Pharmacy & Therapeutics- August 2008 - (Page 456) Monitoring Asthma Control frequent users of health care resources, several studies that employ such data demonstrate the unreliability of these data as a lone predictor of patient outcomes. It is anticipated that combining patient-reported outcomes (PRO) information with claims data will improve asthma management and, sub sequently, patient outcomes. By including PRO information, clinicians can adjust therapies to maintain better asthma control and reduce the risk of exacerbations. Cowie et al. A prospective cohort study of 378 adults in a university asthma program was conducted to assess the value of measuring emergency department (ED) visits as a predictor of future uncontrolled asthma.15 After one year, 73 patients (24.5%) had visited the ED because of asthma exacerbations. On retrospective analysis, these 73 patients demonstrated more self-reported lifestyle restrictions because of asthma, more hospital admissions and ED visits for asthma, and poorer asthma control than those who had not required ED asthma treatment since their entry into the cohort. The factors most strongly associated with subsequent ED treatment included waking at night, regular use of a beta2agonist, and a past history of emergency treatment or admission for asthma. The authors suggested that special attention should be given to patients whose asthma causes lifestyle restrictions and the need for asthma-related hospitalization. Johnson et al. Another group of authors used administrative claims data to compare asthma outcomes among 196 high-risk asthmatic patients, enrolled in an asthma care support program, with a matched cohort.16 The investigators looked at asthma-related hospitalizations, ED visits, and physician office visits to determine the effects of the care support program. Using propensity scores from the claims data, they observed that the number of total hospitalizations, asthmarelated hospitalizations, bed days, and ED visits was lower for the study participants than for the matched cohorts not participating in the program. This suggests that the beneficial effects of monitoring and education in an asthma diseasemanagement program might promote more informed use of these health care resources. We note, however, that models using claims-based data to determine asthma-related outcomes may be limited because of a lack of sensitivity, specificity, and consistency.13,17,18 Grana et al. These authors developed an administrative claims-based, risk-stratification model to identify high-risk and more severely ill asthma patients based on data from 54,573 members enrolled in a large, independent HMO (U.S. Healthcare).17 Using logistic regression, the model sought to predict the probability of asthma-related hospital admissions among five groups with asthma of varying severity. Using pharmacy, laboratory, and specialists claims, the researchers found that this model demonstrated a sensitivity score of 0.70 and a specificity score of 0.71; however, it also lacked consistency.17 When the authors applied logistic regression to data for the following year to predict subsequent claims, the model consistently overestimated the probability of hospital admissions. This study, in a “gatekeeper” HMO model, captured all primary care visits. On the other hand, in Preferred Provider health plans (PPOs) with no mandated primary care physician (PCP), asthma severity might be overestimated for a plan’s members because the reason for a primary care visit might not be captured. This oversight can cause clinical visits for specific diseases to go unnoticed in the claims data. Lieu et al. A model by Lieu et al., which employed health utilization claims data, identified 19% of pediatric patients who were at high risk for asthma-related adverse outcomes, defined as ED visits and hospitalizations.13 This model, based on data from 210,125 children who were members of Northern California Kaiser Permanente, showed a sensitivity of only 49% and a specificity of 84%. The authors concluded that prediction models based on computerized utilization data could identify children with asthma who might have an elevated risk for future ED visits and hospitalizations; however, they also acknowledged that such models can have limited sensitivity and specificity in actual patient populations. Li et al. In another predictive model based on claims data, Li et al. found that a prior hospitalization for asthma was the most predictive factor for future hospitalizations; the relative risk (RR) was 6.5 for a hospital admission if the patient had been admitted in the baseline year.18 The sensitivity of this factor alone in predicting subsequent hospitalizations, however, was low (33%). This risk was more accurately characterized when lung function was assessed during the hospitalization. When lung function was included in the analysis, a previously hospitalized patient whose lung function remained moderately or seriously reduced had a greater than 50% chance of being admitted again the following year. (Moderately or seriously reduced function was defined as a forced expiratory volume in 1 second [FEV1], a forced vital capacity [FVC], or an FEV1 or FVC below 60%.) The authors concluded that further research would be required before such a model can be used in clinical practice. Effectiveness of Claims plus Pharmacy Data None of these claims-based studies demonstrate an R-square for asthma as close to unity as predictive tools of other clinical conditions. The following studies, which analyzed data from various MCOs, suggest that supplementing administrative claims data with pharmacy utilization data might be more useful in identifying patients at high risk for emergency hospital care than medical claims data alone.19–21 Schatz et al. In a retrospective cohort study conducted at two California sites, patients who needed targeted intervention were identified through administrative data sources.19 The investigators stratified patient risk for ED hospital care based on patients’ use of beta-agonists: • Patients considered at high risk for emergency hospital care needed such care in the previous year or used more than 14 beta-agonist canisters or oral corticosteroids. • Patients considered at medium risk used more than 14 beta-agonist canisters or oral corticosteroids but needed no emergency or hospital care. • Patients considered at low risk used fewer than 14 betaagonist canisters, did not use oral corticosteroids, and did not need emergency or hospital care. This simple scheme proved valuable; the model was able to identify patients within each group who were at risk for emergency hospital care. continued on page 463 456 P&T® • August 2008 • Vol. 33 No. 8
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.