Discussion
This is the first study to examine the relationship between case-volume and outcomes in patients admitted to ICUs for AECOPD-related ARF. In the CUB-REA ICU prospective database, 14,440 subjects were admitted for AECOPD-related ARF between 1998 and 2010. During this period, severity, mortality, and the use of NIV increased. Higher case-volume units were associated with an increased use of NIV and a trend toward reduced mortality.
Trends Over Time
The striking increase in the use of NIV over time has already been reported. In our study, it was not accompanied by a decrease in severity. On the contrary, SAPS II scores increased during the study period. The increase in patient mortality did not exceed what would be predicted from SAPS II data despite the increasing use of NIV, suggesting that NIV is not associated with impaired prognosis in these patients. Indeed, it is now well known that, in patients with hypercapnic respiratory failure, NIV improves outcomes, especially survival, by reducing the need for intubation and the prevalence of related complications. In the nineties, it was shown that the use of NIV for AECOPD had changed over time. It was also demonstrated that the success of NIV in part depends on its acceptance and patient's compliance, which in turn depends on the way NIV is administered by the ICU team. In this work, as in our study, the authors reported a constant rate of treatment success despite an increasing patient severity. They concluded that the learning and training process of the ICU team might be of great importance. Although NIV has become a standard first-line treatment of AECOPD, it has been reported that, on some occasions, doctors could be reluctant to admit patients in the ICU. This suggests the need for further encouragement of clinicians to use NIV in this situation.
Altogether, mortality of patients hospitalized for AECOPD ranges from 4% to 30%, depending on the setting and initial patient characteristics. In our study, the increase in patient's severity and mortality over time could in part be related to the growing proportion of patients >80 years. This is related to global demographic trends in the population and to improved long-term survival for COPD patients. In addition, the availability of NIV is likely to be associated with less reluctance of physicians to admit elderly patients to ICUs. Such an increase in the proportion of older patients is quite general in ICUs, and in this study, age greater than 80 years was positively associated with the use of NIV. Furthermore, NIV likely reduces the need for endotracheal intubation, with consequent reduction in the mortality of very old patients with acute hypercapnic respiratory failure.
Factors Associated With Case-volume
Since the early 1980s, many studies have emphasized the relationship between case-volume and outcomes in health services. Two hypotheses have been suggested: 1) physicians (and hospitals) develop more skills if they treat more patients ("practice makes perfect") and 2) physicians (and hospitals) achieving better outcomes receive more referrals and thus accrue larger volumes ("selective referral"). In the case of AECOPD-related ARF and the use of NIV, it is likely that experience improves the skills of ICU teams. Because NIV is a relatively recent therapeutic option, it remains underused, as suggested by the persistent difference in its use between tertiles of case-volume over time. After adjustment for all possible confounding factors, ICU case-volume was significantly associated with both NIV use (which increases with increasing case-volume) and mortality (which decreases as case-volume increases). This is the first study to report such a relationship between case-volume and the use of NIV in AECOPD-related ARF. Variations in the use of NIV between centers could be considered surprising, given the large amount of data available on this technique. However, the first randomized controlled trials on NIV were published in the middle of the nineties and its use for hypercapnic ARF only became generalized during the early 2000s. Thus, it could still be considered a relatively "new" technique for some practitioners. A more likely explanation could be the differences in nurse/bed ratio. This is an important issue because initiation of NIV requires high attention levels and is therefore time consuming, as has been pointed out in other studies. Unfortunately, this ratio (which is officially defined since year 2002 in France) was not available in the CUB-REA database for all the units during the study period. Centers associated with high use of NIV could also be those with more interest in ARF and/or closer collaboration with respiratory physicians, but such information is difficult to gather.
High-volume ICUs were associated with better outcomes in terms of survival than medium- and low-volume ICUs. This observation could be explained by the preferential use of NIV in high-volume units. More generally, high-volume units may also be those with the most experienced staff, more human and material resources, or more quality control programs. Other studies on the links between case-volume and outcomes have been published in the field of critical care with divergent results. Lecuyer et al found that case-volume had a favorable impact on mortality in hematological patients with ARF. Similarly, a decrease of mortality in cardiac or cancer surgery patients has been found in high-volume centers. Recently, a U.S. database analysis also reported an increase in the use of NIV during the same time period, with a concomitant decrease in the need for intubation and in-hospital mortality. In addition, the authors found an increased mortality in the subgroup of patients with NIV failure requiring transition to InV.
Limitations
Our study has some obvious limitations. First, due to its retrospective nature, the reliability of the recorded data depends on the quality of coding by local investigators. However, the CUB-REA database is regularly subjected to audits and tested for coherence to check the validity of data. This methodology does not allow determination of the sequence of NIV and InV administration, i.e., NIV failure requiring intubation or NIV following weaning from InV. Indeed, the impact of postextubation NIV and the impact of NIV as a first-line treatment are two very different issues. In addition, the coding process did not provide clinical data for specifically assessing the severity of AECOPD and the appropriateness of NIV use.
Second, the global increase in case-volume between 1998 and 2010 could modify the link between case-volume and the studied variables. However, multivariate conditional logistic regressions were performed, after matching on the year of admission, which would at least partly reduce the risk of time-related bias.
Third, it must be pointed out that these results apply only to patients treated in the ICUs and not to the use of NIV outside ICU (especially during transport from home to hospital), which has also increased considerably in the last few years. Finally, the increasing severity and the rather stable outcomes may also be due to better ventilators and interfaces.