Abstract

1
Age, Cholesterol Status, and Bypass Surgery
Discriminate as to Whether Patients Develop
Postoperative Atrial Fibrillation
Subjects
Abstract

• Subjects were 152 males and 101 females from The
  Center for Cardiovascular Rehabilitation in Bad
  Schallerbach, Austria A post-operative
  in-patient cardiac rehabilitation facility.

Age, Cholesterol Status, and Bypass Surgery
Discriminate as to Whether Patients Develop
Postoperative Atrial Fibrillation. J.S. Green,
FACSM, R. Berent, S.P. von Duvillard, FACSM, P.
Schmid, and S.F. Crouse, FACSM. Texas AM
University, College Station, TX. Center for
Cardiovascular Rehabilitation, Bad Schallerbach,
Austria. Texas AM University-Commerce,
Commerce, TX. PURPOSE To employ discriminant
function analysis to develop a model which
discriminates between those who will develop
post-operative atrial fibrillation and those who
will not. METHODS The sample included in the
analysis was composed of 152 males (mean
age63.5, mean height 171.7 cm, mean weight
81.4 kg) and 101 females (mean age67.8, mean
height 160.9 cm, mean weight 70.9 kg)
undergoing cardiac rehabilitation at the Center
for Cardiovascular Rehabilitation in Bad
Schallerbach, Austria. All subjects were
in-patients. The SPSS statistical package
version 12.0 was used to develop the discriminant
function model (DF) from both continuous and
nominal cardiac related variables. Included in
the original discriminant model were age (AGE),
gender, whether one had bypass surgery (BYP),
height, weight, time on bypass pump, whether one
had valvular heart disease, and whether one had
high cholesterol (HYPCH). RESULTS A Wilks
Lambda of .873 (plt.0001) indicated the
discriminant model was significant. AGE, BYP,
and HYPCHOL were the only variables that loaded
on the model (plt0.05) with standardized canonical
discriminant function coefficients of .489, .536,
.547 respectively. This indicated that all three
variables contributed about equally to
discrimination. The non-standardized
discriminant coefficients and the corresponding
discriminant function were DF -.046(AGE)
1.11(BYP) 1.13(HYPCHOL) 1.84. The cutoff
score was determined using a weighted function of
the centroid means and was calculated to be
-.160. Cross validation analysis showed that
64.4 percent of cases were classified correctly.
CONCLUSION It was concluded that discriminant
function analysis could be used with reasonable
effectiveness in post surgical cardiac patients
to help determine whether or not they will
develop atrial fibrillation.
J.S. Green1, FACSM R. Berent2 S.P.
vonDuvillard3, FACSM P. Schmid2 and S.F.
Crouse1, FACSM
Methods

• The Discriminant Function program within the
  SPSS statistical
• package (version 12.0) was used to analyze the
  data and to determine if a viable function could
  be obtained.

1Applied Exercise Science Laboratory Department
of Health and Kinesiology Texas AM University

• Original variables entered into the discriminant
  model were
• Age (AGE)
• Gender (GENDER)
• Height (HT)
• Weight (WT)
• Time on Perfusion Pump (TOP)
• Whether or Not a Subject Had Bypass Surgery
  (BYP)
• Whether or Not a Subject Had Heart Valve Disease
  (HVD)
• Whether or Not a Subject Had High Cholesterol
  (HYPCHOL)

Background
2Center for Cardiovascular Rehabilitation Bad
Schallerbach, Austria

• Almassi et al., 1997
• Postoperative atrial fibrillation (AF) is common
  in cardiac surgery
• Predictors of AF included
• Age odds ratio 1.6
• Whether or not the patient had COPD odds ratio
  1.4
• Use of digoxin within two weeks before surgery
  odds ratio 1.4
• Resting pulse lt 80 bpm odds ratio 1.3
• Postoperative systolic blood pressure 120 odds
  ratio 1.2
• Steinberg et al., 1993
• Postoperative AF was associated with
• Left ventricular hypertrophy on resting pre-op
  ECG
• Lower ejection fractions that those that did not
  develop AF
• Longer P-wave durations on the resting post-op
  ECG

Results

• A Wilkes Lambda of .87 ( p lt .0001) indicated
  viable function that was statistically
  significant.
• The variables that loaded in the model were AGE,
  BYP, HYPCHOL, with standardized canonical
  discriminant function coefficients of .49, .54,
  and .55 respectively indicating that the three
  loaded variables all contributed about equally
  to discrimination.
• The Discriminant Function with non-standardized
  coefficients is
• -.046(AGE) 1.11(BYP) 1.13(HYPCHOL)
• The cut-off score was computed using a subject
  number weighted function of the centroid means
  and was found to be -.16
• Cross validation analysis demonstrated that 64
  of cases were classified correctly.

?-10.6
3Texas AM - Commerce Commerce, TX
Rationale Purpose
Conclusions

• To the authors knowledge, few studies exist
  which have used Discriminant Function analysis
  in an attempt to discriminate
• as to whether or not post-operative cardiac
  patients will develop atrial fibrillation
• The purpose of this study was to use
  Discriminant Function Analysis to
• determine if a function could be developed using
  common patient
• demographic and medical information which could
  discriminate as to whether or not post-operative
  cardiac patients will develop atrial
• fibrillation

• The results of this investigation suggest that
  Discriminant Function Analysis could be used as
  a tool to aid in determining which cardiac
  surgical patients will develop post-operative
  atrial fibrillation, however due to a
  significant mis-classification rate, it should
  not be used exclusively.