INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Because asthma can generally be controlled with medications, poor adherence to treatment protocols may be contributing to the recent rise in morbidity and mortality from this disease (1, 2). Although inhaled corticosteroids play a central role in the safe and efficacious control of asthma (1), adherence with these medications is often poor (1, 3, 4). These medications are time-consuming and inconvenient to use, expensive, and offer no immediate perceivable benefit.

Our hypothesis was that socioeconomic factors and certain patient-held health beliefs may be significantly associated with medication adherence. To accomplish this, we examined patient characteristics that may influence adherence with twice-daily inhaled-steroid regimens (5). These characteristics are schematically represented in Figure 1, and include sociodemographic factors (6, 7), health beliefs (5, 8), asthma severity (9, 10), patient knowledge of the function of inhaled steroids (10, 11), and the perceived adequacy of patient-clinician communications (8, 12). The health beliefs considered were perceived benefits of treatment and health locus of control, which reflects beliefs about personal control over disease outcome (5, 8, 13). Also, by monitoring adherence over time, we planned to describe temporal patterns of adherence.

View larger version (23K): [in this window] [in a new window]   Figure 1.   Patient characteristics investigated as influences on inhaled steroid adherence. ED = emergency department, ICU = intensive care unit.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Adults with moderate to severe asthma as defined by the National Heart, Lung and Blood Institute (NHBLI) criteria (2) were recruited from the outpatient clinics of the University of Connecticut Health Center (UCHC) and St. Francis Hospital and Medical Center (SFH). SFH is located in inner-city Hartford and the UCHC is situated in a Hartford suburb. All but one patient had been evaluated at least once by an asthma specialist. The inclusion criteria were: (1) a pre-bronchodilator FEV₁ of less than 80% predicted, obtained within 3 yr of study entry; (2) 15% reversibility following administration of bronchodilator; and (3) regularly prescribed inhaled steroids. All subjects were nonsmokers with less than a 10 pack-yr history of cigarette use, and were free of other significant pulmonary or cardiac diseases.

Protocol

The institutional review boards of the participating institutions approved the study protocol. After informed consent was obtained, patients were interviewed to determine their demographic characteristics, socioeconomic status, asthma severity, and current medications. For this interview a written script with standard questions was used. Following this, a health locus of control questionnaire, as described below, was administered by a trained research assistant. After the interviews, spirometry was performed on all patients according to American Thoracic Society (ATS) criteria (14).

For the 18 patients whose primary language was Spanish, all instructions, interviews, and questionnaires were administered in Spanish by a bilingual research assistant. The Spanish version of the interview was obtained by translating the entire interview into Spanish and then back into English, and comparing the two English versions.

Patients were followed prospectively for 42 d to measure adherence to a prescribed twice-daily inhaled steroid regimen. If necessary, the subject's inhaled-steroid dose schedule was converted to a twice-daily protocol without altering the total daily dose. This change was made to standardize dose scheduling, which itself can alter adherence (15). The inhaled-steroid canister was placed in a metered dose inhaler (MDI) Chronolog (Medtrac Technologies, Inc., Lakewood, CO). This electronic sleeve recorded the time and date of actuation of the medication. Each participant was told that the MDI Chronolog recorded the time and date of MDI actuation, but they were kept blind to specific study hypotheses.

At the end of the 6-wk adherence-measurement period, patients completed a final set of questionnaires designed to examine their knowledge of asthma pathophysiology and the function of their medications. The questionnaire also assessed their perception of the adequacy of patient-physician communication and of the utility of inhaled corticosteroids. Misconceptions expressed by patients were corrected after the questionnaires were completed.

Outcome: Adherence

Adherence was measured by counting actuations of prescribed inhaled corticosteroids as recorded by the electronic device. Actuations between 1:00 A.M. and 1:00 P.M. were considered A.M. dosing; actuations between 1:00 P.M. and 1:00 A.M. were considered P.M. dosing. Adherence was defined by the following formula: (number of recorded actuations  prescribed number of actuations) × 100. The last 5 wk of monitoring were analyzed. The first week was not analyzed because other investigators in other asthma studies have reported an enrollment effect. Adherence initially increases because of sensitivity to being in a study, but declines over time (16). It was anticipated that after the first week following enrollment, adherence would be more uniform.

The mean adherence for each patient was calculated by averaging daily adherence over the 70 dosing periods (5 wk × 7 d × 2 times per day). Poor adherence was defined as a mean adherence of less than 70% of the prescribed dose. This cutoff was chosen because it has been used by other investigators (19), and has been useful in our own clinical experience. When the Chronolog was returned at the end of the 42-d measurement period, it was tested to assure that it functioned properly, in order to distinguish Chronolog malfunction from inhaler nonadherence.

Two patterns of inhaler overuse have been described. Intentional overuse, or "dumping," occurs when subjects purposely actuate the Chronolog and inhaler on multiple occasions over a short time span, perhaps to make up for missing doses (20). Bursts of actuations, or "seizures," due to malfunction in the electronic device also occur, possibly from bumping of the device, washing of the mouthpiece, or impending battery failure (21). Because neither of these patterns reflects true adherence, any number exceeding four excess inhalations per patient within any dosing period was truncated to four inhalations more than the prescribed number. For example, if the prescribed dose was 3 puffs twice daily and the recorded use for one dosing period was 9 puffs, this was truncated to 7 puffs (3 + 4).

Predictors of Adherence

Demographics. Information collected from patients included age, gender, and racial-ethnic-group membership. For this analysis, racial-ethnic-group status was categorized according to the patient's own classification as African-American, Hispanic, Caucasian, or other. Minority was defined as African-American or Hispanic. The patient's primary language was recorded.

Socioeconomic status. Several measures of socioeconomic status were obtained. Educational achievement was described as the number of years of formal schooling completed, and was dichotomized as completion or failure of completion of high school. Household income was categorized as less than $10,000 annually, between $10,000 and $19,000, $20,000 through $29,000, $30,000 through $49,000, and $50,000 or more. Because of the relatively high proportion of individuals in the study who were receiving public assistance, the income variable was dichotomized to more or less than $20,000 per year. Type of health insurance was categorized as commercial, Medicare only, public assistance, or none.

Asthma severity. Since disease severity may influence adherence to medication (5, 8), we included severity variables in our analysis. Asthma severity was represented as lifetime numbers of emergency department (ED) visits, hospitalizations, intensive-care-unit (ICU) admissions, and endotracheal intubations for asthma. Also recorded were markers of recent fluctuations of disease, including the number of days of work missed over the previous year because of asthma, and night awakenings for asthma in the week before study entry. The number of days for which prednisone was required in the 6 mo before study entry, and the patient's current medications, including current use of prednisone, were recorded. The number of prescribed daily puffs as well as the number of micrograms of inhaled corticosteroid were recorded. Micrograms of inhaled corticosteroid were calculated using 40 µg per actuation for beclomethasone, 100 µg per actuation for triamcinolone, 250 µg per actuation for flunisolide, and 220 µg per actuation for fluticasone. (All subjects using fluticasone were using the 220-µg-per-actuation formulation.) Inhaled -agonist use was categorized as regularly prescribed (e.g., a regular regimen of albuterol or salmeterol), or used on an as-needed basis. Spirometry was done according to American Thoracic Society criteria (14), including measurement of FVC, FEV₁, and forced expiratory flow rate at 25% to 75% of capacity (FEF_25-75).

Patient knowledge of the function of inhaled steroids. For patients to take medications adequately, they must understand the medication's purpose (5, 8, 10, 11). Because no validated instrument was available for measuring patient knowledge of the function of inhaled steroids, a questionnaire was developed for this. The patient rated the following statements as definitely true, mostly true, don't know, most likely false, definitely false. These responses were converted to a numerical metric, with higher scores indicating more knowledge. The total score for the seven items could range from 7 (least knowledge) to 35 (most knowledge).

Inhaled steroids make airway muscles relax.

Inhaled steroids work by fighting inflammation.

Inhaled steroids work by fighting infection.

The full effect of inhaled steroids depends on regular use and may take days to work.

Inhaled steroids make you fat.

When you use inhaled steroids your wheeze or chest tightness gets better immediately.

The full effect of inhaled steroids begins immediately.

In addition, patients were asked to propose theories for why medications are not taken as prescribed. They were also asked whether they had ever experienced a severe adverse reaction to a medication that required medical treatment.

Health beliefs.

1. Health locus of control. Health locus of control refers to a patient's beliefs about factors that control health. These include self-actions (internal), the actions of "powerful" others (powerful people), the skill of physicians (doctors), or fate (chance) (13). This concept is based on the tenet that individuals who believe that they can control their own health are more likely to be adherent to a medication regimen than are those who believe that health is due to the actions of others or to luck, chance, or fate. Form C of the MHLC Scales, developed by Wallston and coworkers (13), is an 18-item instrument that is adaptable for any medical condition. Items are graded on a Likert scale of 1 to 5: (strongly disagree, disagree, neutral, agree, strongly agree). Verified with factor analysis (13), six items are used for internality, six for chance, three to examine physician control over one's health, and three for the effects of powerful others. Recently, Dupen and coworkers (22) demonstrated the construct validity and internal consistency of Form C of the MHLC scales in 186 patients. They did not find a hypothesized association between knowledge about asthma and powerful-others score, but a higher powerful-others score was modestly associated with the importance of the relationship with their doctor. The relationship between adherence and health locus of control was not investigated.
2. Attitude toward inhaled steroids. In some asthma patients, better adherence to medication has been attributed to greater patient recognition of the efficacy of the patient's medications (5, 8, 10). Because no validated instrument was available for measuring this concept, a questionnaire was developed. Attitudes toward inhaled steroids were assessed by asking the patient to rate the following statements as definitely true, mostly true, don't know, most likely false, or definitely false. Once again these responses were coded on a numerical metric of 1 to 5, with a higher score associated with greater patient perception of the efficacy of inhaled steroids. Thus, the potential range for this questionnaire is 4 to 20.

Inhaled steroids help you.

Since you have been using inhaled steroids, your asthma has gotten better.

Inhaled steroids' cost does not make their benefits worthwhile.

You find your inhaled steroids easy to use.

Patient-clinician communication. Satisfaction with the patient- physician relationship can affect adherence to medication (5, 12). Because no content-appropriate validated instrument was available, a questionnaire was developed for evaluating this characteristic. The patient's ease of communication with the clinician was examined with the following questions, which were converted to a numerical metric. Patients rated each item according to the choices definitely true, mostly true, don't know, most likely false, and definitely false, with higher scores representing better communication. The total score could range from 6 to 30.

You would feel uncomfortable telling your doctor if you forget to use your inhaled steroids.

It is easy to tell your doctor about problems with your medications.

Your doctor gave you instructions and medications you can begin to use when your asthma starts acting up.

It is easy to reach your doctor or nurse by phone.

When your asthma acts up, you call your doctor for help.

Your doctor will see you right away if your asthma is very bad.

Statistical Analysis

The outcome variable, adherence, was converted to a binary variable (above or below 70%) because its distribution was skewed and not amenable to transformations (logarithmic, square, or square root). For univariate analyses comparing individual predictors to adherence, chi-square tests were used for categorical variables. Because sample sizes were small and not normally distributed, Wilcoxon's rank-sum tests were used for continuous variables. Those showing a significant relationship to adherence in univariate analyses then were entered into a multivariate logistic-regression equation. Because several of the variables within a predictor group (Figure 1) were expected to be highly correlated with each other (e.g., those variables reflecting socioeconomic status), and because the sample size was relatively small, only one representative (the most highly associated with adherence) of each of the groups was entered into the multivariate analysis. Each questionnaire developed for this study was tested for reliability of internal consistency through the use of Cronbach's alpha coefficient. This statistic tests the internal consistency of an instrument, and how much correlation can be expected between a particular scale and all other scales with the same item number measuring the same variable. Reliability is considered acceptable for group comparison if   0.70. Change in adherence over time was estimated from the slope of the adherence-week regression equation. Morning and evening adherence within individuals was determined with paired t tests.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Of 54 patients who gave informed consent to the study, three did not complete the protocol: one did not return despite telephone calls and letters, one refused to maintain a twice-daily regimen of inhaled steroids, and one with a history of emotional problems dropped out shortly after entry, for undisclosed reasons. The Chronolog data for a fourth patient were lost after that patient failed to return to clinic for several weeks. Thus, data from 50 patients, 13 male and 37 female, were analyzed.

The mean age of the study patients was 46 ± 14 yr, with a range of 20 to 81 yr. Additional demographic characteristics of the patients are described in Table 1. Sixteen patients had less than 12 yr of formal education, and 20 were unemployed. Of the 49 patients providing information, 23 had household incomes of less than $20,000, 16 had incomes between $20,000 and $50,000, and 10 had incomes over $50,000. Spanish was the primary language of 18 patients.

Table 2 outlines asthma-severity characteristics of the 50 patients. Twenty-nine had been hospitalized for asthma, 13 at least 20 times. Thirty-eight patients had had ED visits for asthma, of whom 20 reported at least 20 visits. Of 16 patients who had ICU admissions, 13 had been intubated. For the 17 patients taking prednisone at the start of the study, doses ranged from 5 mg to 60 mg daily. Twenty-two patients had missed work or school because of asthma in the year prior to study entry. At Visit 1, two patients were not on a twice-daily inhaled-steroid schedule; one later dropped out, stating preference for a three-times-daily regimen. The dosing schedules were as follows: 2 puffs twice daily (eight patients), 3 puffs twice daily (five patients), 4 puffs twice daily (24 patients), 5 puffs twice daily (five patients), 6 puffs twice daily (four patients), 8 puffs twice daily (two patients), and 10 puffs twice daily (two patients).

Table 3 shows the mean scores of the domains of the health-locus-of-control questionnaire. Cronbach's alpha for the patient-clinician communication questionnaire was 0.74, for the patient knowledge of inhaled-steroids questionnaire 0.73, and for the patient attitudes concerning inhaled-steroids scale 0.51. 

Adherence

Mean adherence decreased by 1.6% per week, (p = 0.003) over the study period, with the greatest decline between Weeks 1 and 2 (Figure 2). Mean adherence over the last 5 wk of the study was 63% ± 38%. Individual patient adherence, partitioned into high- and low-adherence groups, is shown in Figure 3. Fifty-four percent of subjects actuated their inhalers at least 70% of the prescribed number of times and were considered adherent. Four patients did not actuate their inhaled-steroid canister at all during the final 5 wk of the study. Several patterns of inhaler nonadherence are represented in Figure 4. Some reasons suggested by patients for poor adherence were forgetfulness (40%), too busy (30%), asthma not severe enough to warrant inhaled-steroid medication (12%), no appreciable result from taking the medication (6%), too inconvenient (4%).

View larger version (12K): [in this window] [in a new window]   Figure 2.   Change in mean adherence over the study period. The bars indicate standard errors.

View larger version (14K): [in this window] [in a new window]   Figure 3.   Scatter plot of lower adherence compared with higher adherence. This graph indicates that highly adherent individuals were clustered above 70% mean adherence.

View larger version (13K): [in this window] [in a new window]   View larger version (12K): [in this window] [in a new window]   View larger version (13K): [in this window] [in a new window]   View larger version (18K): [in this window] [in a new window]   Figure 4.   Patterns of poor adherence. Patient A (top left), prescribed 4 puffs twice daily, apparently forgot to take the medication 30% of the time. Patient B (top right), also prescribed 4 puffs twice daily, had virtually complete nonadherence with the dosing regimen. Patient C (bottom left) not only apparently forgot to take the medication approximately 27% of the time, but underused the medication (i.e., 1, 2, or 3 puffs) a considerable portion of the time. Patient D (bottom right) underused and overused the medication on numerous occasions.

For 41 of the 50 patients, there was no statistically significant difference in morning as compared with evening adherence according to paired t tests. Six patients had significantly better morning adherence and three patients had significantly better evening adherence.

There were 62 instances (1.5%) of overactuation (actuating the MDI four more times than the prescribed number). One patient's Chronolog recorded 546 actuations on one occasion, 516 on another, and 129 on yet another, and 120 actuations on two occasions each. The truncation rule resulted in changing this patient's adherence score from 616% to 43%. Truncating overuse did not affect the adherence status of any other patient.

Table 4 presents the univariate relationships between the predictor variables and adherence. Age and gender were not significantly associated with adherence. Additionally, none of the clinical asthma-severity indicators or locus-of-control subscales significantly predicted adherence. Socioeconomic factors associated with poor adherence included less than 12 yr of formal education (p < 0.001), minority status (p = 0.007), household income less than $20,000 (p = 0.002), and Spanish as primary language (p = 0.005). Other cognitive factors predicting nonadherence were poor patient-clinician communication (p < 0.001) and less knowledge of inhaled steroids (p = 0.02). A lower inhaled-steroid attitude score approached significance (p = 0.06) (e.g., having less patient confidence in the benefits offered by these medications). A higher chance score on the health-locus-of-control questionnaire was associated with a lower communication score (r = 0.29, p = 0.05).

For the multivariate analysis, education was chosen as the representative of socioeconomic status (Table 4), and the other measures of socioeconomic status were all highly correlated with education (Table 5). When education, patient-clinician communication, and inhaled-steroid knowledge were entered into the logistic regression analysis, less than 12 yr of formal education (OR: 6.72; CI: 1.10 to 41.0), and poor patient-clinician communication (OR: 1.2; CI: 1.01 to 1.55) were independently associated with poor adherence.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Although inhaled steroids are the cornerstone of asthma therapy, patients tend to underuse these medications (1). The purpose of the present study was to identify patient factors associated with poor adherence to twice-daily inhaled-steroid regimens for moderate to severe asthma. Demographics, socioeconomic characteristics, asthma-severity factors, and health beliefs were related to adherence as measured by electronically monitoring inhaler use. Adherence, which gradually declined over the 6-wk study period, averaged 63 ± 38% over the final 5 wk. Lower socioeconomic status was by far the characteristic most highly associated with nonadherence. Of the socioeconomic variables analyzed, the strongest predictor was education level: patients who did not complete high school had poorer adherence. Mean adherence for high-school graduates was 89 ± 32%, and for those who did not complete high school it was 43 ± 51% (p < 0.001).

Lower socioeconomic status has been associated with poor asthma outcome (6, 7), possibly because this group is less likely to receive state-of-the-art care, including inhaled steroids (7, 23). Although our patients were prescribed inhaled steroids and were given these medications without cost, a significant number did not use them as prescribed. Such nonadherence may be another explanation for the increased asthma morbidity seen with lower socioeconomic status. The relationship of socioeconomic status to adherence with asthma therapy has been noted previously. Patients in lower socioeconomic groups were less likely to institute dust-mite-avoidance measures (24), to follow medical advice given upon discharge from the ED (25), and to fill asthma-medication prescriptions (26). To our knowledge, ours is the first study to examine the influence of socioeconomic status on electronically-monitored adherence to inhaled steroid therapy.

After socioeconomic status, the next most significant predictor of adherence was adequacy of patient-clinician communication. This factor remained significant even when included with education in a multivariate model. The quality of communication between patient and clinician has been hypothesized to be extremely important in promoting adherence (1, 12, 27), a finding supported by our study.

Noteworthy was the lack of a significant impact of health locus of control on adherence. The relationship between adherence and locus of control has been examined in adults with many chronic diseases including schizophrenia (28), renal failure (29), cystic fibrosis (30), diabetes (31), hypertension (31, 32), and chronic obstructive pulmonary disease (COPD) (31, 33), and in mothers of children with asthma (34). In some of these groups a significant relationship has been observed (29- 32), but not in others (28, 31, 33, 34). To our knowledge, ours is the first study to evaluate this construct as a predictor of adherence to inhaled-steroid therapy. Although we did not find a significant relationship, health locus of control bears further study in a larger sample of asthma patients because it could offer a point of cognitive/behavioral intervention to improve adherence.

Despite a rather wide range of clinical severity of asthma in our patient group, disease severity was not related to adherence, a finding that contrasts with those in other studies (9, 10). It has been hypothesized that when patients become more symptomatic, adherence improves (9). However, our study design examined adherence across rather than within-patients. Nonetheless, it is surprising that these patients, with severe outcomes in terms of both historical (hospitalizations, ED admissions) and current measures (prednisone requirement, night awakening) often had very poor adherence.

In addition to the overall measurements of adherence, the individual patient patterns of inhaler use deserve comment. Figure 4, Patient A, illustrates what might be considered a typical pattern of nonadherence: either the patient remembers to take the prescribed dose or forgets entirely. The patient on whom the figure is based was prescribed four inhalations twice daily, but adhered to this prescription appropriately only approximately 65% of the time, and apparently forgot to take the medication 31% of the time. In Figure 4, Patient B, the patient hardly used any of the medication during the study period. An obvious limitation of the electronic monitoring used in the study is that this method cannot distinguish simple forgetfulness from purposeful nonuse. The nonadherence pattern in Figure 4, Patient C, is distinctly different. Although the patient remembered to actuate the inhaler most of the time, consistently fewer than the prescribed 4 puffs were taken. This erratic pattern appears to be more intentional (perhaps from steroid phobia) than caused by simple forgetting. The patient illustrated in Figure 4, Patient D, had patterns of under- and overuse. This pattern could be an attempt to compensate for forgotten doses. These qualitative findings emphasize the need for accurate information on how and why patients use their medications.

The generalizability of the results of the study is limited because the sample is a convenience sample and relatively small. Nevertheless, we recruited individuals from high and low socioeconomic levels, from diverse ethnic/racial groups, and with a wide range of disease severity. Additionally, they displayed considerable variability in health-care beliefs. Although females outnumbered males in the sample, there was no significant gender difference in adherence.

Despite the relatively poor adherence found in the study, it is quite possible that we did not enroll the most nonadherent patients. First, we excluded smokers in order to avoid enrolling patients with COPD. By doing this, we may have eliminated some of the more nonadherent asthma patients. Second, all but one of our patients had been treated at least once by an asthma specialist, a factor that may have improved adherence. Third, we were unable to include patients who refused to participate and those who made an initial appointment but did not return: these too were possibly nonadherent. Additionally, we may have improved adherence because patients knew that the Chronologs recorded time and date of inhaler actuation. Nevertheless, such knowledge certainly did not produce optimal adherence in all subjects.

The problems associated with using an electronic monitoring device to measure adherence to medication have been well described by others (19, 20, 21). Despite the possibility that the battery of the device will fail, multiple artifactual actuations will occur because of jostled wires in the device, because the patient does not inhale the medication despite actuating the monitoring device, or because the patient uses another canister of the same medication but one not encased by a monitoring device, the Chronolog instruments have been shown by us and by others to provide better information than patient diaries, which are notoriously unreliable (18, 35).

A further point is that although assessment of patient-clinician communication, patient knowledge of inhaled-steroid function, and patient attitudes toward these medications are important, no validated tools were available for assessing these characteristics. We developed new instruments and used Cronbach's alpha statistic as a way of testing the internal consistency of each of the instruments, and how much correlation we could expect between our scales and all other scales with the same number of items measuring the same variable. Although the alpha statistic was acceptable for the knowledge and communication instruments, it should be improved for the attitude instrument. This last instrument needs to be revised and to have more items added. All three instruments would benefit from validation on other patient populations. The use of such instruments is useful for further study of interventions targeted at improving adherence.

In summary, the present study emphasizes the problem of nonadherence to the use of inhaled steroids, and the importance of educational level and patient-clinician communication in this nonadherence. These results indicate that more time must be allowed for focusing on adherence, especially in those patients having low educational achievement. Additionally, educational materials and written plans for patients must be at the appropriate literacy level. Methods to improve communication between patients and clinicians should be stressed. Also, better understanding of the underlying patient motivation for various patterns of nonadherence will promote better patient-clinician communication and hence better adherence.