Article, Cardiology

Development of a clinical prediction score for congestive heart failure diagnosis in the emergency care setting: The Brest score

a b s t r a c t

Objective: To derive and validate a clinical prediction rule of Acute congestive heart failure obtainable in the emer- gency care setting.

Design: Derivation of the score was performed on a retrospective 927 patients cohort admitted to our Emergency Department for dyspnea. The prediction model was externally validated on an independent 206-patient prospec- tive cohort.

Interventions and Measures: During the derivation phase, variables associated with acute congestive heart failure were included in a multivariate regression model. Logistic regression coefficients were used to assign scoring points to each variable. During the validation phase, every diagnosis was confirmed by an independent adjudica- tion committee.

Results: The score comprised 11 variables: age >=65 years (1 point), seizure dyspnea (2 points), night outbreak (1 point), orthopnea (1 point), history of pulmonary edema (2 points), chronic pulmonary disease (-2 points), myocardial infarction (1 point), crackles (2 points), leg edema (1 point), ST-segment abnormality (1 point), atrial fibrillation/flutter (1 point) on electrocardiography. In the validation step, 30 patients (14.6%) had a low clinical probability of acute congestive heart failure (score <=3), of which only 2 (6.7%) had a proven acute cardiogenic pulmonary edema. The prevalence of acute congestive heart failure was 58.5% in the 94 patients with an interme- diate probability (score of 4-8) and 91.5% in the 82 patients (39.8%) with a high probability (score >=9).

Conclusion: This score of acute congestive heart failure based on easily available and objective variables is entirely standardized. Applying the score to dyspneic adult emergency patients may enable a more rapid and efficient di- agnostic process.

(C) 2016

Introduction

The prevalence of chronic heart failure is elevated and it represents the second-highest Disease burden for length-of-stay in hospitalized pa- tients, especially as regards to a worldwide aging population [1-3].

The diagnosis of congestive heart failure (CHF) is difficult in the Emergency care setting, and fewer than 50% of CHF diagnoses

* Corresponding author at: Reanimation Medicale, CHRU de Brest, La Cavale Blanche, Bvd Tanguy-Prigent, 29609 Brest Cedex, France. Tel.: +33 2 98 34 71 81; fax: +33 2 98

34 79 65.

E-mail address: [email protected] (E. L’Her).

performed in the Emergency Department (ED) are ultimately con- firmed by cardiologists [4-6]. Clinical probability assessment has proved to be an important component of treatment strategies in various pathol- ogies such as pulmonary embolism. Although such strategies were sug- gested for acute heart failure and CHF diagnosis, as in the Framingham study, these approaches did not seem to be adapted to routine emer- gency medicine practice [7,8]. Such difficulties are related to the fact that neither symptoms nor physical findings are specific and sensitive enough for an accurate diagnosis.

As misdiagnosis may be life threatening, the European Society of Cardiology proposed in 2012 an algorithm for the diagnosis of CHF including natriuretic peptide Blood concentration measurement [9]. Although evidence points to a relationship between elevated values and the presence of CHF, natriuretic levels can be influenced by many

http://dx.doi.org/10.1016/j.ajem.2016.08.023

0735-6757/(C) 2016

other diseases or pathophysiological conditions. Moreover, it is impor- tant to establish the diagnosis without waiting for biological results in order to initiate an adequate medical treatment as soon as possible, and thus reduce Morbidity and mortality rates. Yet, even by using natri- uretic peptide values, accurate diagnosis in this setting occurs less than 80% of the time [9-13].

Therefore, we derived a new prediction rule based on simple clinical variables and electrocardiogram (ECG) findings, using a large retrospec- tive cohort of patients admitted to the ED of a University Hospital for acute dyspnea. The validation of this prediction rule was subsequently performed in a distinct prospective cohort of patients.

Methods

The protocol was validated by our local ethics committee and infor- mation was provided to all patients and their relatives.

Retrospective derivation step

Patients and study design

Patients admitted to the Brest University Hospital ED between Janu- ary 1, 2009 and December 31, 2009, were selected from a computerized collection of more than 43 000 medical records. All consecutive adult patients (N15 years) admitted for an acute dyspnea were included. Ex- clusion criteria were 1) unavailability of medical records, 2) uncertain diagnosis at the end of the hospital stay.

All patients underwent the standard diagnosis work-up on a com- puterized medical file, which recorded physiological characteristics, an- amnesis, risk factors, clinical examination. After a clinical diagnosis, the patients received the best possible medical treatment and orientation.

Data available from the computerized records and the subsequent hospital database were collected by means of a standardized case report form. Based on the literature, we identified and collected information on 34 candidate predictor variables (Table 1) [9]. All data had to be repro- ducible, easily and Rapidly obtainable in the emergency care setting. The variables fell into 5 groups: physiological characteristics (n = 8 vari- ables), anamnesis (n = 6 variables), risk factors (n = 6 variables), phys- ical examination (n = 7 variables), ECG abnormalities (n = 7 variables). The ST segment was considered as the flat, isoelectric section between the end of the S wave (J point) and the beginning of the T wave. Abnor- malities of the ST segment were defined by either elevation or depres- sion >=1 mm at the J point in >=2 contiguous leads.

The same 2 Senior emergency physicians retrospectively validated the final diagnosis over the entire sample, and also took into account both the ED and final hospital discharge evaluation.

Score derivation

To obtain a reproducible prediction rule based on easily recordable variables during emergency care, we chose not to use biological and chest X-ray findings. We also chose not to evaluate variables of which more than 5% data were missing (body weight, sputum characteristics, respiratory rate), except for ECG findings given its high potential rele- vance in the emergency care.

We performed univariate analyses to select candidate predictor var- iables for the multivariate model. The univariate relationship between each candidate predictor and CHF were examined by using the Chi-square or the Fisher test for categorical variables and the Student t-test for continuous variables. A 2-tailed P value of less than .05 was considered significant. We then categorized the continuous variables significantly associated with CHF, considering receiver operating char- acteristic (ROC) curve and values distribution among patients without CHF, in order to choose the most relevant cut-off point.

Clinically relevant variables were included in a multivariate logistic regression model. Non-significant variables were removed using a step-by-step procedure, enabling the calculation of a regression coeffi- cient for each statistically significant variable in the final model. We

assigned points for the score according to the regression coefficients, with 1 point corresponding to a value close to the smallest regression coefficient and serving as the least common denominator for assigning point value for the score items. The score was subsequently computed for each patient; we performed a ROC curve analysis and calculated the area under the curve. Finally, we chose the best discriminative cutoff values that allowed identifying (1) a low-probability group with a prev- alence of approximately 10% and (2) a high-probability group with a prevalence of CHF more than 60%. We assessed the predictive accuracy of the score categories by the proportion of patients with CHF in each group.

Prospective score validation

Patients and study design

We validated the score on an independent cohort of patients includ- ed in a prospective observational study. We included 214 consecutive patients admitted to the ED of Brest University Hospital for acute dys- pnea between December 1, 2012 and March 31, 2014.

Once the patient was admitted, the emergency physician in charge of the patient collected physiological characteristics, clinical data, bio- logical (measurement of B-type natriuretic peptide) and paraclinical (ECG, chest X-ray and echocardiogram) results on a standardized file. Echocardiogram recording was strongly encouraged for all patients.

For each patient enrolled in the study, the physician in charge classi- fied the patient as having an acute exacerbation of congestive heart fail- ure or dyspnea from another cause at the end of the standardized recording datafile.

External validation

To adjudicate the actual diagnosis, all medical records pertaining to the data collected in the ED and during hospitalization were reviewed by 3 independent senior physicians (10 years of experience), including an emergency physician, a cardiologist and a pulmonologist.

The physicians met up, were kept blinded as to the prediction model result, and had access to the computerized ED and hospital med- ical records and to all additional information available at the time of hospital discharge, including sequential chest X-rays, brain Natriuretic peptides (BNP) analysis, echocardiographic findings, and response to medical treatment.

The diagnostic of CHF was deemed present when 2 out of 3 physi- cians had the same diagnosis (CHF or dyspnea from another cause).

We then calculated the previously established score for each patient. We assessed the discrimination performance of this score by a ROC curve analysis. We assessed the predictive accuracy of the score by cal- culating the prevalence of CHF in the 3 previously determined clinical probability categories (low, intermediate, and high).

All statistical analysis was performed using SAS software, version 9.3 (SAS Institute Inc, Cary, NC).

Results

Retrospective derivation step

The total study sample consisted of 927 patients. Patients’ general characteristics and collected clinical variables are displayed in Table 1. Patients with CHF were older and more frequently female (P b .001).

Out of the 34 recorded variables, a statistically significant association with the diagnosis of CHF was observed for 27 variables, of which n = 8 regarding physiological characteristics (age, gender, heart rate, systolic and diastolic arterial blood pressure, pulse oximetry, oxygen adminis- tration), n = 5 regarding anamnesis (onset of symptoms at night, sud- den dyspnea, orthopnea, chest pain, dietary change), n = 6 regarding risk factors (Arterial hypertension, atherosclerosis, chronic heart failure, prior episodes of CHF or myocardial infarction, chronic pulmonary dis- ease), n = 6 regarding clinical examination (frothy or salmon pink

Table 1

Univariate analysis of the 919 patients’ characteristics during the derivation phase

Characteristics

Missing values (n, %)

CHF (n,% or SD)

No CHF (n,% or SD)

Cut-off values

P

Physiological characteristics

465 (50.6%)

454 (49.4%)

Sex

0

Male

0

201 (43.2%)

265 (58.8%)

/

/

Female

0

264 (56.8%)

189 (41.2%)

/

/

Mean age (yr.)

0

81.1 (9.3)

73.3 (13.6)

b65, >=65

b.0001

Heart rate (beat/min)

40 (4.4%)

92.2 (23.6)

97.6 (21.7)

b97, >=97

.0004

Systolic blood pressure (mmHg)

43 (4.7%)

149.9 (34.2)

137.9 (33.2)

b138, >=138

b.0001

Diastolic blood pressure (mmHg)

44 (4.8%)

84.4 (19.9)

78.2 (18.7)

b78, >=78

b.0001

Respiratory rate (b/min)

419 (46%)

26.8 (7.9)

27.8 (8)

/

.16

Pulse oximetry (%)

23 (2.5%)

95.1 (4.4)

94.1 (5.3)

b95, >=95

.0018

Additional Oxygen administration

231 (25.1%)

7.2 (4.6)

7 (4.6)

/

.5248

Risk factors

Previous CHF episodes

Yes

0

119 (25.6%)

19 (4.2%)

No

346 (74.4%)

435 (95.8%)

b.0001

Arterial hypertension

0

Yes

318 (63.4%)

199 (43.8%)

No

147 (36.6%)

255 (56.2%)

b.0001

history of AMI

0

Yes

115 (24.7%)

33 (7.3%)

No

350 (75.3%)

421 (92.7%)

b.0001

Chronic heart failure

0

Yes

215 (46.2%)

96 (21.1%)

No

250 (53.8%)

358 (78.9%)

b.0001

Atherosclerosis

0

Yes

337 (72.5%)

209 (46%)

No

128 (27.5%)

245 (54%)

b.0001

Chronic pulmonary disease

0

Yes

105 (22.6%)

266 (58.6%)

No

360 (77.4%)

188 (41.4%)

b.0001

Clinical examination

Frothy or salmon pink sputum

0

Yes

18 (4%)

7 (1.5%)

No

447 (96%)

447 (98.5%)

.040

Pulmonary crackles

0

Yes

402 (86.5%)

197 (43.4%)

No

63 (13.5%)

257 (56.6%)

b.0001

Sibilant respiration

0

Yes

107 (23%)

127 (28%)

No

358 (77%)

327 (72%)

.0956

Wheezing

0

Yes

56 (12%)

179 (39.4%)

No

409 (88%)

275 (60.6%)

b.0001

Pitting leg edema

0

Yes

249 (53.5%)

110 (24.2%)

No

216 (46.5%)

344 (75.8%)

b.0001

Spontaneous jugular veins distension

0

Yes

125 (22.1%)

44 (9.7%)

No

340 (77.9%)

410 (90.3%)

b.0001

Hepatogular reflux

0

Yes

87 (18.7%)

18 (4%)

No

378 (81.3%)

436 (96%)

b.0001

Anamnesis

Night outbreak

Yes

0

194 (41.7%)

63 (13.9%)

No

271 (58.3%)

391 (86.1%)

b.0001

Sudden dyspnea

0

Yes

216 (46.5%)

38 (8.4%)

No

249 ((53.5%)

416 (91.6%)

b.0001

Orthopnea

0

Yes

144 (31%)

26 (5.7%)

No

321 (69%)

428 (94.3%)

b.0001

Chest pain

0

Yes

93 (20%)

28 (6.2%)

No

372 (80%)

426 (93.8%)

b.0001

Weight gain

0

Yes

14 (3%)

6 (1.3%)

No

451 (97%)

448 (98.7%)

.08

Dietary change

0

Yes

25 (5.4%)

1 (0.2%)

No

440 (94.6%)

453 (99.8%)

b.0001

electrocardiogram abnormalities

315 (35%)

Right bundle branch block

315

(continued on next page)

Table 1 (continued)

Characteristics

Missing values (n, %)

CHF (n,% or SD)

No CHF (n,% or SD)

Cut-off values

P

Yes

35 (9.8%)

23 (9.3%)

No

Paced rhythm Yes

315

323 (90.2%)

26 (7.3%)

223 (90.7%)

11 (4.5%)

.86

.16

No Tachycardia

Yes

315

332 (92.7%)

22 (6.1%)

235 (95.5%)

18 (7.3%)

.57

No

atrioventricular block Yes

315

336 (93.9%)

17 (4.7%)

228 (92.7%)

6 (2.4%)

.14

No

Left bundle branch block

315

340 (95.3%)

240 (97.6%)

Yes

67 (18.7%)

20 (8.1%)

.0003

No

Atrial fibrillation/flutter Yes

315

291 (81.3%)

141 (39.4%)

226 (91.9%)

56 (22.8%)

b.0001

No

ST segment abnormalities Yes

315

217 (60.6%)

60 (16.8%)

190 (77.2%)

17 (6.9%)

.0004

No

298 (83.2%)

229 (93.1%)

AMI, acute myocardial infarction.

Thirty-four different variables were monitored on the database during the derivation phase. These variables were predetermined according to literature analysis and were easily repro- ducible and rapidly obtainable in the emergency care setting. Variables fell into 5 categories: physiological characteristics (n = 8), anamnesis (n = 6), risk factors (n = 6), clinical exam- ination (n = 7), electrocardiography abnormalities (n = 7).

Nine hundred and nineteen patients admitted to the emergency department for acute dyspnea were analyzed during the derivation phase, out of 43 000 medical records during, a one year period. We chose to analyze only variables that are easily obtainable in the emergency care setting, which means that biological and radiological findings were excluded. Variables with more than 5% missing data were not evaluated, except for electrocardiographic abnormalities given their potential relevance. For quantitative parameters, the most relevant cut-off values were calculated using ROC curves in order to help build the first version of the score. P <= .05 was considered significant.

sputum, pulmonary crackles, sibilant respiration, wheezing, pitting leg edema, spontaneous jugular veins distension, hepatojugular reflux), n = 3 regarding ECG abnormalities (left bundle branch block, atrial fibrillation/flutter, ST segment abnormalities) (Table 1). All numerical variables were divided into two groups according to the most discrimi- nant cut-off value.

All variables considered significant in the univariate analysis were included in a multivariate logistic regression model, and 11 predictors demonstrated a significant association with CHF diagnosis: age N 65 years, 3 anamnesis variables (night outbreak, sudden dyspnea, orthopnea), 3 risk factors variables (prior episodes of CHF or myocardial infarction, chronic pulmonary disease), 2 clinical examination variables (pitting leg edema, pulmonary crackles), and 2 ECG abnormalities (atri- al fibrillation/flutter, ST-segment abnormalities).

The Brest score was established in 605 patients (while 315 patients had missing electrocardiogram data) (Table 2). The respective cut-off between categories was 3 points or less for the low probability group,

Clinical score“>Table 2

Brest score derivation after the multivariate analysis

Variables

Logistic regression coefficients

P value

Points score

+ 2

Age N65 years

1,05

0,022

1

Sudden dyspnea

1,75

b 0,0001

2

Night outbreak

0,81

0,012

1

Orthopnea

1,27

0,002

1

Prior CHF episode

1,50

0,0005

2

Chronic pulmonary disease

-1,51

b 0,0001

-2

Myocardial infarction

1,08

0,0044

1

Pulmonary crackles

1,88

b 0,0001

2

Pitting leg edema

1,25

b 0,0001

1

ST segment abnormalities

1,08

0,0083

1

Atrial fibrillation/flutter

0,99

0,0002

1

15

Maximal score

Clinically relevant variables were included in a logistic regression model. Non-significant variables were removed using a step-by-step procedure, enabling the calculation of a re- gression coefficient for each significant variable. Score value for each variable was chosen as the closest to the regression coefficient. Maximal score value is equal to 15 points. P <=

.05 was considered as clinically significant.

9 or more for the high probability group and a score of 4 to 8 for the in- termediate probability group. The area under the curve of the derivated score was 0.91 after cross validation.

Diagnosis adequacy was considered equal to 84.8% in the retrospec- tive derivation step (eTable 1 in the online version at http://dx.doi.org/ 10.1016/j.ajem.2016.08.023).

Prospective score validation phase

A total of 214 patients were evaluated, but the score was only com- puted in 206 patients because of diagnostic uncertainty after adjudica- tion for 8 patients (3.7%). The diagnosis of CHF was confirmed in 132 patients (64.1%), and in 74 patients (35.9%), dyspnea was considered to be related to non-cardiac causes. The general characteristics of the pa- tients included in the validation step are depicted in Table 3. BNP mea- surements were missing for only 13 patients (7 in the CHF group and 6 in the non-CHF group).

Clinical score

The area under the ROC curve when the Brest score was used to dif- ferentiate CHF from other causes of dyspnea was 0.86 (Fig. 1). The best cutoff values for the score were 3 or less, in-between 4 and 8, and 9 or more.

Table 3 shows the prevalence of CHF in each clinical probability cat- egory in the validation sample of our study.

The individual risk scores were calculated in our sample and 30 pa- tients obtained a score lower than 3 (14.6%), 94 patients obtained a score between 4 and 8 (45.6%) and 82 patients obtained a score above 9 (39.8%).

BNP values were significantly different between low and high prob- ability scores (Fig. 2; P = .02), but correlation between Brest score and BNP values was low (r = 0.41; P b .001).

Discussion

In the present study, we derived and validated a new model for im- proving the diagnosis of CHF in patients with undifferentiated dyspnea.

Table 3

Results of the prospective validation phase on 206 patients and classification according to probability categories

CHF

No CHF

Number of patients (%)

132 (64.1%)

74 (35.9%)

Mean age (years)

83 +- 8

76 +- 14

Men (n, %)

57 (43.2%)

43 (58.1%)

BNP (pg/ml)

876 +- 640

351 +- 340

(n = 193/206)

Adjudication Congestive Heart /ailure No Congestive Heart /ailure Total

Agreement of all 3 physicians

(n, %)

Agreement of 2 out of 3 physicians

(n, %)

128 (97%) 70 (94.6%) 198 (96.1%)

4 (3%) 4 (5.4%) 8 (3.9%)

Classification

CH/

No CH/

Total

Low probability

2

28

30

(Score 03)

(6.7%)

(93.3%)

(14.6%)

Intermediate probability

55

39

94

(Score 48)

(58.5%)

(41.5%)

(45.6%)

High probability

75

7

82

(Score 915)

(91. 5%)

(8.5%)

(39.8%)

Two hundred fourteen patients were included during the prospective validation phase of the score, but it was computed for only 206 patients because of diagnostic uncertainty after ad- judication for 8 patients (3.8%). For most patients the adjudication process was simple, with an agreement of all 3 physicians in 94% cases. The 3 probability categories of the Brest score depicted fair accuracy.

The Brest score is standardized and relies exclusively on clinical vari- ables or ECG abnormalities, all being easily and rapidly available in the emergency care setting. This score enables the classification of patients into 3 categories of CHF clinical probability with a fair degree of accura- cy, thus allowing clinicians to initiate a medical treatment with a fair de- gree of confidence.

During the derivation step, several clinical variables were considered for inclusion in the model. Some data were not significantly associated with CHF, probably because only a few patients had those characteris- tics. During the validation step, the diagnosis was validated by 3 inde- pendent physicians in 198 cases (96.1%) and in only 8 cases (3.9%) the diagnosis was confirmed by 2 out of 3 physicians. The CHF prevalence was only 5.8% in the low probability category (14.6% of all patients),

58.5% in the intermediate-probability category (45.6% of all patients), and 91.5% in the high-probability category (39.8% of all patients) (Table 3).

The potential advantage of the Brest score is its simplicity, since all items are easily available in ED and already routinely collected. The low and the high probability groups may very well be the best targets for this score, and these 2 groups of patients may be managed with a fair degree of confidence without having to wait for further results of ei- ther chest X-ray or plasma B-type natriuretic peptide measurement.

In 2005 Wang et al conducted a meta-analysis on eighteen studies (from 1966 to July 2005) concerning the diagnostic accuracy of compo- nents of the clinical examination and the simplicity of the investigation in diagnosing patients with dyspnea [14]. Several variables such as

Sensibility

Low

Intermediate

High

AAUC = 0.86

Specificity

Fig. 1. Receiver operating characteristic (ROC) curve for the Brest score The Brest CHF diagnosis score was validated on an independent prospective cohort of 214 patients. An adjudication committee of 3 physicians confirmed every diagnosis. Predictive accuracy of the score was assessed by calculating the prevalence of CHF in the 3 previously determined categories of probability. Low probability: from 0 to 3; Intermediate probability: from 4 to 8; High probability: from 9 to 15. Area under the ROC curve was 0.86.

BNP

(pg/mL)

Brest Score categories

care setting, when time matters and no complementary exams are available. In order to improve diagnosis when the score is within the gray zone, alternative markers of acute heart failure such as chest X- ray, echocardiographic evaluation of systolic and/or diastolic dysfunc- tion [24], or BNP sampling [25,26], may be proposed in several difficult cases to enhance diagnosis performance. However, such explorations and measurements either require time for performance, and/or may not be available such as during prehospital transportation. Moreover, the common occurrence of shock, acute kidney injury, or chronic heart failure limits the usefulness of BNP measurements regarding a general ED population. As an example of application of the score in the emer- gency care setting, while in the CHF high probability group the probabil- ity of an accurate diagnostic classification is higher than 75%, the balance between risks and benefits of CHF medical treatment clearly fa- vors immediate therapy initiation, without any further results. A contrario, a low score value may indicate the need for further clinical ex-

Fig. 2. BNP values according to Brest score categories During the prospective score validation phase, BNP measurements were missing for only 13 patients (7 in the CHF group and 6 in the non-CHF group). BNP values were significantly different between low and high probability scores (P = .02). BNP values are depicted as pg/mL; results are presented according to the 3 different categories: Low probability (from 0 to 3); Intermediate probability (from 4 to 8); High probability (from 9 to 15). A P <=.05 was considered statistically significant.

medical history and anamnesis, clinical examination data, and the accu- racy of chest X-ray or ECG findings were examined. BNP values accuracy was analyzed taking into account renal insufficiency and other con- founding factors [15-17]. According to their meta-analysis, several fea- tures such as medical history, pitting leg edema, night outbreak of dyspnea, orthopnea, or atrial fibrillation confirmation on ECG seemed useful in diagnosing congestive heart failure in adult patients attending the ED for dyspnea, such as observed within our study. More recently, Steinhart et al and Baggish et al derived prediction models by using N- terminal pro-BNP values and clinical variables to improve the diagnosis of CHF [10,18]. Beside their potential benefits and diagnostic accuracy, all these models presented drawbacks as blood sampling results were required for diagnostic validation, tests that are not usually available in the emergency prehospital care setting, at least at the early phase of treatment. In fact, only a few studies have included clinical variables for diagnosis and most of them have also combined the use of chest X- ray such as Carlson et al or Schocken et al in the Boston score, thus prov- ing difficult to use at the early phase of treatment [19,20]. Other predic- tion models such as the one used in the Framingham study, [7] or the classification of Gothenburg [21], have also been proposed to determine the diagnosis of heart failure in various epidemiologic studies, but these models are not to be considered as clinical diagnosis scores as they inte- grate several variables that are not easily available at the bedside. Sever- al patients without a diagnosis of CHF were depicting elevated BNP values. Main reasons for such BNP values elevation were a history of chronic heart or renal failure with high baseline values, without evi- dences for a CHF episode and the occurrence of another pathological process. A contrario, several patients with a clear diagnosis of CHF were depicting low initial BNP values, due to an acute pathological pro- cess and the 2-hours lag period before BNP elevation. This suggests that in such acute cases, clinical examination and anamnesis, combined with the Brest score calculation, may be of greater interest.

Associating some variables to create a diagnostic prediction rule for

the presence or absence of a disease has shown to be of clinical benefit in other clinical settings [22,23]. Based on the AUC, the predictions of the Brest score discriminated well between CHF and dyspnea from other causes in the retrospective derivation as well as in the prospective validation phases (see Fig. 1). The prospective validation of the score provides further evidence that it can be used in two different ways:

(1) to screen a patient population by using the cut off values, as no gold-standard yet exists for CHF; (2) to estimate the probability of CHF over other causes to adjust Treatment protocols in the emergency

amination and the search for a non CHF medical cause for dyspnea, thus limiting medical treatment adverse events and a more accurate diagno- sis process.

Our study should be interpreted in the context of certain limitations. First, the lack of a gold standard for CHF diagnosis may be questioned. The retrospective review of medical records by experienced physicians taking into account clinical variables and the course of the disease have proved to be reliable in various clinical settings [27,28]. Although the de- pendence of reviewers on information derived from medical records may have involved some loss of data, this bias was limited during the prospec- tive validation phase by the standardization of the Clinical monitoring protocol. Excellent inter-operator agreement indicates that the adjudica- tion process allowed reliable differentiation between patients. Second, the Brest score has been internally and externally validated but still re- quires a prospective impact analysis validation [29]. Although this model was created from a cohort of patients admitted to a single center, the prevalence of several variables may not be similar elsewhere, there- fore limiting the clinical applicability of this tool to other ED. Third, as al- ready described in various clinical trials, several conditions may coexist in a single patient [30]. Two patients with a Brest score equal 3 were consid- ered as a false negative according to adjudication; BNP values in these 2 patients were below 400 pg/mL. Seven patients were considered as false positive (5 patients with a score = 9, 2 patients with a score >=10). All of them were elderly patients with concomitant lower respiratory tract infection and a previous severe cardiopathy diagnosis. In these cases, accurate analysis of anamnesis and BNP values (below 400 pg/mL) would have been contributive to a better diagnosis. In our study, 94/206 patients (45.6%) had an intermediate score value (between 4 and 8), the remaining 112/206 patients (54.4%) falling either below or above this gray zone. This however suggests that the present score would be useful in a substantial proportion of patients. In patients with interme- diate values or in suspected cases of false values, BNP sampling and lung ultrasonography may be interesting adjuncts for diagnosis [31].

Conclusion

We developed and validated a simple prediction score for CHF based on variables commonly available in the emergency care setting in order to give the physicians some guidelines in a significant number of dys- pnea cases. The Brest score is easy to compute and capable of classifying patients suspected of having CHF into 3 clinical probability categories (low, intermediate, and high) with a fair amount of accuracy. An exter- nal validation of this score should now be considered.

Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ajem.2016.08.023.

Competing Interest

No support was provided from any organization for the submitted work; the authors have no financial relationships with any organizations

that might have an interest in the submitted work in the previous three years, no other relationships or activities that could appear to have influ- enced the submitted work.

Details of Contributors

ELH guaranties that all authors had a determinant role while setting up the study, as described below.

ELH had the original idea, designed the work, and wrote the article. AB collected and analyzed data, and wrote the manuscript. EN and GLG designed the validation phase, performed statistical analysis, and helped improving the manuscript. PC and CGG analyzed the data and corrected the manuscript.

Transparency Declaration

ELH guaranties that the manuscript is an honest, accurate, and trans- parent account of the study being reported. No important aspects of the study have been omitted. Any discrepancies from the study as planned have been explained.

The study was performed without specific fundings.

The study was performed according to STARD 2015 standards.

Acknowledgements

The authors wish to acknowledge the help of all emergency physicians from Brest University Hospital in collecting the data. The corresponding author had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

References

  1. Tsuyuki RT, Shibata MC, Nilsson C, Hervas-Malo M. Contemporary burden of illness of congestive heart failure in Canada. Can J Cardiol 2003;19:436-8.
  2. Massie BM, Shah NB. Evolving trends in the epidemiologic factors of heart failure: ra- tionale for Preventive strategies and comprehensive disease management. Am Heart J 1997;133:703-12.
  3. Rosamond W, Flegal K, Friday G, Furie K, Go A, Greenlund K, et al. Heart disease and stroke statistics–2007 update: a report from the American Heart Association Statis- tics Committee and Stroke Statistics Subcommittee. Circulation 2007;115:e69-171.
  4. Remes J, Miettinen H, Reunanen A, Pyorala K. Validity of clinical diagnosis of heart failure in primary health care. Eur Heart J 1991;12:315-21.
  5. Fuat A, Hungin AP, Murphy JJ. Barriers to accurate diagnosis and effective manage- ment of heart failure in primary care: qualitative study. BMJ 2003;326:196.
  6. Ansari M, Massi BM. Heart failure: how big is the problem? Who are the patients? What does the future hold? Am Heart J 2003;146:1-4.
  7. McKee PA, Castel WP, McNamara PM, Kannel WB. The natural history of congestive heart failure: the Framingham Study. N Engl J Med 1971;285:1441-6.
  8. Marantz PR, Tobin JN, Wassertheil-Smoller S, Steingart RM, Wexler JP, Budner N, et al. The relationship between left ventricular systolic function and congestive heart failure diagnosed by clinical criteria. Circulation 1988;77:607-12.
  9. McMurray JJ, Adamopoulos S, Anker SD, Auricchio A, Bohm M, Dickstein K, et al. ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure 2012: The Task Force for the Diagnosis and Treatment of Acute and Chronic Heart Failure 2012 of the European Society of Cardiology. Developed in collaboration with the Heart Failure Association (HFA) of the ESC. Eur Heart J 2012;33:1787-847.
  10. Baggish AL, Sieber U, Lainchbury JG, Cameron R, Anwaruddin S, Chen A, et al. A val- idated clinical and biochemical score for the diagnosis of acute heart failure. The ProBNP Investigation of Dyspnea in the Emergency Department (PRIDE) Acute heart Failure Score. Am Heart J 2006;151:48-54.
  11. Colli A, Fraquelli M, Conte D. B-type natriuretic peptide in heart failure. N Engl J Med 2002;347:1976-8.
  12. Tulevski II, Groenink M, Van Der Wall EE, van Veldhuisen DJ, Boomsma F, Stoker J, et al. Increased brain and atrial natriuretic peptides in patients with chronic right ventricular pressure overload: Correlation between plasma neurohormones and right ventricular dysfunction. Heart 2001;86:27-30.
  13. Wuerz RC, Meador SA. Effects of prehospital medications on mortality and length of stay in congestive heart failure. Ann Emerg Med 1992;21:669-74.
  14. Wang CS, FitzGerald JM, Schulzer M, Mak E, Ayas NT. Does this dyspneic patient in the Emergency Department have congestive heart failure? JAMA 2005;294:1949-56.
  15. Maisel AS, Krishnaswamy P, Nowak RM, McCord J, Hollander JE, Duc P, et al. Rapid measurement of B-type natriuretic peptide in the emergency diagnosis of heart fail- ure. New Engl J Med 2002;347:161-7.
  16. McCullough PA, Duc P, Omland T, Nowak RM, Hollander JE, Herrmann HC, et al. B- type natriuretic peptide and renal function in the diagnosis of heart failure: an anal- ysis from the Breathing Not Properly Multinastional Study. Am J Kidney Dis 2003; 41:571-9.
  17. Maisel A. B-type natriuretic peptide measurement in diagnosing congestive heart failure in the dyspneic emergency department patient. Rev Cardiovasc Med 2002; 3:S10-7.
  18. Steinhart B, Thorpe KE, Bayoumi AM, Moe G, Januzzi Jr JL, Mazer CD. Improving the diagnosis of acute heart failure using a validated prediction model. J Am Coll Cardiol 2009;54:1515-21.
  19. Carlson KJ, Lee DC, Goroll AH, Leahy M, Johnson RA. An analysis of physicians’ rea- sons for prescribing long-term digitalis therapy in outpatients. J Chronic Dis 1985; 38:733-9.
  20. Schocken DD, Arrieta MI, Leaverton PE, Ross EA. Prevalence and mortality rate of congestive heart failure in the United States. J Am Coll Cardiol 1992;20:301-6.
  21. Eriksson H, Caidahl K, Larsson B, Ohlson LO, Welin L, Wilhelmsen L, et al. Cardiac and pulmonary causes of dyspnoea-validation of a scoring test for clinical-epidemiolog- ical use: the Study of Men Born in 1913. Eur Heart J 1987;8:1007-14.
  22. Wells PS, Anderson DR, Rodger M, Ginsberg JS, Kearon C, Gent M, et al. Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the models utility with the SimpliRED D-dimer. Throm Haemost 2000; 83:416-20.
  23. Le Gal G, Righini M, Roy PM, Sanchez O, Aujesky D, Bounameaux H, et al. Prediction of pulmonary embolism in the emergency department: the revised Geneva score. Ann Intern Med 2006;144:165-71.
  24. Ommen SR, Nishimura RA, Appleton CP, Miller FA, Oh JK, Redfield MM, et al. Clinical utility of Doppler echocardiography and Tissue Doppler imaging in the estimation of left ventricular filling pressures: a comparative simultaneous Doppler-catheteriza- tion study. Circulation 2000;102:1788-94.
  25. Rana R, Vlahakis NE, Daniels CE, Jaffe AS, Klee GG, Hubmayr RD, et al. B-type natri- uretic peptide in the assessment of acute lung injury and cardiogenic pulmonary edema. Crit Care Med 2006;34:1941-6.
  26. Li G, Daniels CE, Kojicic M. The accuracy of natriuretic peptides (brain natriuretic peptide and N-terminal pro-brain natriuretic) in the differentiation between trans- fusion-related acute lung injury and transfusion-related circulatory overload in the critically ill. Transfusion 2009;49:13-20.
  27. Ware LB, Matthay MA. Clinical practice. acute pulmonary edema. N Engl J Med 2005;353:2788-96.
  28. Schmickl CN, Shahjehan K, Li G, Dhokarh R, Kashyap R, Janish C, et al. Decision sup- port tool for early differential diagnosis of acute lung injury and congestive heart failure in medical critically ill patients. Chest 2012;141:43-50.
  29. McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users’ guides to the medical literature: XXII: how to use articles about clinical decision rules. Ev- idence-Based Medicine Working Group. JAMA 2000;284:79-84.
  30. National Heart, Lung, and Blood Institute Acute Respiratory Distress Syndrome (ARDS) Clinical Trials Network, Wheeler AP, Bernard GR, Thompson BT, Schoenfeld D, Wiedemann HP, et al. Pulmonary-artery versus central venous cathe- ter to guide treatment of acute lung injury. N Engl J Med 2006;354:2213-24.
  31. Picano E, Pellika PA. Ultrasound of extravascular lung water: a new standard for pul- monary congestion. Eur Heart J 2016;37:2097-104.

Leave a Reply

Your email address will not be published. Required fields are marked *