Patient education in asthma can improve outcome but cost-effectiveness has not been widely assessed. The aim of the present study was to carry out a cost-effectiveness analysis of patient education in asthmatics in a 12-month follow-up.
The authors randomly allocated 78 asthmatics to a control or intervention group after they had received ordinary outpatient consultation care. Intervention consisted of two 2-h group sessions followed by 1–2 individual sessions administered by a nurse and a physiotherapist. Self-management was emphasized. Effectiveness was expressed in terms of quality of life and forced expiratory volume in one second (FEV1). Doctor visits, days off work, dispensed pharmaceuticals, hospital admissions, travel costs, time costs and educational costs were recorded.
The control and intervention groups induced mean total costs of Norwegian Krone (NOK) 16,000 and 10,500 per patient, respectively. A 10-unit improvement in St. George's Respiratory Questionnaire total score and a 5% improvement in FEV1 in the intervention group were associated with a saving of NOK 3,400 and 4,500, respectively, compared to the control group. The number needed to educate (NNE) to make one person symptom free was associated with a saving of NOK 12,200. The results were robust to realistic changes in the assumptions upon which they were based.
It could be concluded from this study that patient education in asthmatics in a 12-month follow-up improved patient outcomes and reduced costs.
This work was supported by the Norwegian Medical Association's Fund for Quality Improvement.
The prevalence of asthma in North-America and Europe is about 5% of the adult population with strong evidence for increasing incidence 1. Asthma results in a significant burden being placed on public health budgets in these countries 2–6, and the potential for public savings is high. Health economical evaluations are therefore, strongly warranted 7, 8.
Patient education in asthmatics can improve patient outcome and has been recommended in several clinical guidelines for asthma 9–12. However, previous cost-effectiveness assessments after asthma education are rare and of limited applicability to clinical practice. Kauppinen et al. 13 reported an improvement in the forced expiratory volume in one second (FEV1) at higher costs without performing a sensitivity analysis and Neri et al. 14 described the cost of one day of admission being prevented.
The present asthma education programme has been shown, in a 12 month follow-up, to improve quality of life, level of lung function 15 and compliance to inhaled corticosteroids 16 as well as reducing GP consultation rates and absenteeism from work 17 due to asthma. Based on these data, estimated the cost-effectiveness ratios of patient education choosing St. George's Respiratory Questionnaire (SGRQ) total score, FEV1 and number needed to educate (NNE) to make one person symptom free as clinically important effectiveness measures were calculated.
To the best of the authors' knowledge no previous report on patient education in asthmatics has evaluated how many patients need to be educated in order to obtain a given effect. NNE is equivalent to the number needed to treat (NNT), that has been strongly advocated in clinical trials 18.
The study was conducted in the societal perspective, implying that all costs are accounted for, irrespective of who pays.
Subjects and methods
Seventy-eight consecutive symptomatic asthmatics were included in the study May 1994–December 1995, after having received ordinary consultation care at the outpatient chest clinic at a medium sized somatic hospital in Southern Norway. At inclusion, they gave written signed consent and were then randomly allocated to an intervention group or a control group using random number tables received from an external statistician in sealed envelopes. The patients in theintervention group first received an education programme, and were then followed by their general practitioners (GPs), while the patients in the control group were transferred directly to their GPs (fig. 1⇓). After 1 yr, all the patients were invited to a follow-up examination. The availability and organization of GP care was similar in the two groups.
Eligible asthmatics were 18–70 yrs of age, and did not suffer from any other serious disease. The primary inclusion criterion for asthmatics was a prebronchodilator FEV1 ≥80% of predicted (% pred) value 19. Furthermore, either a positive reversibility test 19, a documented 20% spontaneous variability (peak expiratory flow (PEF) or FEV1) or a positive methacholine test (PD20). A positive reversibility test was defined as ≥20% increase (FEV1 or PEF) from baseline after inhalation of 400 ng salbutamol. The inclusion rate amongst the eligible patients was 92% (78 out of 85).
None of the randomized patients had earlier received organized asthma education. The education programme consisted of two 2-h group sessions and 1–2 h of individual counselling administered both by a nurse and a physiotherapist (fig. 1⇑). Elementary pathophysiology of asthma, asthma drug mechanisms, how to cope with asthma and principles for self-management were covered. An individual self-management plan aimed atencouraging early change of medication during episodes of asthma attacks was issued. The intervention group also received a specially-made 19-page booklet co-ordinating the information given during the educational sessions.
Both Health Related Quality of Life (HRQoL) and symptom data have been recommended as important effectiveness measures in the management of asthma and have been found to correlate well with pulmonary function measures 5. Other less important alternatives of comparison, as previously published 16, 17, could have been tested, but recommendations for cost-effectiveness analyses advise unidimensional designs 20.
Health Related Quality of Life
Both HRQoL and symptom free days have also been advocated in cost-effectiveness analyses 5, 21. HRQoL was assessed at baseline applying four simple questions translated from the Omnibus interviews 22. The answers were dichotomized at the signs “▾” shown later, to reveal outcomes of clinical importance 5, 20. Wording of the questions and the preprinted alternatives for answering were: 1) The last year my chest trouble has become: Worse, unchanged or ▾ better; 2) The last two weeks I have had asthma symptoms during the day: All the time, 2–4 times a day, every day, 2–3 times a week or ▾ more seldom; 3) The last two weeks I have had asthma symptoms during the night: Waken 3 times or more, waken 1–2 times or ▾ do not wake up; 4) My asthma has the last two weeks restricted me from doing what I want: To a great extent, medium extent or ▾ little or no extent. Regarding the responses to these questions, the two groups were comparable at baseline 15. The same four questions were repeated at the 1-yr follow-up.
At the 1-yr follow-up, patients completed the SGRQ 23, a validated and disease-specific quality of life instrument consisting of 76 weighted items. A “Total” score is calculated, providing a global estimation of the patients' respiratory health. The total scores ranges 0–100, a score of 100 indicating maximum disability. A difference of four units indicates a slight clinical effect, while a difference of eight or twelve units indicates moderate or very good clinical effects, respectively. Thus, a difference of 10 units in SGRQ total score between treatment groups at the 1-yr follow-up, is of clinical significance.
Spirometry was performed prior to randomization and at 12 months follow-up by standard methods 19 using a Jaeger MasterLab Body Box (Würzburg, Germany). The patients were told to abstain from bronchodilators for 6 h before testing.
The costing was based on utilization of care and unit costs (lUS$=7 Norwegian Krone (NOK), 1994). Costs, as presented, were rounded to the nearest NOK 100 unless the original cost was <100. Exact costs were used when statistical tests were performed. Patient co-payments were included in the costs. Direct costs in this study are defined as costs borne by the healthcare system, community and family. Indirect costs are defined as productivity loss and time costs borne by the individual, family, society or by the employer.
The cost of asthma education included patient co-payments and reimbursement costs according to the fee schedule of the National Health Insurance (NHI) covering both group sessions and individual sessions. Market prices were applied for the Peak Flow meter, the cost of premises and the patient brochure.
The cost of hospital care for asthma was based on theNorwegian Diagnosis Related Groups (DRG) reimbursement rates 24, assuming that all admissions fell in diagnosis group 4 24, rate number 97 (NOK 10,880 per stay).
The patients reported monthly GP visits due to asthma 17, using standardized registration forms and self-addressed, stamped envelopes. Detailed information on date, time and duration of consultation, name of the doctor, pulmonary function tests and personal costs by the patient was received, and the costs set according to the NHI fee schedule 25. The public block grant for GP clinics was included, at the time NOK 365 100·yr−1, which with an average consultation rate of 4,010 consultations·yr−1 26, assuming each patient consultation amounted to NOK 91. Pulmonary consultant visits were costed according to the NHI fee schedule 25. The cost of pharmaceuticals according to the Anatomical Therapeutic Chemical (ATC) Classification Index, group R03 and H02A B 27, was based on monthly reports from the local pharmacies recording current market prices for dispensed medication.
The cost of time for those employed was set equal to the national hourly wage rate (NOK 106.2) 28. Time costs were applied for both seeking healthcare and asthma education, assuming 2 and 2.5 h (including waiting and travel time) for GP and pulmonary physician visits, respectively. The individual educational sessions at the same day were assumed to last for 1 h (plus 1.5 h travel time) including 40 min individual nurse and physiotherapist sessions, thereby adjusting for 20 min waiting time per session. The cost of lost leisure for those not employed was assumed to be zero, but alternative assumptions were explored in the sensitivity analyses.
The number of days absent from work due to asthma were reported monthly, and valued according to the national average daily wage rate (NOK 567) 28.
Patients' travel costs in relation to medical care and asthma medication were calculated on the basis of bus fares from the patients' residence.
No discounting was undertaken since the costs and consequences were contained within 1 yr.
Cost-effectiveness analyses relate monetary costs to a measure of clinical effectiveness. A cost-effectiveness ratio is the difference in costs between two alternatives to the difference in effectiveness between the same two alternatives (δ costs/δ effects). A positive ratio implies a positive effect at a certain cost. A negative numerator and a positive denominator imply a positive effect at a lower cost, a so-called dominant strategy 29. An adjusted cost-effectiveness ratio relates to the cost per outcome of clinical significance, for example the cost per 10-unit improvement in SGRQ total score or 5% improvement in FEV1.
One-way sensitivity analyses were undertaken to explore how robust the results were to changes in the basic case assumptions, i.e. the assumptions upon which they were based. The low estimate for production loss was calculated to be hourly wage-rate times 70% and the high estimate times 130% 30, thereby accounting for example, for social costs (employer tax etc.)
A threshold analysis on the most important cost assumption (production loss) was undertaken in order to identify the “break even” point, i.e. the value that made the total costs in the two groups equal 31.
Regarding the binary outcomes from the four quality of life questions, the concept of NNT which is the estimated number of patients who need to be treated with a new treatment rather than the standard treatment, for one additional patient to benefit was applied 18. In this study this expression was changed to NNE. NNE is the reciprocal value of Absolute Risk Reduction (ARR) i.e. 1/ARR. ARR=Pi−Pc, where Pi is the proportion with the desirable outcome in the intervention group and Pc is the proportion in the control group. For example the NNE to make one person experience “A better year” is calculated as follows: NNE=1/(0.81−0.43)=1/0.38=2.63.
The 95% confidence interval (CI) for the difference between two population proportions (ARR) were constructed around the difference between the observed proportions Pi and Pc in the two samples of sizes ni and nc. The standard error of this difference=square root [(Piqi/ni)+Pcqc/nc)] where qi=l−pi and qc=l−Pc 18.
The Chi-squared test or Fisher's exact test was applied for binary outcomes, while the Mann-Whitney U-test was applied for the cost data due to skewedness.
A cost/concentration curve shows the cumulative costs plotted against the cumulative number of patients, starting with the patient with the highest costs and ending with the patient with the lowest cost.
All analyses were performed using the SPSS version 8.0 (SPSS Inc, Chicago, USA). The Norwegian Data Inspectorate and the regional ethical committee approved the study.
The study groups were comparable at baseline (table 1⇓), including use of medication as previously reported 16, 96% were on inhalation steroids at randomization and two-thirds of the asthmatics were employed. At follow-up, the SGRQ total score was 16.3 units lower (HRQoL was better) in the educated group, compared to controls (p<0.001). The lung function (FEV1) improved by 6.1% in the intervention group compared to the noneducated (p<0.05, table 2⇓). There was a need to educate 2.6, 2.2, 2.9 and 3.9 persons for one person to experience a better year with regards to the chest condition, symptom free days, symptom free nights and to have no impact in daily life, respectively (all p-values ≤0.02, fig. 2⇓).
The mean direct cost in the educated group was NOK 5,900, the indirect costs NOK 4,600, in total NOK 10,500 (table 3⇓). The corresponding costs for the control group were NOK 4,000, 11,900 and 16,000. Reduced absenteeism from work in the educated group explained most of the difference in total costs. The production loss accounted for 75% and 32% of the total costs in the control and intervention groups, respectively.
There was an uneven distribution of costs. According to the cost/concentration curve (fig. 3⇓), the most costly patient in each group accounted for more than 30% of the total group costs. For the educated group, 20% of the most resource intensive patients accounted for about 60% of the total costs compared to about 75% in the control group. Nine and 23% induced costs higher than NOK 15,000 (p=0.11, Fisher's exact test) in the education and control group, respectively.
According to the baseline case assumptions, the total costs were lower in the educated group than in the control group. This implies that asthma education incurred lower costs and better outcomes. If indirect costs were disregarded, the costs were higher in the educated group (table 3⇑). With all costs included, the cost-effectiveness ratios were NOK −3,400 per 10 unit improvement in SGRQ total score, NOK −4,500 per 5% improvement in FEV1 and NOK −12,200 per symptom free patient (table 4⇓).
The direct and total cost differences did not vary overtly when choosing the most unfavourable assumptions for the variables (table 5⇓). Since production loss had the largest impact on the total cost difference, the corresponding low and high production loss estimates also had the greatest influence on the sensitivity analyses, varying, the total cost difference between NOK 3,000 and 8,000 (table 5⇓). If the cost of production loss was set to 25% of what was assumed in the baseline case analysis, the two patient groups would have the same average costs per patient (NOK 7,000).
The present study indicates that patient education in mild to moderate asthmatics improved health, and at the same time saved costs. A 10-unit improvement in SGRQ total score and a 5% improvement in FEV1 in the intervention group were associated with a saving of NOK 3,400 and 4,500, respectively, compared to the control group during a 12 month follow-up. The NNE to make one person symptom free was 2.2 and for each patient becoming symptom free after patient education there was a concomitant saving of NOK 12,200. A one-way sensitivity analyses indicated satisfactory robustness of the main conclusions.
To the best of the authors' knowledge the present study is the first to assess cost-effectiveness ratios of patient education in asthmatics referring to HRQoL, lung function and NNE and to report better outcomes at lower costs; a negative cost-effectiveness ratio. Generally in medicine, there is a tendency to be accustomed to positive cost-effectiveness ratios; better outcomes at higher costs. Previously, Kauppinen et al. 13 reported a 5.5% improvement in FEV1 at 17% higher costs in the intervention group 12 months after patient education, compared to controls.
The assumption that the last three specific quality of life questions reflect an average of the study period may be opposed, since the authors asked for the occurrence of asthma symptoms and impact on daily life during the last two weeks. However, in the SGRQ, several of the subscore components also relate to the patient's current state. It is likely that the last two weeks represent a typical average for the study period. Furthermore, no differentiation was carried out in the questionnaire between those reporting symptoms <2–3 times per week and those being symptom free. In figure 2⇑ these were all interpreted as being “symptom free”. Thus, the data presented on daily symptoms, might be an overestimate with regard to symptom control, but in clinical practice many physicians would argue that having symptoms <2–3 times per week is a satisfactory symptomatic relief in asthmatics.
A limitation of the present study is that the SGRQ scores were only obtained at the 12-month follow-up as opposed to FEV1 and the four asthma symptoms, which were also recorded at baseline. Hence, it was not known whether there was a drop or increase in SGRQ scores in the control group from baseline to follow-up. The difference in SGRQ scores is reported as an improvement, assuming comparable quality of life at baseline. This assumption is supported by the fact that the intervention and control groups were comparable at baseline for all variables recorded 15–17. Nevertheless, the cost-effectiveness analysis of the SGRQ scores should be interpreted with caution.
Other methodological issues concerning bias and confounding have been thoroughly discussed and statistically tested, when possible, in previous reports from the same study 15–17. The skewed distribution of costs implies that few persons accounted for a large impact on average costs (fig. 3⇑). This observation seems common when considering cost evaluations for asthma 13, 14, 32, 33. For example, within the USA, 10% of asthmatic patients accounted for 44% of the total medical costs of treating the condition 34.
Regarding cost-effectiveness analyses in patient education, an interesting topic would be how to maintain effectiveness while reducing educational costs. The authors were not able to pinpoint which part of the educational package could be excluded in order to obtain this. With more knowledge on how to educate asthmatics in relation to age, asthma severity, intellectual capacity, educational status, personality etc cost-effectiveness ratios of personally designed education programmes, for example based on pre-intervention testing, should become an important topic for future research.
The authors believe the introduction of the concept NNE is quite reasonable and beneficial. NNE has the advantage of conveying both statistical and clinical significance and it identifies the clinical effort required to attain a beneficial outcome. In comparison, the much used odds ratio is not helpful in clinical decision-making. The low NNE-figures reported here are powerful, especially compared to high NNT-figures often reported in trials testing the effects of medication.
Statistically significant differences in costs regarding GP visits and absenteeism from work, but not for the summed cost items (table 3⇑) were found. A possible explanation for this could be that the statistical assumptions for nonparametric tests of comparable distributions of costs in the two groups, were not met. Besides this, statistical power could of course have been better with larger study populations. However, in cost analyses, statistically significant differences are not critical, as the most important factor is whether the precision around the estimates of costs is sufficient for the decision in hand 35. Thus, sensitivity analyses were performed, testing reasonable low and high estimates for the assumptions without overt changes in the main conclusions. The robustness of the results are supported by the threshold analysis, revealing the low cost estimates for absenteeism from work needed (25%), to make total costs comparable in the treatment groups. There is no simple answer as to why the present asthma education study is the first to demonstrate a dominant strategy. The most likely explanation is that the treatment plan, aimed at immediate change of medication at early symptoms/signs of exacerbation, might well influence the difference in production loss.
The authors would like to thank I.S. Kristiansen, University of Odense, Denmark and health economist H. Husbyn, GlaxoWellcome, Oslo, Norway, for their consistent support and guidance through the maturation of this paper.
- Received April 21, 2000.
- Accepted October 12, 2000.
- © ERS Journals Ltd