Abstract
Non-invasive ventilation (NIV) is a clinically proven, cost-effective intervention for acidotic exacerbations of COPD that is recommended by UK national guidelines. This study examines the extent to which these recommendations are being followed in the UK. Between August and October 2003 a national audit of COPD exacerbations was conducted by the Royal College of Physicians and the British Thoracic Society. 233 (94%) UK hospitals submitted data for 7,529 prospectively recruited acute COPD admissions, documenting process of care and outcomes from a retrospective case note audit. They also completed a resources and organisation of care proforma. Nineteen hospitals (8%) reported they did not offer NIV. There was no access to NIV in 92 (39%) intensive care units in 88 (36%), high-dependency units or on general wards of 85 (34%) hospitals. In 74 (30%) NIV was available on all 3 sites. A low pH (<7.35) was noted at some time during admission for 26% (1714/6544) of patients and NIV was administered to 31%. Patients receiving NIV were more often admitted under a respiratory physician, or seen at some stage by a respiratory specialist and had more severe disease (higher PaCO2 (median 9.8 v 7.8 kPa), lower oxygen tension (median 8.8 v 9.8 kPa), higher incidence of peripheral oedema (48% v 39%), of pneumonia (27% v 16%), higher in-hospital mortality (26% v 14%) and at 90 days (37% v 24%) and longer hospital stays (median 9 v 7 days) than those not receiving NIV. Hospitals with least usage of NIV had similar mortality rates to those using NIV more often. A comprehensive NIV service is not available in many hospitals admitting patients with acute respiratory failure secondary to COPD. Access to acute NIV is inadequate and does not conform with NICE and BTS guidelines. These observational audit data do not demonstrate benefits of NIV on survival when compared to conventional management, contrary to results from randomised trials. Reasons for this are unclear but unmeasured confounding factors and poor patient selection for NIV are likely explanations.
INTRODUCTION
COPD exacerbations account for about 10% of all medical emergency admissions and just over a quarter present with acidosis (Citation[1]). There is a high (7.4%) in-patient mortality (Citation[1]). Non-invasive ventilation (NIV) has been established as a cost-effective, life-saving intervention for hypercapnic respiratory failure associated with COPD exacerbations (Citation[2], Citation[3]) and it is a standard recommendation within the National Institute for Clinical Excellence (NICE) (Citation[4]) and British Thoracic Society (BTS) (Citation[5]) guidelines. Trials, in the intensive (Citation[6], Citation[7], Citation[8], Citation[9]), and in non-intensive (Citation[10], Citation[11]), care settings show that NIV reduces endo-tracheal intubation rates, mortality and length of intensive care unit stay. Plant et al. (Citation[11]) demonstrated the benefits of NIV in patients with acidosis (pH < 7.35) and hypercapnia (PaCO2 > 6.0) who had not improved rapidly following initial management with conventional treatment. This randomized controlled trial forms the basis of current recommendations for NIV use during COPD exacerbations in the United Kingdom (Citation[1], Citation[2]).
The Royal College of Physicians/BTS 2003 audit of the management of patients admitted with exacerbations of COPD provides an opportunity to examine the extent to which these recommendations are being followed and to assess whether the results obtained in randomised trials are reproduced in practice. In particular this audit enables the assessment of NIV utilisation and facilitates the evaluation of outcomes.
METHOD
The methods have been described in detail elsewhere (Citation[12]). The project was run by the Clinical Effectiveness and Evaluation unit (CEEu) of the Royal College of Physicians and the British Thoracic Society. There were two audit proformas: first a survey of resources and organisation of care (completed between August and October 2003); second a record of clinical activity and outcomes for 40 consecutive cases per site identified prospectively from 1st September 2003 but audited from the notes retrospectively. (http://www.rcplondon.ac.ukcollege/ceeu/ceeucopdhome.htm)
Data were entered to a bespoke Web-based programme supported by help-note manuals and notes embedded within the audit software. Data collectors included specialist registrars, respiratory nurses, audit staff and data clerks. The outcomes recorded were inpatient death, death within 90 days of admission, and for discharged patients the length of stay and readmission within 90 days of admission.
To gain a measure of the reliability of the data, units were asked to identify a second individual to re-enter clinical data on the first 5 patients included. Then, 190 units submitted 910 cases for analysis. The levels of reliability were generally good with kappa values of 0.60 and higher dominating the results. Most were over 0.80 (very good).
We used random effects logistic regression (STATA8, Stata corporation Texas USA) to adjust for site clustering to obtain p-values in for association of NIV treatment with outcome and to give NIV treatment odds ratios for mortality adjusted for predictors. Random effects linear regression was used to obtain p-values for association of NIV treatment with length of stay outcome transformed as LOG10 (LOS+1).
Table 2 Casemix factors for those patients with a pH less than 7.35 according to the treatment received
RESULTS
Cohort description
Patient and organisational audit data were available for 233/247 (94%) hospital units representing 180/194 (93%) Trusts. Data were collected on 7529 patients in 8013 episodes of care, median 39 per hospital unit, with only the first episode being analysed. Mean (SD) age was 71 (Citation[10]) years and 53% were male. Two-thirds had previously been in hospital or to an early discharge scheme, for COPD.
Over one-third (39%) lived alone, and one-quarter had limited self-care or were bed or chair bound. Co-morbidity included heart disease (38%), stroke (6%), diabetes (10%), loco-motor problems (12%) and other chest problems (11%). In-hospital mortality was 7.4% (542/7313) and mortality within 90 days of admission was 15.3% (1112/7274). For 6771 discharged patients the readmission rate within 90 days of admission was 31.4% (2069/6585), median length of stay was 6 days (mean 8.3) with 40% (2611/6544) staying more than 7 days and 15% (969/6544) more than 14 days.
Non-invasive ventilation (NIV)
Organisation of NIV services was variable. Nineteen hospitals (8%) reported they did not offer NIV and only 74 (30%) had NIV available across the hospital. There was no access to NIV in 92 (39%) of intensive care units, in 88 (36%) of high dependency units (HDU) and on the general wards of 85 (34%) of hospitals.
A low pH (<7.35) was noted at some time during admission for 26% (1714/6544) of patients. NIV treatment was given to 31% (529/1714) with 16/529 also having invasive support, 70 receiving invasive support alone, 1095 receiving neither and unknown for 20. Doxapram was used with NIV support (12 patients), with invasive support (4 patients) or alone (19 patients). When NIV was not given to acidotic patients, the reasons given included that it was not thought appropriate for the specific patient (n = 321), there were no facilities (n = 48), patient refused (n = 13) or the treatment failed (n=3). Frequency of use of NIV was related to availability (). If NIV was unavailable on either wards or HDU, 9.5% of acidotic patients received NIV, rising to 23.2% if it was available on the HDU, and 37.6% if it was accessible on both wards. The range of variability between hospitals ranged from 0% to 100% of acidotic patients being treated with NIV. Neither in-hospital nor 90-day mortality appeared related to frequency of NIV use.
Table 1 Proportion of patients with a pH < 7.35 receiving NIV treatment according to the availability of NIV treatment within their hospital
Outcomes of invasive, non invasive and no ventilatory support amongst those patients with respiratory acidosis
There were minimal differences between the groups in age, gender, co morbidities, previous admissions for COPD, smoking status or FEV1, although %predicted FEV1 was lower (). However, those who received support did have worse blood gas data (lower pH and higher PaCO2), lower oxygen saturation, and were more likely to have a radiographic pneumonia, or peripheral oedema. Patients receiving NIV were more likely to have been admitted under a respiratory physician, or to be seen at some stage by a respiratory specialist. Mortality rates (in hospital and at 90 days) were higher for those who received NIV and highest for those who had invasive ventilatory support (). They were also more likely to have an acidosis on a second set of blood gases. Overall, NIV treatment did not appear to associate with readmission, but did associate with longer stays in hospital.
Table 3 Outcomes for patients with a pH level below 7.35 according to the treatment
The main predictors of inpatient and 90-day mortality in this acidotic group were age, performance score, lowest pH, initial pH, blood urea and serum albumin, oxygen saturation, CXR cancer, weight and peripheral oedema. Recording a history of heart disease was associated with higher in-hospital mortality when NIV was used (36% (66/183) v 20% (69/337)), but little difference (14–15%) when it was not. Covariate adjustment gave in-hospital and 90-day mortality odds for NIV treatment of 2.21 (95%CI 1.58–3.08) and 1.89 (95%CI 1.42–2.52), respectively.
Other analyses
Two other approaches were used to try and unpack these results further. One approach () was to select those patients who would have been eligible for Plant's randomised trial (Citation[11]). From units offering NIV on wards initial respiratory rate, pH and PaCo2 values were known for 66% (3183/4814) and of these 341 had a respiratory rate of >23/min and a mild to moderate respiratory acidosis to match Plant's criteria and one-third of these (110) received NIV. Plant showed consistent differences in favour of NIV whereas the audit results were in the opposite direction and especially for mild acidosis (7.30–7.35). Overall in-hospital mortality for these selected audit patients was 26% compared to 10% in Plant's trial. The second approach () was to use propensity analysis (Citation[13], Citation[14]) for patients with pH<7.35, by which logistic regression methods took all the audit measured predictors of NIV treatment and computed the propensity (predicted probability) for each patient to receive NIV. For patients with characteristics that made NIV support more likely similar mortality outcomes were observed for treated and untreated patients. For patients with characteristics that made NIV support less likely those actually receiving NIV had higher levels of mortality than those not receiving NIV.
Table 4 Outcomes and NIV treatment: comparison using plant trial criteria using statistical summaries as per plant trial
Table 5 Patients with lowest pH < 7.35, treated or untreated with NIV and grouped by their propensity to receive NIV
DISCUSSION
This observational study provides an insight into the variability of management of severe COPD within the UK. Over 94% of acute hospitals submitted organisational data, most likely a representative picture of UK medicine at the end of 2003. While a third of hospitals reported NIV availability on the ward, on HDU and in ITU, 9% of hospitals reported no provision at all within their trusts. It is likely that for a significant number of patients there was no access to NIV. Organisation is important in that access increases as NIV availability increases, but even when available on general wards only 38% of acidotic patients received NIV. While some of the 62% without NIV may have spontaneously improved it is likely that others would have benefited had they been offered the treatment. These 2003 audit data are therefore worrying and pose two challenges—first to management that hospitals must put organisational structures in place to ensure that NIV is widely available and second to clinicians who must be more prepared to refer patients for NIV. We did not collect data on the staffing or training for NIV but given the variable provision of the service, there could be concerns that too is variable and could have impacted on the higher mortality observed in the audit compared with clinical trial data.
The RCT evidence led the NICE guideline to make a clear recommendation that NIV be available and the Plant trial (Citation[11]) demonstrated that NIV can be used effectively on ordinary wards as well as within HDU and ITU settings (Citation[6], Citation[7], Citation[8], Citation[9]). The outcomes of NIV treatment in the 2003 audit were disappointing as it was expected that the audit would confirm the use of NIV in routine clinical practice. The most likely explanation is that those who received NIV had more severe disease that the audit (which had to limit its data collection) did not measure. The Plant (Citation[11]) RCT excluded patients who responded rapidly to bronchodilators and controlled oxygen. The audit cohort included all patients with acidosis and could not differentiate patients with an acidosis that recovered rapidly e.g., secondary to oxygen toxicity (Citation[15]). Such rapid responders would include many with better prognoses, who may not been offered NIV because they were improving, and who appear in the non NIV cohort. Where a second set of blood gases was available, those not receiving NIV were improving faster than those treated with NIV although with a wide scatter. Only a third of the patients who received NIV would have met the Plant trial's entry criteria, suggesting that routine practice has extended the spectrum of patients treated with NIV. Those denied invasive ventilation may have been given NIV as a last resort. The observed in hospital mortality in the audit of 26% for those who received NIV is appreciably higher than the 10% mortality in the Plant trial, and there was more radiographic pneumonia in the NIV cohort.
There is no direct data from the audit indicating whether NIV was the ceiling of care but some indication is given by the fact that only 3% of patients of patients receiving NIV went on to invasive ventilation. In addition, only 4% of patients were intubated and ventilated directly without receiving NIV.
But there are two caveats to using confounding by severity as the dominant explanation. The first caveat is that outcomes of ‘NIV’ relative to ‘no NIV’ were worse among the milder acidotic patients than among the more severe. Also, only a minority of the audit patients would have been eligible for Plants's RCT. Thus as well as it being possible that NIV is offered to some who are “too ill to benefit”, there may be others where NIV is being used beyond the current evidence. This might include some where hypoventilation was not the principle problem or where cardiac or other causes of acidosis were not recognised. The limited retrospective audit data cannot resolve this. If the role of NIV has extended beyond its initial remit and patients outside the BTS guideline criteria are being treated with NIV, it suggests we need more studies to better define the treatment threshold (Citation[16], Citation[17]).
The second caveat is that the third of hospitals that made least use of NIV showed an overall in-hospital mortality for COPD that was no different to the third who used NIV most often. The case mix of COPD admitted to those hospitals (in so far as it was measured) was similar.
Unlike the controlled trial, routine practice implies use of many different types of ventilator, in a variety of settings, and by staff with unknown levels of training and with no regular audit analysis or feedback. This study did not collect data on the staff training, the way the equipment was applied, failure rates, and thus cannot relate this to outcome but it's possible that combinations of inadequacies like these could have led to a higher mortality that observed in an organised clinical trial. Any future national prospective audit should collect data that allows a more accurate assessment of NIV efficacy (e.g. ascertain time of arterial blood gas analysis in relation to admission and NIV initiation, & mean duration of NIV use) and should also assess the quality and nor mere provision of NIV service.
This study has provided clear evidence that NIV therapy is variably available and variably used across the UK. Given the strong guideline recommendations this variability is not acceptable. All hospitals accepting acute emergency admissions should be able to provide NIV. But the audit cannot demonstrate the expected benefit. We do not know if the extrapolation to use NIV in a wider cohort is justified or whether physicians are selecting patients and administering NIV in the most effective manner. It would be helpful to determine what factors should govern the selection of patients for NIV and what outcome variables would be most useful in assessing its use in practice. The data on NIV from this audit are admittedly crude but the results do not reflect the guideline evidence and we need to understand why.
Declaration of interest
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
REFERENCES
- Hubbard R. Burden of lung disease. Thorax 2006; 61: 557–558
- Lightowler J V, Wedzicha J A, Elliott M W, Ram F SF. Non-invasive positive pressure ventilation to treat respiratory failure resulting from exacerbations of chronic obstructive pulmonary disease: Cochrane systematic review and meta-analysis. BMJ 2003; 326: 185–187
- Plant P K, Owen J L, Parrott S, Elliott M W. Cost effectiveness of ward based non-invasive ventilation for acute exacerbations of chronic obstructive pulmonary disease: economic analysis of randomised controlled trial. BMJ 2003; 326(7396)956
- National Institute of Clinical Excellence (NICE) Guidelines for COPD. November, 2003
- British Thoracic Society. Non-Invasive ventilation in acute respiratory failure. Thorax 2002; 57: 192–212
- Kramer N, Meyer T J, Meharg J, et al. Randomized, prospective trial of noninvasive positive pressure ventilation in acute respiratory failure. Am J Respir Crit Care Med 1995; 151: 1799–1806
- Brochard L, Mancebo J, Wysocki M, et al. Noninvasive ventilation for acute exacerbations of chronic obstructive pulmonary disease. N Engl J Med 1995; 333: 817–822
- Celikel T, Sungur M, Ceyhan B, et al. Comparison of noninvasive positive pressure ventilation with standard medical therapy in hypercapnic acute respiratory failure. Chest 1998; 114: 1636–1642
- Martin T J, Hovis J D, Costantino J P, et al. A randomized prospective evaluation of noninvasive ventilation for acute respiratory failure. Am J Respir Crit Care Med 2000; 161: 807–813
- Bott J, Carroll M P, Conway J H, et al. Randomised controlled trial of nasal ventilation in acute ventilatory failure due to chronic obstructive airways disease. Lancet 1993; 341: 1555–1557
- Plant P K, Owen J L, Elliott M W. Early use of non-invasive ventilation for acute exacerbations of chronic obstructive pulmonary disease on general respiratory wards: a multicentre randomised controlled trial. Lancet 2000; 355: 1931–1935
- Price L C, Lowe D, Hosker H, Anstey K, Pearson M, Roberts C M, The U K. National Audit 2003: impact of hospital resources and organisation of care on patient outcome following admissions for acute COPD exacerbation. Thorax 2006; 61(10)837–842, (October)
- Rubin D B. Estimating causal effects from large datasets using propensity scores. Ann Intern Med 1997; 127: 757–763
- Rosenaum P, Rubin D B. The central role of the propensity score in observational studies for causal effect. Biometrika 1983; 70: 41–55
- Plant P K, Owen J L, Elliott M W. One year period prevalence study of respiratory acidosis in acute exacerbations of COPD: implications for the provision of non-invasive ventilation and oxygen administration. Thorax 2000; 55: 550–554, (July)
- Honrubia T, Garcia Lopez F J, Franco N, et al. Noninvasive vs conventional mechanical ventilation in acute respiratory failure: a multicenter, randomized controlled trial. Chest 2005; 128: 3916–3924
- Diaz G G, Alcaraz A C, Talavera J C, et al. Noninvasive positive-pressure ventilation to treat hypercapnic coma secondary to respiratory failure. Chest 2005; 127: 952–960