The New York University Pediatric Heart Failure

Index: A new method of quantifying chronic heart

failure severity in children
Dana Connolly, RN, MS, Monika Rutkowski, MD, Marcelo Auslender, MD, and Michael Artman, MD
diac physiology, clinical presentations,
Objective: The assessment of the severity of heart failure in pediatric pa- and compensatory mechanisms.2,3
tients is handicapped by the subjectivity of diagnostic parameters. This Moreover, the etiologies of heart fail-
study evaluated the feasibility of a new standardized heart failure index, the ure in children are very different from
New York University Pediatric Heart Failure Index (NYU PHFI), to those in adults.4 For these reasons,
quantify the degree of heart failure in a selected pediatric population. Ross et al5 developed a scale that is
useful for grading heart failure severi-
Methods and Results: The index is a weighted, linear combination of
ty in infants from birth to 6 months of
scores based on symptoms, physical signs, and medical regimen. Overall,
healthy children (n = 12) scored very low (0 to 2) on this index. Mean See editorial, p 618.
scores of children (<2 years; mean age, 4.8 months; n = 12) with a left-to-
right shunt lesion declined from 11.4 (SD ± 4.1, P < .001, 2-tailed test) be- age but is not applicable to older chil-
fore surgery to 1.8 (SD ± 1.3) after surgical correction of their cardiac de- dren and adolescents. Thus it is imper-
fects. The average inter-observer correlation coefficient was 0.95 (P < .001), ative to develop a convenient, noninva-
despite a wide range of scores. sive heart failure index for use with
Conclusions: The NYU PHFI appears to be a reliable and convenient pediatric patients of all ages. Such a
instrument for measuring heart failure severity in children. These initial standardized method of determining
results support further testing in broader diagnostic and age groups and the degree of heart failure would great-
ly facilitate outcome analyses, treat-
over longer periods. (J Pediatr 2001;138:644-8)
ment evaluations, prognostic assess-
ments, quality-of-life assessments, and
therapeutic trials in children with heart
Determinations of the degree of heart plied to non-infant pediatric popula- failure. The purpose of this study was
failure severity are essential for clini- tions, none have been validated in chil- to test the validity and reliability of the
cal assessment and treatment evalua- dren. The NYHA classification mea- newly developed New York Universi-
tion. Remarkably, no readily applied sures functional capacity, not heart ty Pediatric Heart Failure Index in a
standardized method of quantifying failure severity. The NYHA classifica- population of children with chronic
heart failure severity exists for pedi- tion is inappropriate for use in pedi- heart failure.
atric patients. Although standard atrics because functional capacity and
index methods from the adult popula- responses to heart failure states differ NYHA New York Heart Association
tion, such as the New York Heart As- in children compared with adults. NYU PHFI New York University Pediatric
Heart Failure Index
sociation Classification,1 have been ap- Children have markedly different car-

From New York University, School of Medicine, Department of Pediatrics, Pediatric Cardiology, New York, New
METHODOLOGY
York. Instrument Development
Supported in part by the Seventh Masonic District Association, Inc.
Submitted for publication July 24, 2000; revision received Nov 8, 2000; accepted Jan 4, 2001. The NYU PHFI is a weighted, linear
Reprint requests: Dana Connolly, PNP, NYU Pediatric Cardiology, 540 First Ave, TWR Suite combination of scores based on physi-
9V, New York, NY 10016. ologic indicators and medical regimen
Copyright © 2001 by Mosby, Inc. (Table). It carries a range of possible
0022-3476/2001/$35.00 + 0 9/21/114020 scores from zero (no heart failure) to
doi:10.1067/mpd.2001.114020 30 (severe heart failure).

644
THE JOURNAL OF PEDIATRICS
VOLUME 138, NUMBER 5
The process of developing this pedi-
atric heart failure index began by creat-
ing an item pool of markers of chronic
heart failure in children. Items were
generated from the pediatric heart fail-
ure literature and reviewed by a panel
of 8 pediatric cardiologists. Items were
divided into signs and symptoms, con-
ventional heart failure medications, and
ventricular pathophysiology. The origi-
nal index produced a score from 0 to
20. Initial reliability and validity testing
was done by using the original version.
Sign and symptom items include
heart failure markers identified
through routine history and physical
examination. First, physical markers of
heart failure identified through diag-
nostic tests were equated with corre-
sponding markers identified by physi-
cal examination alone. For example,
cardiomegaly and pulmonary edema
can be identified and scored by either
x-ray or physical findings. Similarly,
the presence of a gallop by ausculta-
tion was equated with abnormal ven-
tricular function determined by
echocardiogram. A gallop on physical
examination may be associated with
abnormal ventricular function, al-
though patients who present with
grossly abnormal ventricular function
as determined by an echocardiogram
may not have a gallop on physical ex-
amination. Although detected through
different means, these items were
equated so that in the absence of diag-
nostic test results, the PHFI will detect
abnormal ventricular function based
on physical findings if a gallop is pre-
sent. Next, items were adjusted to con-
trol for developmental differences in
clinical presentation. For example,
tachypnea was equated with dyspnea,
activity intolerance with prolonged
feeding time, and failure to thrive in
babies with cachexia in older children.
Finally, items were weighted according
to the degree of clinically associated
heart failure severity and lack of other
attributable causes.
Medical therapy items were included
because the absence of symptoms does
CONNOLLY ET AL
Table. The NYU PHFI
Score Signs and symptoms
+2 Abnormal ventricular function by echocardiogram or gallop
+2 Dependent edema or pleural effusion or ascites
+2 Failure to thrive or cachexia
+1 Marked cardiomegaly by x-ray or by physical examination
+1 Reported physical activity intolerance or prolonged feeding time
+2 Poor perfusion by physical examination
+1 Pulmonary edema by x-ray or by auscultation
+2 Resting sinus tachycardia
+2 Retractions
Hepatomegaly
+1 < 4 cm below costal margin
+2 > 4 cm below costal margin
Observed tachypnea or dyspnea
+1 Mild to moderate
+2 Moderate to severe
Medications
+1 Digoxin
Diuretics
+1 Low to moderate dose
+2 High dose or more than 1 diuretic
+1 ACE inhibitors or non-ACE inhibitor vasodilators or angiotensin
receptor blockers
+1 β-Blockers
+2 Anticoagulants not related to prosthetic value
+2 Anti-arrhythmic agents or ICD
Physiology
+2 Single ventricle
Heart failure severity is determined from signs and symptoms, medical regimen, and ventricular
physiology. A total score is derived by adding scores attributed to each individual criterion.
Scores can range from zero (no heart failure) to 30 (severe heart failure).
ACE, Angiotensin-converting enzyme; ICD, implantable cardiac defibrillator.
not imply the absence of disease. The weight was then added for higher
addition of heart failure medications dosages of titratable medications, such
(increasing the score) compensated for as diuretics. Additional weight was
the decrease in signs and symptoms also added to items associated with
(decreasing the score) to capture pa- clinically severe heart failure, such as
tients whose symptoms have been pal- anticoagulants not related to the pres-
liated medically. Medication items ence of a prosthetic valve, antiarrhyth-
refer to the type and number of current mic therapy, or the presence of an im-
medications prescribed. These items plantable cardiac defibrillator. An
are based on medical therapy for heart implantable cardiac defibrillator was
failure commonly used today.6 Items included as a sign of heart failure, even
include digoxin, diuretics, angiotensin- though a patient may only have a pri-
converting enzyme inhibitors, β-block- mary arrhythmia and not be in heart
ers, anticoagulants, and anti-arrhyth- failure. In such a case, a patient with
mic medications. Each medication item only a primary arrhythmia and im-
was first weighted equally. Additional plantable cardiac defibrillator would
645
 CONNOLLY ET AL
score very low and thus be considered
to have a clinically insignificant degree
of heart failure.
Single ventricle was included as a
physiologic indicator because in such
patients, ongoing changes in the my-
ocardium at the cellular level may not
be evident by clinical or technical mea-
sures. Single ventricle was treated as a
separate entity because all other physi-
ologic indicators of heart failure would
be detected through the other signs
and symptoms comprising the index.
After several iterations and applica-
tions in the clinical setting, it became
apparent that subjective signs and
symptoms such as abnormal ventricu-
lar function, tachycardia, and tachyp-
nea should not be objectively defined
because of variations caused by techni-
cal, environmental, and case-sensitive
differences. The investigators postulat-
ed that the judgment of the clinician
would be the most accurate, practical,
and convenient method of scoring pa-
tients, thereby facilitating utilization of
the instrument.
Sample
This study was conducted after ap-
proval from the NYU human subjects
committee. Subjects were selected
from a sequential population of chil-
dren treated at NYU Medical Center.
This study was aimed primarily at es-
tablishing the reliability of the index.
Thus the first step was to determine
whether the index measured children’s
heart failure signs and symptoms ver-
sus those of other chronic illnesses. By
scoring children before and after
surgery to repair heart failure-induc-
ing lesions, we first sought to establish
that the index was indeed measuring
heart failure. The next step in validity
testing was to examine a larger cohort
of children with varying degrees of
heart failure.
A total of 133 children were exam-
ined in order to assess: (1) initial inter-
observer reliability assessment (n =
11), (2) comparison between children
with (n = 12) and without heart failure
646
(n = 12) and comparison of scores in
the heart failure group before and after
cardiac surgery, and (3) heart failure
scores in a sample of children with
varying degrees of heart failure (n =
98). The mean age was 10.8 years (SD
± 7.6). Ethnicity was diverse, with 36%
white, 13.2% African American, 13.2%
Asian, 25.4% Latin American, and
12.3% belonging to other ethnic
groups. Sixty-three percent of subjects
were male and 37% were female. In the
cardiac group, diagnoses included
valvular obstruction and regurgitation,
post-repair of tetralogy of Fallot, unre-
paired left-to-right shunt lesions, sin-
gle ventricle, cardiomyopathy, and pri-
mary arrhythmia.
Procedures
Inter-observer reliability was initial-
ly assessed in a group of children (n =
11) from 2 days to 16 years of age
(mean age, 3.6 years) with known
heart disease and varying degrees of
heart failure, from none to severe. Ob-
servers included a pediatric cardiology
attending physician, a pediatric cardi-
ology fellow, a pediatric senior resi-
dent, and a pediatric nurse practition-
er. This panel was chosen because a
panel of 8 experienced pediatric cardi-
ologists developed the index and the
investigators needed to establish its va-
lidity for use with observers of various
levels. All observers independently re-
viewed the patient’s charts and inde-
pendently examined and scored each
patient. In assessing inter-observer re-
liability, care was taken to ensure that
each observer would indicate exactly
which items were being included in the
total score and that each observer
would remain blinded to others’ obser-
vations and scoring decisions.
Validity was assessed by comparing
a group of 12 children immediately be-
fore surgery to correct left-to-right
shunt lesions with an age-matched
group of 12 children without heart dis-
ease. Children’s ages ranged from 0 to
2 years, with a mean age of 4.8 months.
Diagnoses in the cardiac group includ-
THE JOURNAL OF PEDIATRICS
MAY 2001
ed complete atrioventricular septal de-
fect, ventricular septal defect, and
patent ductus arteriosus. Four inde-
pendent observers scored the children.
As a further test of validity, patients in
the same heart failure group were then
re-evaluated 3 weeks after surgery to
determine heart failure scores after re-
pair of the causative lesion.
RESULTS
Reliability Assessment
An overall interclass correlation co-
efficient of 0.95 (P < .001) was ob-
tained among subjects by 4 indepen-
dent observers, demonstrating high
reliability of this index in patients with
low, medium, and high scores. These
results are shown in Fig 1.
Validity Assessment
The difference in mean scores be-
tween the group of children immediate-
ly before surgery to correct left-to-right
shunt lesions and an age-matched
group of children without heart disease
was evaluated. Control patients scored
very low (0 to 2), as would be expected
for a valid index. Similarly, children
with large left-to-right shunts awaiting
surgical repair had significantly higher
scores (mean, 11.4; SD ± 4.1; P < .001),
suggesting that the index discriminates
between normal states and heart failure.
Subjects in the heart failure group
were re-evaluated approximately 3
weeks after surgical repair of their left-
to-right shunt lesions. Mean heart fail-
ure scores declined from 11.4 ± 4.1 be-
fore surgery to 1.8 ± 1.3 after surgery
(P < .001, 2-tailed test). Mean heart
failure scores in this sample from each
of the 4 independent observers are il-
lustrated in Fig 2.
Modifications
Based on feedback from those using
the index, minor modifications were
made. First, overlapping growth para-
meter items were combined and items
were re-weighted accordingly. Mean
heart failure scores in the preoperative
THE JOURNAL OF PEDIATRICS CONNOLLY ET AL
VOLUME 138, NUMBER 5

heart failure group were 11.4 with an

SD of only 4.1. In order to detect more
subtle differences, hepatomegaly, degree
of tachypnea or dyspnea, and diuretic
regimen were then further stratified into
degree of severity. The range of scores
was expanded from 0-20 to 0-30.
The revised index was then applied
in a study of quality of life in children
with heart disease and varying degrees
of heart failure.7 During this phase of
the study, each observer was produc-
ing the same scores for each patient,
with 0 to 1 point variability (inter-
observer reliability: r = 0.95, P < .001);
thus, inter-rater assessments were dis-
continued and a single observer’s as-
sessments were used for the analysis.
The distribution of heart failure scores
in 98 children with various types of Fig 1. Inter-observer agreement in varying degrees of heart failure. Heart failure scores for individ-
heart disease demonstrate that scores ual subjects were determined by 4 independent observers (1 nurse practitioner and 3 physicians).
were distributed over a wide range Close agreement is evident among observers over a wide range of scores.
with no artifacts caused by clustering.

DISCUSSION
Heart failure is a common manifesta-
tion of both congenital and acquired
heart disease in the pediatric popula-
tion. Heart failure has been defined as
“disorder of the circulation, not merely
as disease of the heart. Heart failure
develops not when the heart is injured,
but when compensatory hemodynamic
and neurohormonal mechanisms are
overwhelmed or exhausted.”8 Chronic
heart failure syndrome has been de-
fined as “a condition of cardiac pump
dysfunction with activation of compen-
satory responses that ultimately con-
tribute to silent and progressive deteri-
oration (of) myocardial function.”3
Recognized signs and symptoms of
chronic heart failure in children in- Fig 2. Mean heart failure scores before and after surgical repair of left-to-right shunt lesions. Children
with left-to-right shunt lesions were evaluated before (preoperative score) and after (postoperative
clude respiratory distress, dyspnea on score) surgical repair of their cardiac defects. Mean scores for each individual observer are plotted.
exertion, feeding problems, growth re-
tardation, diaphoresis, hepatomegaly,
cool extremities, and poor peripheral toxins.4 Simply establishing the pres- severity in infants and children has
perfusion. Pediatric heart failure may ence of heart failure and determining proven to be very difficult.
be due to a multiplicity of etiologies, its etiology in children are fairly Like the NYHA and Ross classifica-
including congenital lesions, arrhyth- straightforward processes. In contrast, tions,1,5 the NYU PHFI was devel-
mias, infections, tumors, drugs, and however, quantifying heart failure oped by a committee of experts. The

647
 CONNOLLY ET AL
investigators in the Ross classification
study compared the presence and
severity of items with subjective rat-
ings of heart failure to determine
which items were the most strongly as-
sociated with moderate to severe heart
failure.5 Using a similar approach, we
compared scores of patients with se-
vere degrees of heart failure requiring
surgery and found them to be signifi-
cantly higher than scores of patients
who did not require surgery. We did
not seek to include only items that
were most strongly associated with se-
vere heart failure, because pathophysi-
ology and clinical presentation vary
greatly across disease states and ages.
The results of this study of the NYU
PHFI demonstrate high inter-observer
reliability, because scores correlated
strongly among qualified independent
observers over a wide range of heart
failure scores. Although co-morbid
conditions may cause some of the signs
and symptoms on the index, we have
found that such conditions do not sig-
nificantly change scores. Furthermore,
the index accurately discriminated be-
tween pediatric patients with and with-
648
out heart failure and consistently de-
tected changes in chronic heart failure
severity over time.
This study was the first step in vali-
dating the index to determine whether
revisions were necessary and further
validation studies would be feasible.
Comparison of NYU PHFI scores with
neurohormonal indices of heart failure
would further validate the index as a
measure of heart failure severity from a
physiologic standpoint. In order to de-
termine the prognostic value of the
NYU PHFI, the association between
specific ranges of heart failure scores
and morbidity and mortality in children
with heart failure should also be ex-
plored. Although these data were de-
rived from subjects from only one med-
ical center, the results may be
generalizable to other cohorts within
this population because of the demo-
graphically diverse nature of the sub-
jects involved in this study. Still, the
index should be tested in larger, more
geographically diverse samples.
We thank Drs Durismay Balaguru, Dolores
Danilowicz, Abhay Divekar, Brian O’Connor,
and Rajiv Verma for their helpful comments.
THE JOURNAL OF PEDIATRICS
MAY 2001
We also thank the children who participated in
this study.
REFERENCES
1. The Criteria Committee of the New
York Heart Association. Diseases of the
heart and blood vessels: nomenclature
and criteria for diagnosis. 6th ed. Boston:
Little, Brown and Co; 1964. p. 112.
2. Furdon SA. Recognizing congestive
heart failure in the neonatal period.
Neonatal Network 1997;16:5-13.
3. Auslender M. Pathophysiology of pe-
diatric heart failure. Prog Pediatr Car-
diol 2000;11:175-84.
4. Balaguru D, Artman M, Auslender M.
Management of heart failure in chil-
dren. Curr Prob Pediatr 2000;30:1-36.
5. Ross RD, Bollinger RO, Pinsky WW.
Grading the severity of heart failure
in infants. Pediatr Cardiol 1992;13:
72-5.
6. Auslender M, Artman M. Overview of
the management of pediatric heart fail-
ure. Prog Pediatr Cardiol 2000;11:231-
41.
7. Connolly D, Rutkowski M, Auslender
M, Artman M. Quality of life and
heart failure severity in pediatric pa-
tients [abstract]. Pediatr Res 2000;
47:40A.
8. Packer M. Pathophysiology of chronic
heart failure. Lancet 1992;340:88-92.