patterns of travel for rural individuals hospitalized in new york state: relationships between...
TRANSCRIPT
Hogan 29
Patterns of Travel for Rural Individuals Hospitalized in New York State Relationships Between Distance, Destination, and Case Mix.*
Cbristopber Hogan
AlKlRAm: 7he travel patterns of individuals living in rural a m of New Yo& State who were dbcharged from short-term general hoqilals in New Yo& State in 1983 are examined. Counties are used as the g e graphical unit, and rural individuals who c m geographic boundaries to obtain inpatient hospital care are compared with those who receive such care in their own geographic area. Hospitals seruing the rural population of New Yo& are c h i ? into three types: urban, conskting of all h q ' - tals located in M W ; rural refmal centen; and other rural hapituls. Nixt, the rural patients who are admitted to each of these three types of hoqi.- tals are charactmzed in term of dkitance tmveled, case mix, length of stay, and age. Individuals who travel b t y n d the counties adjacent to their county of residence had a hgher case mix index but m les likely to be more than 75 years o[d. mtance traveled and the sqbected cost of care were strongly positively reluted for patients admitted to urban and rural refmal center hwlals, but were only weakly related for other rural ho@als. Find&, comparisons of rural patients in these three types of hospials were Momzed adjusting for DRG mix, a comparison which b relevant to h q ' t a l reimbursements under the Medicare Bqtmztive Pay- ment System. Using sexral measures of illness sewn@, rural patients in urban hoqtlals and rural refma1 center ho.@tals were more severely ill than rural patients in other rural hospitals aJer adjusting for DRG mix. We conclude that somewhat higherpayments to urban hospitals and ru- ral refmal center hoqitals in New Yo& are jus@kd based on the more sewmly illpatients which they treat.
The need to travel for hospital care is a particularly important problem for rural Americans. While residents of urban areas typically mvel under five miles for hospitalization (Earickson, 1970; Shannon, Skinner and Bash- shur, 1973) those residing in rural areas may be 30 minutes travel or more from the nearest hospital (Bosanac, Parkinson and Hall, 1976). Distance, therefore, provides an additional burden for rural individuals requiring hos- pitalization that is not borne by those living in urban areas. Conversely, geographic isolation also affeas hospitals in r u d areas, which offer a broader
%pedal t h a d ~ to Jay~e K W , PhD. and to for c ~ m - All m r s and omisskms mmain the Ip-
sponsibility of the author. The views expressed in this paper are those of the au- thor and 110 of3dal endorsement by the National Center for Health serrvlces Re- search and Health Cam Technology Assessment or the US. Department of Health
tkm should be fomaded to: ChrMopher Hogan, PhD., Hospital Studies plgpiun, National Center for Health serrvlces Research and Health Care Technology Assess-
men@ which g l w l y lmpmved this paper.
and Human services i~ impued~l should be Inkrred -t~ khd~~--
mt, parklawn Ihddhg, R00m lSA-55, 5600 Flsbeft Lane, ROCkdk, MD. 2W57, THE XXlRNALOF RURAL HEALTH
JULY, 1988 VOLUME 4, NUMBER 2
30 7be Journal of Rum1 Health
range of facilities and services than urban hospitals of similar size, but, due to their small size and geographically h t e d market, produce fewer hghly specialized services and pedorm fewer complex procedures than uhan hospitals taken as a whole (Moscovice and Rosenblatt, 1982).
In this paper, the flow of individuals living in rural New York to hospitals in that state is described. In so doing, the following questions are addressed: First, to what extent do individuals in rural New York travel for hospital care? Second, how do the individuals who travel for hospital care differ from those who receive care locally? Third, what sort of hospitals do they choose as their destination? Finally, do the patterns of patient travel provide any information regarding appropriate levels of reimbursement to hospitals?
Previous research on the travel patterns of hospitalized individuals has produced several quite sensible findings. Individuals by and large receive care at hospitals near their residence (McGuirk and Porell, 1%; Studnicki, 1976), but mean travel times to the hospital differ across diagnoses (Mayer, 1983), and more severely ill patients will on average travel fatther to re- ceive care (Cageorge, Roos and Danzinger, 1981; Folland, 1983). Large ur- ban teaching centers tend to draw patients from a larger geographic area than community hospitals (Garnick, Luft, Robinson and Treteault, 1987). Much of this literature, however, has been quite limited in the geographic area studied, or has concentrated on urban travel patterns, or has studied a specific diagnosis or small p u p of diagnoses, or has simply used distance traveled without attempting to differentiate among patient destinations. In an attempt to add to this body of fmdings, hs study concentrates on all rural individuals within a fairly large state, uses the spectrum of diagnoses which are treated in hospitals, and differentiates among types of hospitals to which the rural patient might travel.
In practical terms, this research is a straightforward description of rural patient flows within New York State. In keeping with the descriptive nature of the analysis, no tests of statistical sigrufcance are presented. The county is the geographic unit used, and county boundaty crossings are the meas- ure of travel distance. While hs is admittedly a coarse measure, there is evidence that it works reasonably well for some hospital-level analyses (Garnick, Luft, Robinson and Treteault, 1937). The patients are character- ized using Diagnosis Related Groups (DRGs) and Disease Staging, as well as using age, length of stay, and the number of diagnoses and procedures recorded on the discharge abstract.
Travel Pattems of Rural In- In this section the flow of individuals living in rural arras of New York
to short-term general hospitals in that state is examined. The patient data come from hospital discharge abstracts pmvided by the Statewide Planning and Research Cooperative System (SPARCS Bureau), New York State De- m e n t of Health, and are the universe of discharges from short-term
Hogan 31
Table 1. Cornpadsons of Metropolitan and Nonmetropolitan Areas in New York State on Selected Characteristics.
MSA Non-MSA
Population (1980, millions) 15.8 1.7
Short-term general hospitals (1983)’ 210 60
Beds per thousand population (1983)” 4.52 4.47
Beds per hospital (1983)”’ 342 133
84% Occupancy rate (1983)”’ 87%
Number of hospitals in this data set. ** Calculated from American Hospital Association (1984) and U.S. Department of
Commerce (1982). *** American Hospital Association (1984).
general hospitals in New York State for the year 1983. Since we have data only on individuals hospitalized in New York, we cannot include in the analysis individuals living in New York State but hospitalized elsewhere. Counties are used as the geographic unit throughout the paper, and we use “urban” to refer to counties located inside Metropolitan Statistical Areas, while “rural” refers to counties outside MSAs (see Note 1).
Table 1 offers some characteristics of New York State. About 10 per- cent of the population of New York lives in rural counties, and mral and urban areas of New York have roughly the same number of staffed haspi- tal beds per capita. As with most states, New York rural hospitals are considerably smaller than their urban counterparts, but rural New York hospitals are about 50 percent larger than the national rural average (Ameri- can Hospital Association, 1984). New York also differs from other states in that the occupancy rate of rural hospitals is not much lower than the occupancy rate for urban hospitals, 84 percent versus 87 percent, giving New York the highest rural hospital occupancy rate in the nation (Ameri- can Hospital Association, 1984).
The analysis was begun by establishing three types of hospitals from which rural residents might be discharged: uhan hospitals, rural e f d center hospitals, and other rural hospitals. Urban and rural hospitals are those hospitals located in MSA and non-MSA counties, mpectively. Rural referral centers are those hospitals which have qualified for M e d i a rural referral center status under the cumnt (1987) criteria used by the Health Care Financing Adrmnistration (HCFA). There are 60 rural hospitals in this data set, four of which are rural referral centers.
32 7he Journal of Rural Health
Table 2. Characteristics of Rural New York Residents Discharged from Short-Term General Hospitals in New York, by Type of Hospital (1983).
Destination Hospitals
Rural Other Urban Referral Rural
Hospitals Centers Hospitals (n-210.) (11-41 (n-56)
Discharges 57,746 26,612 214,193
Mean acute care length of stay (days) 7.79 7.21 6.43
Percent of patients over age 75 7.8% 14.7% 16.5%
DRG case mix index" .97 .85 .80
Disease Staging case mix index*'* 1.07 1.02 .98
Most urban hospitals discharged very few rural patients. Only 71 urban hospitals had more than 100 rural patients, and only 36 had more than 500 rural patients.
Administration based on 1984 data.
representative discharge abstract database (Hogan, 1987).
** Calculated using the DRG relative weights published by the Health Care Financing
*** Calculated using Staged Disease ConditiodStage weights derived from a nationally
Table 2 contains data on the number of discharges, acute care length of stay (that is, length of stay before subacute care or discharge), fraction of patients over age 75, and two case mix indices for rural residents receiv- ing care in these three types of hospitals. A substantial portion (19%) of discharges of rural New York residents took place from urban hospitals, and a further 9 percent received care in rural referral centers. Rural patients in urban hospitals had the longest stays while rural patients in rural non- referral hospitals had the shortest stays. However, urban hospitals &d not attract the oldest patients: the fraction of rural patients over age 75 was half as large in uhan hospitals as it was in rural referral centers and other rural hospitals.
Two case mix indices were calculated for comparisons of expeaed costs of treatment, one based on DRGs, the other based on Disease Stag- ing. The indices differ in how they are affected by treatment patterns: the DRG patient categorization depends heavily on pmedures performed in the hospital, while Disease Staging largely ignores procedure information in categorizing patients (Goldfah and Coffey, 1987), The Disease Staging in- dex, therefore, will be less sensitive to variations in the choice of treatment mode (e.g., surgical versus medical), while the DRG index will reflect treat- ment choices and will be more strongly correlated with actual resource use
Tab
le 3
. Cha
ract
eris
tics
of
Rur
al N
ew York
Res
iden
ts D
isch
arge
d fr
om S
hort
-Ter
m G
ener
al H
ospi
tals
3 3
in New Y
ork,
by
Typ
e of
Hos
pita
l and T
rave
l Dis
tanc
e (1
983)
.
Des
tinat
ion H
ospi
tals
Travel
Dis
tanc
e
Rur
al
Oth
er
Urb
an
Ref
erra
l R
ural
H
ospi
tals
Centers
Hos
pita
ls
(n=
210*)
(n==
4)
(n-5
6)
Dis
char
ges
Mea
n ac
ute
care
leng
th o
f st
ay (
days
)
Perc
ent
of p
atie
nts
over
age
75
DRG
cas
e m
ix i
ndex
"
Sam
e co
unty
-
21,3
34 (8
Wh)
19
0,37
1 (8Y
h)
Adj
acen
t cou
nty
40,7
82 (7
1%)
4,56
8 (1
7%)
22,8
88 (1
1%)
Fart
her
16,9
64 (2
9%)
710
( 3%
) 93
4 (<
lo/)
Sam
e co
unty
A
djac
ent c
ount
y Fa
rthe
r
Sam
e co
unty
A
djac
ent c
ount
y Fa
rthe
r
Sam
e co
unty
A
djac
ent c
ount
y Fa
rthe
r
- 7.
21
6.96
6.
79
9.77
9.
84
15.3
%
8.;%
12
.3%
6.
3%
12.0
9'0
6.47
6.
02
7.81
17.1
9'0
12.2
%
8.2%
- .8
2 .8
0 .88
.91
.77
1.18
1.
20
.88
Dis
ease
Sta
ging
cas
e m
ix i
ndex
.**
Sam
e co
unty
-
1.00
.7
5 A
djac
ent c
ount
y .9
9 1.
05
.99
Fart
her
1.25
1.
31
.93
Mos
t urb
an h
ospi
tals
dis
char
ged
very
few
rura
l pa
tient
s. O
nly
71 u
rban
hos
pita
ls h
ad m
ore
than
100
rura
l pat
ient
s, a
nd o
nly 36 h
ad m
ore
than
500
rur
al
patie
nts.
**
Cal
cula
ted
usin
g th
e D
RG r
elat
ive
wei
ghts
pub
lishe
d by
the
Hea
lth C
are
Fina
ncin
g A
dmin
istr
atio
n ba
sed
on 1
984
data
. **
* C
alcu
late
d us
ing
Stag
ed D
isea
se C
ondi
tiodS
tage
wei
ghts
der
ived
fro
m a
nat
iona
lly r
epre
sent
ativ
e di
scha
rge
abst
ract
dat
abas
e (H
ogan
, 19
87).
34 The Jouml of Rural Health
in individual hospitals. Relative weights pubhhed by HCFA were used in calculating the DRG case mix index, while charge information from a M-
tional database was used to calculate relative weights for the Disease Stag- ing categories (see Note 2). Using either index, urban hospitals attracted rural patients with the highest expected costs, while non-referral rural hos- pitals had the lowest expected costs. The differences indicated by Disease Staging, however, are considerably smaller than those indicated by DRGs, showing that some of the ruraVurban difference may be attributable to differences in treatment patterns between rural and urban hospitals.
Table 3 parallels Table 2, but presents travel distance as well as des- tination for rural patients. Referral centers attracted more patients from out- side their own county (20% vs. 11%) than did other rural hospitals. Patients over age 75 tended to be hospitaked locally, with fewer of the oldest patients traveling for hospitalization, and so the low fraction of turd pa- tients over age 75 in urban hospitals seems to be due largely to the barrier which travel represents for the elderly. For urban and rural referral center hospitals, individuals who traveled to the hospital from beyond the coun- ties adjacent to the hospital had higher case mix indices than individuals who did not travel as far. For other rural hospitals, however, the link between distance and case mix was much more tenuous: the DRG case mix index showed no clear pattern, while the Disease Staging case mix index showed a positive relationship between distance traveled and case mix.
These tables demonstrate clearly several aspects of the travel to New York hospitals by individuals living in rural New York. First, 71 percent of hospitalizations took place in the individual's county of residence. Of the 29 percent who crossed county borders for hospitalization, almost two thirds (19%) traveled to an urban hospital for treatment. The three types of hospi- tals differed in the patients which they attracted from a distance: patients who traveled to urban hospitals and rural referral center hospitals fell on average into highercost DRGs than those who did not travel, whde the link between travel and expected treatment cost was much weaker for other rural hospitals. The fraction of patients over age 75 was lower among patients who traveled than among patients who did not, and consequently rural patients in urban hospitals were much less likely to be over age 75 than rural patients as a whole. Thus, while those who traveled tended to fall into more resource-intensive DRGs, one group with high expected costs, the very old, was under-represented among rural patients who traveled for care. F d y , rural referral centers attracted more patients from outside their own county than did other rural hospitals.
A Case-Mlx-AdjUsted Comparison of Rural Patients in Urban, Rural Re6mra.l Center, and Other Rural Hospitals.
The data presented in the previous section indicated that rural patients in rural referral centers and urban hospitals had higher case mix indices
Hogan 35
and longer stays, but were less likely to be more than 75 years old com- pared to rural patients who received care in other rural hospitals. However, these fmdings are not directly relevant to the current policy debate over appropriate payments to hospitals, since hospital payments are typically adjusted for case mix differences (meaning, usually, DRG mix), while the findings above were not adjusted for case mix. In this section, therefore, we perform a case-mix-adjusted comparison of rural patients treated in these three settings.
The urbadrural differential in the Medicare Prospective Payment Sys- tem motivates the particular comparisons pedormed here. Even after adjust- ing for case mix, teaching commitment, and wage levels, rural hospitals have a lower cost per discharge than urban hospitals, and correspondingly, Medicare pays rural hospitals less per discharge even after adjusting for case mix, wages, and teaching. However, an exception is made for hospi- tals quahfying for rural referral center status: referral centers receive the urban standardized payment. It has been argued that these payment differ- entials are reasonable, since the most severely ill rural patients travel to large urban hospitals for treatment, while referral centels presumably treat the same so& of complex cases as urban hospitals do. In the sections below, the notion that the sickest rural patients receive care in urban and rural referral center hospitals is examined.
Standard methods for case mix adjustment were used. Averages were calculated within each DRG for every patient characteristic. An expected value for each characteristic was calculated for any group of patients, using these DRG averages and the number of patients in the group falling into each DRG. Taking length of stay as an example, the expected length of stay for a group of patients estimates what the group’s average length of stay ulould have been if each patient in each DRG had had the average length of stay for that DRG. A difference between the group’s actual length of stay and its expected length of stay shows the extent to which the group differed from the average within DRGs, and does not confound those within- DRG differences with the mix of patients across short- and long-stay DRGs.
A major difficulty in attempting to use discharge abstract data to ana- lyze severity of illness differences between groups of patients is that many measures of patient illness reflect both illness severity and hospital resource use or physician practice patterns. Given the evidence of substantial differ- ences between urban (competitive, teachind hospitals and rural (monop oly, non-teaching) hospitals (Farley, 1985; Goldfarb and Coffey, 1387, Robin- son and Luft, 1987), it is clear that most comparisons will reflect both the more resource-intensive urban style of practice and patient illness severity. Therefore, patients were compared using five different measures which were ordered from most to least sensitive to differences in practice style: length of stay, number of procedures, number of secondary diagnoses, Disease Staging charge weight, and fraction of patients over age 75. Length of stay and number of procedures clearly will reflect differences in the intensity of service in individual hospitals. The number of secondary diagnoses re- corded on the abstract, whde not directly resource-related, does provide an
Table 4. A
ctua
l Cha
ract
e+ls
tics,
Expected
Cha
ract
eris
tics
Based o
n D
RG
Mix, and
Rat
io o
f Act
ual t
o Expected C
hara
cter
isti
cs, f
or R
ural
Res
iden
ts Discharged
from
New
Yor
k H
ospi
tals
, by
Des
tinat
ion H
ospi
tal (
1983
).
Destination
Hos
pita
ls
Ufb
Z3l
l H
ospi
tals
(n
=21
0*)
Rur
al R
efer
ral
Centers
(n-4
)
Oth
er Rural
Hos
pita
ls
0115
6)
Leng
th o
f st
ay (
days
)
Num
ber o
f pr
oced
ures
Num
ber
of se
cond
ary
diag
nose
s
Dis
ease
Sta
ging
cha
rge
wei
ght"
Perc
ent of
patie
nts
over
age
75
7.79
-
=
1.09
7.
14
--
1'
50 -
1.15
1.
30
1.37
- =
1.
01
1.35
7.20
- =
1.
03
7.00
1.19
- =
1.
11
1.07
1.58
- =
1.
13
1.40
3260
=
1.01
32
34
14.7
=
,w
-
14.9
--
.9
0 -
.93
.97
1.38
=
.98
3147
=
,y$)
1.41
3170
16.5
=
I,&
15
.5
s ?? it B
Mos
t ur
ban
hosp
itals
dis
char
ged
very
few
rura
l pa
tient
s. O
nly
71 u
rban
hos
pita
ls h
ad m
ore
than
100
rura
l pa
tient
s, a
nd o
nly
36 h
ad m
ore
than
500
rur
al
E pa
tient
s.
** C
alcu
late
d us
ing
Stag
ed D
isea
se C
ondi
tiodS
tage
wei
ghts
der
ived
fro
m a
nat
iona
lly r
epre
sent
ativ
e di
scha
rge
abst
ract
dat
abas
e (H
ogan
, 198
7).
Hogan 3 7
indcation of the presence of comorbidities and complications. However, it may also reflect differences in hospitals' coding practices and so may be a quite noisy measure of illness severity (see Note 3). The Disease Staging charge weight will not reflect the practice patterns of individual groups of hospitals, since the assignment of patients to Staged Disease Condition/ stage categories is largely independent of procedures performed; yet, to the degree that Disease Staging discriminates between more and less expensive patients within DRGs, this measure will show whether the expected cost of treatment differs across different destination hospitals. Finally, age is clearly exogenous to the hospital's treatment decisions.
Turning now to Table 4, four of the five measures of illness severity indicate that rural patients in urban and rural refepl center hospitals were more severely ill than rural patients in non-referral rural hospitals, even after adjusting for DRG mix. Length of stay and number of procedures, the measures most strongly related to resources consumed, had the largest differences. Rural patients in urban hospitals had 12 percent longer stays and 24 percent more procedures than patients in non-referral rural hospi- tals, while rural patients in rural referral centers had 6 percent longer stays and 19 percent more procedures than patients in rural non-referral hospi- tals. Rural patients in urban hospitals had slightly more-thanexpected sec- ondary diagnoses (3%), while rural patients in rural referral center hospitals had substantially more-thanexpected secondary diagnoses (15%), compared to patients in other rural hospitals. There were only small differences in Disease Staging charge weights, perhaps the most conservative measure of illness severity; rural patients in urban hospitals and rural referral center hospitals had expected costs that were three percent and two percent higher than rural patients in non-referral center hospitals, respectively, adjusting for DRG mix. These small differences could indicate that practice patterns ac- count for most of the measured differences between destination hospitals, or they could show that Disease Staging discriminates poorly between sets of patients within DRGs. Finally, the fraction of patients over age 75 runs contrary to the other measures: these patients were treated primarily in nonreferral hospitals, with urban hospitals having much less than the ex- pected fraction of patients over age 75.
Table 5 presents figures on these five severity measures by destination hospital and travel distance. The first three measures show irregular pat- terns with travel distance. While those who traveled to urban hospitals had longer-thanexpected stays, the same is not true of those who traveled to referral centers and other rural hospitals. For both urban and referral center hospitals, travel distance was strongly associated with aboveexpected num- bers of procedures, but no such relationship held for other rural hospitals. The number of secondav diagnoses also increased with travel distance for referral center hospitals, but not for urban or other rural hospitals. The last two measures, however, reflected very consistent patterns with travel dis- tance: the Disease Staging weight increased slightly with travel distance for all three types of hospitals, while the fraction of patients over age 75 fell far below what was expected as travel distance increased.
LU co
Tab
le 5
. Rat
ios of A
ctua
l Cha
ract
eris
tics
to Expected
Cha
ract
eris
tics
Bas
ed o
n D
RG
Mix,
for
Rur
al
Res
iden
ts D
isch
arge
d fr
om N
ew York
Hos
pita
ls, b
y D
esti
nat
ion
Hos
pita
l and T
rave
l D
ista
nce
(198
3).
Destination
Hos
pita
ls
Tra
vel
Dis
tanc
e
Rur
al
Oth
er
Urban
Ref
erra
l R
ural
H
ospi
tals
Centers
Hos
pita
ls
(n-2
10')
(n
-4)
61-5
6]
Leng
th o
f st
ay (
days
)
Num
ber
of pr
oced
ures
Num
ber
of se
cond
ary
diag
nose
s
Dis
ease
Sta
ging
cha
rge
wei
ght"
Sam
e co
unty
A
djac
ent
coun
ty
Fart
her
Sam
e co
unty
A
djac
ent
coun
ty
Fart
her
Sam
e co
unty
A
djac
ent
coun
ty
Fart
her
Sam
e co
unty
A
djac
ent
coun
ty
Fart
her
- 1.
05
1.04
.9
3 1.
20
.99
- 1.
08
1.14
1.
19
1.20
1.
50
- 1.
10
1.02
1.
26
1.01
1.
37
- 1.
01
1.01
1.
02
1.04
1.
02
.97
.95
1 .oo
.94
.89
.90
.99
.97
.95
.99
.99
1.00
ja 6 c
Frac
tion
of p
atie
nts
over
age
75
Sam
e co
unty
-
1.03
1.
08
i
Adj
acen
t co
unty
.7
4 .8
5 .9
2 .9
, P Fa
rthe
r .5
2 .6
5 .6
3
Mos
t urb
an h
ospi
tals
dis
char
ged
very
few
rur
al p
atie
nts.
Onl
y 71
urb
an h
ospi
tals
had
mor
e th
an 1
00 ru
ral p
atie
nts,
and
onl
y 36
had
mor
e th
an 5
00 ru
ral
patie
nts.
**
Cal
cula
ted
usin
g St
aged
Dis
ease
Con
diti
odSt
age
wei
ghts
der
ived
fro
m a
nat
iona
lly r
epre
sent
ativ
e di
scha
rge
absr
act d
atab
ase
(Hog
an,
1987
).
Hogan 39
In summary, rural patients in urban and rural referral center hospitals appeared to be more severely ill than rural patients in other rural hospitals, even after adjusting for DRG mix. This result was consistent a c m meas- ures which strongly reflected differences in practice patterns (e.g., length of stay and number of procedures) as well as measures which did not reflect practice patterns in individual hospitals at all (e.g., Disease Staging). The only anomalous fmding was that the very elderly showed a low propensity to travel and so tended to be treated largely in rural non-refeml hospitals.
Conclusion and SugsestionS for Future Research This analysis has described the travel of residents of rural New York to
hospitals in that state, using a few measures of location, distance, and case mix. Over 70 percent of all hospitalizations of rural individuals took place in the patient's own county of residence, and of those patients who trav- eled across county borders for hospitalization, almost two thirds traveled to urban facilities. Travelers tended to fall into resource-intensive DRGs and have longer lengths of stay. However, patients over age 75 were much less likely to have traveled for hospitalization.
After adjusting for DRG mix, rural patients in urban and rural referral center hospitals were determined to be more severely ill. While resource- use based measures (such as length of stay and number of procedures) showed large differences after DRG-mix adjustment, even a very conserva- tive measure such as Disease Staging weights showed some difference. These data also demonstrated that the oldest rural patients were hospital- ized primarily in rural hospitals, even after adjusting for differences in DRG mix.
The most consistent finding in this study was that the very elderly do not travel much for hospitalization. This did not seem to matter much from a payment perspective, as travelers were more severely ill than non-travel- ers despite the low propensity of the very elderly to travel. In other con- texts, however, the travel patterns of the very elderly could be a focus of concern. Hospital closure and regionalization of care, for example, may have a more profound impact on the very elderly than on other segments of the population. Policy makers should be aware of ddferences in willing- ness and ability to travel for care when examining changes in the availabil- ity of health care services.
The findings concerning severity of hess are clearly relevant to the question of hospital payments in New York. DRGs do not completely cap ture all differences between the segments of the rural population served in urban, rural referral center, and other rural hospitals. Even when Disease Staging, a very conservative measure of illness severity differences which is nearly independent of the hospital's actual resource use is used, rural pa- tients in rural referral center and urban hospitals should cost between two percent and three percent m o e to treat than rural patients in other rural hospitals, after accounting for DRG mix. Different payment rates to these three classes of hospitals in New York would, therefore, seem to be justi-
40 i%e Journal of Rural Health
lied in a DRG-based reimbursement system. These findings using data from New York State may or may not be
relevant to Medicare reimbursement policy nationwide. The Medicare ur- badrural differential and rural refenal center distinction clearly make sense to the extent that they capture otherwise unmeasured differences in case mix and illness severity. In New York, the applicable Medicare differential is currently less than 3 percent, whereas in other areas the differential is as high as 20 percent. A similar analysis, pedonned in a region of the country where a large urbadrud differential exists, could provide an interesting contrast to this study, and could shed light on the degree to which vari- ations in the udxmhwal differential are justified on severityaf-illness grounds.
NOTES
1. T ~ I S MSA-bgsed definition encompasses heterogeneous regions and hospitals, as discussed by Moscovice and Rasenblatt (1982). However, this definition of rural areas has found widespread use in research and hospital reimbursement and is adopted here.
2. The HCFA DRG weights are reasonably good cost predictors for both Medicare and non- Medicare patients (Hogan and Coffey, 1987). Disease Staging relative weights were cal- culated from 1% charge data in the Hospital Cost and Utilization Project @CUP-2) data- k, a tio on ally representative sample of US. hospitals. DRGs and Disease Staging explain roughly the same ftaaion of the variation in patient charges ( 2 S 4 , but DRGs use 470 categories while Disease Staging uses approximately 2,000 Staged Disease Condition,' "age patient categories (Hogan, 1987).
3. I want to thank one of the anonymous referees for emphasizing this point.
Hogan
REFERENCES
41
American Hospital Association. Hospital Statktics, 1934 .Mition. Chicago: American Hospital Assonation, 19%.
WMC, E.M., Parkinson, R.C. and Hall, D.S. Geographic access to hospital a r e A M minute travel time standard. Medical CaE, 1976, 14, 616-624.
Cageorge, S.M., Roos, L.L. and Danzinger, R. Gallbladder operdtions: A population-based analysis. Medical Caw, 1981, 19, 510.525.
Farickson, R. m e Spatial Behavior of Ho.spi&l Patients. Chicago: University of Chicago De- partment of Geography, 1970.
Farley, D.E. Competition Among Hapitak: Ma* Stmctuw and Its Relation to Uilizution, Costs and Financhl Position. Rockville, Maryland: Hospital Cost and Utilization Project Research Note 7, DHHS Publication (PHS) 85-3353, US. Department of Health and Hu- man %vices, National Center for Health Services Research and Health Care Technology Assessment, 1985.
Folland, S.T. Predicting hospital market shares. Inquiy, 1%3, 20, 34-44. Gamick, D.W. Luft, H.S., Robinson, J.C. and Treteault, J. Appropriate measures of hospital
market areas. Health Seruices Reseanh, 1987, 22, 69-89. Goldfa&, M.G. and Coffey, R.C. Case-mix differences between teaching and nonteaching
hospitals. I q u i y , 1987, 24, 68-84, Hogan, C. Imp-roving the E q l a W o y P o r n of Btiase Staging Using Seconday SLIGSUge
I n f m t i o n . Unpublished manuscript, Hospital Studies Program, National Center for Health Services Research and Health Care Technology hsesrnent, Rockvile, Maryland, 1987.
Hogan, C. and Coffey, R.M. An Empirical Eualuation of Diagnosis Related Cmp for Hospi- talized Medicaw and Blue C m Populations. Unpublished manuscript, Hospital Studies Program, National Center For Health Services Research and Health Care Technology Assessment, Rockville, Maryland, 1987.
Mayer, J.D. The distance behavior of hospital patients: A disaggregated analysis. Social Scl- ence and Medicine, 1983, 17, 819-827.
McGuirk, M.A. and Porell, F.W. Spatial panems of hospital utilization: The impact of distance and time. Inquiy, 1984, 21, 84-95.
Moscovice, I. and Rosenblan, R. 7he Viability of the Rural HapituL A Sptbesis of Findings j b m Health Seruices Research. Rockvie, Maryland: US. Department of Health and Hu- man Services, National Center for Health Services Research, 1382.
Robinson, J.C. and Luft, H.S. Competition and the cost of haspital care, 1972 to 1982. Journal of the A m ' c u n Medical &ociation, 1387, 257, 324-3245,
Shannon, G.W., Skinner, J.L. and Bashshur, R.L. Time and distance: The journey for medical care. International Journal of Health Services, 1973, 3, 237-244.
Studnick, J. The minimization of travel effon as a delineating intluence for urban haspital service areas. International Journal of H d t h .%vices, 1976, 5 , 679-692.
U.S. Department of Commerce. Bureau of the Census. State and M e t w i t a n A m Data Bcnk 1982: A Statistical Ahstract Supplf?ment. Washington, D.C.: US. Depamnent of Com- merce, 1382.