( Received April 3; revision accepted for publication June 23, 1972. )
The Comprehensive Cane Program is supported in part by the following Grants-in-Aid: ( 1 ) Project 919, Health Services and Mental Health Administration, Department of Health, Education, and Welfare. ( 2 ) Projects 72 and 304, State of Ohio Department of Health. ( 3) Project C-29, The National Founda-tion-Manch of Dimes. This evaluation was supported in pant by a stipend to C. W. from Case Western Reserve University School of Medicine.
ADDRESS FOR REPRINTS: (J.C.S.P.) Comprehensive Care Program, Cleveland Metropolitan General hospital, 3395 Scranton Road, Cleveland, Ohio 44109.
PEDIATRICS, Vol. 50, No. 5, November 1972
793
EVALUATION
OF
A TEN-YEAR
EXPERIENCE
IN A
COMPREHENSIVE
CARE
PROGRAM
FOR
HANDICAPPED
CHILDREN
Jane C. S. Perrin, M.D., Edna L. Rusch, R.N., M.S., Janet L. Pray, M.S.W., ACSW, Gregg F. Wright, B.S., and Glen S. Bartlett, M.D.
1’roin the Comprehensive Care Program, Department of Pediatrics, Cleveland Metropolitan General
Hospital, and the Department of Sociology, Case Western Reserve University
ABSTRACT. Chronically disabled patients under
care of a multidisciplinary hospital program were
scored for functional changes by retrospective
chart review. Family functions were also assessed. Professional input time was measured as the num-ber of clinic visits ( physician ) or years of social worker involvement. Correlation coefficients
calcu-lated between a number of variables did not show
significant zero-order correlations between
profes-sional quantitative input and improvement of pa-tient-family functions.
Pediatrics, 50:793, 1972, PROGRAM
EVALUA-TION, COMPREHENSIVE CARE, HANDICAPPED
CHIL-DREN.
A
N interdisciplinary team effort towardthe clinical diagnosis, management,
and training in the ambulatory and
inpa-tient care of physically and mentally
handi-capped children’ was established at
Cleve-land Metropolitan General Hospital in 1961
to eliminate fragmented care of the
chroni-cally ill child and his family. In 1970 to
1971 retrospective evaluation of
profes-sional input and patient outcome was
un-dertaken to develop a tool for ongoing
pro-spective study. The retrospective study was
an attempt to measure change in physical
and social disability.
The Comprehensive Care Program is
constructed on the premises that (1)
coor-dinated medical care for the handicapped
child is superior to fragmented care, (2)
coordinated care is better assured if the
family identifies with a consistent primary physician, (3) attention to family problems
IJy a social worker is an integral part of
treatment for the chronically handicapped
child, and (4 ) the physician and social
worker form the nucleus for
multidiscipli-nary management which they make
avail-able through their joint interaction with
each other, other professionals, and the
family.
Lacking a control group with which to
compare our patients, we sought to
deter-mine whether a relationship exists between
presence of a primary physician and
im-provement of arm and leg function and
in-dependence of the child; and between
pres-ence of a social worker and improvement in
parent-child relationship, school placement,
and school adjustment.
MATERIALS AND METHODS
Charts of 75 patients, a 10% sample of all
patients enrolled in the Comprehensive
Care Program for at least one year during
the period January 1, 1965, through
Decem-her 31, 1969, were randomly selected2 for
review. Detailed chart review was carried
out by three of us (J.C.S.P., E.R., and
J.L.P.) rotating in pairs with the third
per-son independently checking function scores
(see below ) . In virtually all instances the
independent score results were found to be
TABLE I
PATIENT VARIABLES
Paijeni
.
-Characteristics
-
-Paftenl Functzon .
Family Function
Age Arm-leg function (ALF) Parent-child re-lationahip (PCR)
Sex Independence (IND) Marital status (MAR)
Diagnosis Behavior (BHV) Economicatatua (ECON)
IQ School placement (SCP)
Schooladjustinent (SAD)
Ability to pay medical care (PAY)
Scoring criteria for patient function and family function variables are described in Tables IV and V. Appendix 1.
Patient Variables
Patient variables were scored on entrance
to and exit from the program, or as of the
date of chart review for patients still active
in the program. Patient variables are listed
in Table I. Strict definitions were
estab-lished for scoring of patient and family
function. Family function scoring was
based, where possible, on a modification of
an assessment system used by the Santa
Clara County, California, Department of
Welfare.3 Scoring criteria for patient
func-tion and for family function variables
are defined in Tables IV and V in
Appen-dix 1.
Program Variables
Assessment of professional input to a
given patient or family was limited to
quan-titative measurement, and determined the
program variables listed in Table II. The
primary physician (MDA) was the one
sin-gle pediatrician or physiatrist who attended
the patient for a greater number of clinic
TABLE II
PROGRAM VARIABLES
Length of stay in program (STAY) Number of clinic visits (VISITS)
Number of clinic visits attended by primary physician (MDA) and set of three physicians (MDS)
Percent of clinic visits attended by primary physician (% MDA) and set of three (% MD3)
Time on program assigned to social worker (time SW)
Percent of time on program with a social worker (% SW)
visits than did any other physician. The
percent of clinic visits attended by the
pri-mary physician is:
No. clinic visits with MDA
%MDA==
-Total No. clinic visits
x
100The three physicians (MDA plus two
others ) who attended the patient for a
greater total number of clinic visits than
did any other group constitutes the set of
physicians designated MD3, and percent of
visits attended by this set was calculated in
the same way.
Time with the social worker equals that
segment of patient stay during which a
so-cial worker was assigned. Percent social
worker was the percent of total stay that
the patient had contact with the social
worker:
Time social worker assigned
%SW x100
Total patient stay
The number of contacts as well as
to-tal duration of active contact were not
re-corded.
Data Analysis
As a measure of the mutual relationship
among the variables studied, zero-order
correlation coefficients were calculated
be-tween each pair of variables, and from
these, partial and multiple correlation
coef-ficients were calculated for selected subsets
of variables; these coefficients were
evalu-ated at the 1% level of significance. For an
N of 75 cases, a correlation coefficient (r)
of
+0.3
or -0.3 is significant at the 1%level. Details of the statistical analysis are
given in Appendix 2.
RESU LTS
1. Description of patients.
Of the 75 patients, 39 were male and 36
female. Age at entrance ranged from
new-born to 16.3 years, and stay from 1 to 10
years (median, 5 years ) . Most patients had
multiple diagnoses, making the total
num-her of diagnoses more than 75 (Table III);
TABLE III
PATIENT DIAGNOSES
No.
153 14 23
4 14 11 I 10
48
42
36
IMPROVED FUNCTIONS
Fc/r - good (2 --/)
U
Poor -a--good (3-.-/)Poor --fc/r (3 --2)
::J
Not opp//c.---good(O’-/) Not app//c. ---fo/r (0 -‘-2)24
12
falling into the educable and trainable
mentally retarded categories.5 The patients
made 0.85 to 35 clinic visits per year,
me-dian, 7.
2. Patient and family functions.
For each function measured, a matrix
was constructed depicting incoming and
outgoing scores of the 75 patients. Figure 1
illustrates the number of patients improved
in specific functions. The number of
pa-tients with worsening of a function ranged from 0 to 4.
3. Program variables.
Figure 2 shows physician consistency for
patient clinic visits. Only 11 patients were attended by the primary physician for > 50%
of their clinic visits, but 29 were at.
tended by one of the set of three primary physicians for > 50% of their visits.
Fifty-Mental retardation (MR) Seizure disorder
Cerebral palsy (CP)
Emotional, behavior, language disabilities (Emot.)
Missing limbs
Myelodysplasia or arthrogryposis (Myelo.) Blind (Partial/Total)
Blind and deaf
Bladder or bowel complications (UT Coniplic.)
six patients had a social worker for > 50%
of their stay in the program.
4. Correlatior.
Although there were a number of
signifi-cant positive and negative correlations
be-tween patient and program variables, there
It)
N-30
a
w
> 0
0
U) I-0
6
-BHV SCP SAD
-FUNCTION
FIG. 1. Patient and family function improvements. ALF = arm-leg function,
IND = independence, BHV = behavior, SCP = school placement, SAD =
school adjustment, PCR = parent child relationship, MAR marital status,
PERCENT CLINIC VISITS WITH PRIMARY PHYSICIAN (% MDA)
2I
18
15
a PRIMARY SET OF 3 PHYSICIANS (% MD3)
U
“ MOA%M03
fti 11
111I
#{128}
28.0
240
200
20
%# 10 20 30 40 50 60 70 80 90 00
% MDA/MD3
U)
2
4
a-9
o
z
Fic. 2. Consistency of primary physician attendance at patient’s clinic visits.
was no significant zero-order correkition be-tween our measurement of phyrdcian and social worker quantitative input and im-1)rovent of patient or family function.
To probe further for possible factors
in-fluencing physician and social worker input
on patient outcome, partial and multiple
correlations were obtained for specific
sub-sets of variables.
In the first example (Fig. 3) percent of
primary physician time (MDA) and
im-proved patient independence (INDf) did
not show significant zero-order or partial
correlation (controlling for degree of inde-pendence at entry to the program ) .
Corre-lation with percent primary physician time
does occur when all independence scores at
entry plus improved independence are
con-sidered in a multiple correlation matrix,
with or without a variety of other variables
at entry. In Figure 4, a similar result is
illus-trated for the correlations between percent
social worker time (%SW ) and improved
parent-child relationship (PCRt).
DISCUSSION
Positive or negative zero-order
correla-tions do not delineate cause and effect, but
merely indicate that variables are related to
each other. Nonetheless, presence or
ab-sence of correlations in the context of this
study can suggest hypotheses to apply
to-ward improving the direction of such a pro-gram.
Use of partial and multiple correlations
as a statistical tool, however, did not result
in data directly applicable to program
im-provement: in a patient population it is not
possible to control the multiple factors in
combination that must be added for final
correlation of a functional improvement
and professional time input.
The task of demonstrating that
compre-hensive care (defined here as
multidiscipli-nary coordinated care for the patient and
support for the family ) is superior to
un-coordinated care has proved formidable to
many examiners. Lewis#{176}summarizes papers
presenting negative findings, including a
controlled study of three Guatemalan
vil-lages7 demonstrating a lower rate of illness
in the unattended (control) village than in
either the village receiving public health
services or the one receiving a high protein
feeding supplement. Lewis identifies the
search problem as one of identifying
com-pletely all of the input factors and the
re-sulting output they yield.
SPECULATION
In our study the lack of correlation
Correlation IND4
I IND #{227}I-2--3
81 Other Var. K
‘.AGE
I
q) +MR
fCP
I
V3 EMOTIONAL
:
+MYELOI
: +10
tj +p
is-ol
I +ECON +2
I
+ECON I2+3I
+BHV 1+2+3I
+MAR I+2j’-3+OL +ALF I2+3
+STAY
I
+VISITS
I
-.6 -.3 0 +.3 4.6
l-’ic. 3. Correlation of improved patient independence ( IND ) with percent
)rimar physician visits (#{182}NIDA) modified by incoming IND and other
variables. A bar to the right from 0 designates a positive correlation, left from 0 a negative correlation; the 1A significance level of ±0.3 is marked 1))’ (1 broken line. For explanation of abbreviations see Tables I, II and III.
NlII1bers refer to functional eores ( Tables I\’ and V, Appendix 1).
PERCENT
SW
PCR
PCR
I
Zero order Correlation1
PCR4PCR 1-2-3 KCorrelation
797
PERCENT
MDA
IND
+AGE
I
+ STAY
+ VISITS
oj+MR
I
#{247}CP
I
+EMOT IQ
L +BHV 1+2+3
I
I + IND I2+3
I
ECON 1+2I
+ScP 1+0I
I
#{247}MARI2+3+O+MYELO I
L +SAD 1+2+3+10
hi
I
Zero -order CorrelationI
PartialT
k) +
C%J
:
+
-.-...
c #{188}.
I..
-..-.
+
3
c:3
.--.,- 4...
Iji
IL Part/cl
,j#/jyaI
-/,iffil/,-/4
yjjff4-/#7ffArArA
- -.3 0 +
l-I(;. 1. Correlation of iIl)1)roved l)IreIlt-chill relationship ( PCR t ) with
per-CCI)t social \v()rker tulle ( #{182}S\V) modified Iw incoming PCR and other
been attributed to two major groups of factors:
(
1) Patient variable scoring wascategor-ical, and gross change in function was
re-quired for change of score; limited chart
information eliminated the possibility of
scoring for finer increments of functional
improvement, such as degrees of increased
independence or modest behavioral
im-provement. In addition, assessment of
func-tion at only two, often widely separated,
points in time did not allow an evaluation of
rate of improvement in function, nor of how
quickly after enry into the program the
fin-provement occurred.
(
2)
Measure of professional input bytime with little indication of quality or type
of input was unsatisfactory. We did not
know the extent of involvement of the
pni-mary physician or the patient’s reaction to
the physician. During the period a social
worker was assigned, there was usually no
notation of the number or the depth of
con-tacts with a family.
We are in the process of refining the tool
formulated from the present study for a
prospective evaluation which will overcome
many of these limitations. Experience with
the retrospective study suggests that an
on-going program evaluation can be
economi-cally incorporated into a service and
train-ing program of this type.
SUMMARY
Randomly selected charts of a 10%
sam-ple of patients (N 75) were reviewed to
score patient and family functions at entry
and exit from the program (patient
van-ables ) , and to measure professional input
(
program variables ) . Scoring categorieswere defined and graded for patient
func-tions of arm-leg use, independence,
behav-ior, school placement-adjustment; and for
family functions of parent-child
relation-ships, marital status, economic status, and ability to pay for medical care. The
profes-sional input measure was quantitative and
included number-percent of patient’s clinic
visits attended by the primary MD and
length-percent of time in program patient
had a social worker. Zero-order correlations
were determined as a measure of mutual
re-lationship among all patient and program
variables, and partial and multiple
correla-tions were determined among selected
sub-sets of variables. Results: there was no
sig-nfficant zero-order correlation between
professional quantitative input and
im-provement of patient-family function.
Cor-relations were present when a number of
interacting variables were considered
to-gether. Experience with limitations of
retro-spective evaluation will be applied to
de-velop the tool for a prospective evaluation
of changes in disability level of patients
within our program.
Acknowledgment
We are grateful for technical assistance from
Ms. Carolyn Sands and Ms. Marlene Leppelmeier.
REFERENCES
1. Allen, J. E., Lelchuck, L. : A comprehensive care program for children with handicaps.
Amer. J. Dis. Child., 111 :229, 1966.
2. Hald, A. : Statistical Tables and Formulas.
New York: John \Vile’ and Sons, p. 92,
1952.
3. Guide for identifying functional level
devel-oped by Department of Welfare, Santa
Clara County, California, no date.
4. Fisher, B. A., and Yates, F. : Statistical Tables for Agricultural, Biological and Medical
Re-search. New York: Hafnen Publishing Co. Inc., 1957.
5. Diagnosis and Statistical Manual of Mental Disorders ( Second Edition) . Washington,
D.C. : American Psychiatric Association, 14,
1968.
6. LeWiS, C.: Symposium: Does comprehensive cane make a difference? Wlrit is the
evi-dence? Amer. J. Dis. Child., 122:469, 1971. 7. Scrimshaw, N. S., Cuzman, M. A., Flores, M.,
and Cordon, J. E. : Nutrition and infection
field studies, Cuatemalan villages, 1959-1964: V. Disease incidence among preschool
children under natural village conditions, with improved diet, and with medical and public health services. Arch. Environ. Health, 16:223, 1968.
8. Growth and Development, Occupational Then-apv Assistant Training Program. Coliinihiis, Ohio, fl() date.
1. Normal
2.Partial
3. Minimal or none
I). School Placement (SCP) Score
1.Appropriate
. Inappropriate
0. Not applicable
9. Unknown
E. School Adjustment (SAD) Score
1. Good .Fair
3. Poor
0. Does not apply 9. UILkIIOWH
TABLE V
FAMILY FuNcrIoN VARIABLES
A. Parent-child relationship (PCR)
Score Definition
1.No significant Basic needs met, loved,
problems
ft.Moderate problems
799
lished monograph), Department of
Sociol-ogy, Case Western Reserve University,
Cleveland, Ohio, no date.
10. Czannocki, B., and Thiessen, V. : Program MU-PAR (unpublished monograph) Department
of Sociology, Case Western Reserve Univer-sity, Cleveland, Ohio, no date.
1 1. Goldberger, A. S. : Econometric Theory. New
York: John Wiley and Sons, P. 218, 1964.
TABLE IV
PATIENT FUNCTION VARIABLES
APPENDIX I
TABLE IV ((‘onlinued)
A. Arm-leg function (ALF)
Score
9. Unknown
B. Independence (IND). Score
1.Norlnal for age
. Partial dependence 3. Total dependence
0. Does not apply 9. Unknown C. Behavior (BHV)
Score
1.Normal
. Erratic 3. Disturbed
0. Does not apply 9. Unknown
Definition Legs: normal gait with or
without assistive do-vices; normal active leg motion in infants Arms :Age level gross and
fine motor coordination with or without assis-tive devices
Legs: Aml)ulation with or without assistive de-vices, but abnormal gait
Arms: Gross and/or fine ulotor coordination he-low age level
Legs : Nonambulatory even with assistive devices
Arms: Inability to per-form motions required for feeding even with assistive devices
Definition
Ability to perform all ac-tivities of daily living (ADL8) at age level Some ADL at age level Inability to perform any
ADL at age level Less than 1 year of age
Definition
No indication of distur-bance
Inconsistent, impulsive Hyperactive, destructive,
uncontrolled IQ<20
3. Severe problems
9.Unknown
B. Marital Function (MAR) Score
1. No significant problems
. Moderate problems
Definition
III proper school for hand-icap and intelligence Eligible but not in school,
or in wrong type school or class
Less than S years old or too handicapped for any available day school
Definition
No unusual problems Problems containable Excluded from school for
severe problems Not in school or too soon
to tell
realistic expectations. Inconsistent in meeting
needs, overrestrictive-overindulgent Gross neglect, rejecting
Definition
Emotional rapport, share responsibilities, realis-tic tolerance
TABLE V (onLinued)
3. Severe prol)lems
9.Unknown
D. Ability to pay for medical care (PAY)
Score Definition
1.Needs no assistance Wealthy, complete
in-surance coverage, or welfare pays all bill8
. Partial assistance Incomplete third party coverage: Blue Cross, Crippled Children’s Service, etc. 3. No assistance No third party assistance
and bills a severe burden
9.Unknown
or the other or both of the variables of
inter-est, e.g., the relationship between primary
l)hysician time and improvement ill l)Ltient
independence, ignoring all other variables.
A Partial correlation coefficient (r1.23)
pro-vides a symmetric measure of the
relation-ship between two variables when the effect
of a third or additional variables is taken
into account or held constant, e.g., the
rela-tionship between primary physician time and
improvement in patient independence when
incoming independence status is taken into
account. A multiple correlation coefficient
(
R12.3) provides a symmetric measure of thetotal relationship between any variable afld
a group of other variables whose effects are
operating simultaneously, e.g., the
relation-ship between primary physician time and
improvement in patient independence when
the combined effects of incoming
indepen-dence status, age, diagnosis, and other
van-ables of interest are all operating
simulta-neously. The square of the correlation
coefficient (r or R2) provides a measure of
the percent of variance in any one variable
which results from its relationship to any
other variable or group of variables; thus,
an rl.2= OS (the 1% level of significance for
N=7.5) means that 9% [(0.3)2=0.09] of the
variance of X1 results from its relationship
to X2, and vice versa.
For the purpose of the correlation analysis
the categorical (ordinal) function variables
were recoded uS “dummy” nominal
van-ables,” with a child receiving a “yes” code
(X = 1) or a “no” code (X = 0) depending on
whether or not he had received each
applica-ble code (see Tables IV and V for each
variable.)
To say that a correlation coefficient of
+0.3 or -0.8 is significant at the 1%
level is to say that only 1% of the time a
con-relation of this magnitude would occur at
random and be misinterpreted as a true
cor-relation (Type I error).
APPENDIX 2
Statistieal Analyth
Correlation analysis was carried out by
Univac 1108 electronic computer utilizing a
zero-order correlation program MDC#{216}RR
(jissing Data Correlation),9 and a partial
and multiple correlation program, MUPAR
(Multiple and Partial Correlation).’#{176} Both
programs calculate correlation coefficients by
matrix inversion. Programming for the
anal-ysis was carried out by G.W. with the advice
of G.S.B.
A zero-order correlation coefficient (nl.2)
provides a symmetric (nondirectional)
mea-sure of the relationship between two
van-ables, Xi and X2, ignoring the effect of any
other variables that may be influencing one
0.Does not apply
9.Unknown
C. Economic Status (ECON) Score
1. Self-support
2.Unemployed,
assis-tance 3. Severe poverty
Constant friction, pat-tern of separation and divorce
Single parent family
Definition
Employed, income at or above minimal wage level
Receives income from
Welfare, Social Secu-rity, pension, etc. No income, or below