The authors reply

    loading  Checking for direct PDF access through Ovid


We thank Dr. Pepper (1) for the insightful questions regarding our recently published article in Critical Care Medicine on the obesity paradox in the ICU (2). As is correctly pointed out, diagnoses based on clinical criteria and billing data are not always consistent. The admission diagnoses used in our study (2) are not derived from International Classification of Diseases, 9th Edition codes in billing data; they are entered by experienced ICU clinicians trained in the Acute Physiology and Chronic Health Evaluation (APACHE) methodology. In short, the admission diagnoses represent the primary reason for the patient being admitted to the ICU, as understood on the day of admission. However, there are limitations to this. The concept of “septic shock” does not exist as an APACHE admission diagnosis and therefore our analysis could only describe those identified as having sepsis at admission.
Several important questions are also raised regarding how diagnoses and interventions may differ across body mass index (BMI) categories. As noted with the example of sepsis and septic shock, interest in specific diagnoses would likely require a more detailed analysis to standardize ambiguous diagnoses or those which evolve over time. The scope of our study (2) was a population-based study, rather than detailed analyses of specific subpopulations. We agree a subsequent study or studies may be warranted to explore the relationship between BMI, mortality, and nutrition in subpopulations such as those with septic shock or those receiving specific ICU treatments.
Regarding the use of APACHE IV, interventions and resuscitation provided during the first 24 hours should not theoretically attenuate the mortality predictions as APACHE methodology uses the worst values during the day of admission (3). Therefore, any improvement in physiology observed due to resuscitation should not reduce the predicted probability of mortality. Lastly, we agree obtaining complete data for all of the various elements of APACHE IV can be difficult and the 857,875 patients with APACHE scores available reflect those without any missing data, indicating rather robust data collection. Although we believe the proportion of patients without enough data to generate an APACHE IV prediction is relatively modest given the burden of data required, it is certainly possible bias can be introduced. However, as an observational study with an inherent potential for bias, our goal was to generate interest in new areas of research where more specific questions regarding the role of nutrition, nutritional status, and mortality can be explored using study designs which can better minimize bias and eliminate confounding.
Dr. Zhou disclosed work for hire. Dr. Liu disclosed he is an employee for Philips Healthcare, which provided the data for the analysis. Dr. Hassan received funding from Philips Healthcare. Dr. Badawi received funding from Philips Healthcare and ICMed. Dr. Harris has disclosed that she does not have any potential conflicts of interest.
    loading  Loading Related Articles