Patient Factors and Medication Guideline Adherence Among Older Women With Asthma FREE

APPROPRIATE USE of asthma medications reduces morbidity and mortality from asthma^1- 2 and improves quality of life.³ The National Asthma Education and Prevention Program (NAEPP) has distilled the literature on the effectiveness of asthma control medications into severity-based guidelines, which were originally published in 1991⁴ and were updated in 1997.³ Adherence to these guidelines has been noted to be low, ranging from 20% to 72% among various groups in the United States.^5- 12 Studies have suggested that younger age,^5- 6,12 male sex,⁶ minority status,^5- 6 low socioeconomic status,⁷ severe asthma,^6,8 and nonspecialist care^{5- 6,9,12- 13} are associated with nonadherence; however, these studies have been limited by poor response rates,⁵ ecological⁷ and cross-sectional design,^6,9 and small sample size.¹⁰

Most studies on asthma guideline adherence have focused on children and young adults. However, asthma also is an important problem among elderly patients. Nationwide, the largest increases in asthma mortality have occurred among older women.¹⁴ Research in cardiovascular disease suggests that medication adherence may be especially low and have different causes among elderly patients.^15- 16 Similar low adherence seems to be true among elderly patients with asthma^11,17; for example, fewer than one third of 92 nonsmoking, elderly patients with asthma in the Cardiovascular Health Study received inhaled corticosteroids.¹⁸

Clark et al¹⁹ have hypothesized that physical factors, such as comorbid conditions, and a variety of psychosocial factors, including social isolation and caregiving responsibilities, may contribute to poor adherence among older patients with asthma. However, few data are available to assess these ideas. Sin and Tu¹³ recently reported that increasing age, comorbidity, and physician specialty predicted poor adherence among older patients hospitalized for asthma; however, they were unable to evaluate other factors owing to reliance on administrative data.

We therefore assessed overall adherence to NAEPP medication guidelines among older women with asthma in the Nurses' Health Study, a prospective cohort study with high follow-up. We further evaluated the impact of physical factors, such as asthma severity, and psychosocial factors, such as social isolation and caregiving, on adherence to NAEPP medication guidelines among these older women.

In 1976, the Nurses' Health Study enrolled 121 700 married, female registered nurses, aged 30 to 55 years, who resided in 1 of 11 US states and responded to a 2-page questionnaire.²⁰ Participants have been followed biennially via questionnaire to allow evaluation of a large number of lifestyle and psychosocial exposures and various disease outcomes. Follow-up of the original cohort in 1998 was greater than 90%.

All participants were asked biennially about a physician diagnosis of asthma from 1988 onward. A supplemental asthma questionnaire detailing asthma symptoms, medication use, and exacerbation-related health care use was sent in 1998 to all participants who reported a physician diagnosis of asthma through 1996 (N = 10 496) except those who died or withdrew from the study before 1998 (n = 657). Women were included in the present study if they reported a physician diagnosis of asthma on an original form and reiterated a physician diagnosis of asthma on the supplemental form 2 to 10 years later. Women were excluded if they reported a physician diagnosis of chronic obstructive pulmonary disease (COPD) or alternative pulmonary condition or if they declined to provide information on asthma severity or medication use. Validity of the case definition of asthma and a case definition of physician diagnosis of COPD has been previously established in the Nurses' Health Study.^21- 22

Asthma symptoms, severity, and medication use were assessed on the 1998 supplemental questionnaire. Other variables were ascertained by prior questionnaires sent to all participants in the Nurses' Health Study in 1992 (socioeconomic status and confidante items only), 1996, and 1998. The most recent information (ie, from 1998) was used for variables in which inferences were unlikely to be subject to reverse causality. That is, data from 1998 were used for variables such as age and menopausal status that were unlikely to be affected by adherence with asthma medications. Prospectively collected data were used for variables in which inferences might otherwise be attributable to reverse causality. For example, caregiving was ascertained in 1996 and prospectively compared with asthma medication adherence in 1998.

Asthma severity was classified according to NAEPP guidelines,³ modified slightly to match items on the supplemental questionnaire (Table 1). Classification was based on symptoms at the time of the questionnaire; respondents continued to take asthma medications at the time of symptom report. Since all participants in the Nurses' Health Study belong to the same profession and have, in most cases, a similar educational background, variability in socioeconomic status was approximated by participants' husbands' educational attainment. The number of comorbid conditions was calculated as the number of other major diseases reported by the participant, such as cardiovascular disease (including hyperlipidemia), diabetes, and cancer. The number of nonasthma regular medications included daily aspirin or other nonsteroidal anti-inflammatory drug use.

Caregiving was assessed with the question, "Outside of your employment, do you provide regular care for any of the following: disabled or ill spouse, parent, or other person, and non-ill child or grandchild?" Participants were also asked, "How stressful would you say it is to provide care to the individuals mentioned above?"

Subcategories of the Medical Outcomes Study 36-Item Short-Form Health Survey²³ (SF-36) were used to estimate depressive symptoms, vitality, social function, and role limitations due to emotional problems. These subcategories were picked for likely relevance to medication adherence. The SF-36 has been used in clinical and epidemiologic studies and is internally consistent, reliable, and predictive of health outcomes in a variety of populations.^24- 26 Subcategories were assessed and calculated using standard methods.²⁷

Adherence to NAEPP medication guidelines was estimated by medication use in the year preceding administration of the supplemental questionnaire. Participants were asked, "Which medication(s) have you taken for asthma within the past year," followed by a list of medication types along with generic and proprietary names of medications. Table 2 summarizes the NAEPP guidelines as used for classification in this study. Individuals were considered adherent if they reported medication use concordant with their asthma severity step. Long-acting bronchodilator use was defined as long-acting inhaled β-agonist, theophylline, or oral β-agonist. Dose of inhaled corticosteroid was not assessed; therefore, use of any inhaled corticosteroid satisfied our definition of adherence for mild-moderate persistent disease (Table 2). In contrast, the NAEPP guidelines recommend ascending dose of inhaled corticosteroid with worsening asthma severity.

Data are presented as proportions and means ± SDs. Proportions of participants adherent with steps 2 through 4 of the NAEPP guidelines were calculated by dividing the number taking appropriate medications by the number of participants reporting symptoms consistent with a given guideline step. These analyses excluded women who denied a physician visit in the preceding 2 years. Adherence was not estimated for step 1 (mild intermittent) because control medications are not recommended for this step.

The association between guideline adherence and other factors was examined with the χ² test. All analyses were repeated within strata of asthma severity. Clinically relevant variables, such as age, and variables associated with adherence at P≤.10 were evaluated for inclusion in multivariate logistic regression models. Logistic regression was used to model the association between guideline nonadherence and various predictors after adjustment for covariates. Continuous variables were categorized into indicator variables to avoid assumptions of linear relationships in logistic regression models. When bivariate analyses suggested that associations differed within strata of asthma severity, multiplicative interaction terms were tested in the regression models. Results of all interaction testing are reported. Odds ratios are presented with 95% confidence intervals (CIs). All P values are 2-tailed, with P<.05 considered statistically significant. Analyses were performed using SAS statistical software version 6.12 (SAS Institute, Cary, NC).

Of 9839 participants in the Nurses' Health Study who reported a physician diagnosis of asthma and were sent the supplemental asthma questionnaire, 8197 responded and 53 died, yielding overall follow-up of 84%. Among respondents, 2356 also reported COPD or other pulmonary condition. An additional 647 did not reconfirm a physician diagnosis of asthma (usually reporting a one-time episode of "asthmatic bronchitis" associated with an infection), and 87 did not provide adequate information on asthma severity or medication use. After these exclusions, 5107 participants remained in the analysis.

The characteristics of the study population are described in Table 3 and Table 4 (first columns). The mean ± SD age was 63 ± 7 years. Forty-two percent had mild intermittent asthma, 28% had mild persistent, 25% had moderate persistent, and 5% had severe persistent asthma. Ninety-two percent reported a physician visit in the preceding 2 years and 1% denied such a visit; the remaining 7% had characteristics similar to participants who reported a visit and were therefore treated as having had a visit.

Among all participants, use of asthma medications increased with asthma severity (Figure 1) (P<.001). Fifty-one percent of women with mild persistent asthma reported use of inhaled corticosteroids, compared with 53% with moderate persistent and 79% with severe persistent asthma.

Overall adherence to NAEPP guidelines among women with persistent asthma, excluding those who denied a physician visit in the preceding 2 years, was 55% (Table 3, third column). Adherence varied by NAEPP severity classification. The proportion of women with mild and moderate persistent asthma who reported taking the recommended medications was 57% and 55%, respectively, whereas only 32% of women with severe persistent asthma reported medication use consistent with NAEPP guidelines (P<.001). This decrement among the severe persistent group resulted from underuse of long-acting bronchodilators (46% reported inhaled corticosteroid and long-acting bronchodilator use) in addition to underuse of oral steroids (49% reported inhaled corticosteroid and oral corticosteroid use).

Adherence with NAEPP guidelines was inversely associated with age and current smoking (Table 3). Socioeconomic status, measured by husband's educational attainment, was positively associated with adherence, and retired women were less likely to adhere to guidelines than working women.

Asthma characteristics predicted adherence; women who developed asthma before age 18 years were less adherent than women with onset of symptoms after age 45 years, whereas histories of hospitalization and hospitalization in the preceding year were associated with higher adherence (Table 3).

Participants without comorbidities were more adherent than participants with multiple comorbid conditions (Table 3). Other physical factors, such as menopausal status, number of other medications taken, and obesity, were unrelated to adherence.

Few psychosocial factors were associated with adherence, despite the relatively large number of participants in the study. Aspects of social support, included living situation, marital status, religious heritage, and frequency of contact with a confidante, were not associated with adherence (Table 4, third column). Adherence was higher among participants with daily contact with a confidante than with no confidante (44% vs 18%, respectively; P = .01) among participants with severe persistent asthma; however, the interaction term was not statistically significant (P = .27).

Amount of care for an ill spouse was associated with considerable reductions in adherence (P = .02, Table 4). Other job-related potential stressors were not associated with adherence. Categories of well-being on the SF-36 were generally not statistically associated with guideline adherence, with the exception of role limitations due to emotional problems (P = .03).

In multivariate analyses, the strongest predictor of nonadherence was severe asthma (Table 5). Age was positively and monotonically associated with nonadherence. Age and retirement were collinear; retirement predicted nonadherence when age was removed from the model (P<.001). Other asthma characteristics were associated with nonadherence: women with more contact with the health care system because of asthma exacerbations were less likely to be nonadherent with guidelines, whereas women with younger age of onset of asthma symptoms were more likely to be nonadherent.

Socioeconomic status, smoking, and number of comorbid medical conditions remained significant in the multivariate analysis; however, number of hours of caregiving and emotional role were not. The overall explanatory power of the logistic regression model was moderate (c statistic = 0.70). Restricting the analysis to women 60 years and older, lifelong nonsmokers, and confirmed physician visits produced similar results.

In the Nurses' Health Study, a relatively socioeconomically advantaged and medically knowledgeable population of older women, adherence to NAEPP guidelines for the treatment of asthma was low. Overall adherence was comparable to results from a national sample of patients with asthma.¹² Participants with the most severe asthma were least likely to report use of the appropriate array of asthma control medications. Nonadherence was best predicted by asthma characteristics, age, and comorbid medical conditions.

Severe asthma was the strongest predictor of nonadherence in our study, matching findings in younger populations.^6,8 The much lower compliance with NAEPP guidelines among older women with severe asthma is alarming in light of a recent study¹⁴ that found that the greatest increase in asthma mortality in the United States was among older women. Although most participants reported use of inhaled corticosteroids, other long-acting medications were underused. Although we had suspected that low adherence among women with severe disease would be due to reticence on the part of participants and their physicians to use systemic corticosteroids, we found that both systemic corticosteroids and long-acting bronchodilators were underused.

We investigated predictors of adherence to NAEPP guidelines and found that older and retired women were less likely to take appropriate asthma medications compared with their younger, working colleagues. Both older age and retirement predicted nonadherence independent of asthma severity, comorbidity, and other covariates; however, these factors were collinear (overlapping) in regression models. Therefore, the independent effect of age was hard to differentiate from retirement. Age has been previously shown to predict nonadherence with inhaled corticosteroids among Canadians 65 years or older, few of whom presumably were working.¹³

Most physicians recognize that treating medical conditions appropriately becomes harder as the number of conditions increases. Comorbidities were measured and adjusted in the Canadian study and our study, and both studies found a positive association between the number of comorbid conditions and nonadherence. This relationship, however, did not account for the association of age and nonadherence.

Despite the occupational similarity of participants, variation in socioeconomic status, marked by participants' husbands' educational attainment, was independently associated with guideline adherence. This result matches findings from other studies encompassing wider socioeconomic gradients,⁷ but it did not confound the association of age and nonadherence.

Another possible explanation for the observed age gradient is increasing social isolation among older patients.¹⁹ We explored this factor and found that contact with a confidante was only related to adherence among participants with severe persistent asthma. This is a potentially important finding, but it requires confirmation given the lack of statistical significance of the interaction of isolation and asthma severity. Other measures of social support, including marital status and living situation, were not associated with adherence in our study.

Greater caregiving responsibilities also have been hypothesized to affect adherence among older patients.¹⁹ Women caring for ill spouses were more likely to be nonadherent with asthma medications; however, because of the small number of affected participants, this factor did not reach statistical significance in the multivariate model. Caregiving for other individuals, such as grandchildren, was not associated with adherence (data not shown).

Other psychosocial factors did not contribute materially to the association of age and adherence in this study. Role limitation due to emotional problems on the SF-36 was associated with nonadherence in bivariate analyses, but not in multivariate analyses. Depression is often associated with nonadherence²⁸; however, in this study, this effect was not apparent, possibly because we were not able to explicitly separate patient factors from physician practice patterns.

An alternative explanation for the age gradient in care is that older age and retirement are markers of changing insurance status. Most health plans offered to nurses cover prescription medications, whereas Medicare does not currently pay for prescription drugs.²⁹ However, a similar inverse association between age and adherence among elderly patients (≥65 years) hospitalized for asthma was noted in Ontario, where government health insurance covers medication costs for the entire elderly population.¹³ We were therefore not able to account for the age gradient in adherence with information on psychosocial factors and must conclude that physician practice patterns or patient preferences caused the strong gradient in guideline adherence with advancing age.

Interpretation of these results should consider that our criteria for fulfillment of NAEPP guidelines were lenient. A report of medication use at any time in the prior year was sufficient to fulfill criteria, as was use of any dose of inhaled corticosteroids. No penalty was levied for overuse of a medication (eg, short-acting β-agonist) or use of a medication that is not among medications recommended for management of chronic asthma (eg, ipratropium bromide). Also, we underestimated NAEPP asthma severity step by classifying participants by symptoms at treatment and without information on nocturnal symptoms, peak expiratory flow rates, and forced expiratory volume in 1 second.

This study is unique in assessing guideline adherence and its predictors among older women with asthma in a well-characterized cohort with excellent response rates and reliability. A limitation is the cross-sectional assessment of asthma severity and medication use. This design makes it possible that treatment caused women with severe disease to be classified as moderate and so on. However, guideline adherence was similar for moderate persistent asthma and mild persistent asthma, and use of all asthma control medications increased monotonically with severity. Participants in this study are not representative of all older patients with asthma; however, as noted herein, the proportion adherent was similar to that in a representative sample of Americans.

Somewhat paradoxically, participants who reported an urgent office or emergency department visit or a hospitalization were more likely to be adherent with guidelines. Adherence with asthma medication guidelines may have caused emergency department visits and hospitalizations; however, this reasoning seems unlikely given extensive documentation of the beneficial effects of inhaled corticosteroids.^1- 2 A more probable explanation is that such visits highlighted underlying asthma severity, prompting physicians to prescribe the appropriate medications and participants to take them.

In conclusion, asthma is undertreated among older women, even among health care professionals. Undertreatment is particularly marked among older women with the most severe asthma and multiple comorbid medical conditions. Physicians should make greater efforts to evaluate asthma severity and to step-up therapy when indicated. In addition, further research is needed to examine the role of physician prescribing patterns and patient preferences regarding use of asthma medications among older women.

Accepted for publication January 17, 2002.

This study was supported by grants HL-07427, PE-11001, HL-63841, HL-03533, and CA-87969 from the National Institutes of Health, Bethesda, Md.

We thank Karen Corsano, LMS, and Gary Chase, BSc, for invaluable assistance with the implementation of the study and Ichiro Kawachi, MD, PhD, for helpful comments on the manuscript.

Corresponding author and reprints: Carlos A. Camargo, Jr, MD, DrPH, Channing Laboratory, 181 Longwood Ave, Boston, MA 02115 (e-mail: carlos.camargo@channing.harvard.edu).