Response
We thank Dr. Kuo and Chao [1] for raising the issue of advanced directive restrictions in our recent study on noninvasive and invasive ventilation in critically ill patients with acute exacerbation of COPD. [2] In this study, we performed several multivariate analyses to assess the effectiveness of noninvasive ventilation (NIV) compared with invasive mechanical ventilation (IMV) on the intensive care unit and hospital mortality in a large, multicenter database. In a propensity matched analysis, we matched on the propensity of receiving NIV derived from a regression model that used patient demographics, comorbidities, and the SAPS-II score (a measure of acuity). The propensity model also had a hierarchical structure that adjusted for the hospital where the patient received treatment. We had information on advance directive (eg, if the patient had a Do Not Resuscitate (DNR) in approximately two third of the patients. (Table 1) We did not include this variable in our propensity model as we would have had to either impute a large amount of missing data or restrict the sample to patients without missing data—both of which are methodologically problematic. [3] Importantly, we found that the advanced directive restriction variable was balanced in the propensity matched sample (ie, it was no longer statistically different between the NIV and IMV patients). (Table 1) This means it was no longer a confounder and did not bias our results. This highlights a strength of propensity score matching in that the matched samples may be similar in ways that are not directly observable due to matching on a large number of clinically relevant covariates.
Table 1.
Advanced Directive Restrictions and Outcomes Before and After Propensity Matching
| Variable | Initial NIV N = 974 | Initial IMV N = 1603 | P-value | Initial NIV after Propensity Matching N = 389 | Initial IMV after Propensity Matching N = 389 | P-value |
|---|---|---|---|---|---|---|
| Full Code (No Advanced Directive Restrictions) | 597 (61.29%) | 908 (56.64%) | <0.001 | 216 (55.53%) | 229 (58.87%) | 0.421 |
| DNR/Limited Interventions | 87 (8.93%) | 83 (5.18%) | 43 (11.05%) | 33 (8.48%) | ||
| Missing | 290 (29.77%) | 612 (38.18%) | 130 (33.42%) | 127 (32.65%) | ||
| ICU Mortality Rate | 30 (3.08) | 168 (10.48%) | <0.001 | 9 (2.31%) | 23 (5.91%) | 0.011 |
| In-Hospital Mortality Rate | 72 (7.39%) | 258 (16.09%) | <0.001 | 23 (5.91%) | 39 (10.03%) | 0.034 |
DNR = Do Not Resuscitate
CPR = Cardiopulmonary Resuscitation
We agree that patients who have an advance directives that include the patient’s wish to not be intubated, would only receive NIV, and these patients may well be more acutely or chronically ill or differ in other ways from study subjects without such restrictions. These patients are at a higher risk of death, because if the NIV is not successful there is no other way to escalate their care. If we only included patients with no advanced directive in this analysis, our results will be less applicable to a general COPD population and would likely lower the observed mortality in the NIV group. We did take advanced directives into account when assessing NIV failure because the intent of that analysis was clearly different.
In conclusion, we believe our study supports the use NIV as a first line therapy in appropriately selected critically ill patients with COPD with our manuscript’s stated caveats that this is an observational study and not a randomized control trial (RCT). Our findings are supported by other studies in the literature as well. [4] We encourage future research to take advance directive restrictions into account as they can distort multivariate results even when sophisticated propensity score methods are used.[5]
Acknowledgments
Copyright form disclosures: Dr. Stefan received grant support from the National Institutes of Health (NIH) and the National Heart, Lung, and Blood Institute (NHLBI) (Grant 1K01HL114631-01A1 from the NHLBI of the NIH. The project was also supported by the National Center for Advancing Translational Sciences and the NIH [grant number UL1 TR000073]). Dr. Nathanson’s institution consulted for OptiStatim, LLC (Dr. Nathanson’s company, OptiStatim, LLC, has an ongoing consulting agreement with Baystate Medical Center to provide data analysis and related activities for academic studies, including this one). His institution received other support (Dr. Nathanson’s company company, OptiStatim, LLC, provides data analysis and statistical support to the health care industry). Dr. Lindenauer received support for article research from the NIH. His institution received grant support from the NIH.
Contributor Information
Mihaela S Stefan, Baystate medical Center, Springfield, MA.
Brian H Nathanson, OptiStatim LLC, Longmeadow, MA.
Peter K Lindenauer, Center for Quality of Care Research, Baystate Medical Center, Springfield, MA.
References
- 1.Kuo C-C, Chao C-M. The effect of advanced directive restriction. Crit Care Med. 2015 doi: 10.1097/CCM.0000000000001067. in press. [DOI] [PubMed] [Google Scholar]
- 2.Stefan MS, Nathanson BH, Higgins TL, et al. Comparative effectiveness of noninvasive and invasive ventilation in critically ill patients with acute exacerbation of COPD. Crit Care Med. 2015 doi: 10.1097/CCM.0000000000000945. Epub ahead of print. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Harrell FE. Regression modeling strategies with applications to linear models, logistic regression, and survival analysis. New York, NY: Springer-Verlag; 2001. [Google Scholar]
- 4.Carlucci A, Richard JC, Wysocki M, et al. Noninvasive versus conventional mechanical ventilation. An epidemiologic survey. Am J Respir Crit Care Med. 2001;163(4):874–880. doi: 10.1164/ajrccm.163.4.2006027. [DOI] [PubMed] [Google Scholar]
- 5.Nathanson BH. Why Do Not Resuscitate orders matter in comparative effectiveness research. Crit Care Med. 2014;42:2138–2139. doi: 10.1097/CCM.0000000000000450. [DOI] [PubMed] [Google Scholar]