Chest ACCP Career Connection
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Guest Access | Sign In via User Name/Password
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow View responses
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (36)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Marras, T. K.
Right arrow Articles by Chan, C. K.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Marras, T. K.
Right arrow Articles by Chan, C. K.
(Chest. 2000;118:1339-1343.)
© 2000 American College of Chest Physicians

Applying a Prediction Rule To Identify Low-Risk Patients With Community-Acquired Pneumonia*

Theodore K. Marras, MD; Carlos Gutierrez, MD and Charles K. Chan, MD, FCCP

* From the Division of Respirology, Department of Medicine, The University Health Network, University of Toronto, Toronto, Canada.

Correspondence to: Charles K. Chan, MD, FCCP, Division of Respirology, Toronto General Hospital, University Health Network, 10EN-220, 200 Elizabeth St, Toronto ON M5G 2C4, Canada; e-mail: charles.chan{at}uhn.on.ca


   Abstract
TOP
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 Conclusion
 References
 
Study objectives: To study the validity of a recently developed community-acquired pneumonia (CAP) severity prediction rule in estimating mortality, to determine its utility in decision making regarding hospitalization, and to assess factors influencing this decision.

Design: Retrospective chart review.

Setting: Two sites of the University Health Network, the Toronto General and Toronto Western Hospitals, tertiary-care teaching institutions with a sizable primary-care and secondary-care source of referrals, and a total of 900 beds.

Patients: Consecutive patients with CAP admitted between February and June 1996.

Measurements and results: A single trained medical records extractor assembled data to compare our population to that used in developing the CAP prediction rule, in terms of mortality and to assess reasons for hospitalization. Two hundred fifty-five eligible patients were admitted, and 244 charts (96%) were available. Our patients tended to be older, with nearly four times as many residents of chronic care institutions (39% compared with 10%), and had a higher risk class distribution than the published cohort. Risk class-specific mortality was similar in four of five classes. Of the 71 patients in the low-risk classes, 67 had additional reasons for admission; 18 of which were psychosocial (homelessness, substance abuse, or inadequate home supports).

Conclusions: The CAP severity prediction rule estimates mortality well. Admission of low-risk patients was linked to psychosocial and other medical reasons not captured by this rule. The rule can be very useful in assessing the need for hospitalization; however, there remains a significant percentage of patients with a low severity score who may require hospitalization for psychosocial and economic considerations.

Key Words: aged • clinical prediction rules • guidelines • hospitalization • mortality • prognosis • risk • socioeconomic factors


   Introduction
TOP
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 Conclusion
 References
 
Clinical prediction (CP) rules can be important tools in reducing uncertainty and assisting in medical decision making. Guidelines for developing and testing CP rules have been proposed1 and subsequently modified.2 Despite meticulous care in their design and rigorous testing in the clinical trial setting, the broad clinical application of CP rules could sometimes be detrimental, especially to patient groups with special needs, such as those living in poverty. Community-acquired pneumonia (CAP) is a common disorder, leading to > 600,000 hospitalizations annually in the United States.3 Although the bulk of treatment costs for this disease are related primarily to inpatient care,4 5 the enthusiasm to reduce costs in this area by avoiding hospital admissions has been tempered by the substantial mortality associated with CAP (ranging from 8 to 16% in hospitalized patients),6 7 and a lack of published data suggesting who could be safely treated as outpatients.

Mortality prediction rules have been developed to risk-stratify patients with CAP.7 8 The rule developed by the British Thoracic Society8 and subsequently validated by others9 is a very simple and sensitive assessment for identifying seriously ill patients who likely benefit from special medical attention. Fine and colleagues7 have recently attempted to identify low-risk patients who might be safely treated as outpatients. They developed a mortality prediction rule, the pneumonia severity index (PSI), from a database of > 14,000 patients hospitalized with CAP, and used a separate database of > 40,000 patients to subsequently validate the PSI. They stratified patients into five risk classes according to 20 clinical and laboratory variables, and found a clear correlation between mortality and risk class. Further, the patients assigned to risk classes I to III had a mortality of 0.2, 0.6, and 2.7%, respectively, compared to 8% for class IV patients and 30% for class V patients. The authors concluded that all class I patients and many class II and III patients are candidates for outpatient therapy, which could lead to significant cost savings. The PSI has been studied in different settings, finding it to be valid10 but that recalibration may be required when transporting it across populations.11 We sought to apply the prediction rule developed by Fine and colleagues7 to a cohort of patients admitted to our institution to assess its validity in predicting mortality, and its proposed utility in admission decision making.


   Materials and Methods
TOP
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 Conclusion
 References
 
The study involved the Toronto General and Toronto Western Hospitals of the University Health Network (both are urban downtown tertiary-care teaching hospitals affiliated with the University of Toronto). The study population was comprised of a consecutive cohort of adults with CAP admitted to the general medical services of this institution over a 5-month period from February 1, 1996, to June 30, 1996. Disease definitions and complete inclusion and exclusion criteria have been published previously.12 We excluded patients who were known to be infected with the HIV, otherwise immunosuppressed, or suspected to have active tuberculosis. We reviewed the charts of all patients and recorded demographic data, comorbidities, and baseline clinical and laboratory features. We collected data on in-hospital mortality and length of hospital stay. According to the clinical data recorded in the emergency department (ED) assessment and initial laboratory and radiologic evaluation, patients were stratified retrospectively into one of five risk classes according to the PSI.7 To assess for differences between patient populations, the baseline demographics and clinical and laboratory characteristics of our cohort were compared with those in the study by Fine and colleagues.7 Risk class-specific mortality and length of hospital stay were compared between our study population and the published cohort of Fine and colleagues7 to assess how well the PSI predicts mortality. Among patients predicted to be at low risk for mortality (classes I to III), the admission decision was analyzed to assess the potential utility of the PSI in assisting this decision. Admission decisions were made independently of the PSI, since this group of patients was treated in 1996, prior to the publication of the study by Fine and colleagues.7 Ethics approval was obtained from our hospital ethics review board.


   Results
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 Conclusion
 References
 
Two hundred fifty-five eligible patients were admitted to the hospitals over the 5-month study period. Eleven charts (4%) were unavailable for review, leaving 244 patients in the present study cohort. The demographics, comorbidities, physical examination findings, and laboratory variables are presented in Table 1 . The two cohorts were similar in most respects; however, some differences were identified. In our study population, the proportion of nursing home patients was nearly fourfold higher than in the study by Fine and colleagues7 (39% vs 10%, respectively). Thirty-one percent of our cohort, compared with 12% in the study by Fine and colleagues,7 had an altered mental status at presentation. Fewer of our patients had a systolic BP of < 90 mm Hg at presentation (1% vs 11%). Our patients also showed a trend toward more pleural effusions (22% vs 9%). Our patients tended to be sicker, with a lower proportion in risk classes I to III and a higher proportion in risk classes IV and V (Table 2 ).


View this table:
[in this window]
[in a new window]

  		Table 1.. Demographic and Clinical Characteristics*

 

View this table:
[in this window]
[in a new window]

  		Table 2.. Risk Class-Specific Mortality for CAP*

 
Mortality increased directly with risk class designation. The class-specific mortality was similar compared with the study of Fine and colleagues,7 with the exception of class III, where we found a 9.7% mortality rate (3 of 32 patients), compared with 2.7% in the study by Fine and colleagues (Table 2) . Due to the relatively small number of patients in this stratum, the significance of this is unclear. Two of these three patients had advanced dementia and died of their pneumonia. The third patient was a 59-year-old man who suffered a cardiac arrest 12 days after admission. Overall mortality was similar in the two studies (12.7% compared with 10.3%).

Seventy-one of the 244 (29%) patients in our cohort were classified as low risk (class I to III) according to the PSI (Table 2) . A review of these patients’ charts showed clear justification for admission in 67 of 71 cases (94%; Table 3 ). Thirty-five patients had another medical problem, apart from CAP, requiring inpatient care (12 had exacerbations of COPD, 4 had asthma exacerbations, 4 had possible acute coronary syndromes, 4 had congestive heart failure, 4 had painful sickle cell crises, and 1 each with ethanol withdrawal seizures, idiopathic seizures, upper GI bleeding, diabetic ketoacidosis, complex congenital heart disease, bronchiectasis, and possible line sepsis–the latter 3 were admitted for initial parenteral antimicrobial therapy). Eighteen patients had psychosocial issues (homelessness, substance abuse, inadequate home supports) that would preclude outpatient therapy, 16 had failed outpatient therapy for CAP, 13 were hypoxemic (as defined by a pulse oximetry of < 90% or a partial pressure of arterial oxygen of < 60 mm Hg), and 2 had inadequate oral intake. Some patients had more than one of these conditions. The remaining four low-risk patients who were admitted accounted for 1.8% of the total hospitalization days. No patients in risk classes I or II died. The three class III patients who died had all been admitted for psychosocial reasons (inadequate home supports).


View this table:
[in this window]
[in a new window]

  		Table 3.. Admission of Low-Risk Patients

 

   Discussion
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
Conclusion
 References
 
We applied the PSI to a group of patients admitted to our institution with CAP to assess its validity regarding mortality prediction and possible utility in admission decision making. The PSI predicted mortality very well and appears to be able to stratify low-risk (class I to III) from high-risk (class IV to V) subgroups. Our mortality rates were remarkably similar to those predicted by the PSI despite our relatively small sample size. The marked difference in mortality in the class III patients most likely represents a type I error due to our small sample size, but differences between the populations could also play a role. Our patients were older, more likely to reside in a nursing home, and tended to be in a higher risk class. It may be that the PSI, developed in younger, less sick patients, may not be directly applicable to a population with different characteristics, despite the approach of risk class stratification.

Since the study population consisted of those already admitted for treatment, we were unable to evaluate whether some high-risk patients were discharged home. However, even without the assistance of a prediction rule, the medical decision on patients who were admitted was fairly good. Although 29% of our patients were predicted to be at low risk of death from the CAP, as determined by risk class designations of I to III, these patients almost always had clear justification for admission. Other active medical problems requiring inpatient care, failure of outpatient antibiotic therapy, and social circumstances that could jeopardize patient safety were most common.

The social circumstances considered were homelessness (3 cases), substance abuse or other psychologic problems (4 cases), and inadequate home supports (11 cases). Of those who were either homeless or had substance abuse problems, all were < 65 years old and five of seven were men. Of those with inadequate home supports, 10 of 11 were > 65 years old and 7 of 11 were women. Of the 71 low-risk patients, the reason for admission was not apparent in only 4 patients, which accounted for < 2% of the total hospital days. It does not appear that the CP rule would have significantly changed admission decision making in our cohort. Although a high risk of death is a good reason for admission in patients with CAP, we believe there are other cogent considerations for hospitalized care for CAP, some of which are outlined above.

CP rules are designed to provide clinicians with a probability of a disease or outcome in a particular patient situation and thereby assist in clinical decision making. Fine and colleagues7 have developed a well-designed and apparently valid CP rule for mortality prediction in CAP. Its utilization can help in establishing standards of care for CAP and act as a useful educational tool for physicians in training. Some methodologic issues are present but are relatively minor. The collection of data was primarily retrospective. The development of the CP rule was carried out exclusively on patients already admitted to hospital. In the validation process, they prospectively assessed > 2,000 patients who were subsequently treated on an ambulatory or inpatient basis. In addition, they used > 38,000 already hospitalized patients. This methodology may affect applicability to the initial assessment of patients with CAP. Another concern is that no mention is made of blinding to mortality at the time of data collection. The interobserver reliability for history and physical examination variables that are incorporated into the PSI was not discussed, and the operating characteristics of the PSI are not known. The PSI calculation will remain somewhat cumbersome until fully electronic charting becomes standard, which could allow automatic calculation during patient assessment.

A significant limitation of our study is its retrospective nature. We are unable to assess the potential effect of applying this CP rule as the main admission criterion to all patients presenting with CAP. Although it seems that such action may have impacted negatively on many of the low-risk patients in our study, it is unclear how many such patients were assessed and discharged successfully from our ED. Atlas and colleagues13 attempted to prospectively assess the PSI as an aid in deciding which CAP patients can be safely treated without admission. For patients in the low-risk classes (classes I to III), the predicted mortality risk was provided to the ED physicians; for those who were not admitted, a program of visiting nurse services, a free antibiotic, and access to a primary-care physician were offered. Compared with historical controls, the proportion of these patients discharged from the ED increased from 42 to 57%, but 9% (8 of 94) of those discharged were subsequently admitted. Clinical outcomes were no different from those of historical controls, but patients initially treated at home were less satisfied with the initial treatment location than comparable control subjects. Of 826 patients who met screening criteria in the study by Atlas and colleagues,13 70% were excluded for one or a combination of the following reasons: hypoxia, age > 84 years, poor oral intake, recent hospitalization, nursing home residence, homelessness, psychosocial problems believed to potentially compromise treatment adherence, or long-term oxygen therapy. As noted above, we found that many patients with a low predicted mortality risk were likely admitted for these reasons. A further 10% were excluded due to a high PSI risk class designation, leaving 166 potential candidates for outpatient therapy. Overall, about 2% of all admissions for CAP were avoided in that study. Other studies that have studied the PSI include Flanders et al11 and Gonzalez-Moraleja et al.10 These studies and our results, in general, support the use of the PSI in estimating mortality and assisting in clinical decision making.


   Conclusion
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Conclusion
References
 
The PSI appears to be an excellent predictor of mortality, but its use as the principal variable in making the admission decision could compromise patient care, especially in the socioeconomically disadvantaged. It has been shown that uninsured persons have greater difficulty accessing inpatient care,14 fewer procedures, and shorter hospitalizations,15 and are more likely to receive substandard care for medical injury.16 Other authors have found that in Canada, despite a socialized medical system, the poor still have difficulty accessing outpatient medical care.17 We must bear in mind that Fine and colleagues7 state that low-risk patients should be considered as candidates for outpatient treatment, but the available evidence from our study and Atlas and colleagues13 would suggest that a substantial number of patients in the low-risk group would still require hospitalization. Perhaps its current application should be focused on the identification of high-risk patients to assist in their rapid triage and aggressive management, and the education of physicians in training. The role of the PSI in making the admission decision might best be as a component in the algorithm, along with psychosocial and economic considerations, all of which need careful assessment in this process.


   Acknowledgements
 
We wish to thank Drs. Moira Kapral and Howard Leong-Poi, and all the attending staff in the Clinical Teaching Units and housestaff at the Toronto General and Western Hospitals during the academic year 1995 to 1996 for making this project possible.


   Footnotes
 
Abbreviations: CAP = community-acquired pneumonia; CP = clinical prediction; ED = emergency department; PSI = pneumonia severity index

Received for publication October 13, 1999. Accepted for publication March 3, 2000.


   References
TOP
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Conclusion
 References
 

  1. Wasson, JH, Sox, HC, Neff, RK, et al (1985) Clinical prediction rules: applications and methodologic standards. N Engl J Med 313,793-799[Abstract]
  2. Laupacis, A, Sekar, N, Stiell, IG (1997) Clinical prediction rules: a review and suggested modifications of methodologic standards. JAMA 277,488-494[Abstract]
  3. Garibaldi, RA (1985) Epidemiology of community-acquired respiratory tract infections in adults: incidence, etiology, and impact. Am J Med 78,32-37[CrossRef][ISI][Medline]
  4. Niederman, MS, McCombs, JS, Unger, AN, et al (1998) The cost of treating community-acquired pneumonia [abstract]. Am J Respir Crit Care Med 157,A292
  5. La Force, FM (1985) Community-acquired lower respiratory tract: prevention and cost-control strategies. Am J Med 78,52-57[CrossRef][ISI][Medline]
  6. Chow, CW, Lee-Pack, LR, Senathiragah, N, et al (1995) Community acquired, nursing home acquired and hospital acquired pneumonia: a 5 year review of the clinical, bacteriological and radiological characteristics. Can J Infect Dis 6,317-325
  7. Fine, MJ, Auble, TE, Yealy, DM, et al (1997) A prediction rule to identify low-risk patients with community-acquired pneumonia. N Engl J Med 336,243-250[Abstract/Free Full Text]
  8. . British Thoracic Society. (1987) Community-acquired pneumonia in adults in British hospitals in 1982–3: a survey of etiology; mortality, prognostic factors and outcome. Q J Med 239,195-220
  9. Farr, BM, Sloman, AJ, Fisch, MJ (1991) Predicting death in patients hospitalized for community-acquired pneumonia. Ann Intern Med 115,428-436
  10. Gonzalez-Moraleja, J, Sesma, P, Gonzalez, C, et al (1999) What is the cost of inappropriate admissions of pneumonia patients? Arch Bronconeumol 35,312-316[Medline]
  11. Flanders, WD, Tucker, G, Krishnadasan, A, et al (1999) Validation of the pneumonia severity index: importance of study-specific recalibration. J Gen Intern Med 14,333-340[CrossRef][ISI][Medline]
  12. Marras, TK, Chan, CK (1998) Use of guidelines in treating community-acquired pneumonia. Chest 113,1689-1694[Abstract/Free Full Text]
  13. Atlas, SJ, Benzer, TI, Borowsky, LH, et al (1998) Safely increasing the proportion of patients with community-acquired pneumonia treated as outpatients: an interventional trial. Arch Intern Med 158,1350-1356[Abstract/Free Full Text]
  14. Bindman, AB, Grumbach, K, Osmond, D, et al (1995) Preventable hospitalizations and access to health care. JAMA 274,305-311[Abstract]
  15. Weissman, J, Epstein, AM (1989) Case mix and resource utilization by uninsured hospital patients in the Boston metropolitan area. JAMA 261,3572-3576[Abstract]
  16. Burstin, HR, Lipsitz, SR, Brennan, TA (1992) Socioeconomic status and risk for substandard medical care. JAMA 268,2383-2387[Abstract]
  17. Williamson, DL, Fast, JE (1998) Poverty and medical treatment: when public policy compromises accessibility. Can J Public Health 89,120-124[ISI][Medline]



This article has been cited by other articles:


Home page
 Eur Respir JHome page
A. Capelastegui, P. P. Espana, J. M. Quintana, I. Areitio, I. Gorordo, M. Egurrola, and A. Bilbao
Validation of a predictive rule for the management of community-acquired pneumonia
Eur. Respir. J., January 1, 2006; 27(1): 151 - 157.
[Abstract] [Full Text] [PDF]

Home page
 Eur Respir JHome page
M. Woodhead, F. Blasi, S. Ewig, G. Huchon, M. Leven, A. Ortqvist, T. Schaberg, A. Torres, G. van der Heijden, and T. J. M. Verheij
Guidelines for the management of adult lower respiratory tract infections
Eur. Respir. J., December 1, 2005; 26(6): 1138 - 1180.
[Abstract] [Full Text] [PDF]

Home page
 ANN INTERN MEDHome page
J. Carratala, N. Fernandez-Sabe, L. Ortega, X. Castellsague, B. Roson, J. Dorca, A. Fernandez-Aguera, R. Verdaguer, J. Martinez, F. Manresa, et al.
Outpatient Care Compared with Hospitalization for Community-Acquired Pneumonia: A Randomized Trial in Low-Risk Patients
Ann Intern Med, February 1, 2005; 142(3): 165 - 172.
[Abstract] [Full Text] [PDF]

Home page
 ChestHome page
C. H. Goss, G. D. Rubenfeld, D. R. Park, V. L. Sherbin, M. S. Goodman, and R. K. Root
Cost and Incidence of Social Comorbidities in Low-Risk Patients With Community-Acquired Pneumonia Admitted to a Public Hospital
Chest, December 1, 2003; 124(6): 2148 - 2155.
[Abstract] [Full Text] [PDF]

Home page
 ChestHome page
F. W. Arnold, J. A. Ramirez, L. C. McDonald, and E. L. Xia
Hospitalization for Community-Acquired Pneumonia: The Pneumonia Severity Index vs Clinical Judgment
Chest, July 1, 2003; 124(1): 121 - 124.
[Abstract] [Full Text] [PDF]

Home page
 ThoraxHome page
M Woodhead
Assessment of illness severity in community acquired pneumonia: a useful new prediction tool?
Thorax, May 1, 2003; 58(5): 371 - 372.
[Full Text] [PDF]

Home page
 Eur Respir JHome page
P.P. Espana, A. Capelastegui, J.M. Quintana, A. Soto, I. Gorordo, M. Garcia-Urbaneja, and A. Bilbao
A prediction rule to identify allocation of inpatient care in community-acquired pneumonia
Eur. Respir. J., April 1, 2003; 21(4): 695 - 701.
[Abstract] [Full Text] [PDF]

Home page
 Eur Respir JHome page
M. Woodhead
Community-acquired pneumonia guidelines: much guidance, but not much evidence
Eur. Respir. J., July 1, 2002; 20(1): 1 - 3.
[Full Text] [PDF]

Home page
 Eur Respir JHome page
A. Ortqvist
Treatment of community-acquired lower respiratory tract infections in adults
Eur. Respir. J., July 1, 2002; 20(36_suppl): 40S - 53s.
[Abstract] [Full Text] [PDF]

Home page
 JWatch Emergency Med.Home page
Clinical Prediction Rules for Pneumonia Don't Consider the Entire Picture
Journal Watch Emergency Medicine, January 4, 2001; 2001(104): 3 - 3.
[Full Text]

eLetters:

Read all eLetters

Clear justification for admission?
Eddy Lang
Chest Online, 11 Dec 2000 [Full text]
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF) Free
Right arrow Submit a response
Right arrow View responses
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Article Archive
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (36)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Marras, T. K.
Right arrow Articles by Chan, C. K.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Marras, T. K.
Right arrow Articles by Chan, C. K.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS