Data Sources
Yinzhou is located in the coastal city of Ningbo in the eastern part of the People's Republic of China, with a population of approximately 1.6 million. Yinzhou Health and Family Planning Commission established a regional healthcare big data platform since 2005, by integrating electronic healthcare records (EHR) from all local hospitals and community health centers. The platform serves as a repository for EHR-based database across all healthcare facility levels, encompassing individual patient demographic characteristics, healthcare encounter records, lifestyle behaviors, as well as imaging and laboratory test results. The healthcare big data database is further linked with disease registries as well as mortality records administered by the District Centre for Disease Control and Prevention.7 Notably, diagnoses and procedures are recorded using International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM). Personal information and identifiers were anonymized to ensure privacy protection.
Study Design and Population
We conducted a retrospective cohort study by including all adult individuals with newly diagnosed HF (aged 18 years and above at HF diagnosis) between 1st January 2005 and 30th September 2022.
Eligible patients received two or more echocardiograms, with at least a three-month interval between the examinations. Conversely, patients who did not undergo two or more echocardiograms separated by at least three months or were lacking valid LVEF quantification in both echocardiograms were excluded from the study.
The present study was exempt from informed consent requirements due to the deidentified nature of the data. It was approved by the Institutional Review Board (IRB) of First Affiliated Hospital of Ningbo University (Ningbo First Hospital), and conformed to the principles of the Declaration of Helsinki (2021-R075). The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were followed throughout this study and a STROBE checklist was provided (Supplemental table 1).8
Definition of HF Types
Eligible patients with HF were identified based on the presence of the ICD-10 code I50.x, and associated terminologies such as "heart failure" or "cardiac insufficiency" in the recorded diagnosis information. Ultimately, the selected HF cohort comprised patients who had received a single HF diagnosis during their hospitalization or emergency department (ED) visit, or two such diagnoses within a 30-day interval in the outpatient setting. HF patients were categorized into three groups based on their baseline LVEF as follows: HFrEF at an LVEF of ≤ 40%; HFmrEF at an LVEF of 41–49%; and HFpEF at an LVEF of ≥ 50%.
Specifically, the baseline echocardiogram was determined as the earliest record indicative of reduced ejection fraction (EF ≤ 40%). Types and magnitude of EF improvement were determined using the subsequent echocardiogram that was at least three months apart from the baseline. Patients of varying degree of EF improvement were categorized as follows: 1) pHFrEF (subsequent EF ≤ 40%); 2) HFpimpEF, (subsequent EF was 41–49% and had EF improvement < 10%) and 3) HFimpEF (subsequent echocardiogram > 40% and had an absolute EF improvement ≥ 10%).
Baseline demographic and clinical characteristics were captured using diagnosis and procedure codes from 365 days prior to the date of index echocardiogram. Treatment period was defined as the varying time period between baseline and subsequent echocardiogram. The follow-up period was defined as after the 2nd echocardiogram 3 months apart from the baseline one. (Supplemental Fig. 1)
Clinical Outcomes
The primary outcome of the present study was composite outcome consisted of all-cause mortality and/or first HF related readmission. The study also examined an array of secondary outcomes including first all-cause readmission and worsening chronic heart failure (WCHF) events. WCHF was defined as rehospitalization or emergencies in patients with heart failure, discharging from hospitalization or ED, or the occurrence of symptom exacerbation and escalation of diuretic therapy such as intravenous diuretic treatment in outpatient with heart failure.9–11
Covariate Definitions and Measurements
Covariate set included individual sociodemographic variables, such as age, sex, and ethnicity, marital status, and insurance status. Moreover, baseline characteristics, such as age, sex, baseline EF, body mass index (BMI), marital and education status, comorbid conditions including hypertension, hyperlipidemia, diabetes mellitus, atrial fibrillation, stroke, chronic obstructive pulmonary disease (COPD), chronic kidney disease, coronary artery disease, peripheral arterial disease, anemia and myocardial infarction were defined based on International Classification of Diseases, ICD-10 diagnostic codes in the 365 days prior to the baseline echocardiogram. BMI was calculated as weight/height2 (kilograms per square meter). Further, Charlson Comorbidity Index (CCI) was defined Dr. Mary Charlson in 1987, which includes age factor and comorbidities and has been used and validated in other studies.12
Pharmacotherapy administered thought the study period, encompassing a variety of medications such as diuretics, angiotensin converting enzyme inhibitors (ACEi), angiotensin receptor blockers (ARB), angiotensin receptor-neprilysin inhibitors (ARNi), β blockers, mineralocorticoid receptor antagonists (MRA), triple therapy (defined as the combination of ACEi/ARB/ARNi, β blockers and MRA), sodium-dependent glucose transporters 2 (SGLT2), digoxin, and ivabradine. The purpose of this inquiry was to evaluate patients' receipt of GDMT defined as ever-exposure of triple therapy within a defined period.
Laboratory test results (including creatinine, estimated glomerular filtration rate (eGFR), hemoglobin, brain natriuretic peptide (BNP), NT-proBNP, sodium, potassium, and urine protein) were analyzed. The eGFR was calculated using previously published equations: eGFR (mL/min/1.73 m2) = 194 × (serum creatinine) −1.094 × (age)−0.287 (× 0.739, if female).13
Statistical Analysis
Baseline characteristics were summarized using descriptive statistics, with continuous variables expressed as means and standard deviation (SDs) for normally distributed continuous variables, medians and interquartile ranges (IQRs) for skewed continuous variables. Categorical variables presented as frequencies and percentages. In order to compare patient characteristics across different HF groups, statistical tests such as t-tests, χ2 tests, Wilcoxon tests, or rank sum tests were utilized as appropriate. Patients missing laboratory indicators were excluded from the specific indicator calculations.
Ordinal logistic regression models were used to evaluate patient characteristics across pHFrEF, HFimpEF, and HFpimpEF. Baseline individual-level patient characteristics, encompassing age, gender, concurrent comorbid conditions, and the administration of GDMT between initial and subsequent echocardiograms, were incorporated into the multivariable regression models. The findings are presented as odds ratios (OR) accompanied by 95% confidence intervals (CIs) to elucidate associations between key variables. Survival between different HF subtypes was first assessed using Kaplan-Meier curves and differences between HF subtypes was assessed using log-rank test. The association between HF subtypes and clinically meaningful outcomes was estimated in baseline HFrEF population using Cox proportional hazards model with adjustment for age, sex, baseline comorbidities and time-fixed treatment covariates in sequential model. The proportional hazard assumption was verified through visual examination of the scaled Schoenfeld residual.14
R version 4.0.5 (R Foundation for Statistical Computing) was used to conduct all statistical analysis, a two-side P-values < 0.05 considered statistically significant.