Methods

We obtained hospitalization data from the Acute Decompensated Heart Failure National Registry (ADHERE) of patients hospitalized with acute HF. We linked the ADHERE data to Medicare inpatient and denominator files using methods described previously. We included patients ≥65 years of age who had registry hospitalizations for acute HF from January 1, 2001, to March 31, 2006, that were linked to fee-for-service Medicare claims data. Eligible patients lived in the United States at the time of the index admission and were enrolled in fee-for-service Medicare for ≥6 months before the index admission (n = 78,373). We excluded patients who were admitted on an elective basis (n = 5,962). Patients for whom dyspnea severity at admission was not documented in the registry (n = 16,061) and those who did not have dyspnea at admission (n = 7,734) were excluded from the analysis ( Figure 1 ). These excluded patients were generally similar to the study population ( Supplementary Table 1 ). Patients who died in the hospital, left against medical advice, or were discharged or transferred to another short-term hospital or hospice were excluded from the measurement of 30-day readmission and clinical status at discharge. Patients in the postdischarge cohort who enrolled in a Medicare managed care plan during the 30 days after discharge from the index hospitalization were excluded from the measurement of days alive and out of the hospital and Medicare payments.

Consolidated Standards of Reporting Trials diagram of the study population.

The study variable of interest was dyspnea severity on admission (at the time of initial presentation). The dyspnea severity characterization in ADHERE was based on patients’ self-reported symptom severity (dyspnea with moderate activity, dyspnea with minimal activity, or dyspnea at rest) as obtained by the clinicians directly involved in their routine clinical care and as documented in the medical record. A specific research instrument or standardized questionnaire was not used. The outcomes of interest were mortality during the index hospitalization and at 30 days after admission, length of stay and clinical status at discharge, 30-day postdischarge days alive and out of the hospital, readmission (HF and all cause), and Medicare payments. We determined all-cause mortality on the basis of death dates in the Medicare denominator files. Length of stay and in-hospital mortality were based on Medicare claims for the index hospitalization. Clinical status at discharge was categorized as asymptomatic, improved but still symptomatic, or other or unknown, as recorded in the registry on the basis of patient report. Total days alive and out of the hospital in the 30 days after discharge were determined on the basis of the date of death in the Medicare denominator files and hospitalization dates in Medicare inpatient files. We identified readmission on the basis of subsequent inpatient Medicare claims. Readmission for HF was based on subsequent inpatient claims with a primary diagnosis of HF (International Classification of Diseases, Ninth Revision, diagnosis code 428.x, 402.x1, 404.x1, or 404.x3). We calculated time to readmission as the number of days from the index discharge date to the subsequent admission date. Medicare payments in the 30 days after discharge were determined on the basis of payments for inpatient, outpatient, and carrier claims. Payments were reported in 2010 United States dollars with inflation adjustment using the Consumer Price Index for medical care.

We described baseline characteristics of the study population by dyspnea severity at admission using frequencies with percentages for categorical variables and medians with interquartile ranges for continuous variables. We tested for differences between the groups using chi-square tests for categorical variables and Kruskal-Wallis tests for continuous variables. For variables that had low rates of missingness (i.e., <5% of records), we imputed continuous variables to the overall median value, dichotomous variables to “no,” and multichotomous variables to the most frequent categorical value. For variables with >5% missingness (i.e., smoking status, B-type natriuretic peptide level, race, and the ejection fraction), we treated the missing values as a separate category.

We present the observed outcomes by dyspnea severity at admission. For in-hospital mortality and clinical status at discharge, we tested for differences between groups using chi-square tests. For length of stay, days alive out of the hospital at 30 days, and Medicare payments at 30 days, we tested for differences between groups using Kruskal-Wallis tests. For 30-day mortality, we calculated cumulative incidence on the basis of Kaplan-Meier estimates and tested for differences between groups using log-rank tests. For readmission, we calculated cumulative incidence at 30 days on the basis of estimates from the cumulative incidence function, which accounts for the competing risk of mortality, and we tested for differences between groups using Gray tests.

We estimated the unadjusted and adjusted associations between dyspnea severity at admission and the outcomes of interest. In the unadjusted model, dyspnea severity was the only predictor. In the adjusted model, we controlled for baseline covariates, medications at discharge, and the year of the index admission. With the large number of events in each analysis, there was no overfitting problem with the adjustment variables. We used a linear mixed model for days alive out of the hospital at 30 days, Cox proportional-hazard models for 30-day mortality and readmission, generalized linear mixed models with a Poisson distribution and log link for Medicare payments, and a logistic regression model for in-hospital mortality. Finally, we assessed associations between other baseline covariates and dyspnea severity using a generalized logistic regression model. We report the estimated odds ratios associated with each characteristic for comparisons of the dyspnea categories. For all models, significance tests and confidence intervals were based on robust standard errors to account for the clustering of patients by hospital. Because of the large number of comparisons in the analysis, we used a 2-tailed α level of 0.01 to establish statistical significance, and we report 99% confidence intervals. We used SAS version 9.3 (SAS Institute Inc., Cary, North Carolina) for all analyses. The institutional review board of the Duke University Health System approved the study.

Results

Of 48,616 patients with acute HF and dyspnea, 4,022 (8.3%) had dyspnea with moderate activity, 19,619 (40.3%) had dyspnea with minimal activity, and 24,975 (51.4%) had dyspnea at rest. Thus, the group with dyspnea with moderate activity represented a comparatively lower percentage of the study population. Table 1 lists the baseline characteristics of the 3 groups. Patients with dyspnea at rest had more advanced chronic kidney disease (i.e., stages 3 to 5) than patients with dyspnea with moderate activity. The dyspnea at rest group also had the highest percentage of patients with baseline systolic blood pressure ≥140 mm Hg ( Supplementary Table 2 ).

Table 2 lists the observed outcomes on the basis of dyspnea severity at admission. Figure 2 shows the cumulative incidence of 30-day mortality and readmission on the basis of dyspnea severity at admission. There was a graded increase in the cumulative incidence of mortality from patients with dyspnea with moderate activity to patients with dyspnea with minimal activity to patients with dyspnea at rest. We observed similar trends for the readmission outcomes.

Stay updated, free articles. Join our Telegram channel

Join