Unveiling discrepancies in heart failure diagnosis among diabetic patients: insights from hospital and primary care data

Background

Heart failure (HF) is a common and serious complication among patients with diabetes mellitus (DM). However, differences in the diagnostic patterns and clinical characteristics of HF may vary depending on the healthcare setting in which the diagnosis is firstly recorded. Understanding these discrepancies is essential for improving disease identification and continuity of care across healthcare levels.

Purpose

To identify discrepances in heart failure diagnosis among diabetic patients between hospital and primary care registries and analyze the impact of this variability.

Methods

We conducted a cross-sectional, population-based study, embedded in a prospective cohort. A total of 45,328 diabetic patients with a diagnosis of HF were identified either in hospital discharge databases (CMBD-H) or in primary care (PC) electronic health records. Patients were categorized into three groups: (1) HF diagnosis in both sources, (2) only in CMBD-H, and (3) only in PC. Demographic, clinical, and laboratory variables were compared using logistic regression models to estimate odds ratios (OR) with 95% confidence intervals (CI). The project was aproved by the local ethics committee.

Results

Of the total population, 11,691 (25.8%) had HF recorded only in hospital records (CMBD-H), and 33,637 (74.2%) only in primary care (PC).

CMBD-H-diagnosed patients were younger (72.0 ± 11.3 vs. 73.0 ± 11.2 years; p < 0.001), more often men (50.9% vs. 43.9%; p < 0.001), presented lower comorbidity burden and a lower prevalence of atrial fibrillation (4.7% vs. 7.2%; p < 0.001). Conversely, higher albumin/creatinine ratio and serum creatinine levels were observed only in CMBD-H diagnosed patients (p < 0.001). Multivariable models analysing the effect of variables in the incidence of HF showed that age, comorbidity, hypertension, atrial fibrillation, and coronary heart disease were strongly associated with HF in both data sources. Hypertension showed the strongest association (HR = 8.9 in PC vs HR = 12.2 in CMBD-H), followed by atrial fibrillation and coronary heart disease in both databases. Nevertheless, dyslipidaemia and renal dysfunction were found significant predictors only among patients diagnosed at hospital.

Conclusions

Significant differences exist between diabetic patients diagnosed with heart failure in hospital versus primary care. Although key predictors of HF were consistent across care settings, the effect sizes were greater in CMBD-H data, suggesting more severe disease. These findings reveals that data source strongly influences patient characterization. Integrating information from both care levels could improve the accuracy in HF diagnosis and management among patients with DM.For image description, please refer to the figure legend and surrounding text.