© 2006 American Thoracic Society doi: 10.1164/rccm.200609-1389ED
Decision Rules and PneumoniaWhat Are We "Predicting," and for Whom?University of Pittsburgh, Pittsburgh, Pennsylvania In this issue of the Journal (pp. 1249–1256), España and colleagues report a new clinical prediction rule to identify emergency department patients with severe community-acquired pneumonia (SCAP) (1). They defined SCAP as community-acquired pneumonia with at least one of the following outcomes: hospital death, mechanical ventilation, and septic shock. Using prospective and retrospective data from five Spanish hospitals, España and colleagues identified eight clinically available variables that can help identify patients with or at risk for SCAP. This new work is notable for examining a condition difficult to define, by specifically using clinically relevant and objective outcomes relating to mortality. The three-stage derivation and validation methodology is laudable, defining the ability of a new pneumonia rule to identify patients with higher risk of death and selected serious medical outcomes, and also offering useful comparative data on other decision or prognostic rules for severe pneumonia. The variables used in España and colleagues' work overlap with those seen in the comparator prediction scores, emphasizing the apparent validity of the data. There are multiple existing pneumonia decision rules with a varying amount of external validation and postpublication use (2–6). Does a new rule that services only one end of the spectrum, identifying patients at high risk of short-term death and selected nonfatal outcomes, offer an advantage in practice? España and colleagues' rule was not designed to identify low-risk patents, and offers slightly less sensitivity in detecting high-risk patients as compared with an existing rule (the Pneumonia Severity Index [PSI]). España and colleagues' new rule uses fewer variables than the PSI, offering the potential of easier use. Conversely, the new rule uses more variables than the CURB-65 (confusion, urea, respiratory rate, and blood pressure at age 65 years or older), but identifies SCAP better, similar to the tradeoff made between it and the PSI. Current evidence suggests decision rule use is imperfect even when seemingly simple, and implementation strategies are more important than perceived ease when translating decision rules into clinical practice (7, 8). There are a number of larger issues that should be considered. The main value of a pneumonia prediction rule, or any diagnostic or prognostic adjunct, is in addressing the key question of how to identify occult severe illness—that is, severe illness in patients that appear well, but are not. An ideal rule would also identify those who have no occult severe illness, but who will develop it within a specified interval. However, no current pneumonia severity score offers this ability. For all such scores, to qualify for the highest risk categories, patients must meet a number of severity criteria (e.g., hypotension, tachypnea), whereas to qualify for the lowest risk categories, these severity criteria must be absent. Thus, each score may simply formalize what is, or should be, clinically obvious to the bedside clinician rather than offer new patient-specific value. For example, España and colleagues used hypotension (systolic blood pressure < 90 mm Hg) to predict an outcome that includes the same observation (hypotension as a part of septic shock). This circular methodology may have increased the apparent predictive power, albeit modestly. Perhaps it is too much to expect a bedside rule using current common data to deliver the performance we desire. A better approach may be a combination of clinical judgment, clinical severity scoring, and use of novel biomarkers that extend the ability to assess current or future responses to illness. For example, severity scores for sepsis have been created to predict mortality (9), whereas, separately, Aduen and colleagues found that a lactate greater than 4 mmol/L predicted poor outcomes, even in patients with "normal" blood pressure (10). Normotension accompanied by elevated lactate in the setting of infection has been called "cryptic shock," with outcomes as poor as frankly hypotensive septic shock (11). Rivers and colleagues, Annane and coworkers, and others have similarly used lactate to define severe sepsis and septic shock in their landmark trials (12, 13). However, lactate may simply identify occult shock. Identifying impending, but not currently existing, shock remains elusive. For pneumonia, several new biomarker assays are under investigation and may provide additional diagnostic and prognostic value (14). Last, there is a growing body of literature that genetic polymorphisms may also provide independent discriminative information (15). Although decision rules exist to limit the variability and inaccuracies in clinical judgment, no rule can account for rare medical conditions and other factors in deciding care paths, such as social issues and patient desires. Each rule is intended to augment these assessments, and each rule may perform differently when trying to incorporate the prediction into practice. Isn't severe pneumonia just severe sepsis (infection with new organ dysfunction)? In multiple epidemiologic studies, pneumonia is consistently the most common cause of severe sepsis. España and colleagues correctly note there is no universal definition of SCAP. We suggest it may be useful to define severe pneumonia as simply a type of severe sepsis—that is, pneumonia with organ dysfunction. Beyond simplifying the definition, conceptualizing SCAP as severe sepsis may also encourage clinicians to think beyond the pneumonia and respiratory system per se, and prompt consideration of pneumonia's systemic consequences and potential adjunctive therapies. Moreover, this would place SCAP within the conceptual framework of sepsis that has undergone decades of deliberation, consideration, and definition (16). The study by España and colleagues displays the strengths and limits of pneumonia prediction rules. Choosing this rule instead of another pneumonia rule should be based on what is desired—high-risk assessment alone or risk stratification across the spectrum of illness—tempered by the barriers of use that affect all rules in practice. As the science of risk prediction continues to evolve, adjuncts to detect impending organ dysfunction may help better tailor care, and thinking of severe pneumonia as a common form of severe sepsis may be of the most immediate clinical value. FOOTNOTES Conflict of Interest Statement: D.T.H. has received $2,000 in consulting fees and $1,270 plus travel reimbursement from BRAHMS Diagnostica for speaking at a scientific meeting. D.M.Y. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. REFERENCES
Related articles in AJRCCM:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
