Designing a Better “Nest”: Applicable to Preventing Hospital Exposures to Risk Factors for ARDS or just Retrospective Study Design?

The acute respiratory distress syndrome (ARDS) remains a common and important cause of morbidity and mortality in critically ill patients. Despite improved understanding of pathophysiology and numerous clinical trials targeting varied pathways of injury, current treatment is limited to lung-protective strategies of mechanical ventilation and other supportive care measures (1). The majority of patients do not present with ARDS but, instead, develop ARDS after admission (2, 3). In addition, epidemiologic data from Olmstead County, Minnesota suggests rates of ARDS have fallen over the past decade, largely due to reduced rates of “hospital acquired” ARDS suggesting that the development of ARDS may be partially preventable (3). Multiple prior studies have identified risk factors associated with the development of ARDS, such as medical and surgical misadventures, transfusions of blood products, higher tidal volume mechanical ventilation, inappropriate antibiotics, and delayed early goal directed therapy of severe sepsis (4-10).

However, past studies primarily rely on multivariable regression to assess associations of risk factors with ARDS without an ability to control for differences in the baseline risk of developing ARDS, thus limiting our ability to assign causality to individual risk factors. In this issue of Critical Care Medicine, Dr. Ahmed and colleagues have performed a population-based, nested case-control study of consecutive patients who developed ARDS during hospital admission in Olmstead County from 2001-2010 (11). In contrast to prior studies, cases of ARDS were matched to controls who did not develop ARDS based on age, sex, admission for high-risk surgery, sepsis, SpO₂/FiO₂ ratio, and overall risk of developing ARDS as assessed by their Lung Injury Prediction Score (LIPS) (2). Risk factor exposure was ascertained by in-depth chart review by trained study coordinators blinded to the whether patients developed ARDS or survived to hospital discharge. The time frame to screen for exposures was restricted to hospital admission up to six hours prior to the development of ARDS for cases and the equivalent time period for the matched control (unblinded statisticians responsible for matching assigned the relevant time periods). The authors found that medical and surgical misadventures; inadequate empiric and delayed antibiotic delivery; hospital acquired aspiration and risk factors for aspiration (nasogastric tube, difficult intubation, and delirium); and receiving transfusion of red cells, platelets, plasma and cryoprecipitate were more common among cases of ARDS and cases also received greater cumulative fluid infusion and higher tidal volumes among the subset of mechanically ventilated patients. Exposure to antiplatelet agents was more common among controls while opiods, benzodiazepines, antacids and furosemide were more common among cases. Importantly, blinded review of adverse events suggested the roughly 70% of both medical and surgical misadventures were preventable.

However, despite the author’s substantial effort to control for bias, there are inherent limitations to any retrospective study design. Case-control studies are subject to bias in selection, recall and exposure ascertainment. Conclusions are limited by the reliability of the documented medical record. The authors reference prior data showing that the clinical documentation contained the word “iatrogenic” in only 2% of cases admitted to the intensive care unit as a result of an iatrogenic event (12). Many adverse events are likely undocumented but may be preferentially documented when they have more substantial clinical seqeuela. For example, aspiration was identified in 51 cases (12%) but only 1 control. It is possible that patients with a relatively acute status change and an increasing oxygen requirement are far more likely to have aspiration—either correctly or incorrectly—documented in their chart. Similarly, greater fluid infusion and receiving furosemide were both risk factors for ARDS suggesting there was at least some confounding by association or indication for certain risk factors. Also, even with matching, cases were more likely to have shock, an alcohol abuse disorder and hypoalbuminemia at baseline (although, sensitivity analyses adjusting for these baseline differences did not significantly impact results). Finally, the authors appropriately assumed that exposures were likely to be highly correlated and therefore, did not attempt to assess the independent effect of each individual risk factor.

Despite these limitations, this study is unique in that it was able to use the well-validated LIPS (2) to control for baseline risk of developing ARDS. The authors have identified important, and potentially preventable, exposures that may significantly increase rates of ARDS. Importantly, the authors also documented decreasing rates of exposures to the identified risk factors that correlated with falling rates of ARDS over the 10-year study period. The Mayo Clinic in Rochester, where the study was conducted, has been a leader in standardizing care to reduce rates of exposure to potential “secondary hits” on the pathway to ARDS. During the study period, specific protocols were implemented to restrict transfusion of blood products (using computerized order entry with decision support); limit tidal volumes for mechanically ventilated patients (including respiratory therapy driven lung-protective ventilation protocols followed by implementation of a validated automated electronic surveillance and notification system with documented reduced time of exposure to larger tidal volumes (13)); sepsis and pneumonia order sets with computerized order entry and decision support for appropriate antibiotic delivery; increased intensivist staffing in the medical intensive care unit; and the addition of a 24-hour on-site intensivist (3). As with the association between exposure to specific risk factors and ARDS, the temporal association of the declining rates of exposures with implementation of specific protocols does not automatically assign causality and, depending on different systems and practice patterns across institutions, the effectiveness of specific protocols at certain institutions may not be generalizable to other centers. A recent randomized trial from the Hospital of the University of Pennsylvania found that the addition of a nighttime intensivist did not improve the quality or efficiency of care in the medical intensive care unit but the extent to which results from a unit with heavy presence of daytime intensivist and protocolized care processes can be extrapolated to other centers is unclear (14).

Despite some remaining uncertainties, there is now substantial empirical evidence that exposures to several potentially preventable risk factors during a hospital admission can increase the risk of developing ARDS and that the implementation of common sense protocols to improve standardization of care and reduce rates of exposures can substantially reduce the incidence of ARDS. Given the limited options for treating ARDS, committing concerted effort towards prevention seems likely to be a high-yield, low risk and advisable strategy. With their nested case-control study, Dr. Ahmed and colleagues have made a strong argument that designing a better “nest” should apply as much to our hospitals as it does to our case-control studies.