The literature searches retrieved a total of 1480 titles and abstracts (after de-duplication). After screening, 221 references were selected for full-paper screening. After a subsequent detailed review, 81 references were selected as meeting all the inclusion criteria; ESM 7 lists the 140 publications that were excluded after full-paper screening. Nine additional studies were derived from gray literature and citation snowballing. Three of the included studies were each reported in two references (i.e., a conference abstract and a full-text publication). This brought the total to 87 included studies (90 individual references) for data extraction. Figure 1 presents a summary of the searching, screening, and inclusion assessment process in accordance with the PRISMA checklist .
The 87 included studies are summarized in the following section. The studies are described according to the three quality criteria detailed in the Methods section. A complete list of all the studies and their methods are given in ESM 3 (study characteristics) and ESM 4 (applied definition of HF). The results per study are presented in ESM 5.
Study Population and Design
The majority of the studies (n = 69; 79%) assessed the costs associated with HF-related events among exclusively patients with HF, with some defining in more detail the type of HF (n = 28; 32%) [ESM 3, column ‘population’] [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49]. The majority of studies (n = 55; 63%) defined the population by the International Classification of Diseases and Related Health Problems, Ninth Edition (ICD-9) codes (ESM 4). Eight articles referred to both ICD-9 and International Classification of Diseases and Related Health Problems, Tenth Edition (ICD-10), while 32 did not specify the ICD classification (12 used different definitions, 20 did not provide details of definition). The study population sizes had a range from 135  to 11,723,008 . The mean age of the study populations varied from 59 to 84 years.
The majority of studies were retrospective cohort studies (77/87; 89%), and analyses were based on varied healthcare databases (55/77 studies), six of which explicitly referred to as ‘medical records’ and/or ‘chart reviews’ in the studies, administrative claims data (20/77), registry data (1/77), or combined medical records and claims data (1/77) [ESM 3]. Five studies (5/87) used an economic model populated with data from different healthcare databases [45, 52,53,54,55,56]. A cross-sectional design, where costs were assessed at a discrete point in time, was adopted by three studies [57,58,59]. The remaining two studies used information generated alongside a randomized controlled trial [33, 46]. The vast majority were multi-centered studies (74/87; 85%) using data from inpatient settings (n = 41), while 33 studies combined inpatient and outpatient settings.
The perspective of cost analysis was not clearly stated in most of the studies (67/87; 77%). Thus, the perspective of analysis was often based on the researchers’ interpretation of the data. In 40 (45%) studies [22,23,24, 44, 46, 47, 49, 50, 53, 55, 57, 58, 60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86], costs for HF-related events were (or appeared to have been) analyzed from a payer’s perspective. Three studies adopted a healthcare sector perspective [87,88,89]. Only two studies adopted the societal perspective [25, 59], one study adopted the viewpoint of the patient , and one the hospital perspective . The remaining 40 studies (only published as abstracts) left the perspective undefined entirely.
The method of estimating costs was not explicitly stated in the majority of studies (58/87; 67%) (ESM 3). Based on the reviewers’ interpretation, 41 studies employed a bottom-up approach (cost profile is constructed using disaggregated patient-level data), seven used a top-down approach [47, 64, 65, 71, 74, 76, 90] (total expenditure divided by number of patients treated), and one study  used a mixed costing method. Information reported by the remaining studies (n = 38, only published as abstracts) did not allow for conclusions to be drawn on the costing approach.
The earliest cost year used in the studies was 2009 , with the latest cost year being 2017 . Forty-four studies did not report a price year for cost estimates.
The 87 studies were reviewed and costs assigned to the most appropriate cost category, as per Drummond et al. (Table 1) . The majority of studies (85/87; 98%) included hospital inpatient care costs. A sizable proportion of studies reported outpatient care (also known as ambulatory care) costs (36/87; 41%); this included costs for emergency department services, primary care physician office visits, and pharmacy costs. Costs for home care or nursing home/skilled nursing facility were only included by six (7%) studies. Only four studies reported out-of-pocket costs, while none included travel costs or social benefits. Patients’ productivity losses were not reported. One study reported costs of informal caregivers .
The studies were furthermore reviewed for subgroup analyses. In total, 48 (55%) studies presented costs for different patient subgroups, of which 12 performed a multivariable analysis to assess predictors of costs. Table 2 lists the different subgroups analyzed across the 48 studies.
Sources of Heterogeneity
The annual medical costs associated with HF varied greatly because of the presence of comorbidity with a range from $19,537 for patients with HF with hypertension to $77,214 for patients with HF with hyperkalemia [58, 73, 89]. Likewise, the cost of HF hospitalization had a range from $8702 in patients with depression to $40,407 in patients with protein-energy malnutrition [58, 60]. The cost estimates for each comorbidity-related subgroup are presented in ESM 5. The age subgroup analyses showed that the cost of hospitalization was 5–31% higher for patients with HF aged < 65 years than those aged ≥ 65 years [62, 63, 65, 67, 74, 92].
The subgroup analyses across different types of patients with HF showed that the cost per hospitalization episode for patients with HFrEF was 13–100% higher compared with the cost for patients with HF with preserved ejection fraction (HFpEF) [23, 25, 77, 115]. One study investigated costs of chronic vs acute HF, based on cost per discharge, demonstrating higher costs with chronic disease ($17,771 vs 13,976, in 2019 US$) .
The cost of HF hospitalization varied per hospital ownership with higher cost in not-for-profit hospitals, followed by government and for-profit hospitals; however, total charge billed was highest in for-profit hospitals and lowest in government hospitals . Among patients with a liver transplant experiencing a HF event, the HF-related treatment in a transplant center was more costly than the care provided in a non-transplant hospital . The authors concluded that this may be related to selective referral of certain patient phenotypes to transplant centers. Similarly, increased mean hospital charges and length of stay was found for patients with HF with a history of hematopoietic stem cell transplant admitted in teaching hospitals (as compared with non-teaching hospitals) .
Subgroup cost analyses per US region, type of hospital admission, and health plan type were each reported by single studies. A review of the data revealed there is regional variation in the cost of HF hospitalization, with higher costs in the West ($9908 per patient) and Northeast ($9022) compared with other regions of the country (Midwest: $8021; South: $7089) .
The study by Obi et al.  compared annual total medical costs between patients with HF who died to those who survived (stratified by health plan type). Authors found that patients who died incurred substantially higher costs from index to death, as compared with survivors, despite the mean post-index period being 59% shorter compared with survivors (matched cohorts). This trend was observed regardless of health plan type; however, the difference in cost between decedents and survivors was greater for those with commercial plans (2.6-fold higher) than those with Medicare Advantage with Part D plans (1.5-fold higher). The findings from this study offer useful insights on the end-of-life cost burden in patients with HF.
Last, the study by Punekar et al.  evaluated total medical costs associated with HF events in patents with hyperlipidemia, per baseline risk of cardiovascular (CV) disease (high risk, primary prevention, secondary prevention), and the number of HF events (only one, two, or three HF events) over a 2-year follow-up. Authors found that the total medical costs were higher among cases (patients with a CV event) compared with controls (patients without a CV event) in all three cohorts, and costs remained higher throughout the follow-up period for HF events. Mean 2-year total costs among patients with only one HF event and only two HF events were highest for the primary prevention cohort ($75,229 and $106,595, respectively) compared with the secondary prevention ($65,649 and $76,760) and high-risk ($65,010 and $90,246) cohorts. Mean 2-year total costs among patients with only three HF events were highest for the high-risk cohort ($108,319) compared with the primary prevention cohort ($104,347) and secondary prevention cohort ($94,548).
Heart Failure-Related Costs vs All-Cause Costs
Studies were assessed to determine whether published costs were HF related or all cause. Seventy-two (83%) studies estimated costs only attributable to HF events (n = 68), or separate cost estimates were provided for all-cause and HF-related resource use (n = 4) [44, 65, 85, 87]. In the remaining 15 studies, eight (53%) assessed the costs that could be directly attributed to HF by comparing the costs of patients with HF to those with no HF [51, 69, 88, 95, 104, 105, 107, 111]. In the other seven studies [46, 58, 66, 68, 75, 89, 94], it was inferred that all-cause resource use (i.e., no distinction for HF-related resource use) had been included in the analysis.
Comparable Cost Estimates
Comparable cost estimates are summarized in Table 3. Forty-three studies identified from the SLR included a cost year and could be inflated to 2019 US$. Two of these, however, were excluded from the summary of comparable cost estimates because one study reported only informal caregiving costs , while in the other study, patients were identified by the index event of a pharmacy claim rather than an HF event . The 41 remaining studies included in the summary of comparable cost estimates, along with the study-specific cost/price years and the original values are detailed in ESM 5. All studies considered the economic burden of patients with HF in terms of direct medical costs due to inpatient care, outpatient care, and/or medications. Of the 41 studies, 24 only included patients who had been hospitalized for HF, and the remaining 17 included a mix of hospitalized and non-hospitalized patients. The reported costs were subject to large variations, which was, in part, owing to how the cost components were defined in the individual studies or how the resource use was valued, i.e., charges vs costs. Charges refer to the initial individual list prices a US hospital typically sets for the services it provides. They are usually higher than costs, which are the actual expenses incurred by a hospital in providing patient care (including the direct costs of patient care such as nursing, room and board, medicines and supplies; as well as overhead costs for administration).
Total Medical Heart Failure-Related Costs
The annual median medical costs for HF care was $24,383 (IQL $20,713–$40,619) per patient (Table 3). Reported charges were substantially higher than reported costs (mean $98,599) and increased to > $100,000 per patient with HF when all-cause readmissions were included in the cost estimation .
Costs varied with the presence of a comorbidity and with survival. Analyzing the comparable costs by subgroup with a specified comorbidity, the estimated median annual total medical cost was $46,580 (IQL $39,585–$64,664) per patient, almost double the cost estimated for the general HF population ($24,383) (Table 3). The actual reported costs/charges varied greatly with the specified comorbidity. The cost estimates for each patient subgroup are presented in ESM 5.
One study analyzed costs according to survival . The medical cost per patient accrued within 30 days from the date of the earliest qualifying claim with a diagnosis code for HF varied from $1592 for patients who survived to $19,317 for patients who died. Patients with HF who died within 1 year after an index HF encounter incurred markedly higher per-patient-per-month costs than patients in the survivor cohort, with the majority of costs attributable to hospitalizations for both patient cohorts. Likewise, the all-cause medical cost (including inpatient care, outpatient care, and pharmacy costs) for the same population had a range from $3510 (HF survivors) to $25,510 (HF decedents) per patient per month .
Heart Failure-Specific Hospitalization
The median cost for a HF-specific hospitalization was $13,418 (IQL $11,125–$15,667) per patient (Table 3). For patients with co-morbidities (e.g., diabetes mellitus, ischemic heart disease), the median HF-specific hospitalization cost was slightly higher, at $14,015 per patient (IQL $11,769–$20,373).
Olchanski et al.  estimated mean hospitalization costs by category, stratified by diabetes status and HF subtypes (HFrEF and HFpEF). Mean hospitalization costs were higher for patients with diabetes regardless of HFrEF ($16,679) vs HFpEF ($15,301) status at admission compared with those without diabetes (HFrEF, $12,296; HFpEF, $11,828). The authors also reported on the following subcategories: diagnostic tests (range $933–$1197 per patient), room and board expenses ($4110–$6188), therapies ($339–$753), pharmacy ($445–$729), other expenses ($1144–$1364), and total cost per day alive ($3022–$11,828). Patients with HFrEF had 4–9% higher hospitalization costs compared with patients with HFpEF. Room and board expenses were 12–16% higher for patients with HFpEF compared with patients with HFpEF. Similarly, diagnostic and pharmacy costs were 6% and 23% higher for patients with HFpEF, respectively. Therapy costs were 19–31% higher for patients with HFpEF compared with patients with HFrEF. Last, cost per day alive for patients with HFpEF exceeded corresponding costs for patients with HFpEF by 5–10%. The annual median cost for HF-specific hospitalizations was $15,879 (IQL $9444–$20,933) per patient and for all-cause hospitalizations $20,826 (IQL $18,779–$29,045) per patient.
Two studies estimated the average post-discharge cost following a HF hospitalization event [57, 97]. The average 30-day post-discharge cost following a worsening HF admission was estimated at $6283 per patient . It was calculated as the difference between all costs incurred in the 30 days following discharge and the average 1-month cost of a patient with HF . Raju et al.  estimated the monthly HF-related post-discharge costs at $1771 per patient with type 2 diabetes mellitus (T2DM) and at $1543 per patient without T2DM. The monthly all-cause post-discharge costs were estimated at $8722 and $8055, respectively.
One study estimated the cost associated with readmissions for patients discharged following a HF hospitalization . Lahewala et al.  estimated the mean cost of care of readmission over a 30-day period after HF hospitalization in the same hospital at $15,732 per patient and in a different hospital at $25,879 per patient (the associated mean length of stay per patient was 6.1 and 7.5 days, respectively).
Another study evaluated costs for patients with hyperlipidemia with one, two, or three HF hospitalization events . The associated mean 2-year costFootnote 1 range was from $65,010 to $75,229 (one HF event), $76,760–$106,595 (two HF events), and $94,548–$108,319 (three HF events), respectively, depending on pre-index clinical characteristics of the patients (high risk, primary prevention, or secondary prevention). A third study estimated the cost of 30-day all-cause readmission for patients discharged following a HF hospitalization at $7583 .
Outpatient Care Cost
Most of the included studies captured costs for outpatient medication and physician visits within the total medical cost estimated; only seven studies itemized outpatient visit costs separately from other medical costs [25, 34, 46, 57, 64, 76, 86], which had a range from $297  to $3859  per patient. Yoon et al.  estimated the annual average cost per patient per type of outpatient healthcare, including medical/surgical costs ($8487), diagnostic costs ($2178), behavioral costs ($550), other costs ($973), and pharmacy costs ($2904).
Six studies reported the cost of the emergency department services [24, 25, 55, 57, 58, 64]. The median cost per visit was $1441 (IQL $829–$1933). A seventh study estimated the annual HF-related emergency service cost at $212–$291 per patient, averaged over a cohort of 1-year survivors and 1-year decedents post-index HF claim, respectively . Per-patient estimates reflecting only those patients who did visit the emergency department were not provided. The same study estimated the annual average HF-related ambulatory costs at $930 and $1902 per patient among patients with HF who died and those who survived, respectively. The respective cost estimates per patient per month were $18–$93 and $156–$192 for HF-related emergency and ambulatory services, respectively .
Other Medical Costs
The economic burden of home healthcare was estimated only by Echouffo-Tcheugui et al. at $2227 per patient per year . The same study provided an average annual cost estimate of $813 per patient, for a list of services including nursing home, rehabilitation, vision, medical supplies, and dental services. The study reported cross-sectionally on patients with a HF diagnosis (10-year inclusion window), with no requirements for hospital admission.
Costs for Patients with Heart Failure with Reduced Ejection Fraction
Five studies reported costs for patients with HFrEF [23, 25, 34, 44, 77]. The median HF-specific hospitalization cost (including room and board, diagnostics, therapies, pharmacy, emergency department, and overhead costs) was $12,915 (IQL $12,156–$13,664) per stay [23, 25, 77]. One study estimated the mean HF-specific hospitalization cost at $12,915 in patients with T2DM and at $10,103 in patients without T2DM . The same study further estimated the total cost per day of HF-specific hospital stay at $2205 and $2115 in patients with T2DM and without T2DM, respectively.
Two studies estimated charges (rather than costs) for HF-related care [34, 44]. Givertz et al. looked at HFrEF in patients who had been hospitalized and/or required intravenous diuretic because of worsening HF (0.7 hospitalizations per person during the first 30 days after HF onset). The mean medical charges (including dispensed prescriptions and number of private healthcare and hospital-facility visits) accrued within the first 30 days following the event were $31,300; monthly mean charges for the remaining year was $32,524 per patient . Bress et al.  assessed resource utilization and associated charges among a real-world cohort of patients with HFrEF in an academic medical center setting. The median hospital charges for an all-cause, HF-specific, and 30-day all-cause readmissions were $25,545, $26,393, and $31,503, respectively. Index hospitalizations were included in the charge analysis. Cost estimates specific to other types of HF (with preserved ejection fraction [diastolic HF]; acute decompensated HF) were reported in five studies [23,24,25, 49, 72].
The quality of this review is in part limited by the quality of the identified studies. The studies that do not report the cost year of the provided estimates were considered of insufficient quality and were excluded from further quality assessment. Therefore, quality assessment is provided for 43 studies.
Only studies that reported a cost/price year were included in a comparison and analysis of cost estimates. All studies included in the analysis of comparable costs met the quality criteria for economic evaluations as used by the British Medical Journal  and were therefore concluded to be of high quality. An overview of adherence of the studies to each of the items used for the quality assessment is presented in Fig. 2.
All the studies that were subject to the quality assessment (n = 43) clearly stated the research question and defined the target population. The study setting (inpatient/outpatient) and location was well reported in all but one study .
The perspective of the economic analysis was adequately reported and justified in 19 studies (44%); less than half (21/43; 40%) [22, 24, 25, 44, 46, 47, 49, 53, 59, 60, 63, 65, 67,68,69, 74, 81, 87,88,89, 116] did not clearly define the perspective but the available information allowed for interpretation in line with published definitions ; three studies (10%), only published as abstracts, left the perspective completely undefined [34, 90, 97].
Sources of resource use and unit costs, currency and cost year, and the methods of resource use valuation (i.e., cost/charge, cost-to-charge ratio) were well defined in all studies. The majority of studies (26/43; 60%) reported adequately on the methodology of estimating costs [22,23,24, 43, 44, 49, 50, 57,58,59,60, 63, 66,67,68,69,70,71,72, 75,76,77, 85, 87, 88, 116]; in 14 (33%), the costing approach was concluded by the information reported and relevant guidelines by Drummond et al. , although not explicitly stated [25, 45,46,47, 53,54,55, 64, 65, 81, 86, 89, 90, 118]; three studies (10%), only published as abstracts, did not provide any information on the costing methodology [34, 73, 97].
Time horizon was reported by 38 studies, with the majority (33/43; 77%) estimating costs over a relatively short time frame of up to 1 year. Only five studies estimated the costs for a HF event for a period greater than 1 year [44, 46, 71, 75, 88], and only one of them discounted costs incurred after the first year. Time horizon was not applicable for the remaining five economic model studies, from which only de-novo cost inputs were extracted [45, 53,54,55, 74]. In all the studies, the answer to the study question was given, conclusions followed from the data reported, and in all but six studies (14%, only published as abstracts) [34, 49, 73, 86, 90, 97] the conclusions were accompanied by the appropriate caveats.