Pneumonia: Admit or Discharge?
March 1, 1997
Pneumonia: Admit or Discharge?
Abstract & Commentary
Source: Fine MJ, et al. A prediction rule to identify low-risk patients with community-acquired pneumonia. N Engl J Med 1997;336:243-250.
The clearly defined purpose of a recent study by Fine et al was to develop a prediction rule for prognosis that would accurately identify patients with community-acquired pneumonia (CAP) at low risk of death within 30 days of presentation and to assess the accuracy of this rule for clinically significant morbidity (i.e., subsequent readmission to the ICU). First, the prediction rule was generated from over 14,000 adult inpatients with CAP. Patients with HIV and recent hospitalization were excluded from the cohort. The rule was generated in two steps. Step 1 looked only at demographic variables (patient’s age, sex, and nursing home residence), coexisting illnesses (neoplasm, CHF, coronary artery disease, CVA, renal and liver disease), and immediate physical exam findings (vital signs and mental status), data readily available to all clinicians. Non-nursing home residents less than 50 years of age with normal vital signs and mental status and without coexisting illnesses as defined above were assigned Class I status. These patients had a 30-day mortality rate of less than 0.5%.
Step 2 consisted of the analysis of the 14 variables in Step 1 plus seven additional lab and radiographic findings. Abnormal findings were defined as follows: BUN greater than 30 mg/dL, sodium less than 130 mmol/L, glucose greater than 250 mg/dL, hematocrit less than 30%, PO2 less than 60 mmHg or saturation less than 90%, and presence of a pleural effusion. Points were assigned for each category, and the total score placed the patients in risk Classes II through V, representing increasing mortality, with a Class V predicted probability of death of 10% at 30 days.
In the second phase of the study, the authors sought to validate the prediction rule. This was done twice: retrospectively on 38,000 patients hospitalized with CAP, and again, prospectively, on 2300 patients. Based on the data from the prospective cohort, for the 1600 patients assigned to risk Classes I through III, seven patients died. Four of these deaths were due to pneumonia and none were deemed preventable. Among outpatients, the subsequent rate of hospitalization at 30 days ranged from 5% in Class I to 20% in Class IV. None in the Class I through III groups died, and one required an ICU stay. Of the eight outpatients in Class IV or V who were subsequently hospitalized, three died and one was admitted to the ICU.
The authors conclude that they have developed a prediction rule that identified Class I through III patients at minimal risk for death and other significant outcomes from CAP. They suggest that Class I and II patients may be managed at home with oral antibiotics. Class III patients may require a 24-hour observational admission followed by home care, and Class IV and V patients should receive traditional inpatient management. They do hasten to add that their prediction rule does not account for psychosocial factors that would limit home care, nor rarer medical conditions not considered in the Step 1 analysis.
COMMENT BY KATHERINE L. HEILPERN, MD
This is an ambitious and important study, even if the results are not yet "generalized" to clinical practice. Both the study authors and the author of the accompanying editorial caution that the results require randomized, prospective validation.1 And yet, I feel comfortable with one of the conclusions even if it is preliminary.
If one examines only the Class I patients, they truly represent a low-risk group. They are less than 50 years of age, without coronary, cerebrovascular, renal, or hepatic disease. None of these patients, by definition, had a respiratory rate of more than 30 breaths/min, a temperature lower than 35°C or higher than 40°C, a pulse rate greater than 125 beats/min, or a systolic blood pressure lower than 90 mmHg. Our gestalt, in the medical lexicon, is that they are "not sick." I feel comfortable treating these patients with oral antibiotics and close follow-up. Obviously, this should not include the alcoholic, the crack addict, the HIV-positive patient, or the homeless. Prediction rules give us comfort with the statistical odds but should never supplant judgment.
Pneumonia occurs in four million U.S. adults annually and is the sixth leading cause of death in the United States, so this is clearly important.2 Although we are under pressure from insurance companies and managed care organizations, I urge caution before we draw any firm conclusions beyond the limited Class I group (we have seen the devastating effect on families and neonates as we struggle with "boomerang babies").
I will make one predictionwe will hear a lot more about this study in the coming months. This is a study to read and remember as we struggle to keep a lid on health care costs and do the right thing, with minimal risk, for our patients.
References
1. Farr BM. Prognosis and decisions in pneumonia. N Engl J Med 1997;336:288-289. Editorial.
2. Centers for Disease Control and Prevention. Pneumonia and influenza death rates, United States 1979-1994. MMWR Morb Mortal Wkly Rep 1995;44:535-537.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content