We conducted a cross-sectional study of 1995 data on hospital admissions, physician visits and medications dispensed for individual patients by linking 3 British Columbia Ministry of Health administrative databases: Pharmacare, the Medical Services Plan and Hospital Programs. All patients between 5 and 50 years of age for whom at least one prescription for a short-acting β-agonist was filled in 1995 and who were in the Pharmacare database were included in the study. Patients older than 50 years of age were excluded to limit the possibility that β-agonists were being taken for other chronic respiratory illnesses, such as chronic obstructive pulmonary disease and emphysema.
Pharmacare is the BC government's pharmaceutical reimbursement program. It provides comprehensive coverage for all senior citizens (plan A, patients 65 years of age and older), all people receiving social assistance (plan C) and families in the general population whose annual drug expenses total more than $600 (the annual deductible) (plan E). Under plan E, once a family reaches the $600 threshold, members become eligible for reimbursement of drug expenditures, and the complete drug profile of all family members (including drugs prescribed before the family reached the threshold level of expense) is added to the database. Because plan A covers only patients 65 years of age and over (i.e., outside our age criteria), only patients in plans C and E were included in this study.
The Medical Services Plan database contains the billing records of all physicians in the province and was used to ascertain the number of physician visits by each subject in the study.
The provincial Hospital Programs database captures data on all hospital admissions in the province, including urgency, primary diagnosis and length of stay.
Each short-acting β-agonist and inhaled corticosteroid formulation was uniquely identifiable in the Pharmacare database by its drug identification number, which facilitated standardization of doses. To control for differences in potency, strength and formulation, the total amount of each short-acting β-agonist and inhaled corticosteroid obtained by each patient was determined for the year, and these amounts were standardized in terms of number of 200-puff canisters of salbutamol (100 μg per puff) or equivalent dose of beclomethasone dipropionate (micrograms per day) respectively. The multidose inhaler and dry powder formulations of each medication were assumed to be equivalent. All short- acting β-agonists (salbutamol, fenoterol and terbutaline) were considered equipotent, whereas budesonide 40 μg and beclomethasone dipropionate 50 μg were considered equivalent.
The 1995 asthma guidelines9,10
defined asthma control as the use of no more than 3 puffs of short-acting β-agonist per week, which is equivalent to less than one 200-puff canister per year. Patients in this study were stratified into low and high drug use groups according to the standardized number of canisters of short-acting β-agonist obtained and the average standardized daily dose of inhaled corticosteroid. Low use of short-acting β-agonist was defined as the equivalent of no more than 4 canisters in 1 year, a level that allowed for use related to exercise- induced asthma and use of more than one canister concurrently. High use was defined as the equivalent of 9 or more canisters. Low use of inhaled corticosteroids was defined as an average daily dose equivalent to up to 100 μg beclomethasone, and high use was defined as an average daily dose of at least 400 μg. Next, each patient was classified into a subgroup on the basis of appropriate or inappropriate medication use, according to the combination of their use of β-agonist and inhaled corticosteroid. Patients with appropriate medication use were those with 4 or fewer canisters of short-acting β-agonist and at least 400 μg/day of inhaled corticosteroid (low β-agonist and high corticosteroid use). Patients with inappropriate medication use were those with 9 or more canisters of short-acting β-agonist and no more than 100 μg/day of inhaled corticosteroid (high β-agonist and low corticosteroid use). For the purposes of this study, any patient who did not fall into 1 of these 2 subgroups was excluded from the analysis.
For the purposes of this study, prescribing physicians were doctors who prescribed any asthma medication to any of the patients in the study. The number of visits to each prescribing physician by each patient was determined from the Pharmacare database by means of a unique physician identifier that is recorded for each filled prescription of any asthma medication. In addition, the total number of physicians seen by each patient during the year, regardless of whether the patient received a prescription for an asthma medication from that physician (referred to here as “all physicians”), was determined from the claims database of the Medical Services Plan.
The occurrence and frequency of respiratory-related hospital admissions were determined from the Hospital Programs database on the basis of admissions specifying ICD-9 code 08 (diseases of the respiratory system, clinical modification of International Classification of Diseases, 9th revision11
). Only admissions for which the primary reason was a respiratory-related condition were included. The number of patients admitted to hospital, the number of admissions among those who were admitted at least once and the number of admitted patients who required urgent admission (i.e., who were admitted after assessment in the emergency department) were determined.
The primary comparison of interest was between appropriate and inappropriate use of asthma medication. Student t-tests and χ2 tests were applied to evaluate baseline differences between the 2 study groups. Multivariate analyses were then conducted to assess whether appropriate use was an independent predictor of health resource utilization, with adjustment for the effect of age (in years), sex and social status (health plan type). Unadjusted and adjusted relative risks due to inappropriate use for the various outcome variables, along with 95% confidence intervals, were computed using appropriate models.
Logistic regression was used to estimate the relative risks (RRs) of admission to hospital and of urgent admission. The Poisson regression was applied to model the frequencies of admission and of urgent admission. The Poisson regression model was also used to evaluate the pattern of physician visits (the number of unique physicians in the “prescribing physicians” and “all physicians” categories). Gamma regression models (generalized linear models with gamma-distributed outcomes and logarithmic link functions) were developed to estimate the RR values for the mean number of prescriptions per prescribing physician and the mean number of visits per physician.
For each model, the RR was computed as expβ, where β was the coefficient in the regression model. For the Poisson and gamma regression models, the RR is the ratio of the mean outcome for inappropriate users to that for appropriate users. For the logistic regression models (modelling the risk of admission to hospital or emergency admission), the RR is the odds ratio for inappropriate users compared with appropriate users.