
| Two minor and two major criteria | 1 | |||
| Two major criteria | 1 |
The modification of this rule by adding the variable renal failure to the set of major criteria such that one of three of these criteria was required for the definition of severe pneumonia resulted in similar performances (sensitivity 82%, specificity 91%, positive predictive value 65%, negative predictive value 96%) (Table 7). Multivariate analysis including the four major criteria revealed that only the requirement of mechanical ventilation (RR 59.6, 95% CI 10.6 to 338) and the presence of septic shock (RR 5.7, 95% CI 1.4 to 22.9) remained independently associated with death.
Analysis of false-negatives and false-positives. Using the first combined prediction rule (at least two of three baseline [minor] criteria or one of two major criteria), 13 patients were classified as nonsevere although admitted to the ICU (false-negatives), and 16 patients were classified as severe although they were not admitted to the ICU (false-positives).
Of the 13 classified as false-negatives, 10 met one and three no severity criterion (Table 7). Five patients had a PaO2 /Fi O2 < 250 and five multilobar involvement. One of the three patients without severity criterion according to the rule had renal failure and another confusion. Overall, confusion was present in six patients, renal failure in two, and pleural effusion in two. Two of 13 were admitted to the ICU between 24 and 48 h of hospitalization. On the other hand, there was a tendency for the APACHE II score to be lower in the misclassified group as compared with patients correctly classified as severe (17.0 ± 7.5 versus 22.3 ± 9.8, p = 0.08). Only 1 of 13 false-negatives died, and there was also a tendency for a lower mortality in the misclassified as compared with the correctly classified group (41% versus 7%, p = not significant [NS]).
All 16 classified as false-positives had a PaO2 /Fi O2 < 250, including 15 also having multilobar involvement and one systolic blood pressure < 90 mm Hg. Seven patients (44%) had chronic obstructive pulmonary disease (COPD) as comorbid illness. Three patients were bedridden with severe chronic comorbidities, and eventual ICU admission was denied. Nine patients had confusion, four pleural effusion, but none renal failure. The 16 patients rapidly recovered clinically (defervescence and improved gas exchange) after a mean of 39 ± 32 h of treatment. None of these patients died.
The predictive values of those parameters most strongly associated with severe pneumonia but not included in the ATS severity criteria (especially the presence of confusion) were not superior to those included in the prediction rule (Tables 2 and 8. Therefore, the inclusion of these criteria would not significantly improve the performance of the prediction rule.
| Parameter | Sensitivity n (%) | Specificity n (%) | PPV n (%) | NPV n (%) | ||||
|---|---|---|---|---|---|---|---|---|
| Body temperature | ||||||||
| < 38.3° Celsius | 43/64 (67) | 182/331 (55) | 43/192 (22) | 182/203 (90) | ||||
| Presence of dyspnea | 61/64 (95) | 119/331 (36) | 61/273 (22) | 119/122 (98) | ||||
| Presence of confusion | 35/64 (55) | 270/331 (82) | 35/96 (37) | 270/299 (90) | ||||
| Heart rate > 125 beats/min | 11/61 (18) | 296/325 (91) | 11/40 (28) | 296/346 (86) | ||||
| Pco 2 > 44 mm Hg | 21/63 (33) | 246/289 (85) | 21/64 (33) | 246/288 (85) | ||||
| Presence of pleural effusion | 17/64 (27) | 287/331 (87) | 17/61 (28) | 287/334 (86) |
The prospective validation of the criteria for severe CAP proposed by the ATS guidelines revealed the following findings: (1) any given single baseline criteria at admission had low sensitivity and positive predictive value; (2) the proposed ATS definition of severe pneumonia as the presence of at least one of 10 criteria had a high sensitivity but low specificity and positive predictive value (98%, 32%, and 24%, respectively); (3) a modification of the criteria relying on two of three baseline (minor) criteria (systolic blood pressure < 90 mm Hg, multilobar involvement, PaO2 /Fi O2 < 250) or one of two major criteria (requirement of mechanical ventilation and presence of septic shock) had a sensitivity of 78%, a specificity of 94%, a positive predictive value of 75%, and a negative predictive value of 95%.
One of the most remarkable findings of this validation study is the low predictive potential of any given single baseline criterion for defining severe pneumonia. Both criteria reflecting arterial hypoxemia (respiratory rate and PaO2 /Fi O2 ) were insensitive and nonspecific. Both were also only weakly associated with mortality from severe pneumonia. The reasons for that primarily may include the high incidence of pulmonary (and perhaps also cardiac) comorbid illnesses making these criteria less discriminative as markers for respiratory compromise. The two baseline radiographic criteria had a high specificity but remained insensitive and had low positive predictive values. These performances are mainly due to the fact that the correlation of severe pneumonia with the extention of infiltrates visible on chest radiograph was only moderate. The initial blood pressure was highly correlated with severity at the given cutoffs but had an especially low incidence which accounts for its very low sensitivity. Thus, the poor predictive potential of baseline parameters can be explained by the difficulties in adequately assessing acute respiratory failure by clinical and blood gas parameters (especially in patients with cardiopulmonary comorbid illnesses), the limited correlation of radiographic parameters with severity, and a limited incidence, and therefore low sensitivity, of severe hypotension. As expected, the two criteria of requirement for mechanical ventilation and the presence of septic shock most closely reflected severity of pneumonia and also had high predictive values for severity. However, the sensitivity remained low because of the limited incidence of these most severe events. Progressive infiltrates and renal failure both were only moderately correlated with severity, had a limited incidence and therefore a lower performance.
The application of the discriminative rule originally recommended in the ATS guidelines (defining severe pneumonia in the presence of at least one criterion) resulted in a very high sensitivity of 98%. However, in view of 68% false-positives, the definition of severe pneumonia provided did not result in a satisfactory discrimination from nonsevere pneumonia.
The set of criteria proposed by the ATS guidelines provided only six criteria which regularly are available to the clinician at admission, i.e., baseline criteria. One criterion (progressive infiltrates) by definition reflects clinical deterioration and is not available at admission. As expected (19), the remaining three, requirement of mechanical ventilation, septic shock, and renal failure also represented evolutionary criteria in a considerable number of occasions (40%) in our population. Moreover, many hospitals do not have emergency facilities where the patient can safely be treated and observed for a few hours until a final decision about ICU admission is made. Therefore, we thought it was adequate to evaluate separately criteria regularly available at admission (minor criteria) and those assessed at admission or during clinical course (major criteria).
A rule consisting of two out of the three baseline (minor) parameters (PaO2 /Fi O2 < 250, multilobar involvement, and systolic blood pressure < 90 mm Hg) had a moderate performance, but the sensitivity still remained as low as 33%, and the positive predictive value 56%. Thus, this rule was extended to a rule assigning as severe pneumonia also cases that met one of two major criteria (requirement for mechanical ventilation or presence of septic shock). This rule achieved the most favorable performance, with a sensitivity of 78%, a specificity of 94%, and a positive predictive value of 75%. The addition of renal failure as a third major criterion did not increase its operative indices.
One crucial issue in the validation of severity criteria is the reference standard for severe pneumonia. We used admission to the ICU in our hospital to define severe pneumonia. Because decisions about ICU admission are subject to individual clinical judgment, local facilities and policies, this definition clearly is not an irrefutable gold standard. However, severity criteria should be objective parameters derived from good clinical practice and, therefore, it seems logical to validate severity criteria in a specialized tertiary care referral center.
The false-negative group had a lower APACHE II score and mortality as compared with the correctly classified patients, suggesting that the majority of these patients indeed had less severe pneumonia. The mean APACHE II score of correctly classified patients admitted to the ICU was 22, thus meeting the cutoff value of 22 reported as significantly associated with death in severe CAP in the elderly by others (11). Nevertheless, all but one of these misclassified patients were admitted to the ICU because of respiratory failure, and the majority of these patients was misclassified in the absence of a PaO2 /Fi O2 < 250 at admission. Therefore, a future validation study should additionally investigate the value of both parameters reflecting acute respiratory failure (respiratory rate and PaO2 /Fi O2 < 250) after application of a defined amount of supplemental oxygen for a defined period.
No patient died in the false-positive group. Several of these patients probably would have been admitted to the ICU in the absence of the rapid clinical recovery they made. Again, it is evident that serial determinations of parameters reflecting respiratory failure after application of supplemental oxygen should be incorporated into future study designs that assess minor (baseline) criteria. For these reasons, the imperfect specificity of our modified prediction rule is less troublesome because it does define patients who are potential candidates for ICU admission.
The proposed modified rule to define severe pneumonia remains imperfect because the performance relying only on baseline (minor) clinical criteria was limited. Because no further clinical parameters had increased predictive values for severe pneumonia, their addition to a prediction rule would at best result in marginally improved performances at the cost of higher complexity of the rule. Thus, the proposed rule can only serve as a supplement to clinical judgment. In addition to this rule, the clinician should remain alert to potential and frequently underestimated predictors of severe pneumonia, namely the presence of confusion (13, 21) and pleural effusion.
The proposed rule may serve as a useful counterpart to the prediction rule of Fine and coworkers (17). Whereas the latter rule may help to identify patients at low risk for mortality who can be treated safely in an outpatient setting, our rule provides simple clinical criteria for those patients who require ICU admission. On the other had, the proposed rule may serve as an objective criterion to better define severe pneumonia for the purposes of epidemiological research or clinical intervention trials. However, before recommending its use in clinical practice and its application in epidemiological research, the rule needs to be validated in independent multicenter populations. As pointed out previously, the rule may still be significantly improved when respiratory parameters after oxygen supplementation have been assessed.
In conclusion, the ATS guidelines definition of severe pneumonia was highly sensitive but poorly specific and had a low positive predictive value for need to be admitted to ICU. Our suggested modified rule consisting of two out of the three baseline (minor) criteria independently associated with severity (PaO2 /Fi O2 < 250, multilobar involvement, and systolic blood pressure < 90 mm Hg) or one of two major criteria (requirement for mechanical ventilation and septic shock) had a more balanced performance and, if validated in an independent population, may be a more accurate definition of severe CAP.
Supported by Commisionat per a Universitats i Recerca de la Generalitat de Catalunya 1997 SGR 00086, and IDIBAPS Hospital Clinic Barcelona.
| 1. | Torres A., El M., Ebiary, Zavala E., Hernandez C.Severe community-acquired pneumonia. Sem. Respir. Crit. Care Med.171996265271 |
| 2. | Woodhead M. A., McFarlane J. T., Rodgers F. G., Laverick A., Pilkington R., Macrae A. D.Aetiology and outcome of severe community-acquired pneumonia. J. Infect.101985204210 |
| 3. | Pachon J., Prado M. D., Capote F., Cuello J. A., Granacho J., Verano A.Severe community acquired pneumonia. Am. Rev. Respir. Dis.1421990369373 |
| 4. | Torres A., Serra-Batlles J., Ferrer A., Jimenez P., Celis R., Cobo E., Rodriguez-Roisin R.Severe community acquired pneumonia: epidemiology and prognosis factors. Am. Rev. Respir. Dis.1141991312318 |
| 5. | British Thoracic Society Research Committee and The Public Health Laboratory ServiceThe aetiology, management and outcome of severe community-acquired pneumonia on the intensive care unit. Respir. Med.861992713 |
| 6. | Rello J., Quintana E., Ausina V., Net A., Prats G.A three-year study of severe community-acquired pneumonia with emphasis on outcome. Chest1031993232235 |
| 7. | Moine P., Vercken J. B., Chevret S., Castang C., Gajdos P.the French Study Group for Community-Acquired Pneumonia in the Intensive Care UnitSevere community-acquired pneumonia: etiology, epidemiology, and prognosis factors. Chest105199414871495 |
| 8. | Leroy O., Santre C., Beuscart C., Georges H., Guery B., Jacquier J. M., Beaucaire G.A five-year study of severe community-acquired pneumonia with emphasis on prognosis in patients admitted to an intensive care unit. Intens. Care Med.2119952431 |
| 9. | Feldman C., Ross S., Mahomed A. G., Omar J., Smith C.The aetiology of severe community-acquired pneumonia and its impact on initial, empiric, antimicrobial chemotherapy. Respir. Med.891995187192 |
| 10. | Leroy O., Georges H., Beuscart C., Guery B., Coffinier C., Vandenbusche C., Thevenin D., Beaucaire G.Severe community- acquired pneumonia in ICUs: prospective validation of a prognostic score. Intens. Care Med.22199613071314 |
| 11. | Rello J., Rodriguez R., Jubert P., Alvarez B.the study group for severe community-acquired pneumoniaSevere community-acquired pneumonia in the elderly: epidemiology and prognosis. Clin. Infect. Dis.231996723728 |
| 12. | American Thoracic SocietyGuidelines for the initial management of adults with community-acquired pneumonia: diagnosis, assessment of severity, and initial antimicrobial therapy. Am. Rev. Respir. Dis.148199314181426 |
| 13. | British Thoracic Society and the Public Health Laboratory ServiceCommunity-acquired pneumonia in adults in British hospitals in 1982–1983: a survey of aetiology, mortality, prognostic factors and outcome. Q. J. Med.2391987195220 |
| 14. | Farr B. M., Sloman A. J., Fisch M. J.Predicting death in patients hospitalized for community acquired pneumonia. Ann. Intern. Med.1151991428436 |
| 15. | Marrie T. J., Durant H., Yates L.Community acquired pneumonia requiring hospitalisation: 5-year prospective study. Rev. Infect. Dis.111989586599 |
| 16. | Fine M. J., Smith M. A., Carson C. A., Mutha S. S., Sankey S. S., Weissfeld L. A., Kapoor W. N.Prognosis and outcome of patients with community-acquired pneumonia: a meta-analysis. J.A.M.A.275199613341341 |
| 17. | Fine M. J., Auble T. E., Yealy D. M., Hanusa B. H., Weissfeld L. A., Singer D. E., Coley C. M., Marrie T. J., Kapoor W. N.A prediction rule to identify low-risk patients with community-acquired pneumonia. N. Engl. J. Med.3361997243250 |
| 18. | Gordon G., Throop D., Berberian L., Niederman M. S., Bass J., Alemayehu D.Validation of the American Thoracic Society (ATS) guidelines for community-acquired pneumonia in hospitalized patients (abstract). Am. J. Respir. Crit. Care Med.1531996A257 |
| 19. | Ewig S.Community-acquired pneumonia: epidemiology, risk, and prognosis. Eur. Respir. Mon.319961335 |
| 20. | Bone R. C., Balk R. A., Cerra F. B., Dellinger R. P., Fein A. M., Knaus W. A., Schein R. M. H., Sibbald W. J.Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Chest101199216441655 |
| 21. | Neill A. M., Martin I. R., Weir R., Anderson R., Chereshsky A., Epton M. J., Jackson R., Schousboe M., Frampton C., Hutton S., Chambers S. T., Town G. I.Community acquired pneumonia: aetiology and usefulness of severity criteria on admission. Thorax51199610101016 |
Dr. Santiago Ewig was a 1997 research fellow from the Medizinische Universitätsklinik and Poliklinik Bonn, Bonn, Germany.
Dr. Mauricio Ruiz was a 1997 European Respiratory Society (ERS) research fellow from the Hospital Clı́nico de la Universidad de Chile, Santiago de Chile, Chile.
Dr. Francisco Arancibia was a 1997 research fellow from the Insituto Nacional del Tórax (INER y CT), Santiago de Chile, Chile.
