Pneumonia Alphabet Soup


Pneumonia. It’s one of the first conditions we learn to diagnose as medical students. It was probably the cause of the first really sick, septic geriatric patient you saw in residency. Conversely you have also probably sent a fair share of patient’s home with an outpatient course of antibiotics and PCP follow-up.  While determining the appropriate treatment and disposition for patients on the extreme ends of illness severity is quite straight forward; that pesky majority in the middle can be a conundrum at times. Who can go home? Who needs broad spectrum? Who needs step-down? Over the last two decades there has been a smorgasbord of pneumonia related acronyms used in clinical practice to predict severity, guide therapeutics and recommend disposition. During our most recent resident Journal Club, we took a look at a handful of the more familiar acronyms as well as some new ones coming down the pipeline.

Williams J, Greenslade J, Chu K, Brown A and Lipman J. Utility of community-acquired pneumonia severity scores in guiding disposition from the emergency department: Intensive care or short-stay unit? Emergency Medicine Australasia 2018. 30, 538-546


PCP Pneumonia3.jpg

Community-Acquired pneumonia (CAP) is a common and deadly condition seen in Emergency Departments around the world. After the diagnosis is made, the disposition decision of the patient to the appropriate location and level of care can be challenging. On one end, delayed admission to the ICU for severe infections is associated with worse outcomes. On the other hand, unnecessary  admission to the hospital for low severity patients is over 20x more costly compared to outpatient therapy and exposes patients to additional sources of morbidity. A multitude of CAP severity scores have developed over the past decades to help identify the patients that may require higher levels of care. Additionally, the utility of low severity scores have not been thoroughly studied as a predicted tool to identify patients that may be suitable for admission to ED short-stay/observation units. 


Single-center, observational study in an urban, academic, high-volume, tertiary-care center in Australia. Data was gathered prospectively with calculated scores being performed retrospectively. 


  1. Prediction of ICU admission or 30-day mortality

  2. Utility of low scores for prediction of discharge within 48hrs, thus potentially identifying patients suitable for admission to a short-stay or observational unit. 


Participants were enrolled from a parent database of patients admitted to the hospital with presumed infection (of any kind) after initial presentation to the Emergency Department. Enrollment occurred during two non-contiguous time periods totaling 162 weeks.  From the initial database of 9,719 patients, 1,091 were identified to have community acquired pneumonia based on the criteria of demonstrating a lower respiratory tract source WITH a consolidation on CXR. From this pool, patients were excluded if they were hospitalized within the last 30 days, on immunosuppression, reside in a nursing home, or if they were ‘not for resuscitation’ (DNR equivalent in Australia) for a final n=618. 

Two endpoints were determined and defined. 1) Severe Pneumonia was defined as either ICU admission or 30-day mortality. There were not clearly stated requirements for ICU admission. 2) Mild Pneumonia was defined as discharge with 48 hours without death or readmission in 30-days. 

The prospectively gathered patient information from the parent database was then used to retrospectively calculate CAP severity scores on each patient for the following: Pneumonia Severity Index (PSI), CURB-65, CORB, CURXO, SMARTCOP and IDSA/ATS 2007 guidelines. 

Severe Pneumonia category thresholds were determined based on published standards for each severity scoring tool (eg CURB-65≥3). Likewise Mild Pneumonia category thresholds were also determined (eg CURB-65 = 0)

Diagnostic accuracy scores (sensitivity, specificity, predictive values, likelihood ratios and AUROCs) were then calculated using Severe Pneumonia thresholds to predict outcome 1 and Mild Pneumonia thresholds to predict outcome 2. 

PICO Model Summary 

Patients: 618 Australian adults admitted with to the hospital from the Emergency Department with pneumonia (lower respiratory tract infection and consolidation reported on CXR). Exclusion criteria included: recent hospitalization, immunosuppression, nursing home resident and DNR patients)

Intervention: Six previous validated CAP severity scoring tools. Pneumonia Severity Index (PSI), CURB-65, CORB, CURXO, SMARTCOP, and IDSA/ATS minor criteria. 

Comparison: No true control. The diagnostic accuracy statistics (sensitivity, specificity, predictive values, likelihood ratios and AUROCs) for each scoring tool were calculated. Scoring tools were not compared directly to each other. 

Outcomes: 1)Prediction of ICU admission or 30-day mortality 2)utility of score for prediction of discharge within 48hrs


Of the 618 admitted patients who met criteria for CAP, 75 (12.1%) were admitted to the ICU or deceased at 30-days (outcome 1) and 87(14.1%) were discharged within 48 hours (outcome 2). 

The Severe Pneumonia threshold for SMARTCOP ≥3 and CRUXO ≥10 showed optimal sensitivity to predict ICU admission or 30-day mortality. [85% (75-92) and 85% (75-92) respectively]. Meanwhile a CORB score 2-4 and CURB-65 score 3-5 demonstrated the highest specificity for ICU admission or 30-day mortality [93% (90-95) and 94% (92-96) respectively]. 

The Mild Pneumonia thresholds for all six CAP severity scores showed concerns for diagnostic accuracy. SMARTCOP of 0 and CURXO of 0 had the highest specificities, 87% (84-90) and 81% (78-85) respectively. 


  • The six study CAP severity scores had different strengths and weaknesses. 

  • Due to high sensitivity, SMARTCOP ≤ 3 and CURXO≤10 were useful to rule out severe CAP

  • Due to high specificity, CORB (2-4) and CURB-65 (3-5) were useful to identify patients at high risk of ICU requirements. 

  • None of the severity scores demonstrated utility in the use of low scores to predict discharge within 48hrs 


  • External validity - This is a single center study in a county with universal access to healthcare. Comparison to the baseline demographics of Brisbane shows on average a healthier, wealthier and more homogenous patient population compared to most academic Emergency Departments in the United States. Additionally their 30-day mortality rate was 2%, much lower than reported ranges in US based studies ranging from 7.7%-19.2%. 

  • Pneumonia Severity Index - One of the components in the calculation of this score is whether the patient is presenting from a nursing facility. All nursing facility residents were excluded from this study, thus the conclusions  drawn from the study regarding PSI should be taken with a grain of salt. 

  • Missing Data -  Multiple scoring systems require pH and ABG values which were missing from a large percentage of patients (49% and 64.5% respectively). This was addressed with the use of multiple imputations that showed no changes in in primary analyses, though the lack of required information for many scoring systems does raise concern for ability to adequately evaluate each systems predictive value. 

  • Unmeasured confounders - While a vast amount of objective data was gathered on enrolled patients, additional information known to impact a patient’s morbidity and mortality were not adequately discussed in describing baseline characteristics. This includes many social factors such as homelessness, poor follow-up, history of substance abuse, and baseline cognitive function. Admittedly, the majority of these are not assessed in CAP severity scores. However these factors play an important role in a physicians assessment of the patient’s underlaying risks and health. 


Overall this is an ambitious study that attempts to answer an important question facing EM providers across the world. As boarding, throughput, and crowding become more prevalent; there is a growing need for tools that can predict both the sickest patients that need more resources/higher level of care but also the patients that may be suitable for observation status admission. 

This essentially retrospective study was able to re-demonstrate that CURB-65 (which we are quite familiar with in the US) and CORB (common practice in Australia) are good tools to determine which of our patients we already plan to admit may need a higher level of care such as Step-Down or the ICU. Unfortunately, none of the six studied CAP severity scores were able to predict those patients that might be best served in our observations units, with none of them showing adequate diagnostic accuracy in predicting discharge within 48hours. This paper hopefully will serve as a launching point for future study to determine possible prediction tools that accurately identify patients suitable for management in observation units. 

Webb BJ, Dascomb K, Stenehjem E, et al. Derivation and Multicenter Validation of the Drug Resistance in Pneumonia Clinical Prediction Score. Antimicrob Agents Chemother 2016;60(5):2652–63. 


In 2016 new IDSA Guidelines were published establishing treatment recommendations for HAP and VAP and, perhaps more dramatically, abolishing the term HCAP without providing guidance on how to identify patients at risk for pneumonia due to drug-resistant pathogens.  Web et al worked to create a pneumonia scoring system to better identify patients at risk for drug resistent pathogens (DRPs) and ultimately improve antibiotic stewardship. The end product of this work was the DRIP (drug resistance in pneumonia) score. 

The Design

This score was created in 3 parts

  1. Literature review to determine risk factors for pneumonia due to DRPs. Risk factors assessed included prior antibiotic use, prior hospitalization, residence in a long-term care facility, chronic lung disease, immunosuppression, chronic kidney disease, infusion therapy, poor functional status, aspiration risk, diabetes, pneumonia severity, cerebrovascular disease, cognitive impairment, prior colonization with DRP or MRSA, gastric acid suppression, and presence of indwelling catheter. 

  2. These risk factors were assessed using logistic regression and area-under-the-curve analysis in the derivation cohort, a set of patients from a previous pneumonia study with biological diagnosis of pneumonia resulting in the clinical prediction score, DRIP score.

  3. The DRIP score was then validated in a prospective observational study in which the score was not used to affect clinical management.  The DRIP score was compared to HCAP criteria as well as 6 other systems to assess for specificity, sensitivity, and reduction of broad spectrum antibiotic use. 

Pneumonia Definition: The stringent definition used in this study may differ, for better or for worse, from how pneumonia is clinically diagnosed. 

  •    2+ clinical signs/symptoms: temperatures (<36 or > 38), tachypnea, hypoxia, ABG SpO2 < 60mmHg, sputum production, leukocytosis +/- bandemia

  • PLUS   Radiographic evidence of new parenchymal opacity or cavitation

The Cohorts

The Derivation Cohort: 213 patients with microbiological confirmation of pneumonia pathogen enrolled in multiple Utah hospitals from 2011-2012. The mean age of these patients was 63.1 years old with the primary comorbidity being chronic pulmonary disease. Of these patients 25% were determined to have DRPs. 

The Validation Cohort: 218 patients from 4 hospitals in different geographic regions of the United States with a mean age of 65.2 and the primary comorbidity being chronic pulmonary disease. DRPs were recovered in 33% of patients and 49.5% of these patients met criteria for HCAP. In general the validation cohort is sicker than the derivation cohort. 

The DRIP Score

  •  Major Risk factors (2 points): antibiotic use within 60 days, residence in long-term care facility, tube feeding, and prior infection with DRP within past year

  •  Minor Risk Factor (1 point): hospitalization within 60 days, chronic pulmonary disease, poor functional status, gastric acid suppression, wound care, MRSA colonization

  •  A score greater than or equal to 4 signifies high risk for DRP causing a pneumonia. 

How the DRIP Score Performs

In the Derivation Cohort - sensitivity 0.76, specificity 0.91, PPV 0.73, NPV 0.95

  • Antibiotic regimen suggested by DRIP score appropriate for organism recovered: 87%

In the Validation Cohort - sensitivity 0.82, specificity 0.81, PPV 0.68, NPV 0.90

  • The DRIP score is significantly more accurate compared to the HCAP score for identifying patient with DRPs: 81.5% vs 69.5% (p = 0.005)

  • The DRIP score would have decreased unnecessary antibiotic use by 46% based on microorganism recovered.

The Limitations

  • This score was derived from and validated using patients with a microbiologic diagnosis 

  • The score has not been validated in a larger study and no impact study has been done to assess effect on patient-centered outcomes

  • The DRIP score does not allow for specturm oof disease in COPD patients

  • The DRIP score was not used to affect discharge planning and patients who were discharged with DRP pneumonia were not captured in any of these data sets. 

The Take-Aways

  • The HCAP criteria are nonspecific and ultimately result in overuse of broad spectrum antibiotics for treatment of pneumonia. In the era of antibiotic stewardship a better score is needed to determine risk for DRPs. The DRIP score has been shown to be superior to the HCAP criteria in all statistical parameters and thus may offer guidance in lieu of the pending CAP guidelines. 

  • The DRIP score was able to identify more specific risk factors for DRPs than HCAP and significantly improve antibiotic stewardship

  • While the DRIP score is better able to identify at risk patients in the admitted population and better guide antibiotic use, it cannot be used to influence discharge planning or outpatient management of pneumonia

  • The population at highest risk for under-treatment are patients with severe COPD as the DRIP score is a binary scoring system and does not take into account severity of disease. 

Li, J. Ye, H., Zhao, L. B-type natriuretic peptide in predicting the severity of community-acquired pneumonia. World J Emerg Med 2015;6(2):131–6.


Previous research has indicated that B Natriuretic Peptide (BNP) levels are significantly increased in patients with severe sepsis and septic shock. It has been hypothesized that BNP might be used as a reliable indicator of sepsis-induced myocardial inhibition and accompanying proinflammatory cytokine state. A pilot study was performed prior to this study, in which BNP was found to accurately predict the severity of CAP, driving the hypothesis that it might hold promise as a biomarker.


The authors sought “to evaluate the role of BNP in risk stratification of CAP based on clinical observations, laboratory measurements, and chest X-ray examinations.”


This was an observational study of 202 adult patients with suspected CAP presenting to the ED of Fuxing Hospital between 2011 and 2012.  The study cohort included an equal distribution of males and females and an age range of 18 to 96 (mean 81.7 +/- 10.7). Patients were excluded from analysis if they had a history of heart failure, CAD, acute renal failure, ESRD, cirrhosis, hypertensive heart disease, pregnancy, pulmonary hypertension, pulmonary embolism, active tuberculosis, pulmonary fibrosis, nosocomial pneumonia (HCAP), immunosuppressive diseases, primary aldosteronism, and hyperthyroidism. This decision was made to limit the amount of confounders that might affect BNP. Subgroups of high-risk (PSI class IV and V) and low-risk (PSI I-III) were created, as well as subgroups on survival versus non-survival. 

CAP was defined as the presence of an infiltrate on chest radiograph in concert with recently acquired respiratory signs / symptoms (these included presence of fever and leukocytosis or leukopenia).

The enrolled patients had laboratory and hemodynamic parameters measured upon enrollment and were required to have an echocardiogram within 24 hours. The BNP was performed using a quantitative point-of-care laboratory test relying on chemiluminescence immunoassay.


A minority of the patient cohort (27.7%) fell into the low-risk PSI group and the remaining 72.3% were in the high-risk group.

There was a rise in BNP between every PSI group which consistently met the threshold for significance. In other words, the increase in BNP from PSI class I to II was statistically significant, as it was from class II to III, and so forth. The subgroup analysis that split low-risk and high-risk groups noted the same conclusion. The strength of this correlation was respectable, with a correlation coefficient of 0.78. The performance of BNP in identifying high-risk or low-risk patients was plotted on a receiver operator characteristic (ROC) curve, and yielded an area under the curve (AUC) of 0.952. The optimal BNP cut-off to identify high-risk patients was 125pg/mL, with a sensitivity of 89% and a specificity of 95%. 

Finally, the subgroup analysis of survivors versus non-survivors was performed. Within the study population, 16.8% of patients died. There was a statistically significant difference in BNP levels between the survivors and non-survivors, as well as a statistically significant difference in their PSI scores. In this study, in which CRP and WBC were also evaluated as potential biomarkers for CAP mortality, BNP performed better than either of the others (AUC 0.82 for BNP versus 0.78 and 0.79, respectively). Whereas a cutoff of 125pg/mL was identified as the best cut-off for identifying severe CAP, the authors identified a BNP cutoff of 299pg/mL for risk of mortality.


The researchers concede that the BNP performed comparably with the widely-used PSI score, but note that this tool requires more complex integration of data to calculate. They cite another study that demonstrated that only 70% of patients with CAP were classified by PSI in an emergency department setting. As such, it was suggested that BNP could potentially supplant PSI as a biomarker of severity in the emergency department setting to help differentiate patients that would be better served by inpatient treatment. In the absence of cormorbidities, the authors conclude that the clinician should consider inpatient treatment of CAP when the patient’s BNP is greater than 125pg/mL, and that a higher level of inpatient care should be considered when BNP is greater than 299pg/mL due to correlation with mortality.

Is this data sufficiently powerful as to be practice-changing? Many emergency clinicians continue to use other pneumonia severity scores such as CURB-65 or PSI to inform their decision-making. In a patient who is discovered to have pneumonia, the addition of a BNP after-the-fact may not significantly change disposition. However, in those patients who present with undifferentiated respiratory distress or cough and subsequently have a BNP drawn during their initial workup, it is conceivable that the BNP could be interpreted as useful prognostic information once the diagnosis of pneumonia is made.  


  • Williams, et al - Jason Nagle, MD, PGY-3 University of Cincinnati Department of Emergency Medicine

  • Li, et al - Michael Spigner, MD, PGY-3 University of Cincinnati Department of Emergency Medicine

  • Webb, et al - Susan Owens, MD, PGY-3 University of Cincinnati Department of Emergency Medicine

  • Editing and Peer Review - Jeffery Hill, MD MEd