Predicting Death in Patients Hospitalized for Community-acquired Pneumonia Barry M. Farr, MD, MSc; Andrew J. Sloman, MD; and Michael J. Fisch, MD
• Objective: To validate a previously reported discriminant rule for predicting mortality in adult patients with primary community-acquired pneumonia and to determine which factors available at hospital admission predict a fatal outcome among such patients. • Design: Historical cohort study. • Setting: University hospital. • Patients: Adults admitted to the hospital for community-acquired pneumonia. • Measurements: Using stepwise logistic regression, we analyzed prognostic factors (data available at admission and recorded in the medical record) that showed a univariate association with mortality. The predictive values of three discriminant rules were measured to validate the results of a previous study. • Main Results: Of 245 patients, 20 (8.2%) died. Of 42 prognostic factors identified in previous studies, 8 were associated with mortality, but only a respiratory rate of 30/min or more, a diastolic blood pressure of 60 mm Hg or less, and a blood urea nitrogen of more than 7 mmol/L remained predictive in the multivariate analysis. A discriminant rule composed of these three variables was 70% sensitive and 84% specific in predicting mortality, yielding an overall accuracy of 82%. • Conclusion: Tachypnea, diastolic hypotension, and an elevated blood urea nitrogen were independently associated with death from pneumonia in our study, confirming the value of a previously reported discriminant rule from the British Thoracic Society. This rule may be useful in triage decisions because it identifies high-risk patients who may benefit from special medical attention.
Community-acquired pneumonia remains an important cause of mortality in developed nations. In the United States, approximately 924 000 cases of pneumonia occur annually, accounting for 2.8% of all hospital admissions (1). The patient dies in 6% to 24% of cases of community-acquired pneumonia requiring hospitalization, and at least 50 000 deaths per year in the United States are attributed to this illness. Until recently, a rigorous epidemiologic study using multivariate analysis of the prognostic factors associated with mortality in pneumonia had not been done (2-39) (see Appendix Tables 1 and 2). In 1987, the British Thoracic Society (BTS) subjected an extensive list of patient variables to statistical analysis in a prospective study of prognosis in 453 adults with community-acquired pneumonia (10). Multivariate analyses identified three variables that were consistently associated with death—respiratory rate, diastolic blood pressure, and blood urea nitrogen. Using optimal cutoff points of these variables for that data set (respiratory rate of 30/min or more, diastolic blood pressure of 60 mm Hg or less, and blood urea nitrogen of more than 7 mmol/L [19.6 mg/dL]), the investigators formulated an accurate but simple discriminant rule that was considered to be positive when at least two of the three factors were present. In the BTS study, a positive rule was associated with a 21-fold increase in mortality (88% sensitivity and 79% specificity). Such a rule, if confirmed to be predictive, might be useful at the time of admission for rapidly identifying high-risk patients who might require special attention, perhaps in an intensive care unit. The validity of the BTS results regarding prognostic factors and discriminant rules was not tested in an independent cohort of patients with pneumonia, however, and the purpose of our study was to assess the validity of these findings among patients with community-acquired pneumonia who had been admitted to the University of Virginia Hospital.
Methods Sample
Annals of Internal Medicine. 1991;115:428-436. From University of Virginia Health Sciences Center, Charlottesville, Virginia. For current author addresses, see end of text. 428
Patients with primary pneumonia were selected by retrospective review of the hospital records of consecutive cases of pneumonia occurring in a 25-month period (January 1984 through January 1986) at the University of Virginia Hospital. Cases were identified by a computer listing of all patients with a discharge diagnosis of pneumonia. Primary community-acquired pneumonia was defined as an acute respiratory illness contracted in the community and accompanied by a new radiographic infiltrate. The study group ranged in age from 15 to 80 years. Patients were excluded if pneumonia was not the primary reason for hospitalization, if pneumonia was an ex-
© 1991 American College of Physicians
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
Table 1. Pathogens Isolated from Cultures of 245 Patients with Community-acquired Pneumonia* Pathogen
Patients with Isolate in Sputumt
Patients with Isolate in Bloody yi(0/r,\
7.0 mmol/L Diastolic blood pressure < 60 mm Hg Respiratory rate > 30/min
Beta Coefficient
Relative Risk (95% CI)
P Value
2.19 1.30 1.15
8.93 (2.66 to 30.00) 3.67 (1.23 to 10.90) 3.16 (1.07 to 9.31)
< 0.001 0.014 0.028
15 September 1991 • Annals of Internal Medicine Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
• Volume 115 • Number 6
429
Appendix Table 2. Prognostic Factors for Death from Pneumonia Identified in Previous Studies: Factors from Laboratory Studies, Cause and Site of Infection, and Treatment Studies Showing an Associationt Univariate Multivariate
Variable
Statistical Test Not Stated Laboratory study Blood urea nitrogen Elevated > 7 mmol/L > 24 mmol/L Creatinine > 180 umol/L Hematocrit Leukocyte count < 4 x 109/L < 5 x 109/L < 9 x 109/L < 10 x 109/L > 10 x 109/L > 20 x 109/L < 10 x 109/L or > 25 x 109/L Lymphocyte count < 109/L Po2 < 6.6 kPa Decreased P0l Elevated PCOl Decreased blood pH Abnormal liver enzyme levels Decreased serum albumin Albumin < 25 g/L At least two lobes involved Both lungs affected Bronchopneumonia (compared with lobar) Atrial fibrillation Negative antigen skin test Stress hormone profile§ Cause and site of infection Streptococcus pneumoniae Pneumococcus type III Pneumococcus type III, mucoid strain Mycoplasmata Haemophilus influenzae Staphylococci Staphylococci with influenza A Legionella species Gram-negative bacilli Specific type of gram-negative bacilli Pneumonia with serologic evidence of influenza A Pneumonia during influenza epidemic No pathogen isolated Bacteremia Pneumococcal bacteremia Pneumococcal antigenemia Serum level of pneumococcal antigen Bacteremia other than pneumococcal Gram-negative bacteremia Staphylococcal bacteremia Empyema Extrapulmonary focus of infection Therapy and course of illness Antibiotics used within 2 weeks of admission Symptomatic therapy before admission Therapy with digoxin before admission Previous illness treated before admission Other therapy before admission Shorter interval between onset of illness and hospitalization Type of antibiotics used Antibiotics given before day 4 of illness No response to penicillin treatment Failure of hypothermia to remit during treatment
27 29 29
10 10
Studies Showing No Associationt
38 10
4
24
10,31 28 35
10
10
10
27 4 14,24 27 27 29 29 29 27 27 4,14,28,29,35
10 10 10 10
36
31
38
10 10
22 27 36 39 4,6,34 6,14,15,19,20,29
10 33 33 10
6 3,6,11,13,34 5 5 3
10 18,25 3
2 38 3,6,15,16,23,34 6,20,34
17 3 10
18 10 20 2 3 3
33 14,19
26,33 37 10 10 10 10
10
35 7,14 14 31 31
Continued on following page 434
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
trogen was correlated with age, previous illness, elevated liver enzymes, and an elevated leukocyte count, which may explain why these variables did not remain associated with death in the multivariate analysis. Similarly, diastolic hypotension was correlated with a history of confusion and with systolic hypotension. Tachypnea was correlated with tachycardia, underlying lung disease, a history of smoking, male sex, and bacteremia.
mechanical ventilation (P = 0.009). Seven patients who died were not intubated, and rule 1 predicted death in 4 (57%) of them, compared with 31 of the 201 survivors (15%) who were not intubated (P = 0.02). The accuracy of rule 1 in predicting outcome for the nine patients with chronic renal failure (admission blood urea nitrogen range 7 to 27 mmol/L) was also assessed. Death was predicted in three of the eight survivors and was not predicted in the single patient who died.
Discriminant Rules Application of the three discriminant rules assessed by the BTS study (rules 1, 2, and 3 in Table 4) showed that a rule incorporating tachypnea, diastolic hypotension, and a raised blood urea nitrogen level (rule 1) was the most powerful predictor of mortality (Table 4). All three rules showed both high specificity and negative predictive value, but rule 1 was a more sensitive predictor and was associated with a higher relative risk for death than rule 2 (diastolic hypotension, tachypnea, and confusion) or rule 3 (elevated blood urea nitrogen, confusion, P0, < 6.6 kPa, and leukocyte count < 10 x 109/L or lymphocyte count < 1 x 109/L). Although the numbers of patients tested by rules 2 and 3 were smaller (n = 237 and n = 205, respectively) than the number of patients tested by rule 1 (n = 241) (because there were some patients with missing information), all 20 patients who died were in the group tested by rule 2, and 19 of the 20 patients who died were in the group tested by rule 3. Rule 1 identified 14 of these 20 patients, but rules 2 and 3 identified only 7 and 8 of these patients, respectively. Of the 30 patients who were intubated and mechanically ventilated, 13 died. On the basis of data collected at admission, rule 1 predicted death in 10 (77%) of these patients and in 4 of the 17 patients (24%) who survived
Discussion Our historical cohort study of patients with community-acquired pneumonia confirms the predictive value of three prognostic factors identified in a previous study (admission respiratory rate, diastolic blood pressure, and blood urea nitrogen) and the utility of a simple discriminant rule that uses these variables to predict mortality. This rule was first reported by the BTS Research Committee (10), which applied a stepwise logistic regression analysis to clinical variables that had shown a univariate association with mortality. In the BTS study, a multivariate association was shown for abstinence from alcohol, absence of chest pain, absence of vomiting, treatment with digoxin before admission, tachypnea, diastolic hypotension, and an elevated blood urea nitrogen level. A clinically convenient rule was created by dichotomizing the latter three variables at optimal cutoff points as determined from the BTS data set: respiratory rate of 30/min or more, a diastolic blood pressure of 60 mm Hg or less, and a blood urea nitrogen level of more than 7 mmol/L. The presence of any two of these three criteria was associated with a 21-fold greater risk for death in the BTS cohort study and predicted mortality with 88% sensitivity and 79% specificity.
Table 4. Test Performance of Three Discriminant Rules in Predicting Mortality Rule
Positive Predictive Value
Negative Predictive Value
Overall Accuracy
Sensitivity (95% CI)*
Specificity (95% CI)*
Youden Index
Likeli- Relative hood Risk Ratio
vi/n/CZ, \
niny 70) —
* 1. Two or more of the following: blood urea nitrogen > 7.0 mmol/L; diastolic blood pressure < 60 mm Hg; and respiratory rate > 30/min 2. Two or more of the following: diastolic blood pressure < 60 mm Hg; respiratory rate > 30/min; and confusion 3. Three or more of the following: blood urea nitrogen > 7.0 mmol/L; confusion; Po2 < 6.6 kPa; leukocyte count < 10 x 109/L or lymphocyte count < 109/L
14/49 (28.6)
186/192 (96.9)
200/241 (82.3)
14/20 (70.0 [49.9 to 90.1])
186/221 (84.2 [79.4 to 89.0])
0.54
4.4
9.1
7/32(21.9)
192/205 (93.7)
199/237 (84.0)
7/20 (35.0 [14.1, 55.9])
192/217(88.5 [84.3 to 92.7])
0.24
3.0
3.4
8/33 (24.2)
161/172 (93.6)
169/205 (82.4)
8/19(42.1 [19.9 to 64.3])
161/186(86.6 [81.7 to 91.5])
0.29
3.1
3.8
* CI = confidence interval. 15 September 1991 • Annals of Internal Medicine Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
• Volume 115 • Number 6
431
Our cohort study of patients with pneumonia from central Virginia examined the association between death and 42 admission variables, including those identified as prognostic factors in the BTS study (10) and others (2-39). Univariate associations were confirmed for older age, age of more than 60 years, respiratory rate of 30/min or more, decreased diastolic blood pressure, diastolic blood pressure of 60 mm Hg or less, decreased systolic blood pressure, elevated blood urea nitrogen, blood urea nitrogen of more than 7 mmol/L, decreased serum albumin level, leukocyte count of less than 4 x 109/L, and a P0i of 6.6 kPa or less. These results confirmed the association between a higher respiratory rate and mortality that was first reported in the BTS study but failed to confirm the association between mortality and four other factors found to be significant by the BTS study: These factors included abstinence from alcohol, absence of chest pain, absence of vomiting, and digoxin therapy before admission. In our stepwise logistic regression analysis, only a respiratory rate of 30/min or more, a diastolic blood pressure of 60 mm Hg or less, and a blood urea nitrogen of more than 7 mmol/L— the three components of the BTS rule—remained predictive. Other univariate predictors might have remained associated in the multivariate analysis with a larger sample size yielding more deaths and higher statistical power. When applied to our cohort, the BTS rule predicted mortality with 70% sensitivity, 84% specificity, and a likelihood ratio of 4.4. A positive rule was associated with a ninefold greater risk for death from communityacquired pneumonia. The lower sensitivity and relative risk for death associated with the rule in our study relative to its performance in the BTS study are not surprising, because a prognostic model derived from the data of the BTS study would be expected to perform better in that population than in an independent cohort. Nevertheless, the predictive value of the rule in our study confirms its utility as an indicator of prognosis. When the rule was applied to our patients who required mechanical ventilation, a subset of patients showing a strong univariate association with death, the rule predicted death as well as it did in the entire sample (sensitivity, 77% compared with 70%; specificity, 76% compared with 84%). Therefore, the rule appears to be applicable to severely ill patients with community-acquired pneumonia, even those in an intensive care unit. The rule did not perform as well in patients with chronic renal failure. The rule may have a higher falsepositive rate among patients with renal failure; because such patients have a chronically elevated blood urea nitrogen level, the addition of only one more positive variable (that is, tachypnea or diastolic hypotension) would achieve a positive rule. The correlation of these three variables, which are measures of physiologic derangement, with death from pneumonia is not unexpected. An elevated blood urea nitrogen may result from dehydration and hypotension (42); diastolic hypotension results from intravascular depletion, alterations in global vasoregulation, and myocardial depression (43); and tachypnea occurs as the respiratory system adjusts to the presence of local in432
flammation, hypercapnia, acidosis, and, in some cases, hypoxia. The levels of these three variables should therefore reflect the degree of these physiologic imbalances and thus may bear a logical relation to the probability of death. However, the emergence of these variables as independent prognostic factors from a large group of variables using multivariate analyses in two different cohort studies is noteworthy. Nevertheless, although the discriminant rule composed of these three variables appears to be an accurate predictor of outcome in patients with community-acquired pneumonia, it does not necessarily follow that its use will alter that outcome. Indeed, it has been suggested that death from pneumonia mortality is predetermined (14) and that intensive care merely prolongs the interval before death (28). This question has not been tested in a randomized controlled trial, however, and probably will not be because of the ethical implications of withholding intensive care from severely ill patients. Nevertheless, the application of this rule has the potential for improving survival in patients with community-acquired pneumonia. Some investigators believe that specialized care does prevent death from pneumonia (34, 44-46), and the rule may be helpful in the emergency room to rapidly determine which patients with pneumonia need such care. The predictive value of this discriminant rule has not been studied in outpatient clinics, which treat many patients with oral therapy at home who have milder cases of pneumonia, but because the probability of death is considerably lower among such patients, the positive predictive value of the rule among such patients would be expected to be lower. Another study of prognostic factors in communityacquired pneumonia was reported after the completion of our study (47); this other study involved only adult patients with community-acquired pneumonia, including the types of patients excluded from our study, such as patients who were immunocompromised because of metastatic cancer or cancer chemotherapy, patients in whom pneumonia was an expected terminal event, and patients with mycobacterial, postobstructive, or aspiration pneumonia. Given this difference in study design, the finding of underlying neoplastic disease as the strongest predictor of death is understandable. Other independent predictors of death were age of more than 65 years; absence of pleuritic chest pain; mental status changes; vital sign abnormalities; and pneumonia due to Staphylococcus aureus, gram-negative bacilli, bronchial obstruction, or aspiration. The use of variables not available at hospital admission, such as results of sputum and blood cultures, render these findings somewhat less applicable to emergency room triage. Our study has confirmed the predictive value of a previously reported, simple discriminant rule that incorporates respiratory rate, diastolic blood pressure, and blood urea nitrogen level to predict death in patients with community-acquired pneumonia at the time of hospital admission. Such a rule may be of use to the admitting physician for rapidly determining which patients warrant special attention because of a high risk for death.
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
Appendix Table 1. Prognostic Factors for Death from Pneumonia Identified in Previous Studies: Factors from History and Physical Examination* Variable
History Age Older > 40 years > 50 years > 60 years > 65 years > 70 years > 75 years Male sex Employment status^ Social status Any preexisting chronic illness At least two preexisting illnesses Preexisting respiratory illness Chronic bronchitis Chronic obstructive pulmonary disease Preexisting nonrespiratory illness Heart disease or pulmonary edema Rheumatic heart disease Neoplasm Diffuse or metastatic cancer Cirrhosis Preexisting renal or hepatic insufficiency Chronic alcoholism Chronic alcoholism without cirrhosis Acute alcoholism Dementia Immunocompromised Smoker or ex-smoker Absence of alcohol use Present use of hypnotics or tranquilizers Absence of chest pain Presence of chest pain Absence of vomiting Cough Dyspnea Increased sputum production Chills, sweats, and rigors Confusion History of falls Inability to walk Physical examination Admission temperature Admission temperature < 37 °C Hypothermia in first 24 hours Admission temperature > 37.8 °C Admission temperature > 40 °C Greatest temperature attained Higher respiratory rate Respiratory rate > 26/min Higher pulse rate Pulse > 100/min Hypotension (unspecified) Decreased systolic blood pressure Systolic blood pressure < 100 mm Hg Decreased diastolic blood pressure Decreased mean arterial pressure Dullness to percussion Bronchial breathing Crepitations Peripheral edema Elevated jugular venous pressure Confusion
Studies !Showing an Assochitiont Multivariate Statistical Test Univariate Not Stated
2,19,29,33 3 8,19 2,4,6,10,14 2,3,19 7
6,10,35
10,38
30
38
10,14,33 37 10,37
26 28,34 8,31 24
10 4,5,7,14,19,23, 26,29,33
Studies Showing Association
N o
6
10,34 26 10,35,37 10,37
5
32 10 14 12 28
31 38
33 24 4,14,19,28,33,35 24 14
32
37 35 21 10
37
10
10
10
10
10,19,35,37 10 37 37 37 37 37
10 37 10
37 38 10 31 38 37 31 10 10
10,37 37
10 9,27
37 10 10
37 10 38 37 37 37 37 37
27
10,37
37
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6 Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
433
Appendix Table 2. Prognostic Factors for Death from Pneumonia Identified in Previous Studies: Factors from Laboratory Studies, Cause and Site of Infection, and Treatment Studies Showing an Associationt Univariate Multivariate
Variable
Statistical Test Not Stated Laboratory study Blood urea nitrogen Elevated > 7 mmol/L > 24 mmol/L Creatinine > 180 umol/L Hematocrit Leukocyte count < 4 x 109/L < 5 x 109/L < 9 x 109/L < 10 x 109/L > 10 x 109/L > 20 x 109/L < 10 x 109/L or > 25 x 109/L Lymphocyte count < 109/L Po2 < 6.6 kPa Decreased P0l Elevated PCOl Decreased blood pH Abnormal liver enzyme levels Decreased serum albumin Albumin < 25 g/L At least two lobes involved Both lungs affected Bronchopneumonia (compared with lobar) Atrial fibrillation Negative antigen skin test Stress hormone profile§ Cause and site of infection Streptococcus pneumoniae Pneumococcus type III Pneumococcus type III, mucoid strain Mycoplasmata Haemophilus influenzae Staphylococci Staphylococci with influenza A Legionella species Gram-negative bacilli Specific type of gram-negative bacilli Pneumonia with serologic evidence of influenza A Pneumonia during influenza epidemic No pathogen isolated Bacteremia Pneumococcal bacteremia Pneumococcal antigenemia Serum level of pneumococcal antigen Bacteremia other than pneumococcal Gram-negative bacteremia Staphylococcal bacteremia Empyema Extrapulmonary focus of infection Therapy and course of illness Antibiotics used within 2 weeks of admission Symptomatic therapy before admission Therapy with digoxin before admission Previous illness treated before admission Other therapy before admission Shorter interval between onset of illness and hospitalization Type of antibiotics used Antibiotics given before day 4 of illness No response to penicillin treatment Failure of hypothermia to remit during treatment
27 29 29
10 10
Studies Showing No Associationt
38 10
4
24
10,31 28 35
10
10
10
27 4 14,24 27 27 29 29 29 27 27 4,14,28,29,35
10 10 10 10
36
31
38
10 10
22 27 36 39 4,6,34 6,14,15,19,20,29
10 33 33 10
6 3,6,11,13,34 5 5 3
10 18,25 3
2 38 3,6,15,16,23,34 6,20,34
17 3 10
18 10 20 2 3 3
33 14,19
26,33 37 10 10 10 10
10
35 7,14 14 31 31
Continued on following page 434
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
Appendix Table 2.—Continued Variable
Studies Showing an Association! Statistical Test Univariate Multivariate Not Stated
Systolic blood pressure < 90 mm Hg and not responding to treatment Development of renal failure while in hospital Need for higher oxygen concentrations, especially with FI ()l = 111 Use of positive end-expiratory pressure Do-not-resuscitate order at admission Other APACHE II value
Studies Showing No Associationt
31 35 35 28 38 38
* Studies cited represent several different patient populations: patients with community-acquired pneumonia (2-10, 13, 16, 18, 22, 25, 27, 36); patients with severe community-acquired pneumonia (28, 34, 35, 39); elderly patients with pneumonia (31, 36, 37); patients with streptococcal pneumonia (15, 20, 23); patients with pneumococcal bacteremia (14, 19, 24, 26, 29, 30, 33); patients with pneumonia during influenza outbreaks (11, 12, 17); patients in an underdeveloped nation who had pneumonia (31); and male British doctors with pneumonia (21). Some studies that examined the level of physical and laboratory variables as an index of disease severity looked specifically at values obtained at admission or shortly thereafter (6, 9, 10, 14, 27-29, 31, 35, 37-39). + Reference numbers are given. t Employed, unemployed, or retired. § Included adrenaline, noradrenaline, human growth hormone, Cortisol, glucose, free fatty acids, prolactin, dopamine, glucagon, insulin, and adrenocorticotrophin. 1! FI ( ) , = fraction of inspired oxygen.
Acknowledgments: The authors thank Linda C. Wilson and Donna I, Consiglio for manuscript preparation and Kathleen McClure and Jo Ann Peach for technical assistance. Requests for Reprints: Barry Farr, MD, University of Virginia, Health Sciences Center, Box 473, Charlottesville, VA 22908. Current Author Addresses: Dr. Farr: University of Virginia Health Sciences Center, Box 473, Charlottesville, VA 22908. Dr. Sloman: University of Iowa Hospitals & Clinics, Department of Internal Medicine, Iowa City, IA 52242. Dr. Fisch: University of Virginia Health Sciences Center, Box 199, Charlottesville, VA 22908.
References 1. Vital Health Statistics Series: National Hospital Discharge Annual Survey; 1987 Summary. Hyattsville, Maryland: National Center for Health Statistics; 1989. Series 13, no. 99. 2. Mufson MA, Chang V, Gill V, Wood SC, Romansky MJ, Chanock RM. The role of viruses, mycoplasmas and bacteria in acute pneumonia in civilian adults. Am J Epidemiol. 1967;86:526-44. 3. Sullivan RJ Jr, Dowdle WR, Marine WM, Hierholzer JC. Adult pneumonia in a general hospital. Arch Intern Med. 1972;129:935-42. 4. Moore MA, Merson MH, Charache P, Shepard RH. The characteristics and mortality of outpatient-acquired pneumonia. Johns Hopkins Med J. 1977;140:9-14. 5. White RJ, Blainey AD, Harrison KJ, Clark SK. Causes of pneumonia presenting to a district general hospital. Thorax. 1981;36:566-70. 6. Macfarlane JT, Finch RG, Ward MJ, Macrae AD. Hospital study of adult community-acquired pneumonia. Lancet. 1982;2:255-8. 7. McNabb WR, Shanson DC, Williams TD, Lant AF. Adult community-acquired pneumonia in central London. J R Soc Med. 1984:77: 550-5. 8. Marrie TJ, Haldane EV, Faulkner RS, Durant H, Kwan C. Community-acquired pneumonia requiring hospitalization. Is it different in the elderly? J Am Geriatr Soc. 1985;33:671-80. 9. Fedullo AJ, Swinburne AJ. Relationship of patient age to clinical features and outcome for in-hospital treatment of pneumonia. J Geront. 1985;40:29-33. 10. British Thoracic Society and the Public Health Laboratory Service. Community-acquired pneumonia in adults in British hospitals in 1982-1983: a survey of aetiology, mortality, prognostic factors and outcome. Q J Med. 1987;62:195-220. 11. Robertson L, Caley JP, Moore J. Importance of Staphylococcus aureus in pneumonia in the 1957 epidemic of influenza A. Lancet. 1958;2:233-6. 12. Louria DB, Blumenfeld HL, Ellis JT, Kilbourne ED, Rogers DE. Studies on influenzae in the pandemic of 1957-1958. II. Pulmonary complications of Influenza. J Clin Invest. 1959;38:213-65. 13. Holland WW, Tanner EI, Pereira MS, Taylor CE. A study of the aetiology of respiratory disease in a general hospital. Br Med J. 1960;1:1917-22.
14. Austrian R, Gold J. Pneumococcal bacteremia with especial reference to bacteremic pneumococcal pneumonia. Ann Intern Med. 1964;60:759-76. 15. Calder MA, McHardy VU, Schonell ME. Importance of pneumococcal typing in pneumonia. Lancet. 1970;1:5-7. 16. Fekety FR Jr, Caldwell J, Gump D, Johnson JE, Maxson W, Mulholland J, et al. Bacteria, viruses, and mycoplasmas in acute pneumonia in adults. Am Rev Respir Dis. 1971;104:499-507. 17. Bisno AL, Griffin JP, Van Epps KA, Niell HB, Rytel MW. Pneumonia and Hong Kong influenza: a prospective study of the 1968-1969 epidemic. Am J Med Sci. 1971;261:251-63. 18. Dorff GJ, Rytel MW, Farmer SG, Scanlon G. Etiology and characteristic features of pneumonias in a municipal hospital. Am J Med Sci. 1973;266:349-58. 19. Mufson MA, Kruss DM, Wasil RE, Metzer WI. Capsular types and outcome of bacteremic pneumococcal disease in the antibiotic era. Arch Intern Med. 1974;134:505-10. 20. Spencer RC, Savage MA. Use of Counter and rocket immunoelectrophoresis in acute respiratory infections due to Streptococcus pneumoniae. J Clin Path. 1976;29:187-90. 21. Doll R, Peto R. Mortality in relation to smoking; 20 years' observations on male British doctors. Br Med J. 1976;2:1525-36. 22. Burns MW, Devitt L, Bryant DH. Pneumonia in a city hospital. Med J Aust. 1976;2:787-91. 23. Valenti WM, Jenzer M, Bentley DW. Type-specific pneumococcal respiratory disease in the elderly and chronically ill. Am Rev Respir Dis. 1978;117:233-8. 24. Bille J, Glauser MP, Freedman LR. Risk of death in adult pneumococcal bacteremia. In: Lambert HP, Caldwell AD, eds. Pneumonia and Pneumococcal Infections. London, England: London Royal Society of Medicine; 1980. 25. Ebright Jr, Rytel MW. Bacterial pneumonia in the elderly. J Am Geriatr Soc. 1980;28:220-3. 26. Mufson MA, Oley G, Hughey D. Pneumococcal disease in a medium-sized community in the United States. JAMA. 1982;248:1486-9. 27. Macfarlane JT. Adverse prognostic factors in pneumonia. Thorax. 1983:38:231. 28. Hook EW 3d, Horton CA, Schaberg DR. Failure of intensive care unit to influence mortality from pneumococcal bacteremia. JAMA. 1983;249:1055-7. 29. Banks RA, George RC, McNicol MW. Pneumococcal pneumonia with bacteraemia. Br J Dis Chest. 1984;78:352-7. 30. Esposito AL. Community-acquired bacteremic pneumococcal pneumonia. Effect of age on manifestations and outcome. Arch Intern Med. 1984;144:945-8. 31. Allen SC. Lobar pneumonia in Northern Zambia: clinical study of 502 adult patients. Thorax. 1984;39:612-6. 32. Andrews J, Chandrasekaran P, McSwiggan D. Lower respiratory tract infections in an acute geriatric male ward: a one-year prospective surveillance. Gerontology. 1984;30:290-296. 33. Grandsen WR, Eykyn SJ, Phillips I. Pneumococcal bacteremia: 325 episodes diagnosed at St. Thomas's Hospital. Br Med J. 1985;290: 505-8.
15 September 1991 • Annals of Internal Medicine Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
• Volume 115 • Number 6
435
34. Wood head MA, Macfarlane JT, Rodgers FG, Laverick A, Pilkington R, Macrae AD. Aetiology and outcome of severe community-acquired pneumonia. J Infect. 1985;10:204-10. 35. Ortqvist A, Sterner G, Nilsson JA. Severe community-acquired pneumonia: factors influencing need of intensive care treatment and prognosis. Scand J Infect Dis. 1985;17:377-86. 36. Marrie TJ, Durant H. A positive response to any of seven intradermal antigens predicts favorable outcome in patients hospitalized with community-acquired pneumonia. Clin Invest Med. 1988;! 1: 10-5. 37. Starczewski AR, Allen SC, Vargas E, Lye M. Clinical prognostic indices of fatality in elderly patients admitted to hospital with acute pneumonia. Age Ageing. 1988;17:181-6. 38. Daley J, Jencks S, Draper D, Lenhart G, Thomas IN, Walker J. Predicting hospital-associated mortality for medicare patients. A method for patients with stroke, pneumonia, acute myocardial infarction, and congestive heart failure. JAMA. 1988;260:3617-24. 39. Feldman C, Joffe B, Panz VR, Levy H, Walker L, Kallenbach JM, et al. Initial hormonal and metabolic profile in critically ill patients with community-acquired lobar pneumonia. S Afr Med J. 1989;76: 593-6.
436
40. SAS Institute, Inc. SAS User's Guide: Statistics. Version 5. Cary, North Carolina:SAS Institute, Inc.; 1985:956. 41. Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology, The Essentials. 2d edition. Baltimore, Maryland:Williams & Wilkins, 1988:61-4. 42. Crofton J, Douglas A. Respiratory Diseases. 3d edition. Oxford, England: Blackwell Scientific Publications; 1981:190. 43. Benson H, Mohammed A, Adler LN, Abelmann WH. Hemodynamic effects of pneumonia. I. Normal and hypodynamic responses. J Clin Invest. 1970;49:791-8. 44. Rogers RM, Weiler C, Ruppenthal B. Impact of the respiratory intensive care unit on survival of patients with acute respiratory failure. Chest. 1972;62:94-7. 45. Gold MI, Shin BH. Respiratory care in the absence of a respiratory care unit. Chest. 1974;65:388-93. 46. Potgeiter PD, Rosenthal E, Benatar SR. Immediate and long-term survival in patients admitted to a respiratory ICU. Crit Care Med. 1985;13:798-802. 47. Fine MJ, Orloff JJ, Arisumi D, Fang GD, Arena VC, Hanusa BH. Prognosis of patients hospitalized with community acquired pneumonia. Am J Med. 1990;88:5-lN-5-8N.
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
• Objective: To validate a previously reported discriminant rule for predicting mortality in adult patients with primary community-acquired pneumonia and to determine which factors available at hospital admission predict a fatal outcome among such patients. • Design: Historical cohort study. • Setting: University hospital. • Patients: Adults admitted to the hospital for community-acquired pneumonia. • Measurements: Using stepwise logistic regression, we analyzed prognostic factors (data available at admission and recorded in the medical record) that showed a univariate association with mortality. The predictive values of three discriminant rules were measured to validate the results of a previous study. • Main Results: Of 245 patients, 20 (8.2%) died. Of 42 prognostic factors identified in previous studies, 8 were associated with mortality, but only a respiratory rate of 30/min or more, a diastolic blood pressure of 60 mm Hg or less, and a blood urea nitrogen of more than 7 mmol/L remained predictive in the multivariate analysis. A discriminant rule composed of these three variables was 70% sensitive and 84% specific in predicting mortality, yielding an overall accuracy of 82%. • Conclusion: Tachypnea, diastolic hypotension, and an elevated blood urea nitrogen were independently associated with death from pneumonia in our study, confirming the value of a previously reported discriminant rule from the British Thoracic Society. This rule may be useful in triage decisions because it identifies high-risk patients who may benefit from special medical attention.
Community-acquired pneumonia remains an important cause of mortality in developed nations. In the United States, approximately 924 000 cases of pneumonia occur annually, accounting for 2.8% of all hospital admissions (1). The patient dies in 6% to 24% of cases of community-acquired pneumonia requiring hospitalization, and at least 50 000 deaths per year in the United States are attributed to this illness. Until recently, a rigorous epidemiologic study using multivariate analysis of the prognostic factors associated with mortality in pneumonia had not been done (2-39) (see Appendix Tables 1 and 2). In 1987, the British Thoracic Society (BTS) subjected an extensive list of patient variables to statistical analysis in a prospective study of prognosis in 453 adults with community-acquired pneumonia (10). Multivariate analyses identified three variables that were consistently associated with death—respiratory rate, diastolic blood pressure, and blood urea nitrogen. Using optimal cutoff points of these variables for that data set (respiratory rate of 30/min or more, diastolic blood pressure of 60 mm Hg or less, and blood urea nitrogen of more than 7 mmol/L [19.6 mg/dL]), the investigators formulated an accurate but simple discriminant rule that was considered to be positive when at least two of the three factors were present. In the BTS study, a positive rule was associated with a 21-fold increase in mortality (88% sensitivity and 79% specificity). Such a rule, if confirmed to be predictive, might be useful at the time of admission for rapidly identifying high-risk patients who might require special attention, perhaps in an intensive care unit. The validity of the BTS results regarding prognostic factors and discriminant rules was not tested in an independent cohort of patients with pneumonia, however, and the purpose of our study was to assess the validity of these findings among patients with community-acquired pneumonia who had been admitted to the University of Virginia Hospital.
Methods Sample
Annals of Internal Medicine. 1991;115:428-436. From University of Virginia Health Sciences Center, Charlottesville, Virginia. For current author addresses, see end of text. 428
Patients with primary pneumonia were selected by retrospective review of the hospital records of consecutive cases of pneumonia occurring in a 25-month period (January 1984 through January 1986) at the University of Virginia Hospital. Cases were identified by a computer listing of all patients with a discharge diagnosis of pneumonia. Primary community-acquired pneumonia was defined as an acute respiratory illness contracted in the community and accompanied by a new radiographic infiltrate. The study group ranged in age from 15 to 80 years. Patients were excluded if pneumonia was not the primary reason for hospitalization, if pneumonia was an ex-
© 1991 American College of Physicians
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
Table 1. Pathogens Isolated from Cultures of 245 Patients with Community-acquired Pneumonia* Pathogen
Patients with Isolate in Sputumt
Patients with Isolate in Bloody yi(0/r,\
7.0 mmol/L Diastolic blood pressure < 60 mm Hg Respiratory rate > 30/min
Beta Coefficient
Relative Risk (95% CI)
P Value
2.19 1.30 1.15
8.93 (2.66 to 30.00) 3.67 (1.23 to 10.90) 3.16 (1.07 to 9.31)
< 0.001 0.014 0.028
15 September 1991 • Annals of Internal Medicine Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
• Volume 115 • Number 6
429
Appendix Table 2. Prognostic Factors for Death from Pneumonia Identified in Previous Studies: Factors from Laboratory Studies, Cause and Site of Infection, and Treatment Studies Showing an Associationt Univariate Multivariate
Variable
Statistical Test Not Stated Laboratory study Blood urea nitrogen Elevated > 7 mmol/L > 24 mmol/L Creatinine > 180 umol/L Hematocrit Leukocyte count < 4 x 109/L < 5 x 109/L < 9 x 109/L < 10 x 109/L > 10 x 109/L > 20 x 109/L < 10 x 109/L or > 25 x 109/L Lymphocyte count < 109/L Po2 < 6.6 kPa Decreased P0l Elevated PCOl Decreased blood pH Abnormal liver enzyme levels Decreased serum albumin Albumin < 25 g/L At least two lobes involved Both lungs affected Bronchopneumonia (compared with lobar) Atrial fibrillation Negative antigen skin test Stress hormone profile§ Cause and site of infection Streptococcus pneumoniae Pneumococcus type III Pneumococcus type III, mucoid strain Mycoplasmata Haemophilus influenzae Staphylococci Staphylococci with influenza A Legionella species Gram-negative bacilli Specific type of gram-negative bacilli Pneumonia with serologic evidence of influenza A Pneumonia during influenza epidemic No pathogen isolated Bacteremia Pneumococcal bacteremia Pneumococcal antigenemia Serum level of pneumococcal antigen Bacteremia other than pneumococcal Gram-negative bacteremia Staphylococcal bacteremia Empyema Extrapulmonary focus of infection Therapy and course of illness Antibiotics used within 2 weeks of admission Symptomatic therapy before admission Therapy with digoxin before admission Previous illness treated before admission Other therapy before admission Shorter interval between onset of illness and hospitalization Type of antibiotics used Antibiotics given before day 4 of illness No response to penicillin treatment Failure of hypothermia to remit during treatment
27 29 29
10 10
Studies Showing No Associationt
38 10
4
24
10,31 28 35
10
10
10
27 4 14,24 27 27 29 29 29 27 27 4,14,28,29,35
10 10 10 10
36
31
38
10 10
22 27 36 39 4,6,34 6,14,15,19,20,29
10 33 33 10
6 3,6,11,13,34 5 5 3
10 18,25 3
2 38 3,6,15,16,23,34 6,20,34
17 3 10
18 10 20 2 3 3
33 14,19
26,33 37 10 10 10 10
10
35 7,14 14 31 31
Continued on following page 434
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
trogen was correlated with age, previous illness, elevated liver enzymes, and an elevated leukocyte count, which may explain why these variables did not remain associated with death in the multivariate analysis. Similarly, diastolic hypotension was correlated with a history of confusion and with systolic hypotension. Tachypnea was correlated with tachycardia, underlying lung disease, a history of smoking, male sex, and bacteremia.
mechanical ventilation (P = 0.009). Seven patients who died were not intubated, and rule 1 predicted death in 4 (57%) of them, compared with 31 of the 201 survivors (15%) who were not intubated (P = 0.02). The accuracy of rule 1 in predicting outcome for the nine patients with chronic renal failure (admission blood urea nitrogen range 7 to 27 mmol/L) was also assessed. Death was predicted in three of the eight survivors and was not predicted in the single patient who died.
Discriminant Rules Application of the three discriminant rules assessed by the BTS study (rules 1, 2, and 3 in Table 4) showed that a rule incorporating tachypnea, diastolic hypotension, and a raised blood urea nitrogen level (rule 1) was the most powerful predictor of mortality (Table 4). All three rules showed both high specificity and negative predictive value, but rule 1 was a more sensitive predictor and was associated with a higher relative risk for death than rule 2 (diastolic hypotension, tachypnea, and confusion) or rule 3 (elevated blood urea nitrogen, confusion, P0, < 6.6 kPa, and leukocyte count < 10 x 109/L or lymphocyte count < 1 x 109/L). Although the numbers of patients tested by rules 2 and 3 were smaller (n = 237 and n = 205, respectively) than the number of patients tested by rule 1 (n = 241) (because there were some patients with missing information), all 20 patients who died were in the group tested by rule 2, and 19 of the 20 patients who died were in the group tested by rule 3. Rule 1 identified 14 of these 20 patients, but rules 2 and 3 identified only 7 and 8 of these patients, respectively. Of the 30 patients who were intubated and mechanically ventilated, 13 died. On the basis of data collected at admission, rule 1 predicted death in 10 (77%) of these patients and in 4 of the 17 patients (24%) who survived
Discussion Our historical cohort study of patients with community-acquired pneumonia confirms the predictive value of three prognostic factors identified in a previous study (admission respiratory rate, diastolic blood pressure, and blood urea nitrogen) and the utility of a simple discriminant rule that uses these variables to predict mortality. This rule was first reported by the BTS Research Committee (10), which applied a stepwise logistic regression analysis to clinical variables that had shown a univariate association with mortality. In the BTS study, a multivariate association was shown for abstinence from alcohol, absence of chest pain, absence of vomiting, treatment with digoxin before admission, tachypnea, diastolic hypotension, and an elevated blood urea nitrogen level. A clinically convenient rule was created by dichotomizing the latter three variables at optimal cutoff points as determined from the BTS data set: respiratory rate of 30/min or more, a diastolic blood pressure of 60 mm Hg or less, and a blood urea nitrogen level of more than 7 mmol/L. The presence of any two of these three criteria was associated with a 21-fold greater risk for death in the BTS cohort study and predicted mortality with 88% sensitivity and 79% specificity.
Table 4. Test Performance of Three Discriminant Rules in Predicting Mortality Rule
Positive Predictive Value
Negative Predictive Value
Overall Accuracy
Sensitivity (95% CI)*
Specificity (95% CI)*
Youden Index
Likeli- Relative hood Risk Ratio
vi/n/CZ, \
niny 70) —
* 1. Two or more of the following: blood urea nitrogen > 7.0 mmol/L; diastolic blood pressure < 60 mm Hg; and respiratory rate > 30/min 2. Two or more of the following: diastolic blood pressure < 60 mm Hg; respiratory rate > 30/min; and confusion 3. Three or more of the following: blood urea nitrogen > 7.0 mmol/L; confusion; Po2 < 6.6 kPa; leukocyte count < 10 x 109/L or lymphocyte count < 109/L
14/49 (28.6)
186/192 (96.9)
200/241 (82.3)
14/20 (70.0 [49.9 to 90.1])
186/221 (84.2 [79.4 to 89.0])
0.54
4.4
9.1
7/32(21.9)
192/205 (93.7)
199/237 (84.0)
7/20 (35.0 [14.1, 55.9])
192/217(88.5 [84.3 to 92.7])
0.24
3.0
3.4
8/33 (24.2)
161/172 (93.6)
169/205 (82.4)
8/19(42.1 [19.9 to 64.3])
161/186(86.6 [81.7 to 91.5])
0.29
3.1
3.8
* CI = confidence interval. 15 September 1991 • Annals of Internal Medicine Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
• Volume 115 • Number 6
431
Our cohort study of patients with pneumonia from central Virginia examined the association between death and 42 admission variables, including those identified as prognostic factors in the BTS study (10) and others (2-39). Univariate associations were confirmed for older age, age of more than 60 years, respiratory rate of 30/min or more, decreased diastolic blood pressure, diastolic blood pressure of 60 mm Hg or less, decreased systolic blood pressure, elevated blood urea nitrogen, blood urea nitrogen of more than 7 mmol/L, decreased serum albumin level, leukocyte count of less than 4 x 109/L, and a P0i of 6.6 kPa or less. These results confirmed the association between a higher respiratory rate and mortality that was first reported in the BTS study but failed to confirm the association between mortality and four other factors found to be significant by the BTS study: These factors included abstinence from alcohol, absence of chest pain, absence of vomiting, and digoxin therapy before admission. In our stepwise logistic regression analysis, only a respiratory rate of 30/min or more, a diastolic blood pressure of 60 mm Hg or less, and a blood urea nitrogen of more than 7 mmol/L— the three components of the BTS rule—remained predictive. Other univariate predictors might have remained associated in the multivariate analysis with a larger sample size yielding more deaths and higher statistical power. When applied to our cohort, the BTS rule predicted mortality with 70% sensitivity, 84% specificity, and a likelihood ratio of 4.4. A positive rule was associated with a ninefold greater risk for death from communityacquired pneumonia. The lower sensitivity and relative risk for death associated with the rule in our study relative to its performance in the BTS study are not surprising, because a prognostic model derived from the data of the BTS study would be expected to perform better in that population than in an independent cohort. Nevertheless, the predictive value of the rule in our study confirms its utility as an indicator of prognosis. When the rule was applied to our patients who required mechanical ventilation, a subset of patients showing a strong univariate association with death, the rule predicted death as well as it did in the entire sample (sensitivity, 77% compared with 70%; specificity, 76% compared with 84%). Therefore, the rule appears to be applicable to severely ill patients with community-acquired pneumonia, even those in an intensive care unit. The rule did not perform as well in patients with chronic renal failure. The rule may have a higher falsepositive rate among patients with renal failure; because such patients have a chronically elevated blood urea nitrogen level, the addition of only one more positive variable (that is, tachypnea or diastolic hypotension) would achieve a positive rule. The correlation of these three variables, which are measures of physiologic derangement, with death from pneumonia is not unexpected. An elevated blood urea nitrogen may result from dehydration and hypotension (42); diastolic hypotension results from intravascular depletion, alterations in global vasoregulation, and myocardial depression (43); and tachypnea occurs as the respiratory system adjusts to the presence of local in432
flammation, hypercapnia, acidosis, and, in some cases, hypoxia. The levels of these three variables should therefore reflect the degree of these physiologic imbalances and thus may bear a logical relation to the probability of death. However, the emergence of these variables as independent prognostic factors from a large group of variables using multivariate analyses in two different cohort studies is noteworthy. Nevertheless, although the discriminant rule composed of these three variables appears to be an accurate predictor of outcome in patients with community-acquired pneumonia, it does not necessarily follow that its use will alter that outcome. Indeed, it has been suggested that death from pneumonia mortality is predetermined (14) and that intensive care merely prolongs the interval before death (28). This question has not been tested in a randomized controlled trial, however, and probably will not be because of the ethical implications of withholding intensive care from severely ill patients. Nevertheless, the application of this rule has the potential for improving survival in patients with community-acquired pneumonia. Some investigators believe that specialized care does prevent death from pneumonia (34, 44-46), and the rule may be helpful in the emergency room to rapidly determine which patients with pneumonia need such care. The predictive value of this discriminant rule has not been studied in outpatient clinics, which treat many patients with oral therapy at home who have milder cases of pneumonia, but because the probability of death is considerably lower among such patients, the positive predictive value of the rule among such patients would be expected to be lower. Another study of prognostic factors in communityacquired pneumonia was reported after the completion of our study (47); this other study involved only adult patients with community-acquired pneumonia, including the types of patients excluded from our study, such as patients who were immunocompromised because of metastatic cancer or cancer chemotherapy, patients in whom pneumonia was an expected terminal event, and patients with mycobacterial, postobstructive, or aspiration pneumonia. Given this difference in study design, the finding of underlying neoplastic disease as the strongest predictor of death is understandable. Other independent predictors of death were age of more than 65 years; absence of pleuritic chest pain; mental status changes; vital sign abnormalities; and pneumonia due to Staphylococcus aureus, gram-negative bacilli, bronchial obstruction, or aspiration. The use of variables not available at hospital admission, such as results of sputum and blood cultures, render these findings somewhat less applicable to emergency room triage. Our study has confirmed the predictive value of a previously reported, simple discriminant rule that incorporates respiratory rate, diastolic blood pressure, and blood urea nitrogen level to predict death in patients with community-acquired pneumonia at the time of hospital admission. Such a rule may be of use to the admitting physician for rapidly determining which patients warrant special attention because of a high risk for death.
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
Appendix Table 1. Prognostic Factors for Death from Pneumonia Identified in Previous Studies: Factors from History and Physical Examination* Variable
History Age Older > 40 years > 50 years > 60 years > 65 years > 70 years > 75 years Male sex Employment status^ Social status Any preexisting chronic illness At least two preexisting illnesses Preexisting respiratory illness Chronic bronchitis Chronic obstructive pulmonary disease Preexisting nonrespiratory illness Heart disease or pulmonary edema Rheumatic heart disease Neoplasm Diffuse or metastatic cancer Cirrhosis Preexisting renal or hepatic insufficiency Chronic alcoholism Chronic alcoholism without cirrhosis Acute alcoholism Dementia Immunocompromised Smoker or ex-smoker Absence of alcohol use Present use of hypnotics or tranquilizers Absence of chest pain Presence of chest pain Absence of vomiting Cough Dyspnea Increased sputum production Chills, sweats, and rigors Confusion History of falls Inability to walk Physical examination Admission temperature Admission temperature < 37 °C Hypothermia in first 24 hours Admission temperature > 37.8 °C Admission temperature > 40 °C Greatest temperature attained Higher respiratory rate Respiratory rate > 26/min Higher pulse rate Pulse > 100/min Hypotension (unspecified) Decreased systolic blood pressure Systolic blood pressure < 100 mm Hg Decreased diastolic blood pressure Decreased mean arterial pressure Dullness to percussion Bronchial breathing Crepitations Peripheral edema Elevated jugular venous pressure Confusion
Studies !Showing an Assochitiont Multivariate Statistical Test Univariate Not Stated
2,19,29,33 3 8,19 2,4,6,10,14 2,3,19 7
6,10,35
10,38
30
38
10,14,33 37 10,37
26 28,34 8,31 24
10 4,5,7,14,19,23, 26,29,33
Studies Showing Association
N o
6
10,34 26 10,35,37 10,37
5
32 10 14 12 28
31 38
33 24 4,14,19,28,33,35 24 14
32
37 35 21 10
37
10
10
10
10
10,19,35,37 10 37 37 37 37 37
10 37 10
37 38 10 31 38 37 31 10 10
10,37 37
10 9,27
37 10 10
37 10 38 37 37 37 37 37
27
10,37
37
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6 Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
433
Appendix Table 2. Prognostic Factors for Death from Pneumonia Identified in Previous Studies: Factors from Laboratory Studies, Cause and Site of Infection, and Treatment Studies Showing an Associationt Univariate Multivariate
Variable
Statistical Test Not Stated Laboratory study Blood urea nitrogen Elevated > 7 mmol/L > 24 mmol/L Creatinine > 180 umol/L Hematocrit Leukocyte count < 4 x 109/L < 5 x 109/L < 9 x 109/L < 10 x 109/L > 10 x 109/L > 20 x 109/L < 10 x 109/L or > 25 x 109/L Lymphocyte count < 109/L Po2 < 6.6 kPa Decreased P0l Elevated PCOl Decreased blood pH Abnormal liver enzyme levels Decreased serum albumin Albumin < 25 g/L At least two lobes involved Both lungs affected Bronchopneumonia (compared with lobar) Atrial fibrillation Negative antigen skin test Stress hormone profile§ Cause and site of infection Streptococcus pneumoniae Pneumococcus type III Pneumococcus type III, mucoid strain Mycoplasmata Haemophilus influenzae Staphylococci Staphylococci with influenza A Legionella species Gram-negative bacilli Specific type of gram-negative bacilli Pneumonia with serologic evidence of influenza A Pneumonia during influenza epidemic No pathogen isolated Bacteremia Pneumococcal bacteremia Pneumococcal antigenemia Serum level of pneumococcal antigen Bacteremia other than pneumococcal Gram-negative bacteremia Staphylococcal bacteremia Empyema Extrapulmonary focus of infection Therapy and course of illness Antibiotics used within 2 weeks of admission Symptomatic therapy before admission Therapy with digoxin before admission Previous illness treated before admission Other therapy before admission Shorter interval between onset of illness and hospitalization Type of antibiotics used Antibiotics given before day 4 of illness No response to penicillin treatment Failure of hypothermia to remit during treatment
27 29 29
10 10
Studies Showing No Associationt
38 10
4
24
10,31 28 35
10
10
10
27 4 14,24 27 27 29 29 29 27 27 4,14,28,29,35
10 10 10 10
36
31
38
10 10
22 27 36 39 4,6,34 6,14,15,19,20,29
10 33 33 10
6 3,6,11,13,34 5 5 3
10 18,25 3
2 38 3,6,15,16,23,34 6,20,34
17 3 10
18 10 20 2 3 3
33 14,19
26,33 37 10 10 10 10
10
35 7,14 14 31 31
Continued on following page 434
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
Appendix Table 2.—Continued Variable
Studies Showing an Association! Statistical Test Univariate Multivariate Not Stated
Systolic blood pressure < 90 mm Hg and not responding to treatment Development of renal failure while in hospital Need for higher oxygen concentrations, especially with FI ()l = 111 Use of positive end-expiratory pressure Do-not-resuscitate order at admission Other APACHE II value
Studies Showing No Associationt
31 35 35 28 38 38
* Studies cited represent several different patient populations: patients with community-acquired pneumonia (2-10, 13, 16, 18, 22, 25, 27, 36); patients with severe community-acquired pneumonia (28, 34, 35, 39); elderly patients with pneumonia (31, 36, 37); patients with streptococcal pneumonia (15, 20, 23); patients with pneumococcal bacteremia (14, 19, 24, 26, 29, 30, 33); patients with pneumonia during influenza outbreaks (11, 12, 17); patients in an underdeveloped nation who had pneumonia (31); and male British doctors with pneumonia (21). Some studies that examined the level of physical and laboratory variables as an index of disease severity looked specifically at values obtained at admission or shortly thereafter (6, 9, 10, 14, 27-29, 31, 35, 37-39). + Reference numbers are given. t Employed, unemployed, or retired. § Included adrenaline, noradrenaline, human growth hormone, Cortisol, glucose, free fatty acids, prolactin, dopamine, glucagon, insulin, and adrenocorticotrophin. 1! FI ( ) , = fraction of inspired oxygen.
Acknowledgments: The authors thank Linda C. Wilson and Donna I, Consiglio for manuscript preparation and Kathleen McClure and Jo Ann Peach for technical assistance. Requests for Reprints: Barry Farr, MD, University of Virginia, Health Sciences Center, Box 473, Charlottesville, VA 22908. Current Author Addresses: Dr. Farr: University of Virginia Health Sciences Center, Box 473, Charlottesville, VA 22908. Dr. Sloman: University of Iowa Hospitals & Clinics, Department of Internal Medicine, Iowa City, IA 52242. Dr. Fisch: University of Virginia Health Sciences Center, Box 199, Charlottesville, VA 22908.
References 1. Vital Health Statistics Series: National Hospital Discharge Annual Survey; 1987 Summary. Hyattsville, Maryland: National Center for Health Statistics; 1989. Series 13, no. 99. 2. Mufson MA, Chang V, Gill V, Wood SC, Romansky MJ, Chanock RM. The role of viruses, mycoplasmas and bacteria in acute pneumonia in civilian adults. Am J Epidemiol. 1967;86:526-44. 3. Sullivan RJ Jr, Dowdle WR, Marine WM, Hierholzer JC. Adult pneumonia in a general hospital. Arch Intern Med. 1972;129:935-42. 4. Moore MA, Merson MH, Charache P, Shepard RH. The characteristics and mortality of outpatient-acquired pneumonia. Johns Hopkins Med J. 1977;140:9-14. 5. White RJ, Blainey AD, Harrison KJ, Clark SK. Causes of pneumonia presenting to a district general hospital. Thorax. 1981;36:566-70. 6. Macfarlane JT, Finch RG, Ward MJ, Macrae AD. Hospital study of adult community-acquired pneumonia. Lancet. 1982;2:255-8. 7. McNabb WR, Shanson DC, Williams TD, Lant AF. Adult community-acquired pneumonia in central London. J R Soc Med. 1984:77: 550-5. 8. Marrie TJ, Haldane EV, Faulkner RS, Durant H, Kwan C. Community-acquired pneumonia requiring hospitalization. Is it different in the elderly? J Am Geriatr Soc. 1985;33:671-80. 9. Fedullo AJ, Swinburne AJ. Relationship of patient age to clinical features and outcome for in-hospital treatment of pneumonia. J Geront. 1985;40:29-33. 10. British Thoracic Society and the Public Health Laboratory Service. Community-acquired pneumonia in adults in British hospitals in 1982-1983: a survey of aetiology, mortality, prognostic factors and outcome. Q J Med. 1987;62:195-220. 11. Robertson L, Caley JP, Moore J. Importance of Staphylococcus aureus in pneumonia in the 1957 epidemic of influenza A. Lancet. 1958;2:233-6. 12. Louria DB, Blumenfeld HL, Ellis JT, Kilbourne ED, Rogers DE. Studies on influenzae in the pandemic of 1957-1958. II. Pulmonary complications of Influenza. J Clin Invest. 1959;38:213-65. 13. Holland WW, Tanner EI, Pereira MS, Taylor CE. A study of the aetiology of respiratory disease in a general hospital. Br Med J. 1960;1:1917-22.
14. Austrian R, Gold J. Pneumococcal bacteremia with especial reference to bacteremic pneumococcal pneumonia. Ann Intern Med. 1964;60:759-76. 15. Calder MA, McHardy VU, Schonell ME. Importance of pneumococcal typing in pneumonia. Lancet. 1970;1:5-7. 16. Fekety FR Jr, Caldwell J, Gump D, Johnson JE, Maxson W, Mulholland J, et al. Bacteria, viruses, and mycoplasmas in acute pneumonia in adults. Am Rev Respir Dis. 1971;104:499-507. 17. Bisno AL, Griffin JP, Van Epps KA, Niell HB, Rytel MW. Pneumonia and Hong Kong influenza: a prospective study of the 1968-1969 epidemic. Am J Med Sci. 1971;261:251-63. 18. Dorff GJ, Rytel MW, Farmer SG, Scanlon G. Etiology and characteristic features of pneumonias in a municipal hospital. Am J Med Sci. 1973;266:349-58. 19. Mufson MA, Kruss DM, Wasil RE, Metzer WI. Capsular types and outcome of bacteremic pneumococcal disease in the antibiotic era. Arch Intern Med. 1974;134:505-10. 20. Spencer RC, Savage MA. Use of Counter and rocket immunoelectrophoresis in acute respiratory infections due to Streptococcus pneumoniae. J Clin Path. 1976;29:187-90. 21. Doll R, Peto R. Mortality in relation to smoking; 20 years' observations on male British doctors. Br Med J. 1976;2:1525-36. 22. Burns MW, Devitt L, Bryant DH. Pneumonia in a city hospital. Med J Aust. 1976;2:787-91. 23. Valenti WM, Jenzer M, Bentley DW. Type-specific pneumococcal respiratory disease in the elderly and chronically ill. Am Rev Respir Dis. 1978;117:233-8. 24. Bille J, Glauser MP, Freedman LR. Risk of death in adult pneumococcal bacteremia. In: Lambert HP, Caldwell AD, eds. Pneumonia and Pneumococcal Infections. London, England: London Royal Society of Medicine; 1980. 25. Ebright Jr, Rytel MW. Bacterial pneumonia in the elderly. J Am Geriatr Soc. 1980;28:220-3. 26. Mufson MA, Oley G, Hughey D. Pneumococcal disease in a medium-sized community in the United States. JAMA. 1982;248:1486-9. 27. Macfarlane JT. Adverse prognostic factors in pneumonia. Thorax. 1983:38:231. 28. Hook EW 3d, Horton CA, Schaberg DR. Failure of intensive care unit to influence mortality from pneumococcal bacteremia. JAMA. 1983;249:1055-7. 29. Banks RA, George RC, McNicol MW. Pneumococcal pneumonia with bacteraemia. Br J Dis Chest. 1984;78:352-7. 30. Esposito AL. Community-acquired bacteremic pneumococcal pneumonia. Effect of age on manifestations and outcome. Arch Intern Med. 1984;144:945-8. 31. Allen SC. Lobar pneumonia in Northern Zambia: clinical study of 502 adult patients. Thorax. 1984;39:612-6. 32. Andrews J, Chandrasekaran P, McSwiggan D. Lower respiratory tract infections in an acute geriatric male ward: a one-year prospective surveillance. Gerontology. 1984;30:290-296. 33. Grandsen WR, Eykyn SJ, Phillips I. Pneumococcal bacteremia: 325 episodes diagnosed at St. Thomas's Hospital. Br Med J. 1985;290: 505-8.
15 September 1991 • Annals of Internal Medicine Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
• Volume 115 • Number 6
435
34. Wood head MA, Macfarlane JT, Rodgers FG, Laverick A, Pilkington R, Macrae AD. Aetiology and outcome of severe community-acquired pneumonia. J Infect. 1985;10:204-10. 35. Ortqvist A, Sterner G, Nilsson JA. Severe community-acquired pneumonia: factors influencing need of intensive care treatment and prognosis. Scand J Infect Dis. 1985;17:377-86. 36. Marrie TJ, Durant H. A positive response to any of seven intradermal antigens predicts favorable outcome in patients hospitalized with community-acquired pneumonia. Clin Invest Med. 1988;! 1: 10-5. 37. Starczewski AR, Allen SC, Vargas E, Lye M. Clinical prognostic indices of fatality in elderly patients admitted to hospital with acute pneumonia. Age Ageing. 1988;17:181-6. 38. Daley J, Jencks S, Draper D, Lenhart G, Thomas IN, Walker J. Predicting hospital-associated mortality for medicare patients. A method for patients with stroke, pneumonia, acute myocardial infarction, and congestive heart failure. JAMA. 1988;260:3617-24. 39. Feldman C, Joffe B, Panz VR, Levy H, Walker L, Kallenbach JM, et al. Initial hormonal and metabolic profile in critically ill patients with community-acquired lobar pneumonia. S Afr Med J. 1989;76: 593-6.
436
40. SAS Institute, Inc. SAS User's Guide: Statistics. Version 5. Cary, North Carolina:SAS Institute, Inc.; 1985:956. 41. Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology, The Essentials. 2d edition. Baltimore, Maryland:Williams & Wilkins, 1988:61-4. 42. Crofton J, Douglas A. Respiratory Diseases. 3d edition. Oxford, England: Blackwell Scientific Publications; 1981:190. 43. Benson H, Mohammed A, Adler LN, Abelmann WH. Hemodynamic effects of pneumonia. I. Normal and hypodynamic responses. J Clin Invest. 1970;49:791-8. 44. Rogers RM, Weiler C, Ruppenthal B. Impact of the respiratory intensive care unit on survival of patients with acute respiratory failure. Chest. 1972;62:94-7. 45. Gold MI, Shin BH. Respiratory care in the absence of a respiratory care unit. Chest. 1974;65:388-93. 46. Potgeiter PD, Rosenthal E, Benatar SR. Immediate and long-term survival in patients admitted to a respiratory ICU. Crit Care Med. 1985;13:798-802. 47. Fine MJ, Orloff JJ, Arisumi D, Fang GD, Arena VC, Hanusa BH. Prognosis of patients hospitalized with community acquired pneumonia. Am J Med. 1990;88:5-lN-5-8N.
15 September 1991 • Annals of Internal Medicine • Volume 115 • Number 6
Downloaded From: http://annals.org/ by a Upstate Medical University User on 10/03/2013
