Comparative effectiveness of disease-modifying-drugs in elderly patients after incident hospitalization for heart failure

Abstract

Beta-blockers (BB), ACE inhibitors/angiotensin receptor blockers(ACEi/ARBs) and aldosterone antagonists (AA) do represent diseasemodifying drugs and have become the cornerstone of heart failure(HF) pharmacological therapy [1]. However, it is never safe to assumethat treatments of proven efficacy in younger, healthier patients willprovide equivalent benefit in older patients [2] but current guidelinesrarely distinguish the use of these therapies on the basis of age [3].To help clarify these issues a retrospective cohort study was under-taken byidentifyingall patients hospitalized for HF and never hospital-ized for HFin theprevious 10 years, discharged from 1 January 2009 to31December 2010in theEmilia–Romagna, 4.3 million resident inhabi-tants in the Italian region. Inclusion criteria were one of the followingICD9-CM codes as principal discharge diagnosis: 428.0, 428.1, 428.2x,428.3x, 428.4x, 428.9x, 402.01, 402.11, 402.91, 404.01, 404.03, 404.11,404.13, 404.91, 404.93, and 398.91. Exclusion criteria were: age≤75 years,residenceoutsideEmilia–Romagnaregionanddeathwithin90 days of discharge. Accordingly, the eventual HF disease-modifyingdrug was prescribed away from the known clinically “vulnerable”phase regardless of standard medical therapy [4].Theeligiblecohortwascategorizedin6groupsofpatientsbasedonadrugprescriptionwithin90daysfromhospitaldischarge:1)noBB/ACEi/ARBs/AA (“No drug”); 2) single ACEi/ARBs (“only ACEi/ARBs”); 3) singleBB (“only BB”); 4) ACEi/ARBs plus BB (“double therapy”); 5) BB plusACEi/ARBs plus AA (“triple therapy”); and 6) other different combina-tions between BB and/or ACEi/ARBs and/or AA ( “other combinations”).Data were collected from the Emilia–Romagna Region administra-tive healthcare databases, which capture information about health ser-vice utilization from the entire region. Patients' information wasretrievedfromtheregionaldatabaseofhospitaldischarges,theregionaloutpatient and inpatient drug prescription database and the regionalmortality registry. The following ATC classes were used: C07 class(BB); C09A' 'C09B' 'C09C' 'C09D class (ACEi/ARBs); and C03DA class(AA). Individuals were linked across data files with anonymous patientidentifiers and were followed for 1 year.Foreachpatient,informationonage,gender,comorbiditiesandproce-dures was retrieved from the previous 10 years of hospitalization fromindex admission. The study outcomes were 1-year all-cause mortality,HFrehospitalizationandall-causerehospitalizationfromindexdischarge.All the outcomes were calculated crude and adjusted. Continuous vari-ables were expressed as mean ± SD a nd median and compared using aK-sample equality of median test. Categorical variables were expressedas percentages and the Chi-squar e test was used for comparison.IncidenceofeventswasestimatedbytheKaplan–Meiermethodandcompared using the Log-rank test. Observations were censored at thetime of the event or at the end of follow-up. One-year adjusted risk ofevents was evaluated by using Cox's proportional hazards modelsadjusting for all variables reported in Table 1. We used a multivariatestepwise Cox regression analysis for variable selection with entry andstay probability of 0.1 (10%). Results are reported as Hazard Ratios(HR) with 95% confidence intervals.Allstatisticaltestsweretwo-sided(p-valueb0.05wasconsideredtobe significant). All analyses were performed with the SAS 9.1.3 version(Service Pack 4, SAS Institute Inc., Cary, NC, USA) (SAS Institute, Cary,NC,USA)andtheSTATA12.1(STATACorpLP,CollegeStation,TX,USA).Clinical baseline and demographic characteristics of patients areshownin Table1.After1-yearfollow-upperiodfromindexHFhospital-ization,1690(18.1%)deathand5273(56.4%)hospitalreadmissionshadoccurred. The 1-year HF-related hospital readmissions were 1686 (31%of the total hospital readmissions).