You are here

  • Home
  • Archive
  • Volume 10, Issue 2
  • Resource utilisation and outcomes of people with heart failure in England: a descriptive analysis of linked primary and secondary care data – the PULSE study

Article Text

Article menu
Heart failure and cardiomyopathies
Original research
Resource utilisation and outcomes of people with heart failure in England: a descriptive analysis of linked primary and secondary care data – the PULSE study
  1. Stephan Linden1,
  2. Nicholas D Gollop1 and
  3. Ruth Farmer2
  1. 1Boehringer Ingelheim International GmbH, Ingelheim, Germany
  2. 2Boehringer Ingelheim Ltd, Bracknell, UK
  1. Correspondence to Dr Ruth Farmer; ruth.farmer{at}boehringer-ingelheim.com

Abstract

Background Heart failure (HF) is associated with high levels of resource use and mortality, but prior UK studies have not compared outcomes by HF subtype (HF with reduced ejection fraction (HFrEF) and HF with preserved ejection fraction (HFpEF)) in large patient populations. This study investigated healthcare resource utilisation and mortality in patients with HF in England, overall and by HF subtype.

Methods This non-interventional cohort study linked data from the Clinical Practice Research Datalink database to Hospital Episode Statistics inpatient and UK Office for National Statistics mortality data. Patients with a recorded HF diagnosis (new (incident) or existing (prevalent)) based on clinical codes or measures of ejection fraction between 2015 and 2019 were included.

Results Of 383 896 patients identified with HF, 100 224 patients (26%) had a recorded subtype: 68 780 patients with HFrEF (69%) and 31 444 patients (31%) with HFpEF. In total, 918 553 person-years (PY) were included (median follow-up: 2.1 years): 625 619 PY (68%) for unknown HF subtype, 204 862 PY (22%) for HFrEF and 88 017 PY (10%) for HFpEF. Overall, 11% of patients experienced ≥1 HF hospitalisation. After age and sex adjustment, hospitalisations for HF (HHF; including recurrent hospitalisations) and HF-related general practitioner consultations occurred at rates of approximately 80/1000 and 124/1000 PY, respectively, and were highest for patients with HFrEF and unknown subtype. Overall, all-cause and cardiovascular mortality rates were 132/1000 and 49/1000 PY, respectively. Patients with unknown subtype had the highest 1-year and 5-year mortality (20% and 48%), followed by HFrEF (8% and 35%) and HFpEF (6% and 25%).

Conclusions HF is associated with high levels of healthcare resource use, mortality, HHF and comorbidities. Ensuring that patients receive early and appropriate guideline-directed therapies to manage HF and associated comorbidities is likely to improve patient care and reduce the burden of HF on the English healthcare system.

  • Heart Failure, Diastolic
  • Heart Failure, Systolic
  • Electronic Health Records
  • Health Care Economics and Organizations
  • Outcome Assessment, Health Care

Data availability statement

All data relevant to the study are included in the article or uploaded as online supplemental information. The results of all data analyses are contained within the manuscript and online supplemental material.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/openhrt-2023-002467

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Heart failure (HF) is associated with high levels of resource use and mortality, but prior UK studies have not compared outcomes by HF subtype in large patient populations.

    WHAT THIS STUDY ADDS

    • This study shows that classification of HF subtype (HF with reduced ejection fraction (HFrEF) and HF with preserved ejection fraction (HFpEF)) based on ejection fraction coding is suboptimal within the UK healthcare system, with less than one-third of patients in this study having a recorded subtype. It also highlights the high burden of comorbidities, mortality and healthcare resource use (HF-specific and all-cause) among all patients with HF, regardless of subtype.

    • Although HF is not the primary cause of most hospitalisations for patients with HF, rehospitalisation rates for HF are substantial for patients with both HFrEF and HFpEF. Observed HFpEF all-cause mortality rates were three times greater than the average mortality per 1000 persons previously reported in England and Wales for those aged 70–74 years, and even greater for HFrEF. In patients with comorbidities, such as type 2 diabetes and chronic kidney disease, rates of HF-related hospitalisation, and cardiovascular and all-cause mortality were higher again, indicating a high unmet need for these patients.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • This study shows that there are opportunities for primary care education to increase awareness around HF subtypes and differences in management. Increasing the number of patients who receive appropriate guideline-recommended therapies to manage HF and its associated comorbidities is likely to reduce the burden of HF on the English healthcare system.

    Background

    Heart failure (HF) is associated with high levels of resource use and mortality.1 2 HF may present as impaired left ventricular systolic and/or diastolic function, with impaired systolic function predominant in HF with reduced ejection fraction (HFrEF) and impaired diastolic function predominant in HF with preserved EF (HFpEF). International guidelines define HFrEF as an EF ≤40% and HFpEF as EF ≥50% (HF with mildly reduced/borderline EF (HFmrEF) defined as EF 41%–49%).3 4 Current evidence suggests that patient demographics, aetiology, prognoses and responses to therapy differ by HF phenotype.5–8 Despite this, previous UK studies of HF patient demographics, outcomes and resource use conducted in broad populations have not investigated HF by subtype. Additionally, prior studies that have investigated HF by subtype have done so in smaller or non-UK populations, limiting the generalisability of the findings.1 6–14

    Investigating healthcare resource use and clinical outcomes associated with HF subtypes may highlight potential areas for the improvement and optimisation of patient care and primary care education. Therefore, this real-world, non-interventional study aimed to investigate the clinical and resource burden of HF in England (overall and by subtype) by assessing (1) hospitalisation for HF (HHF) rates (per 1000 person-years (PY)), (2) all-cause and cardiovascular (CV) mortality rates and (3) HF-specific primary care resource use.

    Methods

    This non-interventional cohort study linked data from the Clinical Practice Research Datalink (CPRD) Aurum database to Hospital Episode Statistics (HES) inpatient data and UK Office for National Statistics (ONS) mortality data. The study population was English adults (aged ≥18 years) with a diagnosis of HF (incident (new) or prevalent (existing)) recorded in the CPRD or HES database any time prior to 31 December 2019 who were alive and contributed at least 1 day’s data to the CPRD Aurum database between 1 January 2015 and 31 December 2019 (see online supplemental tables S1 and S2 for diagnostic codes).

    The index date for each patient was defined as the latest of the following: 1 January 2015, 1 year after CPRD registration, or the date of the first HF diagnosis in the CPRD or HES databases. Follow-up for resource use and mortality began the day after index and continued until the earliest of the following: 31 December 2019 (end of study), transfer out of the CPRD, death or the last CPRD data collection date. HF cases were defined as prevalent if a patient had a record of HF diagnosis prior to index or as incident if index was the date of first HF diagnosis.

    The look-back period for data collection was a minimum of 1 year prior to the index date, reaching as far back as registration. However, in some cases, primary care data prior to registration were pulled through or captured in registration appointments, providing historical information on comorbidities and clinical events, and allowing for a robust assessment of prevalent versus incident HF at index.15 Treatment data (prescriptions, tests, immunisations, etc) were only available from registration onwards.

    EF categorisation

    At index, patients with HF were categorised according to HF subtype. In this analysis, patients with an EF ≤40% were classified as having HFrEF and those with an EF >40% were classified as having HFpEF. Patients lacking EF subtype data were included and categorised as ‘unknown’ to enable investigation of potential selection bias within subtype recording in primary care.

    Outcomes

    The primary outcomes of the study were resource utilisation (HHF and primary care utilisation; see online supplemental tables S1 and S2 for diagnostic codes) and mortality (all-cause and CV). Secondary outcomes included rates of first HHF (and second/third HHF, if relevant), renal function (creatinine test results or estimated glomerular filtration rate (eGFR)), type/source of HF admission (elective vs non-elective, accident and emergency, general practitioner (GP), consultant or other), all-cause hospitalisation, HHF (code in any diagnostic position), any GP consultation and GP consultation with any HF code.

    Covariates

    At index, a range of demographic and clinical characteristics of the HF cohort were recorded using the look-back period previously described. This included, but was not limited to, age at index, sex, duration of HF, ethnicity, history of comorbidities (eg, type 2 diabetes (T2D), atrial fibrillation, ischaemic heart disease) and use of CV medications (angiotensin-converting inhibitors (ACEi), angiotensin receptor blockers (ARB) and beta blockers).

    Statistical analysis

    The reported analyses are predominantly descriptive in nature, with no formal statistical comparisons made to infer cause and effect. However, some basic adjustments were made to account for possible confounding by age and sex in all analyses since these are strongly associated with many of the outcomes of interest. Specifically, exploratory comparisons of outcomes by HF subtype were adjusted for age and sex, as were all prespecified subgroup analyses, including those for sex and age group (<55 years, 55–64 years, 65–74 years, 75–84 years and 85+ years).

    Baseline characteristics were summarised descriptively. All continuous variable distributions were summarised as mean (SD), median (IQR), minimum, maximum, and 5th and 95th percentiles. Categorical variables were summarised as n (%). Outcome event rates were summarised as the count, observed PY of follow-up, and the observed rate per 1000 PY.

    All primary and secondary event (first or recurrent) outcomes were analysed using the following patient groups: HF overall, coded HFrEF, coded HFpEF and unknown HF subtype. Subtype-specific estimates were obtained by including subtype as a covariate in a negative binomial model to account for overdispersion. Model predictions were used to estimate the expected subgroup rates. Age-adjusted and sex-adjusted estimates were generated by adding these as additional covariates into the model. For all expected rates, 95% confidence intervals (CIs) were calculated. For rates of mortality by prior HHF, number of HHFs was treated as a time-dependent covariate and the number was updated at date of discharge.

    Kaplan-Meier plots for all-cause mortality by HF subtype were generated, and 1-year, 2-year and 5-year survival probabilites were estimated. For CV and non-CV mortality, cumulative incidence functions (CIF) were generated using the stcompet command in Stata V.15 to account for competing risks, and 1-year, 2-year and 5-year mortality rates were estimated.

    For the secondary outcome of eGFR decline, change in eGFR was modelled using a linear mixed model to account for unbalanced repeated measures data—specifying patient as a random intercept term. Time, HF subtype, age at index and sex were included as fixed effects. A quadratic term for time was included in the model to allow for non-linear change in eGFR.

    In the main analysis, all primary and secondary outcomes were analysed in the overall patient population, which included patients with a diagnosis of either incident (new) or prevalent (existing) HF. However, as healthcare resource use and clinical outcomes are likely to differ during the first year post-HF diagnosis, patients were also stratified according to prevalent or incident HF at index for further analysis. These findings are presented in the online supplemental file.

    Additional preplanned subgroup analyses included analyses by age, sex, baseline eGFR status, history of HHF and comorbidities.

    Results

    Overall denominator and overall HF population

    The eligible denominator CPRD adult population comprised 11 414 490 individuals. Of the 383 896 patients identified with HF using data from the combined databases (figure 1), 100 224 patients (26%) had a recorded HF subtype: 68 780 patients (69%) with HFrEF and 31 444 (31%) patients with HFpEF. For the remaining 283 672 patients (74%), HF subtype was classified as unknown.

    Figure 1
    Figure 1

    Inclusion flow chart for overall study population. *Patients whose HF diagnosis was recorded in the HES database only may have received their first diagnosis before 31 December 2019 but after the date at which they exited follow-up from their CPRD practice (either due to deregistration or because of the date at which the practice last contributed data), and hence, they would not be included as HF cases. These patients still contribute to the overall denominator population. CPRD, Clinical Practice Research Datalink; HES, Hospital Episode Statistics; HF, heart failure.

    Patient disposition and follow-up time

    Of the 383 896 individuals in the HF cohort, there were 918 553 PY of follow-up distributed between the HF subtypes (at index) as follows: HFrEF 204 862 PY (22%), HFpEF 88 017 PY (10%) and unknown EF 625 619 PY (68%).

    Mean (SD) follow-up time was 2.4 (1.8) years, and median (IQR) follow-up time was 2.1 (0.7–4.3) years. Approximately 10% of patients had the maximum follow-up period of 5 years. Mean follow-up was shorter for patients with unknown subtypes than for those with HFrEF and HFpEF (2.2 years vs 3.0 years and 2.8 years, respectively).

    Male and female patients with HFpEF were more evenly split (48.2% male, 51.8% female) than patients with HFrEF (65.1% male, 34.9% female; table 1). On average, patients with an unknown subtype were older than those with a diagnosed subtype; this was particularly noticeable in the incident HF cohort (mean age: 76 years for unknown vs 71 years for known subtype) (see online supplemental table S3).

    View this table:
    Table 1

    Patient characteristics and comorbidities

    At index, the prevalence of chronic kidney disease (CKD) and T2D was similar for patients with HFpEF and HFrEF (CKD: 28.6% vs 33.5%; T2D: 23.5% vs 25.9%, respectively).

    Primary outcomes

    HHF and HF-related GP consultations

    Overall, 11% of patients (n=41 763) experienced at least one HHF during follow-up, with a total of 65 107 hospitalisations where HF was the primary diagnosis observed during the study period. Including recurrent hospitalisations, patients experienced HHF at an observed rate of approximately 70.9 per 1000 PY, although estimated crude rates were slightly higher (93.4 per 1000 PY, 95% CI 92.2 to 94.6). After age and sex adjustment, HHF rates were highest for patients with HFrEF (101.6 per 1000 PY, 95% CI 98.8 to 104.5) and unknown subtype (95.0 per 1000 PY, 95% CI 93.5 to 96.5) and lowest for patients with HFpEF (41.8 per 1000 PY, 95% CI 39.7 to 43.9; table 2). In the prevalent HF cohort, observed HHF rates were also higher in patients with HFrEF and unknown subtype (78.4 and 60.5 per 1000 PY, respectively) than in patients with HFpEF (39.4 per 1000 PY). A similar pattern was observed in the incident HF cohort (see online supplemental table S7).

    View this table:
    Table 2

    Observed and estimated (crude and adjusted) event rates for HF-related GP consultations, HHF, all-cause hospitalisations and all-cause GP consultations, overall and by HF subtype

    The estimated crude rate (including recurrent events) of HF-related GP consultations was 124.4 per 1000 PY (95% CI 122.9 to 126.0). Again, rates were highest in patients with HFrEF and unknown subtype, and lowest in patients with HFpEF after adjusting for age and sex (table 2). In the prevalent HF cohort, observed HF-related GP consultations were highest in patients with HFrEF (164.9 per 1000 PY), followed by patients with HFpEF (97.2 per 1000 PY) and unknown subtype (66.9 per 1000 PY). In the incident HF cohort, the highest rate was also observed in patients with HFrEF (228.5 per 1000 PY), though this was followed by patients with unknown subtype (137.1 per 1000 PY) and then HFpEF (86.2 per 1000 PY) (see online supplemental table S7).

    Mortality

    Overall, the all-cause mortality rate was observed to be 131.9 per 1000 PY (table 3). One-year mortality was highest in patients with unknown subtype (20%), followed by HFrEF (8%) and HFpEF (6%). By 5 years, mortality was 48% in those with unknown subtype, 35% in HFrEF patients and 25% for HFpEF (see online supplemental table S4).

    View this table:
    Table 3

    Observed and estimated (crude and adjusted) event rates for all-cause and cardiovascular mortality, overall and by HF subtype

    For patients with HFpEF, the age-adjusted and sex-adjusted all-cause mortality rates rose from 71 per 1000 PY in patients with no prior hospitalisations to 751 per 1000 PY in patients with three or more hospitalisations. Similarly, all-cause mortality rates for patients with HFrEF and unknown HF subtype also increased as the number of prior hospitalisations increased (see online supplemental table S5).

    Overall, the CV mortality rate was observed to be 49 per 1000 PY (table 3). CV mortality accounted for 35%, 45% and 36% of all deaths in the unknown, HFrEF and HFpEF groups, respectively. Unadjusted cumulative incidence estimates showed that CV and non-CV mortality rates increased over time and were consistently highest for patients with an unknown HF subtype (see figure 2, online supplemental table S6).

    Figure 2
    Figure 2

    Unadjusted cumulative incidence functions for cardiovascular and other-cause mortality by HF subtype. Left: cardiovascular mortality Kaplan-Meier survival estimates. Right: other-cause mortality Cox proportional-hazards estimates (estimates at the mean value are for adjusted covariates). Estimates were made allowing for competing risks. HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction.

    Within the prevalent HF cohort, observed all-cause mortality rates were higher in patients with HFrEF and unknown subtype (90.8 and 121.2 per 1000 PY, respectively) than in patients with HFpEF (65.7 per 1000 PY). A similar pattern was observed in the incident HF cohort (see online supplemental table S7). In both the prevalent and incident HF cohorts, CV mortality rates were highest in patients with unknown subtype, followed by patients with HFrEF and then HFpEF (see online supplemental table S7).

    In patients with comorbidities (eg, T2D and CKD), rates of HHF, CV and all-cause mortality in the overall cohort were higher than in those without (see online supplemental tables S8 and S9).

    HF medication use

    As indicated by the baseline characteristics, patients in the HFrEF subgroup had the highest utilisation of HF medications during follow-up, followed by those in the HFpEF and then those in the unknown subgroups. In the HFrEF group, prescription rates for ACEi/ARB and beta blockers were 9.1 and 9.3 per PY, respectively, with prescriptions issued once every 6 weeks on average. Observed prescribing of later-line treatments during follow-up was lower, particularly for those that are less likely to be prescribed and managed in primary care.

    Secondary endpoints

    Mean 1-year and 2-year changes in eGFR (estimated from serum creatinine test results) were slightly larger in the HFrEF and unknown subgroups (see online supplemental figure S1), but the 5-year change was similar in all subgroups (estimated at −6.7 mL/min/1.73 cm2, −6.3 mL/min/1.73 cm2 and −6.2 mL/min/1.73 cm2 for the HFrEF, HFpEF and unknown subgroups, respectively; see table 4).

    View this table:
    Table 4

    Estimated 1-year, 2-year and 5-year absolute change in eGFR in patients with HF with reduced, preserved or unknown ejection fraction

    Subgroup analyses revealed that, in general, poorer outcomes occurred in older age groups, males (after age adjustment), patients with a baseline eGFR <60 (see online supplemental table S11), and those with a history of HHF (in the prior 12 months), T2D or CKD (see online supplemental tables S8, S9 and S12–S14).

    Discussion

    In this large-scale, real-world, UK database study, patients with coded HFrEF, HFpEF or HF with uncoded subtype experienced high rates of all-cause hospitalisation (1654.8 per 1000 PY), GP consultations (22 190 per 1000 PY) and death (131.9 per 1000 PY). Although HF was not the primary cause of the majority of hospitalisations for patients with a recorded HF subtype, HF-related rehospitalisations during the study were substantial (36% of all HHF).

    Over two-thirds of patients with HF (68%) did not have a recorded subtype based on either coding or measurement of EF. Given the high rate of missing data in our analysis, all outcome findings should be interpreted with caution due to the possibility of selection bias.

    The baseline characteristics of patients with unknown HF subtype in our study suggest that patients with HFpEF may be disproportionately under-represented in our analysis. For example, patients with an unknown HF subtype were older than those with recorded HFrEF or HFpEF, whose mean ages were similar in our analysis. As age is a stronger risk factor for HFpEF than HFrEF, independent of sex,16 there may have been a higher proportion of patients with HFpEF in the unknown subtype group. Similarly, the higher proportion of comorbidities in the unknown subtype group is consistent with a potentially higher proportion of non-coded patients with HFpEF, since HFpEF is more commonly associated with comorbidities than HFrEF.17 HFpEF is observed more often in women than men;18 however, the ratio of female to male patients in our analysis was similar (51% vs 49%). In light of these considerations, all conclusions relating to differential outcomes between patients with HFpEF and HFrEF should be interpreted with caution and evaluated in the light of other similar studies to assess their external validity. Importantly, however, lack of information on subtype does not negate the validity of our findings regarding clinical characteristics and outcomes for HF patients, and how recording of subtype in primary care may impact this. For example, if patients with an unknown subtype experience worse outcomes, it may be important to consider why this is the case and what could be done to change this.

    Comparison of primary outcome findings with existing literature

    Two existing CPRD-based studies can provide comparator patient characteristics for the overall HF population in this study.11 12 Both of these studies recorded patient characteristics at the time of HF diagnosis, and hence the relevant comparator group in our study is the incident HF group. Key characteristics show good concordance between these prior studies and the incident HF group in our study, in terms of age, male/female split, body mass index and the prevalence of cardiac and non-cardiac comorbidities (see online supplemental table S3). Although based on a similar data source, the prior studies used the CPRD GOLD database, whereas our study used CPRD Aurum, which is larger and based on English practices only. This suggests that in terms of the overall HF population, the phenotypes of UK HF patients we present here remain similar to those characterised in earlier studies, and our study, therefore, provides a reliable and generalisable description of the typical HF patient.

    Previously published data on hospitalisation rates after chronic HF diagnosis have shown that between 2012 and 2015, during the first year after diagnosis, the age-adjusted rate of HF-specific hospitalisation rose from 172 to 221 per 1000 PY.19 This is higher than the rate estimated for patients with incident HF in this study (115.9 per 1000 PY; see online supplemental table S7). Despite this, the age-adjusted rate from year 2 onwards was lower (97 per 1000 PY; 95% CI 9.1 to 10.3),19 and more comparable to the rates we estimated for the prevalent HHF population.

    For HFpEF specifically, a previous study, using data from the Hull LifeLab database, estimated a HHF rate of 150 per 1000 PY, with a rate of all-cause hospitalisation of 1050 per 1000 PY.20 In our analysis, the age-adjusted and sex-adjusted rate of HHF for patients with HFpEF during follow-up was substantially lower (41.8 per 1000 PY), as was the adjusted rate for patients with prevalent HFpEF at index (48.1 per 1000 PY). However, the rate of all-cause hospitalisation was more comparable (1278 per 1000 PY).

    The differences in the HFrEF and HFpEF rates observed in our study compared with existing trials and HF registries may be partly related to definitions of HHF. Our study only counts hospitalisations for which HF is the primary admission reason (based on the International Statistical Classification of Diseases and Related Health Problems 10 code), which may exclude hospitalisations that are attributed to HF in trials or in registry data where the hospitalisation cause is more closely assessed. The comparable all-cause hospitalisation rates for HFpEF (an outcome which is less prone to nuances in definition) observed in this study and the Hull LifeLab study support this interpretation.20

    All-cause and CV mortality rates from a previous study based on Hull LifeLab data were reported to be 102 per 1000 PY and 43 per 1000 PY, respectively, in patients with HFpEF.20 Rates in both the prevalent (65.7 and 24.8 per 1000 PY, respectively) and incident (44.6 and 14.4 per 1000 PY, respectively) HFpEF subgroups in this study were substantially lower. These differences may be a consequence of selection bias and the high levels of missing data on subtype, as previously discussed. A previous study of 1794 patients with HFrEF attending outpatient cardiology clinics in four hospitals21 observed a 5-year all-cause mortality rate of approximately 40%. Our study observed a similar unadjusted mortality rate of 35% in the HFrEF subgroup overall and 36% in the prevalent HFrEF subgroup (see online supplemental tables S4 and S16). The greater similarity in mortality outcomes between our analysis and existing research in HFrEF versus HFpEF may support the possibility that HFpEF patients are disproportionately represented in our unknown group, with potential differences in characteristics and prognoses between patients with HFpEF who are coded versus those who are not.

    Comparison between HFrEF, HFpEF and unknown subgroups

    After age and sex adjustment, our results suggest that patients with HFpEF experience lower mortality, HHF and HF-related GP consultation rates than patients with HFrEF. For all-cause hospitalisations and GP consultations, the rates were more similar between the two groups. Although this may be clinically plausible, a potential explanation for the difference in HHF rates is that HF is less well recognised as the cause of hospitalisation in patients with HFpEF.

    Similar trends were observed in differences between patients with HFrEF and those with HFpEF for all subgroup analyses performed, although it should be acknowledged that for the all-cause and cause-specific mortality estimates, the plots and survival estimates do not adjust for age or sex; therefore, these results should be interpreted with caution.

    Outcomes for those with unknown subtype were generally poorer than those with recorded subtype, even after adjustments for age and sex. The reason for this is unclear; however, it is likely related to the underlying reason for, or consequence of, these individuals having no subtype recorded (eg, it could lead to suboptimal treatment and care, or patients may have received secondary care only (eg, in emergency situations), so primary care subtype records are lacking).

    Interpretation of other findings

    Despite being broadly consistent (in terms of overall HF) with other HF studies linking primary and secondary care records, the rates of HF as a primary cause of first and recurrent hospitalisation were lower than expected. However, further context is needed to fully interpret this finding. As mentioned, only 11% of patients (41 763) had at least one hospitalisation during follow-up, meaning that 36% (23 344/65 107) of all hospitalisations observed were rehospitalisations during the study period. Our subgroup analysis also clearly demonstrates that risk of HHF is greater in patients who had experienced HHF in the year prior to index versus those who had not (3.5 and 8.8 times higher for patients with HFrEF and HFpEF, respectively; see online supplemental table S15). Therefore, although our data may on average suggest a low rate of rehospitalisation in the overall population, the rate of rehospitalisation following an initial HHF was substantial, regardless of HF subtype.

    In 2015, the overall UK mortality rate for patients aged 70–74 years was reported to be 19.6 per 1000 persons, and 33.8 in those aged 75–79 years.22 In our analysis, mortality rates in the 65–74 years and 75–84 years subgroups were substantially higher than those for all HF subtype groups. Although the age bands used previously do not match the ones used in this study, they span the mean ages of the HFrEF and HFpEF cohorts. Overall, the HFpEF mortality rate was approximately three times higher than the previously reported 70–74 years rate, and the HFrEF and unknown subtypes were even higher.

    CV mortality was the most common cause of death observed in our study, yet it represented only 37% of all deaths occurring during follow-up (45% for HFrEF, 36% for HFpEF), suggesting that many patients with HF die of other complications of comorbidities.23–25 In fact, other research has shown that comorbidities account for much of the excess mortality risk in patients with HF.21 Given that many comorbidities may also be CV in nature, the low proportion of CV mortality could also be driven by data limitations. In this study, we used death certificates to determine the primary cause of patient death. Deaths with CV listed as the primary, but not contributory, cause of death were included to avoid overestimating CV-related death, but this likely led to an underestimation in comparison to situations where more detailed cause of death sources, such as clinical trial adjudication data, are available.

    Generally, the subgroup analysis indicates that the burden of both HFrEF and HFpEF increases with age in those with existing comorbidities, as well as among those with longer HF duration or who have experienced a recent HHF. After age adjustment, outcomes were also observed to be worse in males than in females, although the reverse was true prior to adjustment, suggesting that females diagnosed with HF are generally older than males. Further, the general trend was that outcomes were poorer in those with T2D and those with CKD, even after age and sex adjustment. However, an exception to this trend was observed in the CKD subgroup analysis for kidney function decline. In patients with an existing diagnosis of CKD, or a baseline eGFR of <60, decline in eGFR through the study period was smaller than those without. This may be explained by difference in baseline eGFR: by starting at a lower level, there may be a floor effect. Furthermore, patients with prior history of CKD or with a baseline eGFR <60 may receive additional monitoring and/or treatments to reduce eGFR decline versus patients without.

    Limitations

    As acknowledged above, the main limitations of this study are the high proportion of patients without a recorded HF subtype and the potential for selection bias in subtype recording, particularly against patients with HFpEF. Consequently, our findings may not be generalisable to all patients with HFpEF in the UK. Reasons for the suboptimal recording of HF subtype are unclear but may include inadequate use of echocardiography and other diagnostic tests, and/or poor official recording of measured EF in primary care (discussed further in Bellanca et al26).

    It should also be acknowledged that the definition of HFpEF adopted in our analysis (EF >40%) aligns with the European guideline definition of HFmrEF, rather than HFpEF (EF ≥50%).3 This was intentional, as HFmrEF was still a relatively new concept at the time of the analysis and was therefore assumed to be poorly coded in a clinical setting. Another limitation is that age and sex adjustments could not be performed for all analyses, including the Kaplan-Meier curves for all-cause mortality and CIFs for CV mortality, along with the corresponding 1-year, 2-year and 5-year rates, which may potentially affect the robustness of these results. For the CIFs, due to evidence of non-proportional hazards, the Fine and Gray method,27 which allows covariate adjustment, was used to estimate CIFs separately by subgroup using a function that did not allow for covariate adjustment. As such, these secondary results must be interpreted in the context of the wider analyses that did include age and sex adjustment. Lastly, we acknowledge that resource use and outcomes can vary in the first year postdiagnosis of HF. However, subanalyses stratified by prevalent versus incident HF showed findings that were broadly consistent with the main analysis.

    Conclusion

    Classification of HF subtype (HFrEF or HFpEF) based on coding or measurement of EF is suboptimal within the English healthcare system. All patients with HF, regardless of recorded subtype, experience high rates of comorbidities, mortality and healthcare resource use, including both HF-related and all-cause hospitalisations and GP consultations. Outcomes are worst for patients with unknown subtype, possibly reflecting suboptimal treatment, and for patients with comorbidities such as T2D and CKD. Ensuring that patients receive early and appropriate guideline-directed therapies to manage HF and its associated comorbidities is likely to improve patient care and reduce the burden of HF on the English healthcare system.

    Data availability statement

    All data relevant to the study are included in the article or uploaded as online supplemental information. The results of all data analyses are contained within the manuscript and online supplemental material.

    Ethics statements

    Patient consent for publication

    Ethics approval

    Ethics approval for this study was granted by the Independent Scientific Advisory Committee (ISAC) of the UK Medicines and Healthcare products Regulatory Agency (ISAC number 20_000051) on 21 September 2020. The study was conducted in accordance with the principles established by the Declaration of Helsinki (2013). As this was a non-interventional cohort study based on deidentified patient-level data from healthcare databases reported in aggregate only, subjects were not required to provide informed consent. A waiver of informed consent was provided by the ISAC. The data provider (CPRD) has ethics approval from the Health Research Authority to support observational research using anonymised patient data.

    Acknowledgments

    The authors would like to thank Leana Bellanca of Boehringer Ingelheim Ltd. for her advisory input into the study design and her assistance in data interpretation. Medical writing support was provided by Rose Martin of Nucleus Global, a medical communications agency contracted and funded by Boehringer Ingelheim.

    References

    1. ↵
      1. Conrad N,
      2. Judge A,
      3. Canoy D, et al
      . Temporal trends and patterns in mortality after incident heart failure: a longitudinal analysis of 86 000 individuals. JAMA Cardiol 2019;4:1102–11. doi:10.1001/jamacardio.2019.3593
      OpenUrl
    2. ↵
      1. Richard Hobbs FD
      . Clinical burden and health service challenges of chronic heart failure. Br J Gen Pract 2010;60:611–5. doi:10.3399/bjgp10X515133
      OpenUrlFREE Full Text
    3. ↵
      1. Ponikowski P,
      2. Voors AA,
      3. Anker SD, et al
      . ESC guidelines for the diagnosis and treatment of acute and chronic heart failure: the task force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) developed with the special contribution of the Heart Failure Association (HFA) of the ESC. Eur Heart J 2016;37:2129–200. doi:10.1093/eurheartj/ehw128
      OpenUrlCrossRefPubMed
    4. ↵
      1. Yancy CW,
      2. Jessup M,
      3. Bozkurt B, et al
      . 2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association task force on practice guidelines. J Am Coll Cardiol 2013;62:e147–239. doi:10.1016/j.jacc.2013.05.019
      OpenUrlFREE Full Text
    5. ↵
      1. Hsu JJ,
      2. Ziaeian B,
      3. Fonarow GC
      . Heart failure with mid-range (borderline) ejection fraction: clinical implications and future directions. JACC Heart Fail 2017;5:763–71. doi:10.1016/j.jchf.2017.06.013
      OpenUrlAbstract/FREE Full Text
    6. ↵
      1. Shiga T,
      2. Suzuki A,
      3. Haruta S, et al
      . Clinical characteristics of hospitalized heart failure patients with preserved, mid-range, and reduced ejection fractions in Japan. ESC Heart Fail 2019;6:475–86. doi:10.1002/ehf2.12418
      OpenUrl
    7. ↵
      1. Dulai R,
      2. Sheikh A,
      3. Qureshi A
      . Prevalence, clinical characteristics and outcomes of HF with preserved versus reduced ejection fraction. Br J Cardiol 2016;23. doi:10.5837/bjc.2016.005
    8. ↵
      1. Fonarow GC,
      2. Stough WG,
      3. Abraham WT, et al
      . Characteristics, treatments, and outcomes of patients with preserved systolic function hospitalized for heart failure: a report from the OPTIMIZE-HF registry. J Am Coll Cardiol 2007;50:768–77. doi:10.1016/j.jacc.2007.04.064
      OpenUrlFREE Full Text
    9. ↵
      1. Dierckx R,
      2. Cleland JGF,
      3. Parsons S, et al
      . Prescribing patterns to optimize heart rate: analysis of 1,000 consecutive outpatient appointments to a single heart failure clinic over a 6-month period. JACC Heart Fail 2015;3:224–30. doi:10.1016/j.jchf.2014.11.003
      OpenUrlAbstract/FREE Full Text
    10. ↵
      1. Conrad N,
      2. Judge A,
      3. Canoy D, et al
      . Diagnostic tests, drug prescriptions, and follow-up patterns after incident heart failure: a cohort study of 93,000 UK patients. PLoS Med 2019;16:e1002805. doi:10.1371/journal.pmed.1002805
    11. ↵
      1. Conrad N,
      2. Judge A,
      3. Tran J, et al
      . Temporal trends and patterns in heart failure incidence: a population-based study of 4 million individuals. Lancet 2018;391:572–80. doi:10.1016/S0140-6736(17)32520-5
      OpenUrlCrossRefPubMed
    12. ↵
      1. Taylor CJ,
      2. Ordóñez-Mena JM,
      3. Roalfe AK, et al
      . Trends in survival after a diagnosis of heart failure in the United Kingdom 2000-2017: population based cohort study. BMJ 2019;364:l223. doi:10.1136/bmj.l223
    13. ↵
      1. Koudstaal S,
      2. Pujades-Rodriguez M,
      3. Denaxas S, et al
      . Prognostic burden of heart failure recorded in primary care, acute hospital admissions, or both: a population-based linked electronic health record cohort study in 2.1 million people. Eur J Heart Fail 2017;19:1119–27. doi:10.1002/ejhf.709
      OpenUrlPubMed
    14. ↵
      1. Owan TE,
      2. Hodge DO,
      3. Herges RM, et al
      . Trends in prevalence and outcome of heart failure with preserved ejection fraction. N Engl J Med 2006;355:251–9. doi:10.1056/NEJMoa052256
      OpenUrlCrossRefPubMedWeb of Science
    15. ↵
      1. Lewis JD,
      2. Bilker WB,
      3. Weinstein RB, et al
      . The relationship between time since registration and measured incidence rates in the general practice research database. Pharmacoepidemiol Drug Saf 2005;14:443–51. doi:10.1002/pds.1115
      OpenUrlCrossRefPubMedWeb of Science
    16. ↵
      1. Ho JE,
      2. Enserro D,
      3. Brouwers FP, et al
      . Predicting heart failure with preserved and reduced ejection fraction: the international collaboration on heart failure subtypes. Circ Heart Fail 2016;9. doi:10.1161/CIRCHEARTFAILURE.115.003116
    17. ↵
      1. Borlaug BA
      . Evaluation and management of heart failure with preserved ejection fraction. Nat Rev Cardiol 2020;17:559–73. doi:10.1038/s41569-020-0363-2
      OpenUrlCrossRefPubMed
    18. ↵
      1. Sotomi Y,
      2. Hikoso S,
      3. Nakatani D, et al
      . Sex differences in heart failure with preserved ejection fraction. J Am Heart Assoc 2021;10:e018574. doi:10.1161/JAHA.120.018574
    19. ↵
      1. Lawson CA,
      2. Zaccardi F,
      3. Squire I, et al
      . 20-year trends in cause-specific heart failure outcomes by sex, socioeconomic status, and place of diagnosis: a population-based study. Lancet Public Health 2019;4:e406–20. doi:10.1016/S2468-2667(19)30108-2
      OpenUrl
    20. ↵
      1. Fonseca AF,
      2. Lahoz R,
      3. Studer R, et al
      . Hospitalisation and mortality in patients with heart failure and left ventricular ejection fraction over 40% in a real world setting in the UK. Virtual. HFA Discoveries; 2020. 5–9.
    21. ↵
      1. Drozd M,
      2. Relton SD,
      3. Walker AMN, et al
      . Association of heart failure and its comorbidities with loss of life expectancy. Heart 2021;107:1417–21. doi:10.1136/heartjnl-2020-317833
      OpenUrlAbstract/FREE Full Text
    22. ↵
      1. Our World in Data
      . Death rate by age group in England and Wales. 2021. Available: https://ourworldindata.org/grapher/death-rate-by-age-group-in-england-and-wales?country=~England+and+Wales [Accessed 14 Apr 2021].
    23. ↵
      1. Henkel DM,
      2. Redfield MM,
      3. Weston SA, et al
      . Death in heart failure: a community perspective. Circ Heart Fail 2008;1:91–7. doi:10.1161/CIRCHEARTFAILURE.107.743146
      OpenUrlAbstract/FREE Full Text
    24. ↵
      1. Vergaro G,
      2. Ghionzoli N,
      3. Innocenti L, et al
      . Noncardiac versus cardiac mortality in heart failure with preserved, midrange, and reduced ejection fraction. J Am Heart Assoc 2019;8:e013441. doi:10.1161/JAHA.119.013441
    25. ↵
      1. Nakamaru R,
      2. Shiraishi Y,
      3. Sandhu AT, et al
      . Cardiovascular vs. non-cardiovascular deaths after heart failure hospitalization in young, older, and very old patients. ESC Heart Fail 2023;10:673–84. doi:10.1002/ehf2.14245
      OpenUrl
    26. ↵
      1. Bellanca L,
      2. Linden S,
      3. Farmer R
      . Incidence and prevalence of heart failure in England: a descriptive analysis of linked primary and secondary care data – the PULSE study. BMC Cardiovasc Disord 2023;23:374. doi:10.1186/s12872-023-03337-1
    27. ↵
      1. Fine JP,
      2. Gray RJ
      . A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999;94:496–509. doi:10.1080/01621459.1999.10474144
      OpenUrlCrossRefWeb of Science

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • Contributors RF and NDG conceived and designed the study. RF analysed the data and SL wrote the initial draft. All authors provided input on the study design, data analysis and interpretation of the results; revised the paper critically for important intellectual content; and approved the final version. RF guarantees the overall content of the article on behalf of all authors.

    • Funding Boehringer Ingelheim funded the design of this study, and the collection, analysis and interpretation of data. Boehringer Ingelheim also funded medical writing support.

    • Competing interests At the time of study conduct, all authors were full-time employees of Boehringer Ingelheim. The authors declare no other competing interests.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.