Using Quality-Control Analysis of Peak Expiratory Flow Recordings To Guide Therapy for Asthma

Methods

The data for analysis were collected from patients attending an asthma management and education program. Adult patients with asthma who were enrolled in the John Hunter Hospital Asthma Management Service and who kept peak expiratory flow diaries were eligible for entry into the study. The Asthma Management Service is a standardized education and management program that is offered to adults who have had an emergency presentation to the hospital with asthma. Patients are seen in an ambulatory care setting four to five times in a 3-month period. The program involves visits with a respiratory physician and a nurse educator. Its aims are to optimize asthma control; to provide instruction in asthma management skills, such as inhaler technique; to improve knowledge of asthma and asthma medication; and to instruct patients in the self-monitoring of symptoms and peak expiratory flow. Patients used a mini-Wright peak expiratory flow meter (Clement Clarke International, United Kingdom) and recorded the best of three values obtained before and 15 minutes after inhaled bronchodilator therapy in the morning and in the evening. Values were recorded in a daily diary, which was reviewed at clinic visits and collected when the patient was discharged from the Asthma Management Service. Between scheduled visits, if symptoms worsened, patients could contact a nurse educator to obtain an earlier physician review appointment. At the completion of the program, action plans were written for the patients and the patients were discharged back into the care of the physician who had referred them to the program.

The records of 150 patients registered in the Asthma Management Service were screened for exacerbations of asthma. Exacerbations were defined as new medication courses of increased corticosteroid therapy (oral or high-dose inhaled corticosteroid). These courses were prescribed after assessment by a respiratory physician and used for deteriorating asthma as reflected by an increase in asthma symptoms, an increase in β₂-agonist requirements, and a decrease in lung function. Forty-three exacerbations were identified in 35 patients. Diary data were extracted for three time periods: an exacerbation period, a baseline period, and a pre-exacerbation period. The 3-day exacerbation period comprised the day before a corticosteroid course was started, the day it was started, and the day after it was started. A stable baseline period was identified after a scheduled clinic visit that occurred more than 6 weeks after hospital discharge and at which therapy was not altered. Diary data for the 8 to 10 days after this visit were used as baseline data. A preexacerbation period was defined as the 7-day period immediately before the exacerbation period.

Patient demographic characteristics, treatment details, and diary data were extracted using standardized forms and entered into a computerized database. Predicted peak expiratory flow values were taken from published guidelines [3]. The best peak expiratory flow for an individual person was the highest peak flow recorded in the person's diary during the stable baseline period.

Statistical Analysis

Quality-control analysis was done using the statistical process control procedures in Minitab statistical software, release 8 (State College, Pennsylvania). Control charts were used to study variations in peak expiratory flow over time. A summary statistic, such as the mean peak expiratory flow, was calculated for each sample (day) and plotted over time (in days). Three lines were drawn on the chart: the center line, which was an estimate of the average value of the summary statistic; a lower control limit, which was drawn 3 standard deviations below the center line; and a third line, which was drawn 2 standard deviations below the center line. If a process is in control it is very unlikely (< 3 in a 100 chance) that a point will fall outside the lower control limit. In this study, we used the lower control limit to define significant decreases in peak expiratory flow. Standard quality-control tests were used to identify deviations in the control charts. A single point falling below the lower control limit indicated that a change may have occurred (special cause) and that investigation was needed (test 1). Two other standard tests were used. Test 2 was reached when 2 of 3 points occurred in a row in a zone between 2 and 3 standard deviations from the center line. Test 3 was reached when four of five points in a row fell between 1 and 2 standard deviations from the center line or beyond (Figure 1).

View larger version:

• In this window
• In a new window

Figure 1. The mean daily peak expiratory flow is plotted on the y-axis. The lower control limit (LCL) is defined by quality-control analysis as 3 standard deviations below the mean peak flow value for the baseline period and is shown as a solid line. An asthma exacerbation was detected by conventional methods on day 18. The quality-control action points (tests 1, 2, and 3) were reached several days earlier. The day that the quality-control action point was first reached is shown by the number of that action point. For example, test 1 was first reached on day 16. Test 2 was first reached on day 13, which was 5 days before the detection of exacerbation by conventional methods. An example of a quality-control chart (x-bar chart) of peak expiratory flow recordings from a patient with asthma.

Action points were obtained from published literature and from quality-control analysis. The published action points that were evaluated included a peak expiratory flow of less than 80% of predicted peak expiratory flow; a peak expiratory flow of less than 80% of a patient's best peak flow; a peak flow of less than 60% of predicted peak expiratory flow; a peak flow of less than 60% of a patient's best peak flow; nocturnal waking because of asthma; and use of β₂-agonist therapy more than four times a day. Three tests for special causes were used to define action points from quality-control analysis and were separately applied to peak flow before and after bronchodilator therapy. An action point was defined as a success if it was reached by the patient during the exacerbation period. An action point was defined as a failure if it was reached during the baseline period or if it was not reached during an exacerbation. The successes and failures were totaled for the study group. The success rate is similar to the “sensitivity” of a diagnostic test. The failure rate from each action point during the baseline period is similar to the “false-positive rate,” and its complement is specificity. The McNemar test was used to compare the overall error rate (successes and failures) of two published action points (peak flow less than 60% of that predicted and peak flow less than 80% of that predicted) with the overall error rate from quality-control analysis, test 2. The significance level was P < 0.05.

Results

Thirty-five patients had a total of 43 asthma exacerbations (Table 1). Two patients required hospitalization for their exacerbations. Most patients (69%) received oral prednisolone for management (mean dose, 40 ± 14 mg; mode, 50 mg). Each patient also received increased aerosol bronchodilator therapy. High-dose inhaled corticosteroid therapy was given to 94% of patients and was continued for as long as 14 days. The average dose of inhaled corticosteroid used during the exacerbations was 3.9 ± 1.9 mg of either beclomethasone dipropionate (through pressurized metered-dose inhaler and valved holding chamber) or budesonide (through Turbuhaler [Astra Pharmaceuticals, North Ryde, Australia], a dry-powder metered-dose inhaler).

The performance characteristics of the action points are shown in Table 2 and Figure 2. The action points from published guidelines had varying success rates. An action point consisting of a peak expiratory flow less than 60% of the patient's predicted peak expiratory flow had a poor success rate; it was positive in only 35% of exacerbations. An action point consisting of a peak flow value less than 80% of a patient's best peak flow value improved the success rate to 88%. However, a high failure rate was associated with these action points. Peak flow values less than 80% of the patient's best peak flow occurred during the baseline period in 47% of patients. If these patients were to follow these action plans, they would increase therapy during stable periods.

View larger version:

• In this window
• In a new window

Figure 2. The action point derived using quality-control analysis (test 2) was significantly better than published peak flow points (PEF < 80% of that predicted and PEF < 60% of that predicted; < 0.002). Performance characteristics of published action points (●) and action points derived from quality-control analysis of peak expiratory flow (PEF) records (*).P

Symptom-based action points using nocturnal waking or frequency of β₂-agonist use also did poorly (Table 2). Many patients had nocturnal waking during their stable baseline period (Table 1); thus, use of this action point would lead to unnecessary treatment (a false-positive result). Similarly, β₂-agonists were used more than four times per day by more than half of the patients at some time during the baseline period. These symptom-based action points had poor specificity for an asthma exacerbation.

When quality-control methods were used to define action points, the operating characteristics improved considerably. The action points defined using quality-control analysis were more discriminating for asthma exacerbations than the published action points. Quality-control action points had good success with a low failure rate. Test 2, which showed positive results in 91% of exacerbations, had the best characteristics. The false-positive rate of test 2 was 23%; this was significantly less than the false-positive rate that occurred with an action point that consisted of a peak flow less than 80% of the best peak flow (P = 0.002).

We examined diary data from the pre-exacerbation period to identify whether quality-control analysis could detect the exacerbation at an earlier stage. Test 1 was positive in the pre-exacerbation period in 31 of 43 exacerbations (72%). This test showed positive results an average of 4.5 ± 2.2 days before the exacerbation. Test 2 showed positive results before eight exacerbations (19%) an average of 2.5 ± 1.3 days earlier. Test 3 showed positive results before exacerbations in 26 patients (60%) an average of 3.9 ± 1.6 days before the exacerbation.

Discussion

We have shown that currently recommended action plans poorly detect asthma exacerbations. Although some of the recommended action points were sensitive for the detection of exacerbations, they had high false-positive rates. The operating characteristics of peak flow monitoring were improved considerably by the use of quality-control techniques. This method gave a high success rate for the detection of exacerbations and had a low failure rate. It also allowed the earlier detection of exacerbations.

The quality of our study can be examined by using published standards for the assessment of diagnostic tests [11]. The action points were compared with an independent outcome, which was treatment of an exacerbation. The quality-control test results were not available to the clinicians who managed the patients and therefore were assessed independently of the diagnosis of the exacerbation. Clinicians had access to the patient's peak flow diary and could potentially calculate the action points taken from the published literature themselves. If this bias were operating, it would tend to improve the performance characteristics of the published action points. Because these points did poorly, we do not think that this bias invalidated the results. Rather, our results are an optimistic assessment of currently recommended criteria.

Each patient in our study had recently had an emergency presentation for asthma. This indicates that the spectrum of disease severity in the study sample was restricted to moderate to severe asthma. Because such patients are at high risk for future severe exacerbations or for death from asthma [12], the use of action plans in this population is of crucial importance. Therefore, although the spectrum of disease severity is narrow, the group being studied represents a clinically relevant population in whom action plans should be used. The published action points would probably not do any better in patients with mild exacerbations of asthma; our previous work with mild asthma exacerbations found peak flow changes of more than 20% to be insensitive markers of exacerbations [13]. Quality-control action points should be suitable for patients with mild asthma because they are individualized on the basis of the patient's own degree of peak flow variability.

We defined asthma exacerbations as courses of prednisolone or increased inhaled corticosteroid therapy. This definition was chosen for several reasons. If an action point is successfully applied, then the outcome is a course of increased corticosteroid therapy. Action points are intended to define when such courses of therapy should commence; hence, comparing action points against their intended outcome is appropriate. Such courses can easily be verified from patient or clinic records. Previous studies have defined asthma exacerbations in terms of symptom worsening or increasing airflow obstruction. We avoided using symptoms and peak flow in the definition of exacerbations because the same data would have been used to define the study variable and the outcome variable. This would have confounded interpretation of the results and falsely elevated the performance of action points because the data would have been highly correlated.

Previous studies of peak flow monitoring in asthma have not readily shown this monitoring to be beneficial [6, 10, 14]. This may relate to the way the information was used, particularly to the lack of sensitivity of published action points. We reasoned that individualized action points would be more appropriate. The peak expiratory flow chart resembles data collected to monitor production quality in industry. The patient's own peak flow recordings during a stable period were used to define the inherent variability of his or her disease. In this way, control limits could be set that incorporated the severity of an individual patient's asthma. Quality-control analysis was able to detect more exacerbations and to do so at an earlier stage. The specificity of action points remains a problem. Although fewer false-positive exacerbations occurred with quality-control analysis, the false-positive rate was still about 23%. This needs to be addressed in future studies: The false-positive rate could be decreased by extending the baseline observation period; incorporating symptom severity scores into the definition of action points [13]; using likelihood ratios and clinical decision making to estimate the probability of an exacerbation; or monitoring FEV₁ in addition to peak expiratory flow [15]. Long-term compliance with peak flow monitoring may also be relevant to the clinical application of these methods.

There are many opportunities to apply quality-control analysis to patient management. A health professional could calculate control limits and derive action points for a patient. These action points could then be marked on the patient's flow diary, and the patient could be instructed to increase treatment once the peak flow had passed the control limit. Action points could be calculated in several ways. Besides using a calculator to estimate mean and standard deviation, a health professional could estimate the lower control limit using a nomogram. The range of peak flow values (highest to lowest) during the baseline period is used to estimate variability and to set the lower control limit. There are also opportunities to use a computer (either desktop or hand-held) to enter peak flow data and to automatically set action points. Long-term compliance with peak flow monitoring may be the greatest limitation of these methods.

In summary, individualized action points can be derived for patients with asthma by applying quality-control analysis to peak flow records. These action points are both sensitive and specific for asthma exacerbations. They consistently perform better than published action points, and they allow early detection of an exacerbation. The error associated with published action points brings their usefulness into question. Patients with asthma should receive individualized action plans, and quality-control analysis should be used to define precise action points for use in asthma management plans.

Dr. Wlodarczyk: Newcastle Environmental Toxicology Unit, Royal Club Building, Level 1, Royal Newcastle Hospital, PO Box 664J, Newcastle 2300, New South Wales, Australia.

Dr. Hensley: Department of Medicine, John Hunter Hospital, Locked Bag 1, Hunter Mail Exchange, Newcastle 2310, New South Wales, Australia.