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

doi:10.1378/chest.118.5.1339

Chest

Volume 118, Issue 5, November 2000, Pages 1339-1343

https://doi.org/10.1378/chest.118.5.1339 Get rights and content

Study objectives

To study the validity of a recentlydeveloped community-acquired pneumonia (CAP) severity prediction rulein estimating mortality, to determine its utility in decision makingregarding hospitalization, and to assess factors influencing thisdecision.

Design

Retrospective chart review.

Setting

Two sites of the University Health Network, theToronto General and Toronto Western Hospitals, tertiary-care teachinginstitutions with a sizable primary-care and secondary-care source ofreferrals, and a total of 900 beds.

Patients

Consecutive patients with CAP admitted between February and June1996.

Measurements and results

A single trainedmedical records extractor assembled data to compare our population tothat used in developing the CAP prediction rule, in terms of mortalityand to assess reasons for hospitalization. Two hundred fifty-fiveeligible patients were admitted, and 244 charts (96%) were available.Our patients tended to be older, with nearly four times as manyresidents of chronic care institutions (39% compared with 10%), andhad a higher risk class distribution than the published cohort. Riskclass-specific mortality was similar in four of five classes. Of the 71patients in the low-risk classes, 67 had additional reasons foradmission; 18 of which were psychosocial (homelessness, substanceabuse, or inadequate home supports).

Conclusions

TheCAP severity prediction rule estimates mortality well. Admission oflow-risk patients was linked to psychosocial and other medical reasonsnot captured by this rule. The rule can be very useful in assessing theneed for hospitalization; however, there remains a significantpercentage of patients with a low severity score who may requirehospitalization for psychosocial and economicconsiderations.

Section snippets

Materials and Methods

The study involved the Toronto General and Toronto WesternHospitals of the University Health Network (both are urban downtowntertiary-care teaching hospitals affiliated with the University ofToronto). The study population was comprised of a consecutive cohort ofadults with CAP admitted to the general medical services of thisinstitution over a 5-month period from February 1, 1996, to June 30,1996. Disease definitions and complete inclusion and exclusion criteriahave been published previously.¹²

Results

Two hundred fifty-five eligible patients were admitted to thehospitals over the 5-month study period. Eleven charts (4%) wereunavailable for review, leaving 244 patients in the present studycohort. The demographics, comorbidities, physical examination findings,and laboratory variables are presented in Table 1. The two cohorts were similar in most respects; however, somedifferences were identified. In our study population, the proportion ofnursing home patients was nearly fourfold higher than in

Discussion

We applied the PSI to a group of patients admitted to ourinstitution with CAP to assess its validity regarding mortalityprediction and possible utility in admission decision making. The PSIpredicted mortality very well and appears to be able to stratifylow-risk (class I to III) from high-risk (class IV to V) subgroups. Ourmortality rates were remarkably similar to those predicted by the PSIdespite our relatively small sample size. The marked difference inmortality in the class III patients most

Conclusion

The PSI appears to be an excellent predictor of mortality, but itsuse as the principal variable in making the admission decision couldcompromise patient care, especially in the socioeconomicallydisadvantaged. It has been shown that uninsured persons have greaterdifficulty accessing inpatient care,¹⁴ fewer procedures,and shorter hospitalizations,¹⁵ and are more likely toreceive substandard care for medical injury.¹⁶ Otherauthors have found that in Canada, despite a socialized medical system,the

ACKNOWLEDGMENT

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

References (17)

RA Garibaldi
Epidemiology of community-acquired respiratory tract infections in adults: incidence, etiology, and impact
Am J Med
(1985)
FM La Force
Community-acquired lower respiratory tract: prevention and cost-control strategies
Am J Med
(1985)
J Gonzalez-Moraleja et al.
What is the cost of inappropriate admissions of pneumonia patients?
Arch Bronconeumol
(1999)
TK Marras et al.
Use of guidelines in treating community-acquired pneumonia
Chest
(1998)
JH Wasson et al.
Clinical prediction rules: applications and methodologic standards
N Engl J Med
(1985)
A Laupacis et al.
Clinical prediction rules: a review and suggested modifications of methodologic standards
JAMA
(1997)
MS Niederman et al.
The cost of treating community-acquired pneumonia [abstract]
Am J Respir Crit Care Med
(1998)
CW Chow et al.
Community acquired, nursing home acquired and hospital acquired pneumonia: a 5 year review of the clinical, bacteriological and radiological characteristics
Can J Infect Dis
(1995)

There are more references available in the full text version of this article.

Cited by (86)

Data-driven approaches to generating knowledge: Machine learning, artificial intelligence, and predictive modeling
2023, Clinical Decision Support and beyond: Progress and Opportunities in Knowledge-Enhanced Health and Healthcare
In this chapter, we will review the methodologies of the most commonly used diagnostic and prognostic models in the medical domain and discuss specific strengths and weaknesses of alternative modeling methods. Popular examples of some modeling methods will be discussed, with a focus is on models and modeling methods that have been utilized in practice. We conclude with a discussion on current directions for the field.
Generation of Knowledge for Clinical Decision Support. Statistical and Machine Learning Techniques.
2014, Clinical Decision Support: The Road to Broad Adoption: Second Edition
This chapter begins with historical backgrounds of knowledge generation for clinical decision support systems. It then reviews the methodologies of the most commonly used diagnostic and prognostic models in the medical domain, and discusses specific strengths and weaknesses of alternative modeling methods. Popular examples of some modeling methods are discussed; since the focus is on models that have been utilized in practice, the discussion concentrates on logistic regression models, classification trees, and artificial neural networks. It concludes with a discussion on current directions for the field.
Pneumococcal pneumonia-are the new severity scores more accurate in predictingadverse outcomes?
2013, Revista Portuguesa de Pneumologia
A decisão do local de tratamento é um dos fatores mais importantes na abordagem de doentes com pneumonia adquirida na comunidade. Os scores de gravidade são ferramentas prognósticas validadas para previsão da mortalidade por pneumonia adquirida na comunidade e decisão do local de tratamento.
O objetivo deste trabalho foi comparar o poder discriminatório de 4 scores – os clássicos PSI e CURB-65 e os mais recentes SCAP e COP-SMART – na previsão de eventos adversos: morte, internamento em Unidades de Cuidados Intensivos (UCI), necessidade de ventilação mecânica invasiva ou suporte de aminas vasopressoras em doentes internados com pneumonia pneumocócica.
Foi efetuado um estudo retrospetivo de doentes internados por pneumonia pneumocócica num período de 5 anos.
Os doentes foram estratificados com base nos dados da admissão e foram-lhes atribuídas classes de risco para cada score: baixo, médio e alto risco. Os resultados foram obtidos comparando as classes de baixo risco versus médio-alto risco.
Foram estudados 142 episódios de internamento onde se observaram 2 mortes e 10 doentes necessitaram de ventilação mecânica e suporte de aminas. A maioria dos doentes foram classificados como baixo risco por todos os scores – encontrámos altos valores preditivos negativos para todos os eventos adversos estudados sendo o mais elevado correspondente ao SCAP. Os scores mais recentes mostraram uma maior precisão para prever internamento na UCI e necessidade de ventilação ou suporte por aminas (principalmente para o score SCAP com valores mais elevados da AUC para todos os eventos adversos).
A taxa de todos os eventos adversos aumentou com o agravamento da classe de risco em todas os scores. Os scores de gravidade mais recentes parecem ter um maior poder discriminatório para todos os eventos adversos no nosso estudo, em particular, o SCAP.
The site-of-care decision is one of the most important factors in the management of patients with community-acquired pneumonia. The severity scores are validated prognostic tools for community-acquired pneumonia mortality and treatment site decision.
The aim of this paper was to compare the discriminatory power of four scores – the classic PSI and CURB65 ant the most recent SCAP and SMART-COP – in predicting major adverse events: death, ICU admission, need for invasive mechanical ventilation or vasopressor support in patients admitted with pneumococcal pneumonia.
A five year retrospective study of patients admitted for pneumococcal pneumonia.
Patients were stratified based on admission data and assigned to low-, intermediate-, and high-risk classes for each score. Results were obtained comparing low versus non-low risk classes.
We studied 142 episodes of hospitalization with 2 deaths and 10 patients needing mechanical ventilation and vasopressor support. The majority of patients were classified as low risk by all scores – we found high negative predictive values for all adverse events studied, the most negative value corresponding to the SCAP score. The more recent scores showed better accuracy for predicting ICU admission and need for ventilation or vasopressor support (mostly for the SCAP score with higher AUC values for all adverse events).
The rate of all adverse outcomes increased directly with increasing risk class in all scores. The new gravity scores appear to have a higher discriminatory power in all adverse events in our study, particularly, the SCAP score.
The routine use of the Pneumonia Severity Index in the emergency department: Effect on process-of-care indicators and results in community acquired pneumonia
2013, Enfermedades Infecciosas y Microbiologia Clinica
Evaluación de indicadores de proceso (hospitalización inadecuada, adecuación y precocidad de antibioterapia) y resultado (estancia hospitalaria, reingresos, ingresos UCI y mortalidad) de neumonía adquirida en la comunidad (NAC) al aplicar la guía SEPAR/IDSA.
Estudio observacional retrospectivo de pacientes consecutivos con diagnóstico al alta de NAC en los primeros semestres de 2007 y 2008 (186 y 161 pacientes, respectivamente) atendidos en urgencias de hospital general. Se han analizado las diferencias en los indicadores entre los grupos de pacientes con/sin Pneumonia Severity Index (PSI) calculado, según el año de implantación de la guía, y se compararon con los de 110 pacientes de 2006 según la guía SEQ/ATS.
La guía SEPAR/IDSA ha mejorado los indicadores de proceso: mayor adecuación del ámbito de tratamiento, disminución de ingresos injustificados (del 39,4% en 2006 al 8,5% en 2007 [p < 0,001] y al 17,2% en 2008 [p = 0,005]) y mayor precocidad de antibioterapia. No se han modificado los indicadores de resultado. En 2008 se observó una reducción de la mortalidad en el subgrupo de pacientes con PSI IV-V en los que se calculó el PSI (2,3%), respecto a los que no se calculó (28,3%; p < 0,001), y esta tasa fue inferior a la de los pacientes con criterios de hospitalización según la guía SEQ/ATS (22,7%; p = 0,003).
La guía SEPAR/IDSA ha reducido los ingresos injustificados y en el segundo año de aplicación se ha observado mayor precocidad de la antibioterapia junto a una reducción de la mortalidad en los pacientes con riesgo moderado-alto en los que se calculó el PSI.
To evaluate process-of-care indicators (inappropriate hospitalisation, suitability and early antibiotic treatment) and outcome indicators (length of hospital stay, hospital readmission, ICU admission, and mortality) in the management of community-acquired pneumonia (CAP) when the SEPAR/IDSA guidelines were applied.
An observational retrospective study conducted on patients diagnosed with CAP during the first semester of 2007 and 2008 (186 and 161 patients, respectively) in the emergency unit of a general hospital. Differences in the process-of-care and outcome indicators between 2007 and 2008 (with and without the Pneumonia Severity Index [PSI]) were evaluated. Moreover, the indicators were compared with those obtained in 2006 (110 patients), when the current guidelines were those of SEQ/ATS.
The SEPAR/IDSA guidelines improved the following process-of-care indicators: appropriateness of treatment, unjustified hospital readmission (39.4% in 2006 vs. 8.5% in 2007 [P < .001], and 17,2% in 2008 [P = .005]), and early treatment. However, outcome indicators did not change. In 2008, a decrease in the mortality of the patients of risk classes IV-V in which the PSI had been estimated was observed in comparison with the patients in which the PSI was not estimated (2.3% vs. 28.3%; P < .001). Moreover, the mortality rate of the patients of risk classes IV-V in which the PSI had been estimated was lower than those measured using the SEQ/ATS guidelines (22.7%; P = .003).
SEPAR/IDSA guidelines decreased the unjustified hospital readmission. In the second year of its application an increase in the number of patients who received early treatment, and a decrease of the mortality rate of the patients of risk classes IV-V in which the PSI had been estimated, were also observed.
The effects of incorporating a pneumonia severity index into the admission protocol for community-acquired pneumonia
2012, Journal of Emergency Medicine
Background: Community-acquired pneumonia (CAP) is a common reason for admissions in the emergency department (ED). However, patient disposition is not always standardized. Objective: To evaluate the effect of incorporating a pneumonia severity index (PSI) on admission rates and medical costs in CAP patients presenting to the ED. Methods: From April 2008 to March 2009, CAP patients presenting to the ED were prospectively screened and low-risk CAP patients (PSI I, II, or III) were enrolled (after group). Discharge and outpatient care were recommended for this group in the absence of other medical conditions requiring hospitalization. Data from low-risk CAP patients from May 2003 to October 2006 were also collected for comparative analysis (before group). Results: There were 365 and 174 patients in the before and after groups, respectively. The admission rate of the after group was significantly lower than that of the before group (30.4% vs. 68.2%, p < 0.01). The subsequent admission rates after ED discharge due to CAP were similar (3.2% vs. 7.7%, p = 0.10). The ultimate admission rate in the after group was significantly lower than that in the before group (32.5% vs. 70.7%, p < 0.01). Direct medical costs per patient for the before and after groups were $US 1532 and $US 1186, respectively (p = 0.03). Conclusions: Incorporation of the PSI into the admission protocol for ED patients with CAP significantly reduced the admission rates and medical costs.
Albumin and C-reactive protein have prognostic significance in patients with community-acquired pneumonia
2011, Journal of Critical Care
Citation Excerpt :
In addition, the PSI can stratify the risk of short-term mortality in patients with CAP. In EDs, patients with CAP can be correctly stratified using PSI score, and this reduces admission rates and medical costs significantly [24-27]. Accurate prognostication is especially important in critically ill patients, and many biochemical markers could reflect the severity of diseases.
This study aims to determine the association of commonly used biochemical markers, such as albumin and C-reactive protein (CRP), with mortality and the prognostic performance of these markers combined with the pneumonia severity index (PSI) for mortality and adverse outcomes in patients with community-acquired pneumonia (CAP).
The data were gathered prospectively for patients hospitalized with CAP via the emergency department. Laboratory values, including CRP and albumin, clinical variables, and the PSI were measured. Primary outcomes were 28-day mortality and survival times. Secondary outcome was admission to the intensive care unit, vasopressor use, or the need for mechanical ventilation during the hospital stay.
A total of 424 patients were included. The 28-day mortality was 13.7%. C-reactive protein and albumin were significantly different between survivors and nonsurvivors. In logistic regression analysis, CRP and albumin were independently associated with 28-day mortality (P < .05). Receiver operating characteristic curves showed improved mortality prediction by adding CRP or albumin to the PSI scale. The Cox proportional hazards analysis showed that high serum albumin (≥3.3 mg/dL) had a hazard ratio of 0.5 (95% confidence interval, 0.3-0.9), and high CRP (≥14.3 mg/dL) had a hazard ratio of 2.0 (95% confidence interval, 1.1-3.4). For predicting secondary outcome, adding albumin to PSI increased areas under the curve significantly, but CRP did not.
Albumin and CRP were associated with 28-day mortality in hospitalized patients with CAP, and these markers increased prognostic performance when combined with the PSI scale.

View all citing articles on Scopus

View full text

Chest

Clinical InvestigationsPneumoniaApplying a Prediction Rule To Identify Low-Risk Patients With Community-Acquired Pneumonia

Study objectives

Design

Setting

Patients

Measurements and results

Conclusions

Section snippets

Materials and Methods

Results

Discussion

Conclusion

ACKNOWLEDGMENT

Am J Med

Am J Med

Arch Bronconeumol

Chest

Clinical prediction rules: applications and methodologic standards

N Engl J Med

Clinical prediction rules: a review and suggested modifications of methodologic standards

JAMA

The cost of treating community-acquired pneumonia [abstract]

Am J Respir Crit Care Med

Community acquired, nursing home acquired and hospital acquired pneumonia: a 5 year review of the clinical, bacteriological and radiological characteristics

Can J Infect Dis

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