Discussion
Mortality rates in patients with AECOPD receiving NIV have improved across the two audit periods but remain worse than those reported in RCTs. There are a number of possible reasons for this. Mortality increased with increasing time to NIV following presentation and was particularly high in those receiving NIV >24 hours after admission. Just over half of the patients with AHRF on admisson received NIV within the recommended 3-hour period. In our analysis, patients with acidosis on admission receiving NIV within 3 hours had lower mortality than those receiving NIV between 3 and 24 hours into the admission, although this was statistically non-significant. Those receiving NIV post 24 hours had a much higher mortality (in 2014, for those admitted with acidotic respiratory failure, NIV <3 hours = 12.3%, 3–24 hours= 15.1%, >24 hours = 33.5%) which was statistically significant (p≤0.001).
Second, the charactersitics of patients receiving NIV differ from those included in the RCTs and most notably we highlight a group of patients admitted with NAH who then delvelop acidosis later in the admission and receive NIV at a late stage. This group are not included in the RCTs and have a higher mortality than those presenting with AHRF (24.6% vs 18%, p=0.001) particularly when given NIV later than 24 hours into the admission. This second group of patients who develop acidosis after admission are not well studied and do not have a solid evidence base on which to inform NIV management2 4 18 . What the audit data tell us is that they have similar patient characteristics to those admitted with acidotic hypercapnic respiratory failure except their admission blood gases show a pH within the normal range and their PaCO2 levels are not raised to the same degree as the admission acidotic patients. The higher mortality in this patient cohort is also unexplained but it is possible that the recognition of their deterioration may be delayed as is subsequent NIV use. The average time between first and second ABG is 1.8 hours in the acidotic group and 5.6 hours in the non-acidotic (p<0.001). It is the second ABG that demonstrates the acidotic hypercapnia picture, and the median length of time from the second ABG to NIV administration in this ‘later acidotic’ cohort was 2 hours. It is also possible that some of this group may have been more appropriately managed as end of life rather than with NIV. However, even when this latter group are excluded from the analysis, the mortality of patients admitted with acidotic hypercapnic respiratory failure remains higher (18%) than that of the RCTs (9.9%).
Patient selection is a further potential explanation for both the difference in mortality between the longitudinal data and the discrepency with the RCT results. Previous audit papers have flagged that many COPD patients are treated outside the evidence base of the RCTs.4 17–20 In the UK, nearly two-thirds of patients with COPD treated with acute, ward-based NIV have NIV documented as their ‘ceiling of treatment’ which also may also account for the differences observed.21–23 Patients in both these audits were older (average 71.1years) than those reported in the Cochrane review (66.8 years) of NIV in AECOPD patients admitted with hypercapnic respiratory acidosis.10 Most audit patients had comorbid illness and notably radiological consolidation consistent with pneumonia, known to increase mortaity and excluded from the RCTs, and was present in over a fifth of the patients managed with NIV in the two audits.3 4 10
The mortality rates we report from 2014 are approaching those of the RCTs when only looking at those with acidotic respiratory failure who received NIV within 3 hours of admission (12.3% in 2014 audit vs 9.9% in RCT), but are still poorer. We have also found that patients in real world practice are older, have multiple comorbdities and are more acidotic than those in research trials. It is likely that some of these reported factors may account for the poorer outcomes of patients managed in real life settings compared with the RCTs.
In contrast, it is difficult to hypothesise that these factors account for the improvement in mortality seen between the two audit periods. It may be that better adherence to guideline recommendations not recorded in this audit, for example, NIV pressures used, ventilator asynchrony, managed agitation have made a significant contribution to impove outcomes.24–26 There have been significant improvements in processes of general COPD care between 2008 and 2014 including a higher proprotion of patients receiving antibiotics within 24 hours and the proportion of patients with controlled oxygen prescribed both which have been demonstrated elsewhere to improve outcomes.4 19 27–32 Notably, PaO2 ABG values in the 2014 cohort were lower than in 2008 (table 1).
Third, there has been a significant increase in the number of patients being discharged by a respiratory consultant, from 75% in 2008 to 85% in 2014. Previous data have shown that mortality rates in both in-patient and at 90 days were lower in units with more respiratory staff/1000 beds.33 Units with 4 or more respiratory consultants were associated with the best outcomes.33
There are a number of strengths and weaknesses in this paper. Audits are not designed to answer research questions and the reasons for the reduction in mortality observed in the 2014 audit and the difference in reported mortality from the RCTs and the audits are not certain. The 2008 and 2014 data sets had some significant differences, so that exact comparisons cannot be made and the latter audit did not include Scotland or Northern Ireland. The interpretation of the audit proforma and the accuracy of some of the data collected in different hospitals is questionable as it is not subject to the usual rigor of research studies hence we have focused on trends. Another limitation is that a proportion of patients were excluded in the audit, whom did not receive NIV at all, and the audit data are not descript enough to explain these exclusions, which may have a bearing on the outcomes.
However, these data do reflect real life clinical practice wheras clinical trials tend to both add factors that are not reproducible outside of studies and exclude more complex patients as subjects. The combined numbers included in this analysis are greater than those of the combined RCTs cited in the Cochrane review.10 The application of NIV to real life clinical settings is also an opportunity to observe the translation of research into practice and to report clinical practice challenges that can inform further research. Particulary with regard to the cohort of patients admitted with NAH, and whether closer monitoring and/or more aggressive treatment of hypercapnia improves outcomes in this group.18 34 Data were not collected on any previous admissions with AECOPD that required treatment with acute NIV. Current data for the use of domiciliary NIV for persistant hypercapnia in COPD patients are mixed, with some trials reporting a mortality benefit at 1 year.35–37 Further research is needed in this area to improve the outcomes for those with NAH on admission.
This study suggests that real world clinical practice still falls short of recommended standards of time from presentation to application of NIV. This, together with the application of NIV to a significant proportion of patients who were exluded from the RCT evidence base, may well explain the discrepency in observed mortality. We suggest that further research should be conducted to replicate real world clinical practice including the inclusion of older subjects with comorbdities and presentations with severer acidosis. For clinicians in practice, more timely application of NIV to those admitted with AHRF and closer monitoring, with clear treatment escalation plans and appropriate ceilings of care for those admitted with NAH, should help identify deteriorating patients requiring earlier NIV, while ensuring those better managed through an end of life pathway are not subject to uncessary interventions of low value.