HF diagnoses were obtained from either a) the ED database (International Classification of Primary Care, second edition, K77) or b) the Minimum Data Set for Admission and Emergency Care (International Classification of Diseases, 10th Revision, clinical codes I11.0*, I13.0*, I13.2*, and I50*).14,15 The 3 LVEF categories are defined as follows: HFrEF for LVEF ≤ 40%, HFpEF for LVEF ≥ 50%, and HFmrEF for LVEF between 41% and 49%.7

Information on events and HCRU was obtained from the hospital's patient management information system, which combines electronic health records and administrative data. Mortality and HCRU data for the year after ED or hospital discharge were obtained from the Community of Madrid Center for Basic Strategic Health Care Information (CIBELES), which records vital status of the population and all-cause mortality data. CIBELES forms part of the information system of the Madrid Regional Health Service (SERMAS).

The index discharge event was defined as the ED or hospital discharge prompting the inclusion of the patient in the analysis. An ED visit was defined as any medical attention in the ED. A hospital stay was defined as any time spent in a hospital unit lasting beyond midnight on the day of admission. Rehospitalizations included readmissions for any cause after the index hospitalization. Admissions subsequent to direct discharge to home from the ED were classed as first admissions, not readmissions. Outpatient visits were defined as any medical, nursing, rehabilitation, or day-hospital consultation. Follow-up time points were 30 days and 1 year after the index discharge.

For cost calculations and the representation of clinical results and HCRU, we used the COHERENT model.13,16 COHERENT is a graphical system for the visual display and comparison of composite endpoints. The system consists of a hierarchical code that defines a mutually exclusive list of potentially relevant clinical situations, including patient clinical status (alive or dead) and location (at home, in the ED, admitted to hospital, or in an outpatient consultation). COHERENT is linked to a cost calculator that computes resource utilization over the observation period for each patient and for the whole study population.16

Costs were calculated as described previously,16,17 using data provided by the Hospital Universitario 12 de Octubre accounts department. The cost of each event was calculated by summing the patient-level unit costs of each product or activity (eg, accommodation, hospital food service, drugs, devices), the unit costs of each product or activity included in the hospital service catalogue (eg, analysis, diagnostic and therapeutic interventions, surgery times, stays in the postoperative recovery unit), and residual costs not directly imputable to the patient or event. Residual costs were linked to the event cost by indirect imputation. The costing of the different patient care processes was based on the patient-level hospital-activity costs study initiative of the Spanish Hospital Costs Network (RECH), of which our hospital is a member.17

Categorical variables are presented as absolute frequency (No.) and relative frequency (%), whereas continuous variables are presented as mean±standard deviation or as median [interquartile range]. For continuous variables, differences were compared by the Student t and Mann-Whitney U tests, analysis of variance, or the Kruskal-Wallis test. For categorical variables, differences were compared by the chi-square or Fisher exact test. Costs are presented as absolute spending in euros and as the percentage of the total cost as a function of follow-up time, patient care unit, and clinical event, and as the median cost per patient and day. Multiple logistic regression models were used to identify variables that predict higher cost in the total cohort and in the 3 LVEF groups. Statistical calculations were performed in R version 4.0.3 (The R Foundation for Statistical Computing, Austria). Differences were considered statistically significant at P <.05.

The study complied with the Declaration of Helsinki and was approved by the local ethics committee.

Between January 1 and December 31, 2018, 93 962 medical visits made by 66 551 patients were recorded. A primary diagnosis of HF was recorded in 1378 patients, with LVEF records available for 1174 (85.2%). Of these patients, 3 (0.3%) died in the ED, 365 (31.1%) were discharged, and 806 (68.7%) were admitted. An additional 127 patients without a primary diagnosis of HF in the ED had had received the diagnosis at the time of discharge after hospitalization, and LVEF records were available for 113 (90.0%) of these patients (figure 1 and figure 1 of the supplementary data). The study cohort thus included 1287 patients, of whom 190 (14.7%) had HFrEF, 146 (11.4%) HFmrEF, and 654 (71.2%) HFpEF. Baseline patient characteristics are shown in table 1. The patients with HFpEF were on average older, and a higher proportion of them were women. The HFpEF patient group also had less ischemic heart disease and a higher prescription rate of prognosis-improving drugs and diuretics. There were no between-group differences in the distribution of hypertension, diabetes, atrial fibrillation, or other comorbidities.

Figure 1.

Central illustration. Of 1287 patients with a primary diagnosis of HF included over a 1-year period, 365 were discharged directly from the ED, 919 were discharged after hospitalization, and 3 died. HFrEF patients discharged from the ED had a higher rate of hospital admission during the follow-up period. In the total cohort, HFrEF patients incurred higher costs whether they were discharged directly from the ED or after hospitalization. HF, heart failure; HFpEF, heart failure with preserved ejection fraction; HFmrEF, heart failure with mildly reduced ejection fraction; HFrEF, heart failure with reduced ejection fraction. gr1.

(0.33MB).

Of the 919 hospitalized patients, 157 (17.1%) had HFrEF, 108 (11.8%) had HFmrEF, and 654 (71.2%) had HFpEF. Baseline characteristics in LVEF each group were similar to those of the full cohort (table 1 of the supplementary material). The HFrEF group had a higher rate of admission to the cardiology unit than the HFmrEF or HFpEFgroups (77 [49.1%], 39 [36.1%], and 89 [13.6%], respectively; P <.001) but had a low rate of admission to the internal medicine unit (72 [45.8%], 69 [63.9%], and 555 [84.9%]; P <.001). There were no between-group differences in the length of stay in the index hospitalization (median 7 [5-10], 7 [5-10], and 7 [4-10] days; P=.35).

The baseline characteristics of patients discharged from the ED (n=365) were similar to those of the hospitalized group (table 2 of the supplementary material). The mean length of stay in the ED was 0.7±0.7 days, with no differences between LVEF groups.

Among the hospitalized patients, in-hospital and 30-day mortality rates were similar for the 3 LVEF groups, but 1-year mortality was significantly lower in hospitalized HFrEF patients than in their HFmrEF and HFpEF counterparts (17.2, 25.9, and 26.7%; P <.001) (figure 2). However, LVEF was not an independent predictor after adjusting for age, sex, and cardiovascular risk factors (P=.58).

Figure 2.

Clincal results stratified by LVEF category. A) Clinical results of hospitalized patients at 30-days and 1-year of follow-up, stratified by left ventricular ejection fraction. B) Clinical results of patients discharged from the emergency department at 30-days and 1-year follow-up, stratified by left ventricular ejection fraction. ED, emergency department; HF, heart failure; HFpEF, heart failure with preserved ejection fraction; HFmrEF, heart failure with mildly reduced ejection fraction; HFrEF, heart failure with reduced ejection fraction. gr2.

(0.42MB).

No between-group differences in hospital readmissions were observed at 30-day and 1-year follow-up. Hospital-discharged patients with HFpEF spent a higher proportion of the observation year at home than their counterparts in the other LVEF groups (figure 3). Hospital-discharged patients with HFrEF attended more outpatient consultations, whereas visits to the ED were more frequent among HFmrEF and HFpEF patients.

Figure 3.

Clinical results, health care resource utilization, and costs (COHERENT) for hospitalized patients. A) 30-day follow-up. B) 1-year follow-up. Hospital indicates first admission during follow-up; ED indicates first treatment in the emergency department. HF, heart failure; HFpEF, heart failure with preserved ejection fraction; HFmrEF, heart failure with mildly reduced ejection fraction; HFrEF, heart failure with reduced ejection fraction. gr3.

(0.67MB).

Among patients in the ED group, the 30-day mortality rates for patients with HFrEF, HFmrEF, and HFpEF were 3.1%, 0.0%, and 2.1% (P=.60), and the 1-year rates were 15.6%, 10.5%, and 16.9% (P=.59). There were no between-group differences in the number of repeat ED visits at 30 days and 1 year.

HFrEF and HFmrEF patients in the ED group had higher 30-day and 1-year hospitalization rates than their counterparts in the hospital-discharged group (HFrEF, 8.0% vs 13.4% at 30 days and 75.0% vs 46.5% at 1 year; HFmrEF, 21.1% vs 12.0% at 30 days and 68.4% vs 50% at 1 year; P <.05). Moreover, ED-discharged patients with HFrEF were hospitalized for a higher proportion of the follow-up period (10 [4-19] days) than those with HFmrEF (6 [0-16] days) or HFpEF (0 [0-11]; P <.001). ED-discharged patients had higher rates of new ED treatment at the 30-day and 1-year follow-ups, independently of LVEF category. Among ED-discharged patients, HFrEF patients had more outpatient visits (11 [3-21]) vs 11.5 [5-18] for HFmrEF and 7 [3-12] for HFpEF; P=.009) (table 3 and figure 3 of the supplementary material).

Whether discharged from the ED or after an index hospitalization, patients with HFrEF were more frequently hospitalized for a cardiovascular cause during follow-up, whereas patients with HFpEF were hospitalized mostly for noncardiovascular causes (table 4 of the supplementary data).

The median annual cost per patient was €5008 [€2747-€9590] for patients discharged after hospitalization and €3889 [€2590-€6269] for patients discharged from the ED (P <.001). For hospital-discharged patients, the total cost was highest for the HFpEF group (€4 474 176), reflecting its size, followed by the HFrEF group (€1 640 162) and the HFmrEF group (€867 351). Nevertheless, HFrEF was associated with a higher median annual cost per patient (€6321 [€3335-€12 796]), compared with €6170 [€3189-€10 484] for the HFmrEF group and €4636 [€2609-€8977] for the HFpEF group (P <.001). The median daily cost per patient was considerably higher for patients with HFrEF (€17.3 [€9.1-€35.1] vs €16.9 [€8.7-€28.7] for HFmrEF and €12.7 [€7.1-€24.6] for HFpEF; P <.001) (table 2).

Costs at the 30-day and 1-year follow-up were lower for patients discharged to home from the ED; the total cost for these patients at 1 year was €1 780 974, with 81.9% of this cost (€1 460 194) attributable to hospitalizations (193 first hospitalizations and 181 readmissions) and 13.9% attributable to 1161 ED visits, with no differences between LVEF groups (table 3). The highest costs for patients discharged from the ED were incurred by those with HFrEF, with a median annual cost per patient of €4763 [€2076-€17 155] and a median daily cost per patient of €13.0 [€5.70-€47.00]; the equivalent costs for patients with HFmrEF were €3900 [590-8.013] and €10.70 [€1.60-€22.00] and for those with HFpEF the costs were €3812 [€259-€5486] and €10.40 [€0.70-€15.00] (all comparisons P <.05).

To identify predictors of higher total cost at 1-year of follow-up in the full cohort (relative to the median for our population), we conducted a multivariate analysis including LVEF group, age> 70 years, sex, diabetes, dyslipidemia, hypertension, ischemic heart disease, atrial fibrillation, chronic kidney disease, and pulmonary disease. In this analysis, HFrEF, HFmrEF, and ischemic heart disease were associated with higher cost. Female sex was associated with lower cost in the HFrEF group but not in the HFmrEF or HFpEF groups. As in the full cohort, ischemic heart disease was associated with higher cost in all the LVEF groups (table 4). The multivariate analysis was repeated to differentiate between HF with LVEF <50% and LVEF ≥ 50% (table 5 of the supplementary data).

Our findings may not be applicable in other settings due to differences in patient care between centers and health care systems, for example in the rate of discharge to home from the ED, length of stay, and readmissions. Moreover, it may not be possible to extrapolate the absolute costs to other health care funding systems. In our view, highlighting the differences between health care funding models provides an opportunity to identify areas for improvement. Although the Minimum Data Set data has inherent limitations, comparison with medical records has validated the use of administrative data as a tool for estimating health service outcomes. However, the level of precision achieved with this approach for studies based on the primary diagnosis is unlikely to be achievable for secondary diagnoses; this is likely to result in an underestimation of comorbidities, which may explain the absence of expected between-group differences in our analysis.