evaluation methods in disease management studies
TRANSCRIPT
Dis Manage Health Outcomes 2008; 16 (5): 365-373EXPERT OPINION 1173-8790/08/0005-0365/$48.00/0
© 2008 Adis Data Information BV. All rights reserved.
Evaluation Methods in DiseaseManagement Studies2004–07
Thomas Wilson,1,2 Martin MacDowell,1,3 Patricia Salber,1,4 Gary Montrose1,5 and Carolyn Hamm1,6
1 Population Health Impact Institute, Loveland, Ohio, USA2 Trajectory Healthcare, LLC, Loveland, Ohio, USA3 UIC College of Medicine at Rockford, Rockford, Illinois, USA4 Universal American Corp., Houston, Texas, USA5 Montrose Healthcare Strategies, LLC, Denver, Colorado, USA6 Walter Reed Army Medical Center, Washington, DC, USA
A variety of program evaluation designs are available to assess the impact of disease or care managementAbstractprograms, which can make it difficult to compare outcomes of different interventions. The need to compareprograms has resulted in consideration of standardizing evaluations of disease management programs; however,recommendations on the conduct of such evaluations have not been widely adopted. The purpose of this article isto examine the consistency of study characteristics of disease management peer reviewed evaluations over a3-year period (1 January 2004–28 February 2007) and to suggest questions that must be answered to ensure basictransparency of methods and metrics.
Study designs vary considerably among the current literature on evaluations of disease managementinterventions involving US health plans (25 studies). The mechanism for defining the intervention populationswere not consistent, even among interventions focused on a single disease, and evaluations employed bothadministrative and clinical data. The current literature included both randomized (n = 10) and non-randomizedstudies (n = 15). The referent population varied among the non-randomized studies, and included data from thepre-intervention period and both concurrent and historical control groups. The outcome metrics used in theevaluations included mortality and readmission rates, as well as time to readmission and various cost parameters.The majority of reviewed studies corrected for the confounding variables of age and sex, and a high proportioncorrected for a range of other confounding factors.
In conclusion, the evaluations of disease management programs in the literature cannot be consideredstandardized. To increase the transparency and validity of disease management intervention evaluations, werecommend consideration of five basic questions regarding intervention descriptions, intervention population,referent population, outcomes metrics, and confounding variables. Standardization on such basic parameters is anecessary step towards being able to assess the quality and validity of evaluations. Such standardization isessential for comparing the effectiveness of alternative programs, and to enable data-driven value-basedpurchasing decisions.
A 2003 article commissioned by the Disease Management management: follow-up, case-control, cross-sectional, ecological,Association of America (DMAA) listed seven different study quasi-experimental, benchmark, and post-only.[1] In 2005, thedesign types that could be used to examine the impact of disease
366 Wilson et al.
DMAA Outcomes Guide[2] listed 12 different ‘study design char- In 2006, a study was commissioned by CorSolutions, a diseaseacteristics’: management company, and performed by the RAND Corporation.
It recommended “in the absence of a control group … the use of1. randomized controlled trial (RCT)benchmark data as rough estimate of secular trend and statistical2. pre-post with the patient as their own controlestimates to the degree possible.”[9] Also in 2006, Sexner et al.[10]
3. pre-post historical control design with concurrent controlrecommended a “retrospective cohort pre-post baseline populationgroupbased savings analysis” where results are “compared with the4. pre-post historical control with actuarial adjustmenttrend of the non-diseased population … the most appropriate5. pre-post historical control with no actuarial adjustmentavailable comparison group.” In 2006, Linden[11] recommended6. pre-post multiple baseline (with randomized controls)that the condition-specific admission rate over an extended base-7. benchmark studyline be used as a reference in the evaluation of disease manage-
8. pre-post time seriesment interventions measuring condition-specific admissions. In
9. post-only control groupDecember 2007, the American Academy of Actuaries recommen-
10. post-only/no control groupded that “causality” was not the goal, but “correlation” or associa-
11. cross-sectional (no comparison) tion was “acceptable to most buyers,” and recommended an actua-12. simple time-series (e.g. run charts). rial method, with a strong stress on equivalence.[12] These reviews
In a 2007 book published by the DMAA,[3] multiple methods did not examine the level of standardization of evaluation today.for evaluating disease management programs were described, Other reviews of the disease management literature have beenbased upon the disciplines of health services research, epidemiolo- conducted, but these have primarily focused on the value ofgy, econometrics, and actuarial sciences. disease management, not the methods used to determine that
These articles highlighted the multiplicity of design options value.[13-20]
available to assess the impact of disease management. ManyThere is a need to review the current level of standardization
industry experts are concerned with this lack of standardizationamong disease management impact studies. The purpose of this
and are working to develop a standard method to enable diseasepaper is to assess the processes and methods used to assess
management programs to be compared side by side.[4] The purposeoutcomes in peer-reviewed, US-based disease management evalu-
of this article is to examine the consistency of disease managementation studies performed with the involvement of a health plan and
peer reviewed evaluations over a 3-year period. This is an essentialpublished between 2004 and 2007. The specific outcomes/effects
initial step in considering the variations present in the publishedof the disease management programs will not be examined.
literature. The goal is to develop a framework for assessing diseaseTo this end, we performed a search in MEDLINE using the keymanagement evaluations that is robust to the variety of disease
search criteria ‘insurance claims’ or ‘administrative claims’ ormanagement interventions and evaluation methods. Valid evalua-‘chronic illness care’ or ‘chronic care model’ or ‘population-basedtion findings should be used when comparing program effective-management’ or ‘disease-management’ or ‘disease management’ness and outcomes.or ‘care management’ and ‘return on investment’ or ‘ROI’ orThis push toward methods standardization was started in ear-‘economic’ or ‘financial’ or ‘savings’ amongst articles publishednest in 2003 when the disease management company Americanbetween 1 January 2004 and 31 December 2006. Abstracts wereHealthways (now known as Healthways) and Johns Hopkinsreviewed by three panelists (P. Salber, M. MacDowell, and T. Wil-University suggested that a “pre-post study was sufficient at thisson) and articles were excluded if they did not involve a healthtime” (the inclusion of an actuarial referent was recommended inplan, focused only on prescription-based disease management,the appendix).[5] This model was never widely adopted[6] and otherfeatured interventions that were managed by employers, or de-approaches have been recommended. The most recent attempt wasscribed studies that were not performed in the US. This resulted inby the DMAA itself, which, in December 2006, recommendedthe selection of 23 articles.[21-43] Two additional articles that met“use of pre-post study design that incorporates, whenever possible,the criteria above and were published between 1 January 2007 andan equivalent, concurrent comparison group” and suggested sever-28 February 2007, were added to the review at a later date.[44,45]al more detailed recommendations including “risk adjustment”Each article was reviewed by M. MacDowell, C. Hamm, andand the use of “non-chronic trend” as a reference.[7] This book was
updated with more detail and clarification in 2007[8] and 2008.[4] T. Wilson.
© 2008 Adis Data Information BV. All rights reserved. Dis Manage Health Outcomes 2008; 16 (5)
Evaluation Methods in Disease Management Studies 367
The goal was to describe the level of standardization and rather benchmarks (as recommended by RAND), articles, or non-variability of the study characteristics listed below. The consensus standard comparisons (e.g. the pre- intervention period in a pre-opinion regarding the evaluation design characteristics was post study) or articles, or non-standard comparisons (e.g. the pre-reached by three reviewers (M. MacDowell, C. Hamm, and period in a pre-post evaluation design).T. Wilson).
1.5 Confounding Factors1. Evaluation Design Characteristics
These are factors, which may or may not be measured, thatDescriptions of the following characteristics must be present in could be independently related to the outcome metric but are not
order to achieve transparency of methods and metrics. part of the intervention. Common confounding variables are ageand sex. These factors can confound a causal relationship. Here is
1.1 Intervention a typical example: if the age distribution in the interventionpopulation was 40% elderly, while in the concurrent referent theThis is what is done to the population. This is not an element ofage distribution was 60% elderly, the better outcomes in thethe evaluation design; it is rather the thing the study is evaluating.intervention population could be totally or partially a result of theyounger population in the intervention group, rather than the1.2 Intervention Populationintervention itself. Other potential confounding factors include
This is the group of people in which the disease management differences between the intervention population and the referent inprogram is implemented. benefit design, disease severity, the presence of co-morbid condi-
tions, the propensity of participants to take care of their health, and1.3 Outcomes Metrics
provider mix. Importantly, it is unlikely that all confoundingfactors can be taken into account in any study. Thus, positive (orThese are the targeted measures that the intervention is de-negative) results from disease management studies should neversigned to impact. Although the term includes the word ‘outcomes,’be thought of as ‘proof.’ Confounders are variables that are onlythese measures may or may not be caused by the intervention.useful when there is a referent other than the pre-interventionIndeed, a basic requirement of an evaluation is the use of these inperiod in a simple pre-post study in a single cohort of patients.the baseline period, to compare with the post period in the inter-
Each evaluation design characteristic was examined by threevention population only, to the pre-period in the external referentreviewers (M. MacDowell, C. Hamm, and T. Wilson).(if available).
2. Characteristics of the Reviewed Disease1.4 ReferentManagement Evaluations
This is what the intervention population is compared with and,ideally, is designed to estimate values for the outcome metrics in
2.1 Interventionsthe intervention population in the absence of the intervention.Broadly defined, referents include control groups, expert opinion,
The majority of the 25 articles that met our selection criteria forthe defined time period for the ‘pre-intervention’ period in a pre/
disease management interventions were ‘telephonic nurse man-post-intervention study (involving either the same individuals or a
agement,’ ‘monthly group meetings,’ ‘senior life management,’different group of people), historical control groups from a time
‘care advocacy,’ ‘immunization reminders,’ ‘home-based teach-period prior to that of the intervention population, and the concur-
ing,’ ‘interventions to enhance fitness,’ ‘interventions to enhancerent period to the intervention population.
care management,’ and ‘collaborative care.’ See table I for aThe referent can be chosen by an experimental method (e.g.
complete description of interventions.randomization) or a non-experimental method (e.g. people whoopt out, staggered roll-out, an employer group that did not partici- 2.2 Intervention Population Definition:pate in disease management, a matched control group, a previousperiod in time for a patient population). We have used the term There was a large number of different disease/condition states‘referent,’ rather than the more common ‘reference population’ as represented in the 25 articles. These included congestive heartthere are some referents in place that are not populations, but failure (CHF; n = 7), diabetes mellitus (n = 2), asthma (n = 3),
© 2008 Adis Data Information BV. All rights reserved. Dis Manage Health Outcomes 2008; 16 (5)
368 Wilson et al.
© 2008 Adis Data Information BV. All rights reserved. Dis Manage Health Outcomes 2008; 16 (5)
Tab
le I
. D
escr
iptio
n of
cite
d ar
ticle
s re
view
ed (
as d
escr
ibed
by
the
auth
ors
of t
he a
rtic
les)
Stu
dyD
esig
n/na
ture
Mai
n at
trib
utes
Inte
rven
tion
Out
com
e m
etric
sC
onfo
undi
ng m
etric
sof
RC
Tof
stu
dypo
pula
tion
Mar
tinR
CT
Med
icar
e,S
enio
r lif
e/ca
reS
F-6
, re
sour
ce u
se m
easu
red
by a
dmis
sion
rat
esA
ge,
sex
(all
othe
rs w
ere
outc
omes
, e.
g. S
F)
et a
l.[21]
(indi
vidu
al)
CH
F,
etc.
man
agem
ent
and
bed-
days
per
tho
usan
d pe
r ye
ar,
mem
ber
satis
fact
ion,
and
cos
ts m
easu
red
by p
aid
clai
ms
Tin
kelm
anN
on-R
CT
Ast
hma
Mul
tiple
Tot
al c
osts
Age
, se
xan
d W
ilson
[22]
inte
rven
tion
DM
Gal
brea
thR
CT
CH
FM
ultip
leA
ll-ca
use
mor
talit
y, E
F,
% o
n gu
idel
ine-
base
d R
x,A
ge,
sex,
rac
e, m
arita
l sta
tus,
fun
ctio
nal s
tatu
s,et
al.[2
3](in
divi
dual
)in
terv
entio
n D
Mco
sts,
sta
ysB
P,
med
ical
his
tory
, pu
lse,
pha
rmac
othe
rapy
DeB
usk
RC
TLo
w-r
isk
Nur
se c
ase
Tim
e to
reh
ospi
taliz
atio
n, r
easo
ns f
or r
e-A
ge,
sex,
rac
e, m
arita
l sta
tus,
edu
catio
n le
vel,
et a
l.[24]
(offi
ces)
CH
Fm
anag
emen
t,ho
spita
lizat
ion
HF
his
tory
(pr
evio
us d
iagn
osis
of
HF
, ba
selin
ete
leph
onic
NY
HA
cla
ss I
–II,
III–I
V),
cau
se o
f H
F (
coro
nary
dise
ase,
hyp
erte
nsio
n, v
alvu
lar
dise
ase,
unkn
own)
, sy
mpt
oms
befo
re h
ospi
taliz
atio
n(a
ngin
a, a
typi
cal c
hest
pai
n, e
xert
iona
l sho
rtne
ssof
bre
ath,
cou
gh,
paro
xysm
al n
octu
rnal
dys
pnea
,pe
riphe
ral e
dem
a, o
rtho
pnea
)
Sac
kett
Non
-RC
TP
rena
tal
Pre
nata
lLB
W,
cost
sN
one
et a
l.[25]
prog
ram
with
case
man
agem
ent
Vill
agra
Non
-RC
TD
iabe
tes
Mul
tiple
Cos
t (in
patie
nt,
outp
atie
nt,
prof
essi
onal
ser
vice
s,A
ge,
sex
and
Ahm
ed[2
6]in
terv
entio
n D
MR
x dr
ugs,
oth
er),
use
(da
ys,
ER
, vi
sits
,ad
mis
sion
s),
qual
lity
indi
cato
rs (
HbA
1c,
AC
E,
dila
ted
retin
al e
xam
, m
icro
albu
min
, lip
id,
toba
cco
use)
Ber
gR
CT
All
mem
bers
Dire
ct m
ail
Influ
enza
/pne
umon
ia a
dmis
sion
/ER
, in
fluen
zaA
ge,
sex,
hou
seho
ld s
ize,
mem
ber
mon
ths,
dua
let
al.[2
7](in
divi
dual
)m
arke
ting
ofva
ccin
atio
ns p
neum
onia
vac
cina
tions
, es
timat
edco
vera
ge,
cond
ition
pre
vale
nce
(eig
ht c
omm
onim
mun
izat
ion
cost
/sav
ings
(m
ultip
lyin
g th
e es
timat
ed a
bsol
ute
chro
nics
)di
ffere
nce
in u
tiliz
atio
n an
d ut
iliza
tion
orva
ccin
atio
n)
Ber
gN
on-R
CT
CH
FT
elep
honi
c D
MA
nnua
l cos
t, $U
S (
med
ical
, R
x).
Med
ical
ser
vice
Dem
ogra
phic
(m
onth
s pr
e, m
ean)
and
co-
et a
l.[28]
utili
zatio
n (a
nnua
lized
rat
e pe
r 10
00),
pre
scrip
tion
mor
bidi
ty (
% C
AD
, C
OP
D,
hype
rten
sion
, di
abet
esdr
ug u
se (
annu
al d
ays
supp
ly p
er p
erso
n),
mel
litus
, ar
thrit
is,
dem
entia
, de
pres
sion
)pr
escr
iptio
n dr
ug u
se b
asel
ine
(% w
ho h
adpr
escr
iptio
n)
Sco
ttR
CT
≥11
outp
atie
ntM
onth
ly g
roup
Per
pat
ient
clin
ic v
isits
, ph
arm
acy
fills
, ho
spita
lA
ge,
sex,
mar
ital s
tatu
s, %
with
a c
hron
icet
al.[2
9](in
divi
dual
)vi
sits
mee
tings
adm
issi
ons,
hos
pita
l obs
erva
tion
adm
issi
ons,
cond
ition
(14
list
ed),
no.
of
Rx,
mob
ility
lim
itatio
n,pa
tient
s an
dho
spita
l out
patie
nt v
isits
, pr
ofes
sion
al s
ervi
ces,
depr
essi
on,
% o
f m
emor
y pr
oble
ms,
live
alo
ne,
doct
ors
emer
genc
y vi
sits
, sk
illed
nur
sing
fac
ility
no b
asic
AD
L de
ficits
, no
adv
ance
d A
DL
defic
its,
adm
issi
ons,
hom
e he
alth
vis
its;
cost
(by
ser
vice
no h
ouse
hold
AD
L de
ficits
, fa
ir or
poo
r he
alth
line)
, Q
OL
(mul
tiple
)st
atus
Con
tinue
d ne
xt p
age
Evaluation Methods in Disease Management Studies 369
© 2008 Adis Data Information BV. All rights reserved. Dis Manage Health Outcomes 2008; 16 (5)
Tab
le I
. C
ontd
Stu
dyD
esig
n/na
ture
Mai
n at
trib
utes
Inte
rven
tion
Out
com
e m
etric
sC
onfo
undi
ng m
etric
sof
RC
Tof
stu
dypo
pula
tion
Ros
tR
CT
Dep
ress
ion
Enh
ance
d ca
reD
ays
free
of
depr
essi
on,
prog
ram
and
out
patie
ntN
one
liste
det
al.[3
0](g
roup
s)m
anag
emen
tco
sts,
pat
ient
tim
e an
d tr
ansp
orta
tion
cost
s,Q
ALY
s
Cat
ovN
on-R
CT
Ast
hma
Hom
e-ba
sed
Hos
pita
l adm
its,
ER
adm
itsS
ex,
race
, ag
e, r
egio
n of
cou
ntry
, ur
ban/
rura
let
al.[3
1]te
achi
ngs
Hud
son
Non
-RC
TC
HF
Rem
ote
ER
vis
its,
inpa
tient
adm
issi
ons,
re-
adm
issi
ons
Non
e (s
trat
ified
by
sex
and
age)
et a
l.[32]
mon
itorin
g
van
Von
noN
on-R
CT
CH
FM
ultip
leC
HF
-rel
ated
hos
pita
l sta
y or
out
patie
nt v
isit,
tot
alA
ge,
sex,
con
ditio
ns (
rheu
mat
ic,
hype
rten
sive
,et
al.[3
3]in
terv
entio
n D
Mco
st d
iffer
ence
s pr
e-po
st (
adju
sted
for
leng
th o
fis
chem
ic,
pulm
onar
y he
art
dise
ases
), d
rugs
prog
ram
par
ticip
atio
ns)
(dio
xin,
diu
retic
s, e
tc),
use
d nu
rsin
g ho
me,
use
dho
me
care
, no
. of
dia
gnos
is c
odes
Del
aron
deR
CT
Ast
hma
Mul
tiple
Cha
nges
in a
sthm
a m
edic
atio
n us
e, p
hysi
cian
Age
, se
xet
al.[3
4]in
terv
entio
n D
Mof
fice
visi
ts,
ER
vis
its,
hosp
italiz
atio
ns,
and
QO
L
Dun
agan
RC
TC
HF
Nur
se,
Mor
talit
y, r
e-ad
mis
sion
or
ED
vis
it, a
ny r
e-A
ge,
sex,
rac
e, m
arita
l sta
tus,
hel
p liv
ing
atet
al.[3
5](in
divi
dual
s)te
leph
onic
adm
issi
on,
prim
ary
HF
rea
dmis
sion
, an
yho
me,
hea
lth w
ith d
aily
livi
ng,
isch
emic
etio
logy
read
mis
ion
or d
eath
of H
F,
LVE
F,
NY
HA
cla
ssifi
catio
n, A
CE
reg
imen
,di
scha
rge
crea
tinin
e, C
harls
on s
core
, S
F-1
2,B
ack
Inve
ntor
y S
cale
, M
LHF
Q (
phys
ical
and
men
tal)
Haw
kins
Non
-RC
TS
peci
al n
eeds
Ris
kM
embe
r sa
tisfa
ctio
n, a
dmits
, LO
S,
ED
tot
al c
osts
Non
eet
al.[3
6]ch
ildre
nas
sess
men
tan
d ca
sem
anag
emen
t
Sch
wer
ner
Non
-RC
T14
chr
onic
,M
ultip
leP
MP
M,
tren
d14
chr
onic
, co
mpl
ex c
ondi
tions
et a
l.[37]
com
plex
inte
rven
tion
DM
cond
ition
s
Wis
eN
on-R
CT
Pop
ulat
ion
Med
ical
PM
PM
(al
l, IP
, O
P,
Rx,
Pro
f),
qual
ity (
HE
DIS
:A
ge,
sex,
hea
lth p
lan
bene
fits
(est
imat
ed),
et a
l.[38]
and
DM
HbA
1c t
estin
g an
d co
ntro
l, ey
e ex
am,
lipid
exa
m,
num
ber
of h
ealth
con
ditio
ns,
prev
alen
ce o
f he
alth
LDL
<13
0, n
ephr
opat
hy m
onito
ring)
cond
ition
s
Dal
yR
CT
Chr
onic
ally
Cas
eH
ospi
taliz
atio
n re
-adm
issi
on,
re-a
dmitt
ed w
ithin
Age
, A
PA
CH
E s
core
, G
lasc
ow,
pre-
hosp
ital
et a
l.[39]
(indi
vidu
al)
criti
cal i
ll >
3m
anag
emen
t48
hou
rs,
deat
h at
re-
adm
issi
on,
days
to
first
re-
adm
issi
on c
o-m
orbi
d sc
ore,
med
icat
ion
pre-
days
adm
issi
on,
tota
l re-
hosp
italiz
atio
n da
ys;
disc
harg
eho
spita
l sco
re,
leng
th o
f ho
spita
l sta
y, s
ex,
race
,m
echa
nica
ldi
spos
ition
(lo
ng-t
erm
acu
te c
are,
reh
abili
tatio
n,pr
imar
y IC
D-9
dia
gnos
isve
ntila
tion,
nurs
ing
hom
e, d
eath
), d
isch
arge
sta
tus
(with
post
-dis
char
geox
ygen
, tr
ache
otom
y, v
entil
ator
, co
gniti
veim
pairm
ent)
; S
F-8
sco
re p
hysi
cal f
unct
ioni
ng a
ndm
enta
l fun
ctio
ning
, pa
tient
AD
L/A
DL
scor
e
Con
tinue
d ne
xt p
age
370 Wilson et al.
© 2008 Adis Data Information BV. All rights reserved. Dis Manage Health Outcomes 2008; 16 (5)
Tab
le I
. C
ontd
Stu
dyD
esig
n/na
ture
Mai
n at
trib
utes
Inte
rven
tion
Out
com
e m
etric
sC
onfo
undi
ng m
etric
sof
RC
Tof
stu
dypo
pula
tion
Sid
orov
[40]
Non
-RC
TC
HF
Mul
tiple
Ass
essm
ent
of L
VH
, pr
escr
ibed
an
AC
E o
rN
one
inte
rven
tion
DM
cont
rain
dica
tion
(LV
EF
140
%),
pre
scrib
ed a
β-
adre
noce
ptor
ant
agon
ist;
inpa
tient
day
s, E
Rvi
sits
, to
tal c
laim
s, E
R c
laim
s; f
inan
cial
(out
patie
nt,
inpa
tient
, C
HF
-rel
ated
cla
ims)
Gry
pma
Non
-RC
TD
epre
ssio
nC
olla
bora
tive
Util
izat
ion
of s
ervi
ces,
ann
ualiz
ed c
osts
,A
ge,
sex
et a
l.[41]
Car
ede
pres
sion
sco
res
(bas
elin
e an
d 6-
mon
th a
ndM
anag
emen
t:pe
rcen
tage
exp
erie
ncin
g 50
% im
prov
emen
t in
IMP
AC
Tde
pres
sion
) an
tidep
ress
ant
use,
tre
atm
ent
cont
acts
Sha
nnon
RC
TH
igh
use,
>65
CA
pro
gram
:U
tiliz
atio
n in
ser
vice
s (p
rimar
y ca
re p
hysi
cian
,A
ge,
sex,
mea
sure
s of
hea
lth s
tatu
s (n
ot s
how
n)et
al.[4
2](in
divi
dual
)ye
ars
refe
rral
s to
spec
ialis
t, ho
spita
l adm
issi
ons,
hos
pita
l day
s,ho
me-
and
ER
)co
mm
unity
-ba
sed
serv
ices
Sw
eene
yN
on-R
CT
Life
-lim
iting
Inte
nsiv
eH
ome
care
day
s, h
ospi
ce d
ays,
reh
abili
tatio
nA
ge,
sex,
bro
ad I
CD
-9 c
ateg
orie
s (t
op 6
of
16:
et a
l.[44]
diag
nosi
spa
tient
-cen
tere
dda
ys,
ER
vis
its,
hosp
ital d
ays,
hos
pita
lne
opla
sms,
circ
ulat
ory,
dig
estiv
e sy
stem
, ne
rvou
sm
anag
emen
tad
mis
sion
s; in
patie
nt a
dmis
sion
s (n
ause
a an
dan
d se
nse,
ill-d
efin
ed,
inju
ry,
and
pois
onin
g)vo
miti
ng,
anem
ia,
fluid
dis
orde
r an
d de
hydr
atio
n,fe
ver,
nut
ritio
nal d
efic
ienc
y, m
etab
olis
m,
anxi
ety)
,m
orta
lity/
surv
ival
Ngu
yen
Non
-RC
TD
iabe
tics
Enh
ance
dH
ospi
tal a
dmis
sion
s (a
ll, d
iabe
tes
rela
ted)
,A
ge,
sex,
HbA
1c le
vels
, ch
roni
c di
seas
e bu
rden
,an
dfit
ness
pro
gram
annu
al p
rimar
y ca
re v
isits
(al
l, di
abet
es r
elat
ed),
prev
entio
n sc
ore
(pre
vent
ion
scor
e is
the
sum
of
Ack
erm
an[4
5]he
alth
care
cos
ts (
tota
l, pr
imar
y ca
re,
inpa
tient
the
num
ber
of t
imes
a s
ubje
ct r
ecei
ved
colo
nco
sts)
(no
te:
stra
tifyi
ng v
aria
bles
: le
vel o
fca
ncer
scr
eeni
ng,
[feca
l occ
ult
bloo
d te
st o
rpa
rtic
ipat
ion
in e
xerc
ise
prog
ram
). N
ote:
ave
rage
flexi
ble
sigm
oido
scop
y],
a sc
reen
ing
inpa
tient
cos
ts f
or m
embe
rs w
ho h
ad in
patie
ntm
amm
ogra
m,
pros
tate
can
cer
scre
enin
g, a
nco
sts
influ
enza
vac
cine
, or
a p
neum
ococ
cal v
acci
nedu
ring
the
2 ye
ars
imm
edia
tely
pre
cedi
ng t
hein
dex
date
)
Fire
ston
eN
on-R
CT
CA
D,
CH
F,
Mul
tiple
Qua
lity
mea
sure
s (L
DL
test
, st
atin
Rx,
Dis
ease
pre
vale
nce
et a
l.[43]
asth
ma,
inte
rven
tion
DM
antih
yper
tens
ive
Rx)
clin
ic v
isits
(ph
ysic
ian,
ER
,di
abet
esad
mis
sion
s, d
ays,
Rx
cost
s)
AD
L =
act
iviti
es o
f da
ily li
ving
; A
PA
CH
E =
Acu
te P
hysi
olog
y an
d C
hron
ic H
ealth
Eva
luat
ion;
BP
= b
lood
pre
ssur
e; C
A =
car
e ad
voca
te;
CA
D =
cor
onar
y he
art
dise
ase;
CH
F =
cong
estiv
e he
art f
ailu
re; C
OP
D =
chr
onic
obs
truc
tive
pulm
onar
y di
seas
e; D
M =
dis
ease
man
agem
ent;
ED
= e
mer
genc
y de
part
men
t; E
F =
eje
ctio
n fr
actio
n; E
R =
em
erge
ncy
room
;H
bA
1c =
one
type
of g
lyco
syla
ted
hem
oglo
bin;
HE
DIS
= H
ealth
Pla
n E
mpl
oyer
Dat
a an
d In
form
atio
n S
et; H
F =
hea
rt fa
ilure
; IC
D-9
= In
tern
atio
nal C
lass
ifica
tion
of D
isea
ses,
ver
sion
9; IM
PA
CT
= I
mpr
ovin
g M
ood-
Pro
mot
ing
Acc
ess
to C
olla
bora
tive
Tre
atm
ent;
IP =
inpa
tient
; L
BW
= lo
w b
irth
wei
ght;
LD
L =
low
-den
sity
lipo
prot
ein;
LO
S =
leng
th o
f st
ay;
LV
EF
=le
ft ve
ntric
ular
eje
ctio
n fr
actio
n; L
VH
= l
eft
vent
ricul
ar h
yper
trop
hy;
ML
HF
Q =
Min
neso
ta L
ivin
g w
ith H
eart
Fai
lure
que
stio
nnai
re;
NY
HA
= N
ew Y
ork
Hea
rt A
ssoc
iatio
n; O
P =
outp
atie
nt; P
MP
M =
per
mem
ber
per
mon
th; P
rof =
pro
fess
iona
l; Q
AL
Y =
qua
lity-
adju
sted
life
-yea
r; Q
OL
= q
ualit
y of
life
; RC
T =
ran
dom
ized
con
trol
led
tria
l; R
x =
pre
scrip
tion;
SF
-6=
Out
com
es S
tudy
36-
Item
Sho
rt-F
orm
Hea
lth S
urve
y.
Evaluation Methods in Disease Management Studies 371
chronic diseases (coronary artery disease, CHF, diabetes, and for a complete description of outcome metrics employed in theasthma; n = 1), total health plan population (without regard to studies.diagnosis; n = 3), depression (n = 2), high utilizers (n = 2), severechronic disease (n = 2), life-limiting diagnoses (n = 1), discharged 2.5 Confounding Factorswith ventilator (n = 1), prenatal (n = 1), and special needs chil-dren (n = 1). Five of the articles listed no confounding variables. Four stud-
The criteria used to define the intervention population were not ies adjusted results for differences in the age and sex of compara-consistent among evaluations of programs for a single disease. For tor groups (all were non-randomized); the other 16 adjusted resultsexample, among studies of CHF interventions, some defined the for differences in other variables, as well as age and sex.study population using the diagnostic codes (i.e. International Among the studies with non-randomized designs, the con-Classification of Diseases, version 9 [ICD-9]) from administrative founding variables that were adjusted for included the number ordata, with or without health plan membership requirements or type of co-morbidities, HEDIS metrics (that did not require chartutilization history (regarding inpatient stays, emergency room review – as is done in the HEDIS hybrid method, which utilizesvisits, etc.). In addition, some used other factors (e.g. the ability to both administrative claims data and chart review to derive thecommunicate with the patient’s primary care provider) as a secon- numerator and denominator of metrics), and prevalence of specificdary criteria for selection. The randomized trials generally used national drug codes (NDC), number of prescriptions, coverageclinical parameters (e.g. left ventricular ejection fraction) to define type, benefit design, and the length of time in the plan/program.the intervention population. There was no commonly applied Other confounding variables that were not assessed using adminis-standard for selecting populations. trative data included household size, marital status, educational
level, the presence/absence of left ventricular hypertrophy, labora-2.3 Referent Types tory parameter values, left ventricular ejection fraction, and New
York Heart Association class. See table I for a complete descrip-Ten of the 25 studies used randomization to select the referent
tion of confounding factors in the studies reviewed.population; randomization was mostly at the individual level(n = 7),[21,23,27,29,34,35,39,42] while two studies randomized physician
3. Conclusions and Recommendationsoffices, rather than individuals.[24,30] One study[34] employedrandomization for one sub-study and analyzed a combined popula-
Clearly, we do not see any trend towards the standardization oftion that included both the intervention group and referents for the
intervention populations, outcome metrics, or even a commonother sub-studies.
definition of ‘disease management.’ within current evaluations ofThe other 15 studies employed non-randomized designs.
disease management programs. Thus, comparison of outcomesTwo[32,36] used the pre-intervention period as the primary referent
between different interventions would appear to be very difficult,for the intervention period, three[25,41,43] used an historical referent
if not impossible.(two of these used more than one design), ten used a concurrent
While we await a standard evaluation design, the best choicereferent (three of these used more than one de-
for value-based purchasing will be to use a basic list of criteriasign).[22,26,28,31,33,37,38,40,44,45] None used expert opinion; however,
when assessing the methods used in a disease management evalua-one study did cite an expert to adjust for benefit design differences
tion. Such a list – if standardized – would at least enable the studybetween referent and intervention populations.
methods to be compared as is done in ‘level of evidence’ scoring inthe evidenced-based medicine area.[46-50] A standardized evalua-2.4 Outcome Metricstion is essential for two reasons: (i) to assess level of confidence inthe evaluation design (and to compare this across programs); andOutcome metrics included mortality, admission, re-admission,(ii) to assess level of confidence in the results.time to re-admission, and cost (total, admission, out-patient, pre-
scription). These were often related to specific diagnoses. Other Disease management evaluation studies – published or not –metrics included days free of depression, the percentage of low may include any number of technical and non-technical elements,birth-weight babies among all births, quality-of-life survey out- but the foundational criteria needed to assess the validity of thesecomes, patient satisfaction survey outcomes, and Healthcare Ef- studies is transparency of methods and metrics. Based upon thefectiveness Data and Information Set (HEDIS) metrics. See table I simple criteria we used to assess the methods used in a representa-
© 2008 Adis Data Information BV. All rights reserved. Dis Manage Health Outcomes 2008; 16 (5)
372 Wilson et al.
Disease Management Outcomes Summit; 2002 Nov 7–10; Palm Springs (CA).tive list of 25 articles, five simple questions that would provideNashville (TN): American Healthways Inc., 2003
increased transparency are listed below: 6. Glabman M. ‘Take my word for it’: the enduring dispute over measuring DM’seconomic value. Managed Care April 2006 [online]. Available from URL:1. Is the intervention used in the disease management programhttp://managedcaremag.com/archives/0604/0604.dmvalue.html [Accessedclearly stated?2008 Nov 27]
2. Is the population in which the intervention is implemented 7. Disease Management Association of America. Outcomes guidelines report. Vol I.Washington, DC: DMAA, 2006clearly defined?
8. Disease Management Association of America. Outcomes guidelines report. Vol II.3. Is the referent used in the disease management evaluationWashington, DC: DMAA, 2007
clearly defined? 9. Mattke S, Bergamo G, Balakrishnan A, et al. Measuring and reporting the perform-ance of disease management programs. Santa Monica (CA): RAND, 2006 Aug4. Are the outcome metrics used to compare the intervention and
10. Sexner S, Baker K, Gold D. Guidelines for analysis of economic returns fromreferent group clearly defined and consistently assessed?health management programs: the art of health promotion. Art Health Prom
5. Are key confounding factors identified, measured, and proper- 2006 Jul/Aug: 1-17
11. Linden A. What will it take for DM to demonstrate a return on investment? Newly taken into account when making a comparison between theperspectives on an old theme. Am J Manag Care 2006; 12: 217-22
outcome metrics obtained for the intervention and referent groups?12. American Academy of Actuaries’ Disease Management Work Group. Disease
Other groups have suggested checklists for disease manage- management: a public policy practice note. Washington, DC: American Acade-my of Actuaries, 2007ment, and we encourage the reader to review these more compre-
13. Weingarten SR, Henning JM, Badamgarav E, et al. Interventions used in diseasehensive lists. They range from the general to the specific: general management programmes for patients with chronic illness: which ones work?
Meta-analysis of published reports. BMJ 2002; 325: 1-8checklists have been offered by Mathematica,[51] the National14. Ofman JJ, Badamgarav E, Henning JM, et al. Does disease management improveManaged Health Care Congress (now The Health/Care Evaluation
clinical and economic outcomes in patients with chronic diseases? A systematicworkgroup),[52] the DMAA,[53] and the American Heart Associa- review. Am J Med 2004; 117: 182-92
tion.[54,55] Specific checklists from health services research[56] and 15. Holtz-Eakin D. An analysis of the literature on disease management programs.Congressional Budget Office, 2004 Oct [online]. Available from URL: http://actuarial sciences[57] are also available. However, using these morewww.cbo.gov/ftpdocs/59xx/doc5909/10-13-DiseaseMngmnt.pdf [Accessed
comprehensive lists are not useful unless some detail is presented 2008 Oct 28]
16. Krause D. Economic effectiveness of disease management: a metaanalysis. Dis– made transparent – regarding the five issues we discuss.Manage 2005; 8: 114-34An article is in preparation that will examine the quality and
17. Goetzel RZ, Ozminkowski RJ, Villagra VG, et al. Return on investment in diseasestrength of these 25 studies using more sophisticated quantitative management: a review. Health Care Financ Rev 2005; 26: 1-19
scoring criteria. This article will take the reader beyond the basics 18. Wilson TW, MacDowell M, Montrose G. The prerequisite importance of judgingthe potential usefulness of workplace health promotion studies: beyond “Pub-listed here to enable a more detailed look at the very importantlished in a Peer-Reviewed Journal” and “Strength of Study Design”. J Health
issues related to equivalence and statistical significance. Productivity 2006; 1: 23-8
19. Pelletier KR. A review and analysis of the clinical and cost-effectiveness studies ofcomprehensive health promotion and disease management programs at the
Acknowledgments worksite: update VI 2000–2004. J Occup Environ Med 2005 Oct; 47 (10):1051-8
20. Mattke S, Seid M, Sai M. Evidence for the effect of disease management: is $1This article was funded in part by an unrestricted educational grant frombillion a year a good investment? Am J Manag Care 2007; 13: 670-6the Health Industry Forum at Brandeis University provided to the Population
21. Martin DC, Berger ML, Anstatt DT, et al. A randomized controlled open trial ofHealth Impact Institute.population-based disease and case management in a Medicare Plus Choicehealth maintenance organization. Prev Chronic Dis 2004; 1 (4): A05 [online].Available from URL: http://www.pubmedcentral.nih.gov/articlerender.fcgi?arReferences tid=1277945 [Accessed 2008 Nov 24]
1. MacDowell M, Wilson TW. Framework for assessing causality in disease manage-22. Tinkelman D, Wilson S. Asthma disease management: regression to the mean or
ment programs [commissioned paper by Disease Management Association ofbetter? Am J Manag Care 2004 Dec; 10 (12): 948-54
America]. Washington, DC: DMAA, 200323. Galbreath AD, Krasuski RA, Smith B, et al. Long-term healthcare and cost2. Outcomes Consolidation Steering Committee. Disease management program
outcomes of disease management in a large, randomized, community-basedguide: including principles for assessing disease management programs. Wash-population with heart failure. Circulation 2004; 110 (23): 3518-26ington, DC: DMAA, 2005: 33
24. DeBusk RF, Miller NH, Parker KM, et al. Care management for low-risk patients3. Lindner A, Wilson T, Berg G, et al. Perspectives in disease management evalua-with heart failure: a randomized, controlled trial. Ann Intern Med 2004; 141 (8):tion: health services research, epidemiology, econometrics, and actuarial. In:606-13Hay J, Rindress D, editors. Measuring disease management. Yardley (PA):
25. Sackett K, Pope RK, Erdley WS. Demonstrating a positive return on investment forDMAA, 2007a prenatal program at a managed care organization: an economic analysis.4. Disease Management Association of America. Outcomes guidelines report. Vol. 3.J Perinat Neonatal Nurs 2004 Apr-Jun; 18 (2): 117-27Washington, DC: DMAA, 2008
5. American Healthways and Johns Hopkins University. Standard outcomes metrics 26. Villagra VG, Ahmed T. Effectiveness of a disease management program forand evaluation methodology of disease management programs. 2nd Annual patients with diabetes. Health Aff 2004; 23 (4): 255-66
© 2008 Adis Data Information BV. All rights reserved. Dis Manage Health Outcomes 2008; 16 (5)
Evaluation Methods in Disease Management Studies 373
27. Berg GD, Thomas E, Silverstein S, et al. Reducing medical service utilization by 43. Firestone B, Bartlett J, Selby J. Can disease management reduce health care costsencouraging vaccines: randomized controlled trial. Am J Prev Med 2004 Nov; by improving quality? Health Aff 2004; 23 (6): 63-7527 (4): 284-8
44. Sweeney L, Halpert A, Waranoff J. Patient-centered management of complex28. Berg GD, Wadhwa S, Johnson AE. A matched-cohort study of health services patients can reduce costs without shortening life. Am J Manag Care 2007; 13:
utilization and financial outcomes for a heart failure disease-management 84-92program in elderly patients. J Am Geriatr Soc 2004 Oct; 52 (10): 1655-61
45. Nguyen HQ, Ackerman R. Impact of a managed-Medicare physical activity benefit29. Scott JC, Conner DA, Venohr I, et al. Effectiveness of a group outpatient visit on health care utilization and costs in older adults with diabetes. Diabetes Care
model for chronically ill older health maintenance organization members: aJan 2007; 30 (1): 43-8
2-year randomized trial of the cooperative health care clinic. J Am Geriatr Soc46. Centre for Evidence-Based Medicine. Levels of evidence [online]. Available from2004 Sep; 52 (9): 1463-70
URL: http://www.cebm.net/index.aspx?o=1025 [Accessed 2008 Nov 27]30. Rost K, Pyne JM, Dickinson LM, et al. Cost-effectiveness of enhancing primary
47. Agency for Healthcare Research and Quality. US Preventive Services Task Forcecare depression management on an ongoing basis. Ann Fam Med 2005 Jan-Feb;3 (1): 7-14 [online]. Available from URL: http://www.ahrq.gov/clinic/uspstf/usp
stopics.htm [Accessed 2008 Nov 27]31. Catov JM, Marsh GM, Youk AO, et al. Asthma home teaching: two evaluationapproaches. Dis Manag 2005; 8 (3): 178-87 48. Consolidated Standards of Reporting Trials (CONSORT) [online]. Available from
URL: http://www.consort-statement.org/?o=1011 [Accessed 2008 Nov 27]32. Hudson LR, Hamar GB, Orr P, et al. Remote physiological monitoring: clinical,financial, and behavioral outcomes in a heart failure population. Dis Manag 49. The Cochrane Collaboration [online]. Available from URL: http://www.cochrane.2005; 8 (6): 372-81 org/ [Accessed 2008 Nov 27]
33. vanVonno CJ, Ozminkowski RJ, Smith MW, et al. What can a pilot congestive 50. Des Jarlais DC, Lyles C, Crepaz N, the TREND Group. Improving the reportingheart failure disease management program tell us about likely return on invest- quality of nonrandomized evaluations of behavioral and public health interven-ment? A case study from a program offered to federal employees. Dis Manag
tions: the TREND statement. Am J Public Health 2004; 94 (3): 361-62005 Dec; 8 (6): 346-60
51. Chen A, Brown R, Archibald N, et al. Best practices in coordinated care.34. Delaronde S, Peruccio DL, Bauer BJ. Improving asthma treatment in a managed
Mathematica 2000 [online]. Available from URL: http://www.mathematica-care population. Am J Manag Care 2005 Jun; 11 (6): 361-8
mpr.com/publications/PDFs/bestpractices.pdf [Accessed 2008 Nov 27]35. Dunagan WC, Littenberg B, Ewald GA, et al. Randomized trial of a nurse-
52. Wilson TW, Gruen J, Thar W, et al. Assessing return on investment of defined-administered, telephone-based disease management program for patients with
population disease management interventions. Jt Comm J Qual Saf 2004; 30heart failure. J Card Fail 2005 Jun; 11 (5): 358-65(11): 614-21
36. Hawkins MR, Diehl-Svrjcek B, Dunbar LJ. Caring for children with special53. Fitzner K, Siderov J, Fetterolf D, et al. Principles for assessing disease managementhealthcare needs in the managed care environment. Lippincotts Case Manag
outcomes. Dis Manage 2004; 7: 191-2012006 Jul-Aug; 11 (4): 216-23
54. Faxon DP, Schwamm LH, Pasternak RC, et al. American Heart Association’s37. Schwerner H, Mellody T, Goldstein AB, et al. Evaluating the impact of a diseaseExpert Panel on Disease Management. Improving quality of care throughmanagement program for chronic complex conditions at two large northeast
health plans using a control group methodology. Dis Manag 2006 Feb; 9 (1): disease management: principles and recommendations from the American34-44 Heart Association’s Expert Panel on Disease Management. Circulation 2004
Jun 1; 109 (21): 2651-438. Wise CG, Bahl V, Mitchell R, et al. Population-based medical and diseasemanagement: an evaluation of cost and quality. Dis Manag 2006 Feb; 9 (1): 55. Krumholz HM, Brindis RG, Brush JE, et al. Standards for statistical models used45-55
for public reporting of health outcomes. Circulation 2006; 113: 456-6239. Daly BJ, Douglas SL, Kelley CG, et al. Trial of a disease management program to 56. Linden A, Roberts N. A user’s guide to the disease management literature:
reduce hospital readmissions of the chronically critically ill. Chest 2005 Aug;recommendations for reporting and assessing program outcomes. Am J Manage
128 (2): 507-17Care 2005; 11 (2): 81-90
40. Sidorov J. Reduced health care costs associated with disease management for57. Duncan I, Owen R, Dove H. Testing actuarial methods for evaluating disease
chronic heart failure: a study using three methods to examine the financialmanagement savings outcomes. Society of Actuaries, 2006 [online]. Availableimpact of a heart failure disease management program among medicare advan-from URL: http://www.soa.org/research/files/pdf/Paper_8_TestingActuaritage enrollees. J Card Fail 2006 Oct; 12 (8): 594-600al%20 Methods.pdf [Accessed 2008 Nov 27]
41. Grypma L, Haverkamp R, Little S, et al. Taking an evidence-based model ofdepression care from research to practice: making lemonade out of depression.Gen Hosp Psychiatry 2006 Mar-Apr; 28 (2): 101-7
Correspondence: Dr Thomas Wilson, Population Health Impact Institute,42. Shannon GR, Wilber KH, Allen D. Reductions in costly healthcare service utiliza-
10663 Loveland-Madeira Rd. #210, Loveland, OH 45140, USA.tion: findings from the Care Advocate Program. J Am Geriatr Soc 2006 Jul; 54(7): 1102-7 E-mail: [email protected]
© 2008 Adis Data Information BV. All rights reserved. Dis Manage Health Outcomes 2008; 16 (5)