Results

Seventy seven patients fulfilled the criteria of cases. Three were excluded because they normally lived outside the Auckland region and two because they were unable to complete the interviewer administered questionnaire, one because of severe mental retardation and one because of severe schizophrenia. One was excluded because of poor English proficiency. Three potential cases were not interviewed; all left hospital after less than 48 hours and two discharged themselves. There were no refusals.

Two hundred and thirty nine matched hospital controls were recruited, representing 1–4 controls per case. Two were excluded because they normally lived outside the Auckland regions, six were excluded because of poor English proficiency, three because of mental retardation, and there was one refusal. Twelve potential controls were not interviewed because of rapid discharge from hospital, three of whom discharged themselves. One hundred subjects were recruited in the community comparison group; a further 45 declined to participate but did not differ from the total group in terms of basic demographic characteristics.

The sociodemographic data of the patients are presented in table 1. Analysis of SLTA cases and matched hospital controls showed that both age and sex were risk factors for SLTA. The risk of SLTA increased with age (odds ratio (OR) 1.04 per year, 95% confidence interval (CI) 1.01 to 1.07) and was less for women (OR 0.36, 95% CI 0.20 to 0.68). These variables were therefore controlled for in all further analyses. There were no differences in ethnicity or socioeconomic indicators between cases and hospital matched controls. However, the community comparison group were more likely to be European (p<0.01) and less socioeconomically disadvantaged (p<0.05) than the cases and matched hospital controls.

Asthma morbidity is also shown in table 1. Cases with SLTA were more likely than matched hospital controls ever to have had previous SLTA (OR 2.04, 95% CI 1.20 to 3.45) and to have been admitted to hospital for acute asthma in the last year (OR 1.86, 95% CI 1.09 to 3.18). The community comparison group had markedly lower morbidity indices than either cases or controls (p<0.001 for all).

The community comparison group was assessed to have less severe asthma than cases or hospital controls (table 2) but there were no differences between cases, hospital controls, or the community comparison group in terms of chronic prescribed drug management of asthma except for the use of oral theophylline (table 2). There were no differences between cases and hospital controls for indicators of quality of ongoing asthma specific medical care (table 2). Data on barriers to health care are shown in table 3. There were no differences between cases and hospital controls in terms of: (1) physical accessibility of health care, (2) financial barriers to health care (both general and asthma specific), and (3) quality of the doctor-patient relationship (either in terms of total factor score (data not shown) or for the individual items). (“Doctor” was defined as the medical professional whom patients considered most important in the management of their asthma.) The community comparison group had fewer financial barriers to health care but less favourable attitudes to their doctor (not statistically significant) (table 3).

Data pertaining to presentation and management of the index attack for cases and matched hospital controls are presented in table 4. As expected, cases had evidence of more severe asthma at presentation. Cases were less likely to present initially to general practitioners and were more likely to present to emergency departments or to call an ambulance (p<0.01). Difficulties encountered by the patient in the management of the index attack are shown in table 4. Cases were more likely to have perceived panic during the attack but were less likely to be concerned about time off work. Otherwise there were no differences in these parameters.

Use of oral theophylline (ongoing and in the two weeks before the index attack) was associated with a higher risk of SLTA (OR 2.14, 95% CI 1.35 to 3.68) while use of inhaled corticosteroids in the two weeks before the index attack (as opposed to prescribed inhaled corticosteroids) was associated with a reduced risk of SLTA (OR 0.69, 95% CI 0.47 to 0.99) compared with hospital controls. However, cases were less likely to have previously run out of their usual medications (48% vs 28%, p<0.01). There were no differences between cases and controls with regard to access to urgent after hours care nor in their previous management of a nocturnal attack (table 5).

There were no differences in self-management knowledge scores or self-management behaviour scores between cases and hospital controls. Cases were less likely to use standard self-management strategies, but only the lower rate of use of a peak flow meter reached statistical significance (table 5). Self-management error in terms of delayed use or non-use of oral steroids during the index attack was associated with an increased risk of SLTA (OR 2.09, 95% CI 1.02 to 4.47). Delayed or non-summoning of an ambulance was not associated with an increased risk of SLTA.

Discussion

In comparison with those admitted with acute asthma, patients with SLTA are relatively more likely to be men. However, women predominate in most statistics of asthma morbidity and mortality, making up about 60% of deaths and SLTA1 ,32-37 and a greater proportion of hospital admissions (up to 75%18), although not all series of asthma death6 and of near fatal asthma38 ,39 have shown a predominance of women. We have recently reported a predominance of men in those presenting with rapid onset asthma and who are more likely to have SLTA.29 This may be a partial explanation for the relative predominance of men in the SLTA cases. However, sex is not a particularly useful clinical criteria for identifying patients at increased risk of SLTA. Similarly, although there is a statistically significant increased risk of SLTA with increasing age, even within the limited age range studied, an odds ratio of 1.04 per year age is not likely to represent a useful discriminator for identifying patients at increased risk of SLTA or death. These factors were controlled for in all further analyses.

Previous SLTA and recent admission to hospital were risk factors for SLTA whether the comparison was made with matched hospital controls or the community comparison group. This is consistent with previous results from our group4-6 and the case-control study of Turner et al,39 which was smaller but of similar design to the current study. Risk stratification, even within the group admitted to hospital with acute severe asthma, can therefore be undertaken on the basis of previous SLTA and recent admission to hospital. This is of considerable relevance to all doctors managing patients with acute exacerbations of asthma.

Poor quality and/or fragmented ongoing health care is a potential remediable factor in asthma death6 ,9 ,10 ,13-17 and life threatening asthma.18 ,32 ,38 One of the postulated reasons for the dramatic decline in asthma morbidity and mortality in New Zealand over the last decade is an improvement in the quality of, and access to, primary health care—both ongoing and emergency care.11 ,12 In a previous cross sectional study of patients admitted to hospital with acute asthma we found evidence of generally good quality, ongoing, asthma specific medical care18 but we hypothesised that poor quality care might still be a risk factor for SLTA. However, our results showed that, compared with matched hospital controls, cases with SLTA did not differ in any of the indices of quality of ongoing medical care. Data on quality of health care were not available for the community group and the possibility still exists that deficiencies in health care may operate to increase the risk of an acute attack and admission to hospital, although not increasing the risk of SLTA per se.

Barriers to health care may take a variety of forms—for example, physical inaccessibility, financial barriers, and attitudinal factors. Inaccessibility to health care is one of the reasons why asthma deaths and SLTA are more likely to occur outside usual working hours.1 Jones and Bentham40 showed that geographical barriers to acute hospital services were an independent risk factor for asthma death, although this association was based on population statistics rather than on data of individuals with asthma. In the current study none of the geographical or organisational factors differed significantly between SLTA cases and hospital controls. Indeed, the data in table 3 argue against an important role for physical barriers in both groups. Analysis of the reasons for difficulty in managing the index attack (table 4) suggests that non-availability of a telephone or a car was not an issue for >90% of patients in both groups. However, the current study was undertaken in an urban region, albeit a sprawling one, and thus the results may not be applicable to other regions, particularly if they contain a substantial rural population.

The results of this study do not support the contention that financial barriers to health care are a risk factor for SLTA when comparison is made with hospital controls. This does not imply that financial barriers are not important and operational in patients with asthma, merely that they influence both of these groups in a similar fashion. This is shown in table 3 by the considerable differences between the community group and both cases and hospital controls in terms of general and asthma specific financial factors. We have recently demonstrated a very close relationship between hospital admissions, SLTA and asthma death and an index of socioeconomic deprivation (SED) in the Auckland region.41 There was a stronger influence of SED on the more severe undesirable adverse outcomes. Although there are no differences between cases (SLTA) and hospital controls in any of the indicators of socioeconomic status, both groups had evidence of significant socioeconomic disadvantage (consistent with results of a previous cross sectional study of patients admitted with acute asthma18).

Successful management of a chronic illness such as asthma requires the establishment of a therapeutic alliance and partnership between the patient and doctor. Deficiencies in this relationship have adverse effects on the acquisition of self-management knowledge and patient self-management behaviour during an acute attack.19 ,20The doctor's level of sensitivity to the emotional needs of the asthmatic patient may influence clinical decisions; doctors too much in tune with the patient's psychological distress tended to overtreat while doctors who paid inadequate attention to such issues discharged inappropriately early.42 In this study all indications were that there was generally an excellent relationship between the patient and the doctor regarded as most important in the management of their asthma, and no evidence that the quality of the doctor-patient relationship was an identifiable risk factor for SLTA.

Factors related to patient self-management strategies shortly before and during the index attack were significantly associated with the risk of SLTA, those with SLTA being (1) less likely to have used inhaled corticosteroids in the two weeks before the index attack, (2) more likely to have used oral theophylline, (3) less likely to have monitored peak flow, and (4) less likely to have appropriately used oral corticosteroids during the acute attack. In part, these results are consistent with our previous results which showed a high rate of errors in the management of severe attacks of asthma, the errors being generally made by the patient rather than the doctor and more likely to occur in relation to strategies that may abort the attack or be potentially life saving.20 We have also shown considerable disparity between the patients' knowledge of what to do in the event of an acute attack and what is acutely done—that is, their behaviour.19 The patients' self-management knowledge, their behaviour during an acute attack, and self-management errors are all influenced by a variety of socioeconomic and health care factors.19 ,20 ,25 Thus, while deficiencies in quality of ongoing asthma care have been cited as a cause of SLTA/asthma death,1-10 efforts directed towards improving patient self-management of acute exacerbations of asthma may be more likely to produce improved outcomes in terms of reductions in asthma mortality and severe morbidity, and should be the focus of intervention strategies and be an essential component of asthma education.43

The results of this study raise issues about the perceived role of the general practitioner in the management of acute severe asthma. Cases with SLTA were more likely to present directly to an emergency department or to call an ambulance; only 29% presented initially to the general practitioner compared with 51% of hospital controls. Similar reliance on emergency departments was noted in the study by Turner et al.39 This behaviour would seem entirely appropriate in the context of a more severe attack of asthma. However, few attacks leading to hospital are rapid (<6 hours)30 and the behaviour during the index attack is reflected in the responses to questions on the previous management of nocturnal exacerbations of asthma; only 4% of those who presented with SLTA and 7% of hospital controls had telephoned their general practitioner during previous acute episodes. Perhaps most disturbing is the fact that, on previous occasions, over 80% had not summoned any form of help nor sought any advice. Previous studies have shown that, in the event of an acute attack of asthma, there were delays in summoning emergency services20 and also delays in attendance at the emergency department; the latter was indicated by more severe asthma on presentation in Auckland than in Toronto.44 This underuse of primary health care, particularly after hours when there may be organisational and other barriers to community care, was one of the factors thought to contribute to asthma morbidity and mortality in New Zealand.1 ,9-12 These data seem to reflect a perception by the patient of a lack of role for the general practitioner in the management of acute asthma. This may be appropriate in the event of a rapid onset or very severe attack (along with the lower rate of use of a peak flow meter), but may also be influenced by the perceived availability of general practitioners after hours and of other barriers to acute health care (including financial, as primary health care is delivered on a fee-for-service basis in New Zealand). Such patient behaviour needs to be considered when providing advice to individual patients and when devising acute asthma management guidelines. Unfortunately, a similar situation seems to exist for patients admitted to hospital with chronic obstructive pulmonary disease (COPD) as only 25% had sought medical help from their general practitioner in the two weeks before admission.45

One of the major aims of asthma education is to increase patients' self-management knowledge and skills,19 ,25 although this does not necessarily translate into improved self-management behaviour.19 ,20 In this study there were no differences between cases and controls in the total scores for self-management knowledge (table 5) although scores for both groups were higher than those obtained from a community based comparison group (data not presented). This suggests that educational strategies directed solely at increasing patient knowledge are unlikely to reduce the risk of SLTA. A critical component of self-management educational initiatives is instruction on action to be undertaken in the event of an acute attack, specifically the indications for use of oral corticosteroids. Delay or non-use of oral corticosteroids was a risk factor for SLTA, thus reinforcing the importance of that single piece of advice in attempts to reduce serious asthma morbidity.

Inhaled corticosteroids are the linchpin of preventive therapy in adult asthma; these agents modulate the inflammatory response in airways,46-49 reduce the level of airway responsiveness,50-52 improve various parameters of asthma control,52-54 and reduce asthma morbidity.55Although we have previously shown that the use of all forms of asthma drugs was associated with increased risk of death or SLTA, and that confounding by severity was present, it was only for higher doses of inhaled steroids that the risk ratio (RR) for SLTA or death fell below 1 after adjustment for multiple severity indicators.5Other pharmacoepidemiological studies have reported similar findings of a reduced risk of adverse outcomes in association with the use of inhaled corticosteroids.56 ,57 Although the rate of use of inhaled corticosteroids was much higher than in the Canadian study of Turner et al,39 this study showed that it was recent use (within two weeks) and not prescribed use that was associated with the reduced risk of a serious adverse event. In view of the similarities in most other respects between the cases and matched hospital controls, these data suggest that recent use of inhaled corticosteroids may prevent SLTA. The importance of the distinction between prescribed and actual use of medications is also highlighted.

In pharmacoepidemiological studies theophylline has been consistently associated with an increased OR for serious adverse events5 ,56 ,57 and, even after adjusting for multiple severity markers, the OR remained significantly raised.5While this may be due to residual confounding (with theophylline use being a marker of poorly controlled asthma for whatever reason or inappropriate medical management), direct or indirect effects of the drug are not excluded. Whatever the mechanism and whatever the indication for administration, the results of this study support the contention that the use of oral theophylline does indicate a patient at increased risk of a serious adverse event.

The lack of information in the literature regarding the possible impact of panic by the patient during the acute attack reflects the general paucity of research in the area of patient behaviour, especially at times when self-management decisions are most crucial. High rates of anxiety, not necessarily related to acute exacerbations, are well documented in those with severe asthma19 ,43 and have been shown to be associated with higher rates of hospitalisation and greater use of “as required” medication and oral steroids.58It has been suggested that symptoms of panic/fear occurring during an attack of asthma may lead to more inappropriate action by the patient.59 ,60 While heightened awareness may help to focus the patient on making correct self-management decisions, a high level of anxiety/fear may incapacitate the patient and lead to no action rather than incorrect but deliberate action, precisely what was found to occur in this study and in previous studies.19 ,20

In this study patients admitted to hospital with an acute attack of asthma were used as controls, thus allowing interview of both cases and controls using the same procedure and controlling to a large extent for the impact of the acute exacerbation, drug treatment, and the effect of hospitalisation. Matching, apart from date of attack and hospital of presentation, was not undertaken as it was argued that some of the features traditionally matched for may be significant risk factors. As it turned out, this was entirely appropriate as both age and sex were statistically significant risk factors and were controlled for in all subsequent analyses. While the use of patients admitted acutely to hospital to a large extent controls for severity of asthma4 ,5 and the effects of acute treatment, the use of hospital controls may represent “over-matching”—that is, the cases and controls were both derived from the same subgroup of asthmatic patients and hence were unlikely to differ in measurable ways. However, the major consideration in the choice of controls was the highly relevant clinical aim to distinguish those at highest risk of SLTA (and death) from the larger group of patients admitted acutely to hospital, rather than distinction from the much larger population of community based asthmatics. In other words, we aimed to distinguish between risk factors for SLTA and risk factors for hospitalisation for acute asthma. Most patients who die of asthma or experience SLTA have had contact with hospital services in the previous year and thus present an opportunity for intervention.4-6 Furthermore, any intervention strategies to reduce morbidity and mortality would be likely be hospital based, at least initially.

It has been argued that patients with SLTA to some extent represent a population of survivors and are thus a highly biased sample.8 However, during the conduct of this study there were only four asthma deaths in this age range in the region and thus the cases in this study are a reasonably unbiased population of all patients having very severe attacks in the region.

Suitable controls in a case-control study should accurately reflect that population from which the cases were derived. However, obtaining a random sample of patients having an acute exacerbation of asthma in the community presented insurmountable logistical difficulties, not only because of the range of sources of acute health care in New Zealand but also because a substantial proportion of patients do not obtain medical help for an acute attack which occurs outside usual working hours (table 5). The conduct of an epidemiological study of asthma prevalence22 provided the opportunity to obtain a random sample of community based asthmatic subjects to provide normative data for some of the instruments used and for comparison with the cases and controls. This comparison reinforced the fact that community based asthmatic subjects differ markedly from those who are admitted to hospital with acute asthma.

Although not a primary aim, this study provides data on risk factors for admission to hospital for acute asthma; these included non-European ethnicity, socioeconomic deprivation (either general or directly related to asthma, confirming our previous results18 ,40), more severe asthma, previous SLTA, and recent admission to hospitals or presentation at the emergency department. This highlights the fact that risk factors for death (and SLTA) may differ according to the control subgroup studied, an important consideration in discussions relating to controlling for severity in the debate about the relationship between fenoterol and asthma deaths.4-6

Thus, in comparison with patients admitted to hospital with acute severe asthma, those with SLTA are (1) indistinguishable on sociodemographic criteria (apart from relative male predominance), (2) are more likely to have had previous SLTA or hospital admission in the last year, (3) have similar quality ongoing asthma care, (4) have no evidence of physical, economic, or other barriers to health care, and (5) have demonstrable deficiencies in management shortly before and during the index attack in terms of being more likely to have been using oral theophylline, less prior use of inhaled corticosteroids, less likely to have monitored PEF and to have appropriately used oral steroids during the attack.

These findings have major implications for asthma education and management. Following an acute attack, particularly one of sufficient severity to necessitate admission to hospital, it is important to review carefully the patients' self-management strategies, to identify errors particularly of omission or delays in undertaking appropriate action, and to institute corrective advice. Educational strategies, while not losing sight of the need for good quality ongoing care, need to focus more on providing individual patients with rational, relevant, and realistic advice on self-management of acute exacerbations and this advice needs to be repeatedly reinforced and refined—the “5 Rs” of asthma education.