Simplification of a registry-based algorithm for ejection fraction prediction in heart failure patients: Applicability in cardiology centres of the Netherlands

Elisa Dal Canto; Alicia Uijl; N Charlotte Onland-Moret; Sophie H Bots; Leonard Hofstra; Igor Tulevski; Folkert W Asselbergs; Pim van der Harst; G Aernout Somsen; Hester M den Ruijter

doi:10.1371/journal.pone.0310023

. 2024 Nov 5;19(11):e0310023. doi: 10.1371/journal.pone.0310023

Simplification of a registry-based algorithm for ejection fraction prediction in heart failure patients: Applicability in cardiology centres of the Netherlands

Elisa Dal Canto ^1,^2,^*, Alicia Uijl ^2,^3,⁴, N Charlotte Onland-Moret ², Sophie H Bots ⁵, Leonard Hofstra ^3,⁶, Igor Tulevski ^3,⁶, Folkert W Asselbergs ³, Pim van der Harst ⁷, G Aernout Somsen ⁶, Hester M den Ruijter ¹

Editor: Satoshi Higuchi⁸

PMCID: PMC11537407 PMID: 39499710

Abstract

Background

Left ventricular ejection fraction (EF) is used to categorize heart failure (HF) into phenotypes but this information is often missing in electronic health records or non-HF registries.

Methods

We tested the applicability of a simplified version of a multivariable algorithm, that was developed on data of the Swedish Heart Failure Registry to predict EF in patients with HF. We used data from 4,868 patients with HF from the Cardiology Centers of the Netherlands database, an organization of 13 cardiac outpatient clinics that operate between the general practitioner and the hospital cardiologist. The algorithm included 17 demographical and clinical variables. We tested model discrimination, model performance and calculated model sensitivity, specificity, positive and negative predictive values for EF ≥ vs. <50% and EF ≥ vs. <40%. We additionally performed a multivariable multinomial analysis for all three separate HF phenotypes (with reduced, mildly reduced and preserved EF) HFrEF vs. HFmrEF vs. HFpEF. Finally, we internally validated the model by using temporal validation.

Results

Mean age was 66 ±12 years, 44% of patients were women, 68% had HFpEF, 17% had HFrEF, and 15% had HFmrEF. The C-statistic was of 0.71 for EF ≥/< 50% (95% CI: 0.69–0.72) and of 0.74 (95% CI: 0.73–0.75) for EF ≥/< 40%. The model had the highest sensitivities for EF ≥50% (0.72, 95% CI: 0.63–0.75) and for EF ≥40% (0.70, 95% CI: 0.65–0.71). Similar results were achieved by the multinomial model, but the C-statistics for predicting HFpEF vs HFrEF was lower (0.61, 95% CI 0.58–0.63). The internal validation confirmed good discriminative ability.

Conclusions

A simple algorithm based on routine clinical characteristics can help discern HF phenotypes in non-cardiology datasets and research settings such as research on primary care data, where measurements of EF is often not available.

Introduction

Heart failure (HF) is a complex syndrome with high morbidity and mortality and its prevalence continues to rise [1]. Guidelines recommend the use of left ventricular ejection fraction (EF) assessed by echocardiography to categorize HF into three phenotypes: HF with preserved EF (HFpEF, EF≥50%), HF with mildly reduced EF (HFmrEF, EF = 40–49%) and HF with reduced EF (HFrEF, EF<40%) [2]. Because HF management differs according to the EF-based phenotype [2], accurate identification and categorization of patients is of utmost importance. HFpEF is currently the most common HF phenotype [3] and its diagnosis can be challenging [1]. Indeed HFpEF remains often undetected as symptoms and signs are not specific especially at rest, and affected subjects are therefore not correctly identified and promptly referred to specialized care. Moreover, treatment options are currently limited [4, 5].

Electronic health records (EHRs) comprise a wealth of information about patients diagnosis and treatment and they are widely used in research and clinical care, including HF care [6]. They have the potential to facilitate either HF research by for instance allowing wide screening for clinical trials and creation of registries, and might help to reduce variation in HF management thereby improving patient outcomes. However, EHRs sometimes lack imaging data and particularly information on EF is often not reported, thereby hampering ascertainment of the HF phenotype and identification of those with HFpEF.

A multivariable algorithm based on demographical and clinical characteristics has been developed using data of the Swedish Heart Failure Registry (SwedeHF) to predict EF in patients with HF [7]. The algorithm was externally validated in the CHECK-HF registry in The Netherlands, showing high discriminating power between HF phenotypes (C-statistic: 0.78 [95% confidence interval (CI) 0.77–0.78] for EF ≥50% and 0.76 (0.75–0.76) for EF ≥40%. Nevertheless, given the varying characteristics of HF populations, additional external validations would be essential to determine the algorithm generalizability and its ability to provide accurate predictions in distinct clinical context. Accordingly, our aim was to provide a further external validation of the HF algorithm in patients with chronic HF from the Cardiology Centers of the Netherlands (CCN) database. However, given the characteristics of our study population including the presence of missing data in some of the predictors, we derived and assessed the applicability of a simplified version, and thus more widely applicable, of the original model.

Materials and methods

Study population

The CCN is an organization of 13 cardiac outpatient clinics that operate between the general practitioner (GP) and the hospital [8]. Patients are referred to CCN on suspicion of cardiovascular disease (CVD) by their GP and undergo an initial standardized diagnostic workup that includes echocardiography, ultrasound of the carotid arteries, exercise stress test, electrocardiography, laboratory tests and a consult with a nurse where anthropometrics and information on symptoms, medical history and medication use are collected. If needed, patients can be referred for further diagnostic workup to a hospital or invited to CCN for follow-up visits.

Among a total of 109,151 CCN patients, we selected and analyzed those diagnosed with HF between June 2007 and February 2018 (index dates), for whom a quantified EF was available at the time of diagnosis (S1 Fig). We defined HFrEF as EF <40%, HFmrEF as EF between 40% and 49%, and HFpEF as EF ≥50%. Patients missing information on demographical, anthropometrical and biochemical predictors included in the diagnostic algorithm were excluded if more than three months had passed between the assessment of EF and the assessment of such predictors (S1 Fig). A timeframe longer than three months from EF assessment was considered acceptable for information on co-morbidities and medication use. EF was assessed with biplane Simpson in 7% of cases and 93% of patients with Teich method.

The medical research ethics committee of the University Medical Center Utrecht approved this study (proposal number 17/359). Patient consent for publication was not required. The Cardiology Centers of the Netherlands data were made available under implied consent and transferred to the University Medical Center Utrecht under the Dutch Personal Data Protection Act. This study used data collected during the regular care process and did not subject participants to additional procedures or impose behavioral patterns on them. Finally, to the purpose of the present analysis, the CCN data were accessed between February 2022 and January 2024.

Statistical analysis

Patients were stratified according to the HF phenotype and comparisons between the groups were carried out by analysis of variance for continuous variables and chi-squared for categorical variables. The main multivariable model included 22 variables: age, sex, NT-proBNP plasma level, New York Heart Association functional class, mean arterial pressure, heart rate, body mass index, estimated glomerular filtration rate (eGFR), history of ischemic heart disease, anemia, chronic obstructive pulmonary disease, diabetes, atrial fibrillation, hypertension, valvular heart disease, malignant cancer, device therapy, use of renin angiotensin system inhibitors (including ACE-inhibitors and angiotensin receptor blockers), beta-blockers, mineralocorticoid receptor antagonists, digoxin and diuretics. A simpler model excluded NT-proBNP and New York Heart Association class [7]. In the present analysis, we tested the applicability of a further simplified model which additionally excluded chronic obstructive pulmonary disease, malignant cancer and device therapy, as this information was unavailable in most patients. Missing values were present in a proportion ranging from 1.42% for the presence of hypertension to 39.17% for anemia (S1 Table) and imputed by using multiple imputation by chained equations (n = 10 imputations, mice package from R statistical software version 1.3.1093). Results were then pooled using Rubin’s rule. We performed logistic regression to fit the model and used area under the receiver operating curves to discern model discrimination. Model discrimination was assessed for EF ≥ vs. <50% and EF ≥ vs. <40%. We assessed model performance with C-statistics. We calculated sensitivity, specificity, positive and negative predictive values along with 95% CI. Secondly, we used a multinomial logistic model to separately predict HFpEF, HFmrEF, and HFrEF (HFrEF was used as reference). In this case model C-statistics was calculated for all pairs of categories using the conditional risk method [9].

Finally, internal validation of the algorithm was carried out by using temporal validation. The dataset was split into training and testing sets based on the time of HF diagnosis, with 75% of the data used for training and the remaining 25% for testing, ensuring that the model was trained on historical data and evaluated on future observations. Logistic regression models were trained using the training set, and predictions were made on the testing set. Model accuracy was assessed using C-statistics and bootstrapping techniques were employed to estimate the 95% CIs. For the multinomial logistic model predicting HFpEF, HFmrEF, and HFrEF (with HFrEF as the reference), accuracy was calculated for all pairs of categories using the conditional risk method.

Results

A total of 4,868 patients with HF were included (44% women, aged 66 ±12 years). Fifty-five percent of patients had hypertension, 15% had diabetes, and 20% had atrial fibrillation. Sixty-eight percent of patients had HFpEF, 17% had HFrEF, and 15% had HFmrEF (Table 1).

Table 1. Baseline characteristics of the cardiology centers of the Netherlands population.

	Overall	HFrEF	HFmrEF	HFpEF	p^a
Demographics
n	4868	827	719	3322
Age (>75 vs <75 years)	65.8 (12.2)	68.3 (11.8)	67.4 (12.6)	64.8 (12.1)	<0.001
Sex (female vs male)	2137 (43.9)	266 (32.2)	294 (40.9)	1577 (47.5)	<0.001
Clinical variables
MAP ≥ 90 mmHg, n (%)	917 (85.8)	217 (76.4)	171 (85.5)	529 (90.4)	<0.001
Heart rate ≥ 70 bpm, n (%)	2605 (63.8)	428 (74.0)	347 (63.1)	1830 (62.0)	<0.001
SBP, mmHg	147.5 (25.4)	137.0 (23.9)	148.3 (25.5)	152.3 (24.7)	<0.001
DBP, mmHg	86.1 (14.8)	84.8 (15.6)	86.9 (16.6)	86.4 (13.8)	0.232
BMI, n (%)					0.003
< 18.5 kg/m2	284 (26.4)	90 (31.0)	66 (32.2)	128 (22.0)
18.5–24.9 kg/m2	9 (0.8)	2 (0.7)	1 (0.5)	6 (1.0)
25–29.9 kg/m2	450 (41.8)	117 (40.3)	92 (44.9)	241 (41.4)
≥ 30.0 kg/m2	334 (31.0)	81 (27.9)	46 (22.4)	207 (35.6)
eGFR, n (%)
> = 90 mL/min/1.73m2	1438 (47.4)	222 (40.1)	177 (39.4)	1039 (51.1)	<0.001
60–89.9 mL/min/1.73m2	1043 (34.4)	182 (32.9)	172 (38.3)	689 (33.9)
59.9–30 mL/min/1.73m2	494 (16.3)	125 (22.6)	89 (19.8)	280 (13.8)
<30 mL/min/1.73m2	60 (2.0)	24 (4.3)	11 (2.4)	25 (1.2)
Ischemic heart disease, n (%)	405 (8.3)	75 (9.1)	65 (9.0)	265 (8.0)	0.447
Anemia	263 (8.9)	58 (11.7)	56 (12.8)	149 (7.4)	<0.001
Atrial fibrillation	969 (19.9)	225 (27.2)	181 (25.2)	563 (16.9)	<0.001
Diabetes	741 (15.5)	153 (19.1)	115 (16.3)	473 (14.4)	0.003
Hypertension	2649 (55.2)	326 (40.7)	358 (50.7)	1965 (59.7)	<0.001
Valvular disease	884 (18.2)	192 (23.2)	165 (22.9)	527 (15.9)	<0.001
Therapy
RAAS agents	3044 (64.2)	576 (70.4)	473 (66.8)	1995 (62.1)	<0.001
Beta-blockers	2543 (53.7)	540 (66.0)	435 (61.4)	1568 (48.8)	<0.001
Diuretics	2589 (54.6)	582 (71.1)	438 (61.9)	1569 (48.8)	<0.001
MRA	755 (15.9)	287 (35.1)	141 (19.9)	327 (10.2)	<0.001
Digoxin	340 (7.2)	107 (13.1)	70 (9.9)	163 (5.1)	<0.001

Open in a new tab

^a The p value refers to the comparisons between the groups, performed by analysis of variance (ANOVA) for continuous variables and chi-squared for trend for categorical variables.

Abbreviations: MAP: mean arterial pressure; BMI: body mass index; eGFR: estimated glomerular filtration rate; RAAS: renin-angiotensin-aldosterone system: MRA: mineralocorticoid receptor antagonists.

The performance of each predictor of the diagnostic algorithm to predict EF ≥50% vs. <50% and EF ≥40% vs. <40% is presented in Table 2. The strongest predictors for both EF ≥50% and EF ≥40% were female sex and presence of arterial hypertension. Use of diuretics, of mineralocorticoid receptor antagonists and presence of atrial fibrillation were the strongest predictors for EF <50% and EF <40%. Some predictors such as body mass index and advanced age, that were positively associated with EF ≥50/40% in the original study, did not show any significant associations in this analysis.

Table 2. Multivariable logistic prediction models predicting EF ≥ 50% vs. EF < 50% and EF ≥ 40% vs. <40%.

Variables	LVEF ≥50%		LVEF ≥40%
	OR (95% CI)	P value	OR (95% CI)	P value
Intercept	3.33 (1.83–7.08)	<0.001	6.30 (3.06–13.0)	<0.001
Age (>75 vs <75 years)	0.83 (0.68–1.01)	0.060	0.85 (0.6–1.07)	0.1287
Sex (female vs male)	1.81 (1.57–2.07)	<0.0001	2.08 (1.75–2.49)	<0.0001
MAP (≥ 90 vs < 90 mmHg)	1.00 (0.99–1.01)	0.7632	1.00 (0.99–1.01)	0.2979
Heart rate (≥ 70 vs < 70 bpm)	0.81 (0.70–0.94)	0.0047	0.68 (0.57–0.82)	<0.0001
BMI
< 18.5 vs 18.5–24.9	1.39 (0.44–4.37)	0.5614	1.11 (0.31–4.02)	0.8649
25–29.9 vs 18.5–24.9	1.07 (0.83–1.36)	0.5825	1.14 (0.88–1.49)	0.3095
≥ 30.0 vs 18.5–24.9	1.13 (0.89–1.43)	0.2999	1.13 (0.83–1.54)	0.4704
eGFR (mL/min/1.73m2)
60–89.9 vs > = 90	0.85 (0.72–1.01)	0.0637	0.91 (0.73–1.13)	0.4026
59.9–30 vs > = 90	0.75 (0.57–0.97)	0.0298	0.73 (0.52–1.01)	0.0598
<30 vs > = 90	0.52 (0.30–0.90)	0.0192	0.44 (0.25–0.78)	0.0056
Ischemic heart disease	0.86 (0.68–1.08)	0.1334	0.86 (0.65–1.14)	0.3079
Anemia	0.92 (0.69–1.23)	0.5767	1.10 (0.81–1.51)	0.5244
Atrial fibrillation	0.65 (0.55–0.76)	<0.0001	0.66 (0.55–0.81)	<0.001
Diabetes	0.81 (0.67–0.97)	0.0288	0.76 (0.61–0.96)	0.0228
Hypertension	1.86 (1.62–2.14)	<0.0001	2.11 (1.78–2.51)	<0.0001
Valvular disease	0.74 (0.62–0.87)	<0.0001	0.82 (0.66–1.00)	0.0503
RAAS agents	0.89 (0.76–1.04)	0.1377	0.87 (0.72–1.05)	0.1446
Beta-blockers	0.69 (0.60–0.79)	<0.0001	0.72 (0.60–0.86)	0.0003
Diuretics	0.68 (0.58–0.79)	<0.001	0.66 (0.53–0.81)	<0.0001
MRA	0.41 (0.34–0.53)	<0.0001	0.37 (0.30–0.46)	<0.0001
Digoxin	0.81 (0.62–1.05)	0.1048	0.83 (0.62–1.09)	0.1863

Open in a new tab

MAP: mean arterial pressure; BMI: body mass index; eGFR: estimated glomerular filtration rate; RAAS: renin-angiotensin-aldosterone system: MRA: mineralocorticoid receptor antagonists.

The model discriminated adequately for both EF ≥50% and EF ≥40% with a C-statistics of 0.71 (95% CI: 0.69–0.72) and of 0.74 (95% CI: 0.73–0.75) respectively (Figs 1 and 2). For EF <50% and EF <40% the model discriminately equally well with C-statistics of 0.71 (95% CI: 0.69–0.72) and of 0.74 (95% CI: 0.73–0.75) respectively (Figs 3 and 4).

The model had the highest sensitivities for EF ≥50% (0.72, 95% CI: 0.63–0.75) and for EF ≥40% (0.70, 95% CI: 0.65–0.71) (Table 3).

Table 3. Sensitivity, specificity, positive and negative predictive values of the logistic prediction models.

	LVEF ≥50%	LVEF ≥40%	LVEF < 50%	LVEF < 40%
Sensitivity	0.72 (0.63–0.75)	0.70 (0.65–0.71)	0.60 (0.57–0.68)	0.67 (0.65–0.72)
Specificity	0.60 (0.57–0.68)	0.67 (0.65–0.72)	0.72 (0.63–0.75)	0.70 (0.65–0.71)
Positive predictive value	0.80 (0.79–0.81)	0.91 (0.90–0.92)	0.50 (0.47–0.52)	0.31 (0.29–0.32)
Negative predictive value	0.50 (0.47–0.52)	0.31 (0.29–0.32)	0.80 (0.79–0.81)	0.91 (0.90–0.92)

Open in a new tab

The results of the multinomial model are shown in Table 4. HFrEF was the reference category. Female sex and presence of arterial hypertension were the strongest predictors for HFmrEF. Predictors for HFpEF were the same as those for HFmrEF, but the associations were much stronger. C-statistics calculated for all pairs of outcome categories were similar to the logistic models for EF ≥50% or EF ≥40%, with C-statistics of 0.71 (95% 0.69–0.72) for HFmrEF vs HFrEF and of 0.74 (95% 0.72–0.76) for HFmrEF vs HFpEF. However, the discriminative performance for predicting HFpEF vs HFrEF was only moderate, with a C-statistic of 0.61 (95% CI 0.58–0.63).

Table 4. Multinomial logistic prediction models predicting HFmrEF vs HFrEF and HFpEF vs HFrEF.

Variables	HFmrEF		HFpEF
	OR (95% CI)	P value	OR (95% CI)	P value
Intercept	0.71 (0.27–1.83)	0.4739	5.75 (2.72–12.1)	<0.001
Age (>75 vs <75 years)	0.96 (0.71–1.28)	0.7822	0.81 (0.63–1.04)	0.0978
Sex (female vs male)	1.56 (1.25–1.95)	<0.0001	2.27 (1.89–2.72)	<0.0001
MAP (≥ 90 vs < 90 mmHg)	1.00 (0.99–1.01)	0.2900	1.00 (0.99–1.01)	0.3625
Heart rate (≥ 70 vs < 70 bpm)	0.70 (0.55–0.89)	0.0034	0.68 (0.56–0.82)	<0.0001
BMI
< 18.5 vs 18.5–24.9	0.71 (0.10–4.57)	0.7119	1.23 (0.32–4.70)	0.7524
25–29.9 vs 18.5–24.9	1.14 (0.81–1.60)	0.4359	1.14 (0.86–1.51)	0.3414
≥ 30.0 vs 18.5–24.9	1.05 (0.70–1.55)	0.8294	1.16 (0.85–1.58)	0.3531
eGFR (mL/min/1.73m2)
60–89.9 vs > = 90	1.06 (0.81–1.37)	0.6755	0.88 (0.70–1.10)	0.2483
59.9–30 vs > = 90	0.85 (0.57–1.28)	0.4441	0.69 (0.49–0.97)	0.0354
<30 vs > = 90	0.60 (0.25–1.28)	0.1707	0.41 (0.22–0.74)	0.0003
Ischemic heart disease	0.96 (0.67–1.37)	0.8223	0.83 (0.63–1.12)	0.2406
Anemia	1.27 (0.87–1.86)	0.1767	1.03 (0.74–1.46)	0.8212
Atrial fibrillation	0.89 (0.69–1.13)	0.2958	0.61 (0.50–0.74)	<0.0001
Diabetes	0.83 (0.62–1.12)	0.2385	0.74 (0.58–0.94)	0.0151
Hypertension	1.54 (1.23–1.91)	0.0001	2.31 (1.94–2.76)	<0.0001
Valvular disease	1.04 (0.81–1.34)	0.7535	0.75 (0.61–0.93)	0.0080
RAAS agents	0.92 (0.72–1.17)	0.4911	0.85 (0.70–1.04)	0.1146
Beta-blockers	0.95 (0.76–1.19)	0.6654	0.67 (0.56–0.80)	<0.0001
Diuretics	0.85 (0.65–1.10)	0.2077	0.62 (0.51–0.77)	<0.0001
MRA	0.51 (0.39–0.67)	<0.0001	0.33 (0.26–0.42)	<0.0001
Digoxin	0.93 (0.65–1.32)	0.6801	0.78 (0.58–1.05)	0.7738

Open in a new tab

MAP: mean arterial pressure; BMI: body mass index; eGFR: estimated glomerular filtration rate; RAAS: renin-angiotensin-aldosterone system: MRA: mineralocorticoid receptor antagonists.

Models were internally validated by temporal validation, showing good discriminative performance, with a C-statistics of 0.72 (95% CI: 0.71–0.73) for EF ≥/< 50% and of 0.69 (0.68–0.70) for EF ≥/< 40% (S2 Table).

Discussion

In this study, we adapted a diagnostic algorithm originally designed for research purposes to predict EF among patients with HF in the Netherlands. This newly derived algorithm was applied to EHRs data obtained from Dutch cardiac screening centers, which represents a real-world clinical setting. Our findings demonstrated that the simplified version of the original algorithm performed adequately in predicting EF. As initially intended, this algorithm can be effectively utilized retrospectively on research data that have been collected on HF patients to ascertain the EF phenotype, thus serving as a valuable tool for further studies and investigations.

It’s important to underscore the algorithm’s primary domain of applicability: research-focused environments such as primary care data, non-cardiology research settings, and broader healthcare datasets where information about HF phenotypes is sporadic or unavailable. On the other hand, we realize that in clinical practice the use of echocardiography is essential and widely available for the diagnosis and management of HF [2, 10, 11]. However, our study focuses on research applications rather than replacing clinical assessments. While echocardiography remains the gold standard for EF assessment in clinical settings, our algorithm still provide a valuable research tool. Its integration into research initiatives has the potential to improve the accuracy of HF studies, especially in scenarios where EF measurements are lacking, such as EHRs. Furthermore, the utility of the algorithm might apply to other healthcare sources such as ICD-10 claims datasets which also often lack detailed clinical information.

This simplified version of the algorithm showed slightly worse performance compared to the performance demonstrated in the derivation and validation cohorts [7]. This might be due to the different characteristics of the study populations, with patients from the CCN database being younger, mostly affected by HFpEF, and healthier with lower prevalence of comorbidities such as diabetes and atrial fibrillation [8] compared to the SwedeHF and CHECK-HF. On the other hand, the aim of our study was to assess the algorithm’s robustness in a population that differs in terms of age and health status from the derivation and validation cohorts. This enhanced our understanding of its generalizability and strengthened its potential for widespread use.

Additionally, we used fewer predictors than the original model in our analysis, and this has likely influenced our findings. Female sex and arterial hypertension were confirmed as strong predictors of HFpEF, while eGFR<30 mL/min/1.73m² and use of mineralocorticoid receptor antagonists were shown to be predictors of HFrEF. However, other predictors yielded different associations compared to those of the original study: atrial fibrillation predicted HFrEF rather than HFpEF, while some predictors such as body mass index and advanced age did not show significant associations.

The algorithm showed the highest sensitivities when used for the identification of patients with EF ≥50% and patients with EF ≥40%, of 0.72 and of 0.70 respectively. On the other hand, sensitivities for EF < 40% and EF<50% were not satisfactory. Finally, the associations with HFmrEF in the multinomial model were overall weak, probably because of a limited number of participants classified as HFmrEF in our study population. These findings and disparities are possibly due to changes in the distribution of patient characteristics observed when HFmrEF was combined with HFrEF and HFpEF, implying that the HFmrEF group should not be combined with neither of the other groups, being a separate and distinct clinical entity [12].

Previous studies have developed diagnostic algorithms to predict EF in HF populations [13, 14]. One of these algorithms was developed from Medicare claims and subsequently externally validated in a sample of commercial insurance enrollees, demonstrating good accuracy [15]. However, this algorithm did not include laboratory test values and might therefore be more suitable in the context of insurance claims databases, in which this information is often missing.

Study limitations

Some limitations of the present analysis should be acknowledged. Firstly, we had missing data on several predictors, which is a common issue in EHRs. Accordingly, we evaluated the performance of a model that included 17 out of the 22 original variables. It is important to note that this newly derived model was not tested neither validated in the original manuscript. However, this allowed us to demonstrate that a model containing fewer variables that are commonly available in clinical practice might be more widely applicable and enable to discriminate HF phenotypes, showing particular good performance in identifying HFpEF patients. Furthermore, we internally validated this model through temporal validation, which confirmed its good discriminative ability. Secondly, EF was often assessed only qualitatively and not quantitatively in the CCN database, which has resulted in a relatively small sample size. Among the cases with quantitatively assessed EF, in 93% of the cases the Teich method was used, which is less accurate than the biplane Simpson method, and tend to estimate the EF [10]. This might have resulted in the misclassification of some HFrEF patients as HFmrEF, potentially explaining the lower ability of the multinomial model to identify HFmrEF. Finally, for our analyses we deemed a timeframe of less than three months between the determination of EF and the assessment of HF medications to be acceptable. We acknowledge that the dosage of these medications may be subject to adjustments over time as part of HF management, and discontinuation and initiation of these drugs might as well occur at notable rates. For instance, the EVOLUTION-HF study reported that 33–38% of patients discontinued ACE-inhibitors and angiotensin receptor blockers within 12 months [16]. However, these medications still represent the cornerstone of HF treatment and are typically prescribed and monitored closely in clinical practice, especially in the initial stages following diagnosis.

Conclusions

Our investigation suggests that this simplified version of the algorithm shows promise in predicting EF, indicating its potential for retrospective utilization in research endeavors involving HF patients. By providing a means to characterize the EF phenotype, this algorithm could serve as a useful tool for guiding future studies within the HF domain, and particularly in the context of EHRs, where direct measurements of EF are not routinely available, such as in primary care settings.

Supporting information

S1 Table. Proportion of missing values (%) in each variable of the algorithm among the included patients.

(DOCX)

pone.0310023.s001.docx^{(13.9KB, docx)}

S2 Table. Internal validation of the models.

(DOCX)

pone.0310023.s002.docx^{(13.4KB, docx)}

S1 Fig. Flow chart of the study population selection.

(TIF)

pone.0310023.s003.tif^{(77.6KB, tif)}

Acknowledgments

The authors thank the other investigators, the staff, and the participants of the Cardiology Centers of the Netherlands for their valuable contributions.

List of abbreviations

HF: heart failure
EF: ejection fraction
HFpEF: Heart failure with preserved ejection fraction
HFrEF: Heart failure with reduced ejection fraction
HFmrEF: Heart failure with mildly reduced ejection fraction
EHRs: Electronic health records
CCN: Cardiology Centers of the Netherlands
GP: general practitioner
CVD: cardiovascular disease
eGFR: estimated glomerular filtration rate

Data Availability

The CCN database are not publicly available and cannot be shared outside the University Medical Center Utrecht's infrastructure due to ethical and data protection constraints. More specifically, the raw data contain potentially identifying and sensitive patient informationis and are kept by the data manager. The CCN is subject to the Dutch General Data Protection Regulation Implementation Act (Uitvoeringswet Algemene Verordening gegevensbescherming) which governs the processing of personal data. Data can however be made available upon reasonable request. Proposals for possible collaborations should be addressed to Dr Leonard Hofstra (L.Hofstra@cardiologiecentra.nl) or Prof Hester den Ruijter (H.M.denRuijter-2@umcutrecht.nl). Alternatively, research proposals or questions can be sent to the CCN Research and Innovation Manager Sebastiaan Blok (s.blok@cardiologiecentra.nl).

Funding Statement

This study was supported by the Dutch Cardiovascular Alliance in the form of a RECONNEXT grant [2020B008] and by the Fondation Leducq in the form of a grant [16CVD02] to HMdR.

References

1.Groenewegen A, Rutten FH, Mosterd A, Hoes AW. Epidemiology of heart failure. Eur J Heart Fail. 2020;22(8):1342–56. Epub 2020/06/03. doi: 10.1002/ejhf.1858 ; PubMed Central PMCID: PMC7540043. [DOI] [PMC free article] [PubMed] [Google Scholar]
2.McDonagh TA, Metra M, Adamo M, Gardner RS, Baumbach A, Bohm M, et al. 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure. Eur Heart J. 2021;42(36):3599–726. Epub 2021/08/28. doi: 10.1093/eurheartj/ehab368 . [DOI] [PubMed] [Google Scholar]
3.Nair N. Epidemiology and pathogenesis of heart failure with preserved ejection fraction. Rev Cardiovasc Med. 2020;21(4):531–40. Epub 2021/01/04. doi: 10.31083/j.rcm.2020.04.154 . [DOI] [PubMed] [Google Scholar]
4.Solomon SD, McMurray JJV, Claggett B, de Boer RA, DeMets D, Hernandez AF, et al. Dapagliflozin in Heart Failure with Mildly Reduced or Preserved Ejection Fraction. N Engl J Med. 2022. Epub 2022/08/27. doi: 10.1056/NEJMoa2206286 . [DOI] [PubMed] [Google Scholar]
5.Anker SD, Butler J, Filippatos G, Ferreira JP, Bocchi E, Bohm M, et al. Empagliflozin in Heart Failure with a Preserved Ejection Fraction. N Engl J Med. 2021;385(16):1451–61. Epub 2021/08/28. doi: 10.1056/NEJMoa2107038 . [DOI] [PubMed] [Google Scholar]
6.Kao DP, Trinkley KE, Lin CT. Heart Failure Management Innovation Enabled by Electronic Health Records. JACC Heart Fail. 2020;8(3):223–33. Epub 2020/01/14. doi: 10.1016/j.jchf.2019.09.008 ; PubMed Central PMCID: PMC7058493. [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Uijl A, Lund LH, Vaartjes I, Brugts JJ, Linssen GC, Asselbergs FW, et al. A registry-based algorithm to predict ejection fraction in patients with heart failure. ESC Heart Fail. 2020;7(5):2388–97. Epub 2020/06/18. doi: 10.1002/ehf2.12779 ; PubMed Central PMCID: PMC7524089. [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Bots SH, Siegersma KR, Onland-Moret NC, Asselbergs FW, Somsen GA, Tulevski, II, et al. Routine clinical care data from thirteen cardiac outpatient clinics: design of the Cardiology Centers of the Netherlands (CCN) database. BMC Cardiovasc Disord. 2021;21(1):287. Epub 2021/06/12. doi: 10.1186/s12872-021-02020-7 ; PubMed Central PMCID: PMC8191101. [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Van Calster B, Vergouwe Y, Looman CW, Van Belle V, Timmerman D, Steyerberg EW. Assessing the discriminative ability of risk models for more than two outcome categories. Eur J Epidemiol. 2012;27(10):761–70. Epub 2012/10/12. doi: 10.1007/s10654-012-9733-3 . [DOI] [PubMed] [Google Scholar]
10.Lang RM, Badano LP, Mor-Avi V, Afilalo J, Armstrong A, Ernande L, et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging. Eur Heart J Cardiovasc Imaging. 2015;16(3):233–70. Epub 2015/02/26. doi: 10.1093/ehjci/jev014 . [DOI] [PubMed] [Google Scholar]
11.Heidenreich PA, Bozkurt B, Aguilar D, Allen LA, Byun JJ, Colvin MM, et al. 2022 AHA/ACC/HFSA Guideline for the Management of Heart Failure: Executive Summary: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. J Am Coll Cardiol. 2022;79(17):1757–80. Epub 2022/04/06. doi: 10.1016/j.jacc.2021.12.011 . [DOI] [PubMed] [Google Scholar]
12.Savarese G, Stolfo D, Sinagra G, Lund LH. Heart failure with mid-range or mildly reduced ejection fraction. Nat Rev Cardiol. 2022;19(2):100–16. Epub 2021/09/08. doi: 10.1038/s41569-021-00605-5 ; PubMed Central PMCID: PMC8420965. [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Desai RJ, Lin KJ, Patorno E, Barberio J, Lee M, Levin R, et al. Development and Preliminary Validation of a Medicare Claims-Based Model to Predict Left Ventricular Ejection Fraction Class in Patients With Heart Failure. Circ Cardiovasc Qual Outcomes. 2018;11(12):e004700. Epub 2018/12/19. doi: 10.1161/CIRCOUTCOMES.118.004700 . [DOI] [PubMed] [Google Scholar]
14.Bovitz T, Gilbertson DT, Herzog CA. Administrative Data and the Philosopher’s Stone: Turning Heart Failure Claims Data into Quantitative Assessment of Left Ventricular Ejection Fraction. Am J Med. 2016;129(2):223–5. Epub 2015/11/01. doi: 10.1016/j.amjmed.2015.10.003 . [DOI] [PubMed] [Google Scholar]
15.Mahesri M, Chin K, Kumar A, Barve A, Studer R, Lahoz R, et al. External validation of a claims-based model to predict left ventricular ejection fraction class in patients with heart failure. PLoS One. 2021;16(6):e0252903. Epub 2021/06/05. doi: 10.1371/journal.pone.0252903 ; PubMed Central PMCID: PMC8177622. [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Savarese G, Kishi T, Vardeny O, Adamsson Eryd S, Bodegard J, Lund LH, et al. Heart Failure Drug Treatment-Inertia, Titration, and Discontinuation: A Multinational Observational Study (EVOLUTION HF). JACC Heart Fail. 2023;11(1):1–14. Epub 20220907. doi: 10.1016/j.jchf.2022.08.009 . [DOI] [PubMed] [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0310023.r001

Decision Letter 0

Gianluigi Savarese

19 Mar 2024

PONE-D-24-04912Simplification of a Registry-Based Algorithm for Ejection Fraction Prediction in Heart Failure Patients: Applicability in Cardiology Centers of the NetherlandsPLOS ONE

Dear Dr. Canto,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by May 03 2024 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Gianluigi Savarese

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please note that PLOS ONE has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, all author-generated code must be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse.

3. We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match.

When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

4. In this instance it seems there may be acceptable restrictions in place that prevent the public sharing of your minimal data. However, in line with our goal of ensuring long-term data availability to all interested researchers, PLOS’ Data Policy states that authors cannot be the sole named individuals responsible for ensuring data access (http://journals.plos.org/plosone/s/data-availability#loc-acceptable-data-sharing-methods).

Data requests to a non-author institutional point of contact, such as a data access or ethics committee, helps guarantee long term stability and availability of data. Providing interested researchers with a durable point of contact ensures data will be accessible even if an author changes email addresses, institutions, or becomes unavailable to answer requests.

Before we proceed with your manuscript, please also provide non-author contact information (phone/email/hyperlink) for a data access committee, ethics committee, or other institutional body to which data requests may be sent. If no institutional body is available to respond to requests for your minimal data, please consider if there any institutional representatives who did not collaborate in the study, and are not listed as authors on the manuscript, who would be able to hold the data and respond to external requests for data access? If so, please provide their contact information (i.e., email address). Please also provide details on how you will ensure persistent or long-term data storage and availability.

5. Please remove your figures from within your manuscript file, leaving only the individual TIFF/EPS image files, uploaded separately. These will be automatically included in the reviewers’ PDF.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: In this manuscript, Dr. Del Canto and colleagues evaluated the generalizability of an algorithm for ejection fraction prediction based on the Swedish Heart Failure Registry on a cohort of patients from the Cardiology Centers of the Netherlands database.

Overall, the article is nicely written, but some issues need to be addressed by the authors:

- Page 4: Considering the high proportion of patients with Teich method as EF assessment, the authors may perform a sensitivity analysis only in patients with EF assessed via Simpson. Even If in the original publication (Reference 7) the method for EF assessment is not reported, a high difference is present between the Teich and Simpson methods in evaluating EF (PMID: 25712077).

- Page 5: While I understand that many missing data and difference between registry may lead to different models, the authors are evaluating a model with 17 variables from the original 22 variables model, which was not tested and/or validated in the original manuscript. This should be clearly stated in the limitation section, or the authors may test this simplified model in the original cohorts for validation.

- Page 5: I would suggest the authors to use 10 imputations in mice instead of 5 and to report in a supplementary table the percentage of missing data for the evaluated variables.

- Page 6: The authors may consider just reporting the model values on sensitivity/specificity etc. in a separate table

- Page 8: I would remove this section: “Its integration into research initiatives holds the promise of enhancing the comprehensiveness and accuracy of HF studies, especially in scenarios where EF measurements are lacking. While the algorithm's complexity might make it better suited for research applications, its insights can indirectly impact clinical care. By contributing to a deeper understanding of HF phenotypes and their associations with various clinical variables, the algorithm indirectly enriches the broader clinical knowledge base. This can, in turn, inform the development of simpler clinical tools that are tailored for quick assessment in various healthcare settings.”

- Page 10: Please tone down the conclusion section

- Why are the figures reported two times?

- Table 1: The reported p-value is from what test? Do RAAS agents include ARNi?

-Figures 1a and 1b: Please change “>=” with “≥”.

- Please also note that the text requires careful proofreading since there are several grammar lapses.

Reviewer #2: Dal Canto et al evaluated a simple model aiming to identify ejection fraction categories from other variables. Such a model, if well-performing, would be valuable for research purposes since EF is often lacking in ICD-code data and EHR. The model the authors evaluated was a simplified version of a model that was previously developed in SwedeHF and validated in CHECK-HF (doi: 10.1002/ehf2.12779). When applying a simplified version of this algorithm in the Cardiology Centers of Netherlands (CCN) database, the corresponding C-indices were 0.71 and 0.74, respectively. The manuscript is overall well-written. I have the following comments for the authors.

- Throughout the manuscript, the wording can give the impression that this was an external validation of the model developed in SwedeHF. E.g. page 3, background “[…] additional external validations are essential […]” and page 7, discussion “[…] allowing us to validate its performance […]”. Although the variables were selected from a previously validated model, this was not a validation of that original model, but rather the derivation of a new model and no validation. This could be clarified throughout the manuscript.

- In context of comment 1, did the authors perform/consider doing any validation procedures, e.g. temporal validation?

- The background/discussion centres on the relevance of a EF-predicting model for EHR data – the authors might consider expanding also on the utility in ICD-10 / claims databases which also typically lack EF.

- Patients were excluded based on several parameters, including missingness of data. I suggest to include a patient selection flowchart.

- Missing data: I suggest to provide details on missing data on a per-variable level. Also, with 38% missing for anemia (which was included in the model) it is likely that 5 imputations is too few.

- “The associations with HFmrEF were overall weak, probably because of a limited number of participants with HFmrEF in our study population”. This should be moved to discussion.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PLoS One. 2024 Nov 5;19(11):e0310023. doi: 10.1371/journal.pone.0310023.r002

Author response to Decision Letter 0

13 May 2024

Submission ID: PONE-D-24-04912

Manuscript title: Simplification of a Registry-Based Algorithm for Ejection Fraction Prediction in Heart Failure Patients: Applicability in Cardiology Centers of the Netherlands

Response to editor and reviewers

We would like to thank the reviewers and editor very much for carefully reviewing our manuscript. Their review helped us to improve the manuscript. We have addressed all comments below, point-by-point. In addition, we have revised the manuscript (changes are marked in blue text).

Reviewer 1

In this manuscript, Dr. Dal Canto and colleagues evaluated the generalizability of an algorithm for ejection fraction prediction based on the Swedish Heart Failure Registry on a cohort of patients from the Cardiology Centers of the Netherlands database. Overall, the article is nicely written, but some issues need to be addressed by the authors:

1. Page 4: Considering the high proportion of patients with Teich method as EF assessment, the authors may perform a sensitivity analysis only in patients with EF assessed via Simpson. Even If in the original publication (Reference 7) the method for EF assessment is not reported, a high difference is present between the Teich and Simpson methods in evaluating EF (PMID: 25712077).

Reply: We appreciate the reviewer’s suggestion to perform a sensitivity analysis focusing on patients with EF assessed via the Simpson method. However, due to the small proportion of patients (only 341) with EF measured using this method, conducting a meaningful sensitivity analysis is not feasible. Nevertheless, we acknowledge the potential impact of using different EF assessment methods on the classification of heart failure phenotypes. The use of the Teich method, which often overestimates EF, may have led to some misclassification, particularly in distinguishing between HFmrEF and other phenotypes. This might help to explain why the association with HFmrEF in the multinomial model were overall weaker. We have revised the discussion section to acknowledge this limitation and its potential implications for our analysis: “Among the cases with quantitatively assessed EF, in 93% of the cases the Teich method was used, which is less accurate than the biplane Simpson method, and tend to estimate the EF. This might have resulted in the misclassification of some HFrEF patients as HFmrEF, potentially explaining the lower ability of the multinomial model to identify HFmrEF. (page 10).”

2. Page 5: While I understand that many missing data and difference between registry may lead to different models, the authors are evaluating a model with 17 variables from the original 22 variables model, which was not tested and/or validated in the original manuscript. This should be clearly stated in the limitation section, or the authors may test this simplified model in the original cohorts for validation.

Reply: thanks for this comment, indeed this represents an important limitation of our analysis. We now stated this more clearly in the limitations section as: “Firstly, we had missing data on several predictors, which is a common issue in EHRs. Accordingly, we evaluated the performance of a model that included 17 out of the 22 original variables. It is important to note that this newly derived model was not tested neither validated in the original manuscript.” Furthermore, in response to comment number 1 from reviewer 2, we revised introduction and discussion to point out that the model derived and tested in our study indeed represents a new model which differs from the original one. By doing so we made it clear that our study does not represent an external validation of the original model. However, in this revised version of the paper we have conducted internal validation of the model through temporal validation, that confirmed its good discriminative ability. We believe that this procedure has improved the robustness of our findings.

3. Page 5: I would suggest the authors to use 10 imputations in mice instead of 5 and to report in a supplementary table the percentage of missing data for the evaluated variables.

Reply: Indeed we agree on the fact that having ten imputation might be more suitable to our data, and revised our approach accordingly, by incrementing the number of imputations. The results that we present in this revised version are based on the new analysis and on the newly imputed dataset. There were some minor changes; in particular, eGFR <30 mL/min/1.73mq is no longer one of the strongest predictor of EF<50% and 40%, while use of diuretics, of MRA and presence of atrial fibrillation are now the variables more strongly associated to EF<40% and EF<50%. There were no changes in the predictors of EF≥50% and ≥40%. For the multinomial model, age > 75 years old is no longer significantly associated to HFpEF. We revised the tables and the results section accordingly: “The strongest predictors for both EF ≥50% and EF ≥40% were female sex and presence of arterial hypertension. Use of diuretics, of mineralocorticoid receptor antagonists and presence of atrial fibrillation were the strongest predictors for EF <50% and EF <40%. (page 7)”

4. Page 6: The authors may consider just reporting the model values on sensitivity/specificity etc. in a separate table.

Reply: we agree on this and added the model sensitivity, specificity, positive and negative predictive values in a table, which now is Table 3. We kept only one sentence in the text to report the highest sensitivities: “The model had the highest sensitivities for ejection fraction ≥50% (0.70, 95% CI: 0.63–0.71) and for ejection fraction ≥40% (0.63, 95% CI: 0.61–0.75) (page 7).”

5. Page 8: I would remove this section: “Its integration into research initiatives holds the promise of enhancing the comprehensiveness and accuracy of HF studies, especially in scenarios where EF measurements are lacking. While the algorithm's complexity might make it better suited for research applications, its insights can indirectly impact clinical care. By contributing to a deeper understanding of HF phenotypes and their associations with various clinical variables, the algorithm indirectly enriches the broader clinical knowledge base. This can, in turn, inform the development of simpler clinical tools that are tailored for quick assessment in various healthcare settings.”

Reply: We left this section out and replaced with a single more concise sentence. Moreover, this section was further revised to address comment number 3 of reviewer 2 on the potential utility of the algorithm in ICD-10 claims datasets (page 8). “Its integration into research initiatives has the potential to improve the accuracy of HF studies, especially in scenarios where EF measurements are lacking, such as EHRs. Furthermore, the utility of the algorithm might apply to other healthcare sources such as ICD-10 claims datasets which also often lack detailed clinical information.”

6. Page 10: Please tone down the conclusion section

Reply: thank you for your feedback, we have revised this section to convey the findings of our study in a more balanced and measured manner, aligning with the overall tone of the manuscript: “Our investigation suggests that this simplified version of the algorithm shows promise in predicting EF, indicating its potential for retrospective utilization in research endeavors involving HF patients. By providing a means to characterize the EF phenotype, this algorithm could serve as a useful tool for guiding future studies within the HF domain, and particularly in the context of EHRs, where direct measurements of EF are not routinely available, such as in primary care, or other non-cardiology health settings such as GP settings.”

7. Why are the figures reported two times?

Reply: we removed one recurrence of the figures.

8. Table 1: The reported p-value is from what test? Do RAAS agents include ARNi?

Reply: The reported p value in table 1 refers to the comparisons between the heart failure phenotypes groups, which was performed by analysis of variance (ANOVA) for continuous variables and chi-squared for trend for categorical variables. We reported this information now as a footnote to the table and in the methods section. RAAS agents include ACE-inhibitors and angiotensin receptors blockers but not angiotensin receptor-neprilysin inhibitors (ARNi). That is because we created a variable that reflects the original RAAS agents variable of the algorithm, which did not include ARNi as well. We clarified this in the text on page 5.

9. Figures 1a and 1b: Please change “>=” with “≥”.

Reply: we revised the figures accordingly.

10. Please also note that the text requires careful proofreading since there are several grammar lapses.

Reply: thanks for this feedback. We carefully reviewed and proofread the manuscript to correct any grammar mistakes.

Reviewer 2

Dal Canto et al evaluated a simple model aiming to identify ejection fraction categories from other variables. Such a model, if well-performing, would be valuable for research purposes since EF is often lacking in ICD-code data and EHR. The model the authors evaluated was a simplified version of a model that was previously developed in SwedeHF and validated in CHECK-HF (doi: 10.1002/ehf2.12779). When applying a simplified version of this algorithm in the Cardiology Centers of Netherlands (CCN) database, the corresponding C-indices were 0.71 and 0.74, respectively. The manuscript is overall well-written. I have the following comments for the authors.

1. Throughout the manuscript, the wording can give the impression that this was an external validation of the model developed in SwedeHF. E.g. page 3, background “[…] additional external validations are essential […]” and page 7, discussion “[…] allowing us to validate its performance […]”. Although the variables were selected from a previously validated model, this was not a validation of that original model, but rather the derivation of a new model and no validation. This could be clarified throughout the manuscript.

Reply: We appreciate the reviewer's feedback and have revised the relevant sections of the manuscript to address this concern. Our original aim was to provide a further external validation of the original model. However, given the issue of missing data in some of the predictors, we derived and tested a new and simplified model, which included 17 out of the 22 original variables. In our revised version we clarified this in both the introduction as: “Accordingly, our aim was to provide a further external validation of the HF algorithm in patients with chronic HF from the Cardiology Centers of the Netherlands (CCN) database. However, given the characteristics of our study population including the presence of missing data in some of the predictors, we derived and assessed the applicability of a simplified version, and thus more widely applicable, of the original model” (page 4). And in the discussion as: “In this study, we adapted a diagnostic algorithm originally designed for research purposes to predict EF among patients with HF in the Netherlands. This newly derived algorithm was applied to EHRs data obtained from Dutch cardiac screening centers, which represents a real-world clinical setting. Our findings demonstrated that the simplified version of the original algorithm performed adequately in predicting EF.” Furthermore, in response to comment number 2 of reviewer 1, we revised the limitation section to acknowledge the fact that this new model we derived was not validated in the original manuscript (page 10): “Firstly, we had missing data on several predictors, which is a common issue in EHRs. Accordingly, we evaluated the performance of a model that included 17 out of the 22 original variables. It is important to note that this newly derived model was not tested neither validated in the original manuscript.

2. In context of comment 1, did the authors perform/consider doing any validation procedures, e.g. temporal validation?

Reply: After careful consideration of the reviewer's feedback, we performed temporal validation to assess the performance of the derived model. This involved splitting the dataset into training and testing sets based on time, with 75% of the data used for training the model and 25% for testing. Logistic and multinomial regression models were trained using the training set, and predictions were made on the testing set to evaluate model accuracy. This is carefully explained in the methods section. Results are presented in the results section as well as in Supplementary Table 2. Validation showed good discriminative performance of the model with a C-statistics of 0.72 (95% CI: 0.71 – 0.73) for EF ≥ 50% and of 0.69 (0.68 – 0.70) for EF ≥40%. We would like to thank the reviewer for this comment as we believe that the inclusion of temporal validation enhances the robustness and generalizability of our findings.

3. The background/discussion centers on the relevance of a EF-predicting model for EHR data – the authors might consider expanding also on the utility in ICD-10 / claims databases which also typically lack EF.

Reply: thanks for this comment. We agree that ICD-10 / claims databases might represent an additional application for the algorithm. We now have acknowledged this in the discussion section as: “Its integration into research initiatives has the potential to improve the accuracy of HF studies, especially in scenarios where EF measurements are lacking, such as EHRs. Furthermore, the utility of the algorithm might apply to other healthcare sources such as ICD-10 claims datasets which also often lack detailed clinical information.” (page 9).

4. Patients were excluded based on several parameters, including missingness of data. I suggest to include a patient selection flowchart.

Reply: in this revised version we have included a Flow chart of the Study Population selection (S1 Figure), which clarifies the steps that have been taken in this regard.

5. Missing data: I suggest to provide details on missing data on a per-variable level. Also, with 38% missing for anemia (which was included in the model) it is likely that 5 imputations is too few.

Reply: We provided the information on the amount of missing values for each variable of the algorithm in S1 Table (Supporting information). Furthermore, we incremented the number of imputations to 10, also in response to comment number 3 of reviewer 1.

6. “The associations with HFmrEF were overall weak, probably because of a limited number of participants with HFmrEF in our study population”. This should be moved to discussion.

Reply: we moved this statement to the discussion (page 9), which has been carefully revised in response to several comments from both reviewers.

Attachment

Submitted filename: Response to Reviewers.docx

pone.0310023.s004.docx^{(27.2KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0310023.r003

Decision Letter 1

Satoshi Higuchi

30 Jul 2024

PONE-D-24-04912R1Simplification of a Registry-Based Algorithm for Ejection Fraction Prediction in Heart Failure Patients: Applicability in Cardiology Centres of the NetherlandsPLOS ONE

Dear Dr. Canto,

Please submit your revised manuscript by Sep 13 2024 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

We look forward to receiving your revised manuscript.

Kind regards,

Satoshi Higuchi

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: No

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Reviewer #1: I thank the authors for replying and integrating my comments.

As it is I have no further comments for the authors.

Reviewer #2: I thank the authors for addressing the previous comments. I have the following comments remaining:

1. The numbers in Table 3 seem inconsistent. For example, since LVEF >=50% has 0.70 sensitivity, shouldn't I expect LVEF <50% to have 0.7 specificity? (since TP and FN for LVEF>=50% should be the same as TN and FP for LVEF<50%). Rather, it seems the LVEF<50% column contains the FNR (1-specificity) instead of sensitivity and the FPR (1-sensitivity) instead of specificity. There are similar apparent inconsistencies for PPV and NPV in Table 3 and also for the C-statistics in Supplementary Table S2.

2. The omission of ARNi from the model strikes me as strange. Of the drugs that could be included in the RASi-variable (ACEi, ARB, and ARNi), ARNi is likely the far best at distinguishing between the EF phenotypes. Moreover, ARNi technically includes an ARB. How much this omission affects model performance is difficult to speculate without knowing the time period of enrollment and how many patients used ARNi. I highly suggest to consider including it in the RASi variable, as well as to disclose ACEi/ARB and ARNi separately in the characteristics table.

3. Please clarify in the methods between which dates these patients had their index dates. It says CCN data were accessed between Feb 2022 and Jan 2024, but I interpret this literally as data access and not when the patients presented in the clinic.

4. Page 11: "While we acknowledge that the dosage of these medications may be subject to adjustments over time as part of HF management, it is unlikely that the discontinuation or initiation of these drugs would occur significantly long after the initial diagnosis of HF. This is because these HF medications represent the cornerstone of HF treatment and are typically prescribed and monitored closely in clinical practice." This sentence seems at odds with observations from several health care systems. E.g. in EVOLUTION-HF (10.1016/j.jchf.2022.08.009) 33-38% of patients discontinued ACEi/ARB within 12 months.

5. I suggest to change the term "mid-range" EF to "mildly reduced" EF to align with currently accepted terminology.

6. Conclusion: Just a minor comment but was the distinction between primary care and GP settings made on purpose? "[...] where direct measurements of EF are not routinely available, such as in primary care, or other non-cardiology health settings such as GP settings."

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

**********

PLoS One. 2024 Nov 5;19(11):e0310023. doi: 10.1371/journal.pone.0310023.r004

Author response to Decision Letter 1

2 Aug 2024

Submission ID: PONE-D-24-04912R1

Manuscript title: Simplification of a Registry-Based Algorithm for Ejection Fraction Prediction in Heart Failure Patients: Applicability in Cardiology Centers of the Netherlands

Response to editor and reviewers

We would like to thank the reviewers and editor very much for carefully reviewing our manuscript. We appreciate their detailed review and the opportunity to improve our manuscript. We have addressed all comments below, point-by-point. In addition, we have revised the manuscript (changes are marked in blue text).

Reviewer #1: I thank the authors for replying and integrating my comments.

As it is I have no further comments for the authors.

Reply: We are pleased to know that our rebuttal addressed the reviewer’s concerns satisfactorily. Thank you for your thorough review and valuable feedback on our manuscript.

Reviewer #2: I thank the authors for addressing the previous comments. I have the following comments remaining:

Reply: Thank you for your valuable feedback. We have carefully reviewed your comment regarding the inconsistencies in Table 3 and the C-statistics in Supplementary Table S2. We acknowledge the importance of accurately presenting these metrics, and apologize for the errors in reporting these results. Upon re-evaluation of our calculations and the analysis of both LVEF >= 50% and LVEF < 50%, as well as LVEF >= 40% and LVEF < 40%, we corrected Table 3 and Supplementary Table S2. The corrected values now ensure the accuracy of our diagnostic performance metrics.

Reply: Thank you for your insightful feedback regarding the inclusion of ARNi in the RAAS agents variable. Our decision of not including them was based on the fact that the original algorithm developed in the Swedish Heart Failure Registry did not include ARNi due to the timeframe of the study and the adoption of these medications during that period. Specifically, the first prescriptions of ARNi in Sweden began in April 2016, and the study from A. Uijl included patients between 2000 and 2012.

In light of your comment, we have additionally checked our dataset from the Cardiology Centers of the Netherlands about the use of ARNi, but we did not find any mention of it. Upon consultation with the cardiologists at CCN, it was confirmed that regular prescription of ARNi, began only around 2021, while our analysis includes participants from 2007 to 2018.

Therefore, because our study period predates the widespread use of ARNi, incorporating them in the RAAS agents variable is unfortunately not feasible due to the lack of data.

Finally, we would like to point out that ARNi is now being recognized for its benefits across all HF phenotypes, including HFpEF and HFmrEF, as demonstrated in the 2023 PARAGLIDE trial. We understand the potential impact of ARNi and will consider this in future analyses, as more recent data becomes available.

Reply: we agree with the reviewer that this is an important information and should be clearly stated in the manuscript. Our analysis includes patients who were diagnosed with HF between June 2007 and February 2018. We have added this information in the methods section on page 4.

Reply: We appreciate your observation regarding the discontinuation rates of ACEi/ARB in HF patients, as highlighted in the EVOLUTION-HF study. We recognize that our statement may not align with these observations. Accordingly, we have revised this paragraph as follows, and included a reference to the EVOLUTION-HF study (page 11).

“We acknowledge that the dosage of these medications may be subject to adjustments over time as part of HF management, and discontinuation and initiation of these drugs might as well occur at notable rates. For instance, the EVOLUTION-HF study reported that 33-38% of patients discontinued ACE-inhibitors and angiotensin receptor blockers within 12 months [16]. However, these medications still represent the cornerstone of HF treatment and are typically prescribed and monitored closely in clinical practice, especially in the initial stages following diagnosis.”

5. I suggest to change the term "mid-range" EF to "mildly reduced" EF to align with currently accepted terminology.

Reply: we have revised this accordingly.

Reply: We acknowledge that the terms "primary care" and "GP settings" can be used interchangeably, and the distinction we made was not intentional. Both terms refer to the settings where general practitioners provide initial and ongoing patient care. We have revised the conclusion to use consistent terminology to avoid confusion: "…where direct measurements of EF are not routinely available, such as in primary care settings."

Attachment

Submitted filename: Rebuttal letter.docx

pone.0310023.s005.docx^{(21KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0310023.r005

Decision Letter 2

Satoshi Higuchi

23 Aug 2024

Simplification of a Registry-Based Algorithm for Ejection Fraction Prediction in Heart Failure Patients: Applicability in Cardiology Centres of the Netherlands

PONE-D-24-04912R2

Dear Dr. Canto,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. If you have any questions relating to publication charges, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Satoshi Higuchi

Academic Editor

PLOS ONE

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #2: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #2: Yes

**********

6. Review Comments to the Author

Reviewer #2: I thank the authors for thoroughly addressing my previous comments. I have no further remarks to be considered.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

**********

PLoS One. doi: 10.1371/journal.pone.0310023.r006

Acceptance letter

Satoshi Higuchi

4 Sep 2024

PONE-D-24-04912R2

PLOS ONE

Dear Dr. Canto,

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now being handed over to our production team.

At this stage, our production department will prepare your paper for publication. This includes ensuring the following:

* All references, tables, and figures are properly cited

* All relevant supporting information is included in the manuscript submission,

* There are no issues that prevent the paper from being properly typeset

If revisions are needed, the production department will contact you directly to resolve them. If no revisions are needed, you will receive an email when the publication date has been set. At this time, we do not offer pre-publication proofs to authors during production of the accepted work. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few weeks to review your paper and let you know the next and final steps.

Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

If we can help with anything else, please email us at customercare@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Satoshi Higuchi

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Table. Proportion of missing values (%) in each variable of the algorithm among the included patients.

(DOCX)

pone.0310023.s001.docx^{(13.9KB, docx)}

S2 Table. Internal validation of the models.

(DOCX)

pone.0310023.s002.docx^{(13.4KB, docx)}

S1 Fig. Flow chart of the study population selection.

(TIF)

pone.0310023.s003.tif^{(77.6KB, tif)}

Attachment

Submitted filename: Response to Reviewers.docx

pone.0310023.s004.docx^{(27.2KB, docx)}

Attachment

Submitted filename: Rebuttal letter.docx

pone.0310023.s005.docx^{(21KB, docx)}

Data Availability Statement

[pone.0310023.ref001] 1.Groenewegen A, Rutten FH, Mosterd A, Hoes AW. Epidemiology of heart failure. Eur J Heart Fail. 2020;22(8):1342–56. Epub 2020/06/03. doi: 10.1002/ejhf.1858 ; PubMed Central PMCID: PMC7540043. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0310023.ref002] 2.McDonagh TA, Metra M, Adamo M, Gardner RS, Baumbach A, Bohm M, et al. 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure. Eur Heart J. 2021;42(36):3599–726. Epub 2021/08/28. doi: 10.1093/eurheartj/ehab368 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref003] 3.Nair N. Epidemiology and pathogenesis of heart failure with preserved ejection fraction. Rev Cardiovasc Med. 2020;21(4):531–40. Epub 2021/01/04. doi: 10.31083/j.rcm.2020.04.154 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref004] 4.Solomon SD, McMurray JJV, Claggett B, de Boer RA, DeMets D, Hernandez AF, et al. Dapagliflozin in Heart Failure with Mildly Reduced or Preserved Ejection Fraction. N Engl J Med. 2022. Epub 2022/08/27. doi: 10.1056/NEJMoa2206286 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref005] 5.Anker SD, Butler J, Filippatos G, Ferreira JP, Bocchi E, Bohm M, et al. Empagliflozin in Heart Failure with a Preserved Ejection Fraction. N Engl J Med. 2021;385(16):1451–61. Epub 2021/08/28. doi: 10.1056/NEJMoa2107038 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref006] 6.Kao DP, Trinkley KE, Lin CT. Heart Failure Management Innovation Enabled by Electronic Health Records. JACC Heart Fail. 2020;8(3):223–33. Epub 2020/01/14. doi: 10.1016/j.jchf.2019.09.008 ; PubMed Central PMCID: PMC7058493. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0310023.ref007] 7.Uijl A, Lund LH, Vaartjes I, Brugts JJ, Linssen GC, Asselbergs FW, et al. A registry-based algorithm to predict ejection fraction in patients with heart failure. ESC Heart Fail. 2020;7(5):2388–97. Epub 2020/06/18. doi: 10.1002/ehf2.12779 ; PubMed Central PMCID: PMC7524089. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0310023.ref008] 8.Bots SH, Siegersma KR, Onland-Moret NC, Asselbergs FW, Somsen GA, Tulevski, II, et al. Routine clinical care data from thirteen cardiac outpatient clinics: design of the Cardiology Centers of the Netherlands (CCN) database. BMC Cardiovasc Disord. 2021;21(1):287. Epub 2021/06/12. doi: 10.1186/s12872-021-02020-7 ; PubMed Central PMCID: PMC8191101. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0310023.ref009] 9.Van Calster B, Vergouwe Y, Looman CW, Van Belle V, Timmerman D, Steyerberg EW. Assessing the discriminative ability of risk models for more than two outcome categories. Eur J Epidemiol. 2012;27(10):761–70. Epub 2012/10/12. doi: 10.1007/s10654-012-9733-3 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref010] 10.Lang RM, Badano LP, Mor-Avi V, Afilalo J, Armstrong A, Ernande L, et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging. Eur Heart J Cardiovasc Imaging. 2015;16(3):233–70. Epub 2015/02/26. doi: 10.1093/ehjci/jev014 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref011] 11.Heidenreich PA, Bozkurt B, Aguilar D, Allen LA, Byun JJ, Colvin MM, et al. 2022 AHA/ACC/HFSA Guideline for the Management of Heart Failure: Executive Summary: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. J Am Coll Cardiol. 2022;79(17):1757–80. Epub 2022/04/06. doi: 10.1016/j.jacc.2021.12.011 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref012] 12.Savarese G, Stolfo D, Sinagra G, Lund LH. Heart failure with mid-range or mildly reduced ejection fraction. Nat Rev Cardiol. 2022;19(2):100–16. Epub 2021/09/08. doi: 10.1038/s41569-021-00605-5 ; PubMed Central PMCID: PMC8420965. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0310023.ref013] 13.Desai RJ, Lin KJ, Patorno E, Barberio J, Lee M, Levin R, et al. Development and Preliminary Validation of a Medicare Claims-Based Model to Predict Left Ventricular Ejection Fraction Class in Patients With Heart Failure. Circ Cardiovasc Qual Outcomes. 2018;11(12):e004700. Epub 2018/12/19. doi: 10.1161/CIRCOUTCOMES.118.004700 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref014] 14.Bovitz T, Gilbertson DT, Herzog CA. Administrative Data and the Philosopher’s Stone: Turning Heart Failure Claims Data into Quantitative Assessment of Left Ventricular Ejection Fraction. Am J Med. 2016;129(2):223–5. Epub 2015/11/01. doi: 10.1016/j.amjmed.2015.10.003 . [DOI] [PubMed] [Google Scholar]

[pone.0310023.ref015] 15.Mahesri M, Chin K, Kumar A, Barve A, Studer R, Lahoz R, et al. External validation of a claims-based model to predict left ventricular ejection fraction class in patients with heart failure. PLoS One. 2021;16(6):e0252903. Epub 2021/06/05. doi: 10.1371/journal.pone.0252903 ; PubMed Central PMCID: PMC8177622. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0310023.ref016] 16.Savarese G, Kishi T, Vardeny O, Adamsson Eryd S, Bodegard J, Lund LH, et al. Heart Failure Drug Treatment-Inertia, Titration, and Discontinuation: A Multinational Observational Study (EVOLUTION HF). JACC Heart Fail. 2023;11(1):1–14. Epub 20220907. doi: 10.1016/j.jchf.2022.08.009 . [DOI] [PubMed] [Google Scholar]

PERMALINK

Simplification of a registry-based algorithm for ejection fraction prediction in heart failure patients: Applicability in cardiology centres of the Netherlands

Elisa Dal Canto

Alicia Uijl

N Charlotte Onland-Moret

Sophie H Bots

Leonard Hofstra

Igor Tulevski

Folkert W Asselbergs

Pim van der Harst

G Aernout Somsen

Hester M den Ruijter

Roles

Abstract

Background

Methods

Results

Conclusions

Introduction

Materials and methods

Study population

Statistical analysis

Results

Table 1. Baseline characteristics of the cardiology centers of the Netherlands population.

Table 2. Multivariable logistic prediction models predicting EF ≥ 50% vs. EF < 50% and EF ≥ 40% vs. <40%.

Fig 1. Discrimination plot displaying ROC curve for logistic model EF cut-off ≥50%.

Fig 2. Discrimination plot displaying ROC curve for logistic model EF cut-off ≥40%.

Fig 3. Discrimination plot displaying ROC curve for logistic model EF cut-off <50%.

Fig 4. Discrimination plot displaying ROC curve for logistic model EF cut-off <40%.

Table 3. Sensitivity, specificity, positive and negative predictive values of the logistic prediction models.

Table 4. Multinomial logistic prediction models predicting HFmrEF vs HFrEF and HFpEF vs HFrEF.

Discussion

Study limitations

Conclusions

Supporting information

Acknowledgments

List of abbreviations

Data Availability

Funding Statement

References

Decision Letter 0

Gianluigi Savarese

Roles

Author response to Decision Letter 0

Decision Letter 1

Satoshi Higuchi

Roles

Author response to Decision Letter 1

Decision Letter 2

Satoshi Higuchi

Roles

Acceptance letter

Satoshi Higuchi

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases