Consistent with previous studies [2, 8, 9, 11, 22], we found that lifetime costs of HF care are high, with hospital care and LTC costs accounting for most of these costs. The hospital care costs in the first year after diagnosis were approximately 2.5 times higher than the average costs in all other years for males and females. In addition, costs in the last years of life were nearly three times higher than those incurred > 10 years prior to death, which is also in line with previous studies [8, 19,20,21,22]. The relationship between TTD and cost was most notable for hospital care and LTC costs.

We compared our findings with published studies from various countries in terms of long-term predictions of costs associated with HF. These studies similarly identified hospitalization and LTC as major cost drivers and reported comparable trends in disease burden and reduced life expectancy, supporting the consistency and general plausibility of our findings [2, 8, 11, 22]. Our study’s findings demonstrated that differences in life expectancy play a crucial role in determining lifetime costs. Similar to findings from earlier studies [2, 9, 12, 19], as patients with HF live longer, the lifetime costs of HF increase significantly owing to the chronic nature of the disease and frequent readmissions. The growing need for postdischarge care, ongoing medical treatments, LTC, and the management of comorbidities contribute to a substantial rise in costs over the remaining survival of patients with HF. Of particular interest is the finding that, while annual costs are heavily influenced by comorbidities, lifetime costs are less affected by them owing to the shorter life expectancy of individuals with more comorbidities. This demonstrates that prevalence-based cost-of-illness studies estimating annual costs can give a different picture of the cost-drivers than incidence-based cost-of-illness studies estimating lifetime costs.

Our study identified important associations between income level and lifetime costs in HF. While patients with lower incomes tend to have shorter lifespans, their lifetime LTC costs exceed those of higher-income patients. A possible explanation for this could be that women with a lower income level become widows earlier than women with a higher income level, requiring them to move to an LTC facility when they reach advanced stages of HF [23,24,25]. Patients with HF with a higher income have higher lifetime hospitalization costs than those with a lower income, not only owing to their higher life expectancy but also likely owing to their higher demand for and better access to high-quality healthcare [26,27,28]. Previous research also showed that the probability of living in a nursing home for individuals with a lower income is three to four times higher compared with individuals with a higher income in the Netherlands [29]. Higher income can enhance the ability to secure alternative forms of care and adapt one’s home environment, influencing decisions about long-term care. For instance, older individuals with higher incomes may afford better housing or modifications to support at-home care, reducing the need for institutional care and promoting independent living [26, 29, 30].

It should be noted that the analysis is conducted from the perspective of the payer. While we cannot entirely exclude the presence of minor private copayments, it is important to clarify that our analysis refers primarily to healthcare covered under the basic health insurance package, where out-of-pocket spending in the Netherlands is exceptionally low by international standards. Although some copayments may occur, such as for certain medications, medical devices, or noncontracted providers, these are generally modest and often arise after insurers initially reimburse the provider and subsequently recover the patient’s deductible or copayment. For LTC, copayments are income-based [30,31,32,33].

Although income at older ages may be less indicative of current material resources compared with income in midlife, it remains a meaningful proxy for long-term socioeconomic status (SES). In the Netherlands, supplementary pensions, which are strongly linked to prior earnings, form a substantial part of retirement income, making post-retirement income a relevant SES indicator. Moreover, by using individuals’ relative income position rather than absolute income, we account for the narrowing of income differences after retirement while still capturing key differences in SES [31, 32].

The sex-specific patterns in healthcare spending we observed—higher LTC, GP, and rehabilitation costs for female patients, alongside higher hospital care costs for male patients and similar medication costs across sexes—are consistent with findings in the Netherlands and other countries. Previous Dutch studies indicate that female individuals use more formal LTC services than male individuals, partly owing to longer life expectancy. Aaltonen et al. (2022) found that females used more formal home care and residential care, while males relied more on informal care provided by spouses. This reliance on formal care contributes to higher LTC costs for females [34].

One of the strengths of this study is its utilization of a large, nationally representative dataset of over 20,000 patients with HF capturing nearly all Dutch hospitals. In addition, the study employs a comprehensive definition of HF as both a primary and secondary diagnosis, increasing the likelihood of capturing the vast majority of all patients with HF diagnosed for the first time in 2015. One of the unique contributions of this study is the detailed inclusion of LTC costs alongside healthcare costs over a patient’s lifetime, demonstrating their impact on total HF costs. Furthermore, the breakdown of costs into categories (i.e., hospital care, LTC, and other costs) and different subgroups of patients on the basis of age, sex, comorbidities, and income level allows for identifying subgroups that incur the highest costs and may need more attention.

This study has several limitations that should be considered when interpreting the results. First, newly diagnosed patients with HF were identified on the basis of hospitalization in 2015, excluding those diagnosed through outpatient settings or general practitioners without hospital admission. However, research shows that HF is frequently diagnosed during hospital stays, suggesting that a large proportion of cases were captured [35]. Second, we defined newly diagnosed patients as those who had been hospitalized for HF in 2015 but had not been hospitalized in the 2 years prior to 2015. As a result, patients who were diagnosed with HF before 2015 and who did not experience hospitalization between 2013 and 2015 may not have been included in our study. Considering that one in four patients with HF is readmitted within 30 days of discharge and approximately half are readmitted within 6 months [36], we assumed that a 2-year window was adequate to identify newly diagnosed HF cases. Moreover, as certain medications can be prescribed for multiple diagnoses, this introduces uncertainty in the identification of specific conditions—we minimized potential overestimation of multimorbidity by applying a 2-year continuous prescription criterion [20, 37, 38]. Another limitation of this study is the identification of patients with HF on the basis of the ICD codes for reasons of hospital admission. Recent studies have shown that the use of different ICD codes may lead to a possible over- or underestimation of HF diagnoses [9]. In addition, HF is often associated with many underlying diseases, making it challenging to identify the correct cases of HF using only the ICD codes. For instance, Lee et al. [39] demonstrated that defining HF by only including hospital admission with HF as the primary diagnosis reduced the number of HF cases by 46% compared with including both primary and secondary diagnoses (with the same ICD codes). Their analysis showed that 75% of patients with HF as a secondary diagnosis had a primary diagnosis related to HF, such as hypertension or angina pectoris [2, 39, 40]. Given these findings, defining HF only as the primary diagnosis may lead to underestimating the number of people with HF. Therefore, our study included people with HF as either the primary or secondary diagnosis for hospital admission. A further limitation is the absence of patient-reported outcomes, such as the New York Heart Association (NYHA) class or the Kansas City Cardiomyopathy Questionnaire (KCCQ) scores, which could have enhanced our ability to capture disease severity and its impact on long-term costs. These measures were not available in the administrative dataset used in our study. One of the limitations of this study is the lack of detailed clinical data on disease severity and biomarkers (e.g., B-type natriuretic peptide [BNP] and N-terminal-prohormone BNP [NT-proBNP]), as the dataset included primarily demographic, administrative, and claims data. Other studies have demonstrated the value of incorporating such clinical parameters into burden-of-HF studies, improving both risk stratification and cost prediction (e.g., refs. [41,42,43,44]). However, the scale and representativeness of the data used in our study offer important real-world insights into the economic burden of HF in the Netherlands.

Another limitation of this study is that only the number of chronic conditions was included in the model rather than the specific types of comorbidities. This approach does not capture the differential impact of conditions such as type 2 diabetes mellitus or chronic kidney disease, which are known to significantly influence HF progression and associated costs. Furthermore, some cardiovascular comorbidities, such as atrial fibrillation and ischemic heart disease, were not explicitly included in the model, which may limit the clinical specificity of our findings.

A key limitation of our analysis is the crude definition of TTD, which was based on calendar years rather than exact dates of healthcare use. This approach was necessary owing to data constraints; we had access to exact dates of death but only to spending data by calendar year without information on the timing of spendings within the year. Although this may introduce some imprecision in our estimations, this method is commonly used in other studies on the basis of claims data (e.g., refs. [45, 46]). Importantly, as this definition was applied consistently across all groups in our regression models and in our estimations, we believe that it does not introduce differential bias in our estimates of lifetime costs.

It should also be mentioned that while the methodological framework used in this study is generalizable and aligns with previous (international) studies [2, 8, 22], the cost estimates are specific to the Dutch healthcare context. Differences in health system structure, pricing, patient demographics, and treatment patterns may limit direct transferability of results to other settings. The Dutch healthcare system offers universal coverage and is recognized for being well structured and providing high-quality primary, secondary, and rehabilitation care with relatively little copayment. The primary care sector is quite extensive since GPs act as gatekeepers to secondary care. Another specific characteristic that is relevant to this study is the relatively large public funding of the LTC sector, whether provided in a nursing home or in the form of home care. Hence, the results of this study may only be generalizable to high-income countries with similar healthcare systems.

Finally, our cost estimates based solely on healthcare and LTC costs may underestimate the true economic burden of this condition. For example, a study by Delgado et al. [47], which estimated the economic burden of informal care costs per patient, revealed that informal care constituted the largest proportion of total costs, ranging from 59.1% to 69.8%. We expect these costs to be lower in the Netherlands than in many other European countries, as LTC, whether provided in nursing homes or at home, is largely publicly funded.