Introduction
Introduction
Obesity has been shown to be associated with elevated levels of health services utilization[1
- 9]. However, to the knowledge of the present authors there has been but one population-representative
study of adults since 1990 in Germany that has assessed the impact of overweight on
utilization [10]
[11]. In other words, there seems to be no non-clinical research other than that study
which has examined associations of obesity and utilization among adults while adjusting
for predisposing and enabling factors known to be relevant to health care utilization
[12]
[13]
[14]
[15]. In that study [10]
[11], which analyzed data of the German Federal Health Survey 1998 [16], obesity was positively associated with number of visits to general practitioners
(GP), but neither with number of visits to any physician nor internal specialists,
nor the number of kinds of specialists visited.
However, a number of features of that study - which besides is seminal for German
health services research - merit further investigation when looking at obesity as
a raison d’être for utilization. First, body mass - as defined by the body mass index
(BMI, i. e. weight in kg/[height in m]²) - was dichotomized so that the obese (BMI
≥ 30) were contrasted to those non-obese. As earlier analyses have shown, however,
adults in obesity class 2 or 3 (35 ≤ BMI < 40 or BMI ≥ 40) represent a population
different from those in class 1 (i. e. 30 ≤ BMI < 35), for instance regarding physical
health-related quality of life [17].
Second, Thode et al. deliberately focused on the utilization of outpatient services,
arguing that theoretically - due to the general regulations of the German health care
system - many of the factors they scrutinized (above all predisposing and enabling)
should not substantially influence utilization of inpatient services. However, as
obesity can be regarded to be a proxy to numerous, especially chronic diseases, it
does seem appropriate to throw a glance at inpatient care of obese adults in this
context, and compare it to that of their normal weight counterparts.
Third, Thode et al. deliberately circumvented issues of high utilization, presumably
doing so more for reasons of time and space (specifically, i. e., given the normative
issues they rightfully note to be involved here) than for reasons of irrelevance.
However, especially extreme obesity may in fact be a major factor with regard to extraordinarily
high levels of utilization. Thus, an explorative look at the association of obesity
and high utilization may be no less than justifiable.
In sum, the present study for Germany aims to answer the following research questions:
-
Do adults in different classes of obesity utilize outpatient health services more,
and/or more frequently, than those in normal weight range?
-
Do adults in different classes of obesity utilize inpatient health services more,
and/or for longer periods of time (i. e. more inpatient days), than those in normal
weight range?
-
Is (especially extreme) obesity associated with high utilization of out- and inpatient
health services?
Methods
Methods
Study population and sampling
The KORA Survey S4 1999/2000 is a cross-sectional, population-representative health
survey of the resident population aged 25 to 74 in the Augsburg region (i. e. Augsburg
city, plus the two adjacent administrative districts). A sample of n = 6.640 individuals
were invited to participate from Augsburg and a random sample of 16 out of 70 communities
in the two districts. In sampling, within each of the 17 communities a random sample
within each of 10 equal strata by sex and age was drawn from the registration office.
Of the n = 4.261 participating in this main survey (response rate: 67 %), a random
sample of n = 1.186 with 30 nearly balanced strata by sex, age and BMI (BMI < 25,
25 ≤ BMI < 30, BMI ≥ 30) was drawn for a three-wave computer-aided telephone interview
(CATI) follow-up after two, four, and six months. In all, n = 947 participated in
all waves (response rate: 80 %). In all, fieldwork lasted from October 1999 to August
2001, and on average ranged over seven and a half months for any participant.
Of the n = 947 participants, five with a BMI < 18.5 kg/m2 were excluded for reasons of cell count and probable underweight-specific health
problems. Table [1] shows the resulting analysis sample (n = 942) by cross-tabulating BMI categories
with all potential confounders considered in the analyses. For analysis, the following
stratification is used according to WHO definitions [18]
[19] (BMI in kg/m2): normal weight 18.5 ≤ BMI < 25, preobese 25 ≤ BMI < 30, obesity class 1 30 ≤ BMI <
35; and obesity classes 2 - 3 BMI ≥ 35.
Table 1 Distributions of potential confounders by groups of body mass index (BMI)[1] (n = 942, KORA Survey S4 1999/2001, Sub-study “Costs of illness related to obesity”)
|
normal weight
(n = 304) |
preobese
(n = 324) |
obese class 1
(n = 233) |
obese classes 2 - 3
(n = 81) |
TOTAL |
| women |
163 (53.6 %) |
164 (50.6 %) |
111 (47.6 %) |
54 (66.7 %) |
492 (52.2 %) |
| men |
141 (46.4 %) |
160 (49.4 %) |
122 (52.4 %) |
27 (33.3 %) |
450 (47.8 %) |
| 25 - 35 years of age |
65 (21.4 %) |
59 (18.2 %) |
39 (16.7 %) |
15 (18.5 %) |
178 (18.9 %) |
| 35 - 45 years of age |
59 (19.4 %) |
63 (19.4 %) |
51 (21.9 %) |
14 (17.3 %) |
187 (19.9 %) |
| 45 - 55 years of age |
56 (18.4 %) |
68 (21.0 %) |
44 (18.9 %) |
25 (30.9 %) |
193 (20.5 %) |
| 55 - 65 years of age |
61 (20.1 %) |
69 (21.3 %) |
50 (21.5 %) |
14 (17.3 %) |
194 (20.6 %) |
| 65 - 75 years of age |
63 (20.7 %) |
65 (20.1 %) |
49 (21.0 %) |
13 (16.0 %) |
190 (20.2 %) |
| upper social class |
60 (19.7 %) |
57 (17.6 %) |
28 (12.1 %) |
10 (12.3 %) |
155 (16.5 %) |
| middle social class |
196 (64.5 %) |
208 (64.2 %) |
134 (57.8 %) |
46 (56.8 %) |
584 (62.1 %) |
| lower social class |
48 (15.8 %) |
59 (18.2 %) |
70 (30.2 %) |
25 (30.9 %) |
202 (21.5 %) |
| private health insurance |
52 (17.2 %) |
45 (14.2 %) |
29 (12.6 %) |
9 (11.3 %) |
135 (14.5 %) |
| statutory health insurance |
251 (82.8 %) |
271 (85.8 %) |
202 (87.4 %) |
71 (88.8 %) |
795 (85.5 %) |
| urban place of residence |
144 (47.4 %) |
138 (42.6 %) |
104 (44.6 %) |
33 (40.7 %) |
419 (44.5 %) |
| rural place of residence |
160 (52.6 %) |
186 (57.4 %) |
129 (55.4 %) |
48 (59.3 %) |
523 (55.5 %) |
|
Notes: Sex and age were stratification dimensions in sampling besides BMI for the
present analysis sample (see text), and sex, age and place of residence in the main
survey from which the present sample was drawn; thus, cross-tabulations with BMI may
in no way be viewed as reflecting the situation in the population. 1Definition of
BMI groups see text.
|
While sex, age and place of residence are basically equally distributed merely due
to the fact that they served as stratification dimensions in sampling (besides BMI),
the most conspicuous difference in this context pertains to the fact that among those
in obesity classes 2 - 3, women with 66.7 % represent a comparably strong majority.
Furthermore, while no great differences relate to the distribution of statutory vs.
private health insurance within body mass categories, respondents from the lower socio-economic
echelon are more strongly represented in obese (class 1: 30.2 %, classes 2 - 3: 30.9
%) than in nonobese groups (normal weight: 15.8 %, preobese: 18.2 %).
Measures
The indicators of health services utilization used in the present study were assessed
via self-report in each of the CATI follow-ups. As an indicator of out-patient utilization,
the numbers of visits to GP in the three eight weeks-periods preceding the follow-ups
were summed up as an estimator of the number of visits within the study period. The
items read as follows: “How often did you visit a physician in the last eight weeks?”,
and - for each of the visits - “Which medical field did that physician belong to?”.
Likewise, inpatient utilization was assessed by adding up the numbers of days spent
in hospital. Here, the following items were used: “Have you stayed in hospital overnight
(i. e. inpatient) during the last eight weeks?”, and “In sum, how many days have you
spent in hospital for inpatient care during the last eight weeks?”.
Obesity
Body weight and height were assessed within the anthropometric examination in the
main survey. Calibration of measuring instruments was ensured by weekly or daily inspections
using standard weights or resistors, as appropriate. Body mass was indexed for each
participant as ([weight in kg]/[height in m]²). Overweight groups were defined following
WHO classifications (see above) [18]
[19].
Predisposing and enabling factors
Sex, age and place of residence were known in advance for each participant due to
the sampling procedure. Social class was indicated by the revised index by Helmert
which is based on education, occupational status, and income (for details, see [20]), the assessment of which followed national recommendations [21]. Kind of sickness fund was assessed by asking participants whether their fund was
statutory (German ‘GKV’) or private (‘PKV’).
Statistical analysis
Following a descriptive analysis, two-part models [22] were performed for each of the two utilization parameters. In each model, the first-step
equation models the probability that respondents reported any relevant utilization
at all, employing a logistic model (procedure LOGISTIC in STATA/SE 8.1 for Windows).
In contrast, the second-step equation models the frequency of utilization among users,
employing a zero-truncated negative binomial model (procedure trnbin0 in STATA/SE
8.1 for Windows). This approach is appropriate here because counts are examined with
no possibility of having zero values; as coefficients, incident rate ratios (IRR)
are reported describing changes in outcome associated with a one unit increase in
regressors.
Besides, to elucidate the conditions of high utilization, multinomial regression models
were employed in which on the side of each regressand three levels were distinguished:
no utilization at all (= reference group), some utilization, and high utilization
(procedure NOMREG in SPSS 12.0.1 for Windows). The latter was defined as the upper
5 % of the distribution in case of GP visits, and as the upper 2.5 % in case of days
spent in hospital. For reasons of relevance to the research questions put forward
above, and for ease of presentation, only statistics for the ‘high vs. no’-contrast
will be reported in the Results-section.
In each of the models, the four BMI-groups as described before defined the focal regressor,
using ‘normal weight’ as the reference group. As covariates, sex, age, lower and middle
vs. upper social class, statutory vs. private health insurance, and rural vs. urban
place of residence (i. e. living in one of the two adjacent districts vs. Augsburg
city) were entered into each model.
Results
Results
Table [2] presents the observed distributions of the number of visits to GP and days spent
in hospital by the predisposing and enabling factors selected for the analysis, and
by BMI-groups. Proportions of respondents who had visited a GP at all were virtually
the same in women and men, but higher among the elderly and those from the lower social
class. While more than half of those whose sickness cover was statutory had visited
a GP at all, only about a third of the privately insured had done so. Furthermore,
more people living in a rural than an urban area had attended a GP. As for obesity,
contrary to expectation not those in classes 2 - 3 have the highest rate of respondents
with at least one GP visit, but those in obesity class 1 (64.2 %).
Table 2 Visits to general practitioners (GP) and inpatient days over half a year, by sex,
age, social class, health insurance, place of residence, and obesity (four BMI-groups[1], KORA-Survey S4 1999/2001, Sub-study “Costs of illness related to obesity”)
|
number of visits to general practitioners |
number of days in hospital (inpatient) |
|
% ≥ 1 |
within ≥ 1 |
% ≥ 8 |
% ≥ 1 |
within ≥ 1 |
% ≥ 7 |
|
|
mean |
median |
min-max |
(upper
5 pctiles) |
|
mean |
median |
min-max |
(upper
2.5 pctiles) |
| women |
53.9 % |
3.32 |
2.0 |
1 - 18 |
4.9 % |
4.9 % |
10.41 |
6.5 |
1 - 49 |
2.4 % |
| men |
53.2 % |
3.27 |
2.0 |
1 - 20 |
5.1 % |
6.0 % |
12.41 |
7.0 |
1 - 41 |
3.3 % |
| 25 - 34 years of age |
43.3 % |
2.65 |
2.0 |
1 - 18 |
1.1 % |
5.6 % |
5.70 |
4.5 |
1 - 15 |
1.7 % |
| 35 - 44 years of age |
45.9 % |
2.52 |
2.0 |
1 - 09 |
2.7 % |
5.9 % |
8.09 |
6.0 |
2 - 30 |
2.7 % |
| 45 - 54 years of age |
46.1 % |
3.20 |
2.0 |
1 - 17 |
5.2 % |
3.6 % |
13.42 |
9.0 |
1 - 40 |
2.6 % |
| 55 - 64 years of age |
63.2 % |
3.60 |
3.0 |
1 - 20 |
5.2 % |
5.7 % |
16.36 |
13.0 |
2 - 49 |
3.1 % |
| 65 - 74 years of age |
68.6 % |
3.97 |
3.0 |
1 - 17 |
10.6 % |
6.3 % |
13.75 |
9.0 |
1 - 41 |
4.2 % |
| upper social class |
43.2 % |
3.76 |
3.0 |
1 - 18 |
5.1 % |
5.2 % |
8.25 |
6.5 |
1 - 20 |
2.6 % |
| middle social class |
52.5 % |
3.11 |
2.0 |
1 - 20 |
4.6 % |
5.6 % |
10.69 |
7.0 |
1 - 49 |
2.9 % |
| lower social class |
65.0 % |
3.50 |
3.0 |
1 - 14 |
6.0 % |
5.0 % |
16.60 |
14.5 |
1 - 40 |
3.0 % |
| private health insurance |
34.1 % |
3.28 |
2.0 |
1 - 17 |
3.0 % |
4.4 % |
11.00 |
6.0 |
2 - 35 |
2.2 % |
| statutory health insurance |
56.7 % |
3.31 |
2.0 |
1 - 20 |
5.4 % |
5.7 % |
11.53 |
7.0 |
1 - 49 |
3.0 % |
| urban residence |
47.0 % |
3.16 |
2.0 |
1 - 17 |
4.6 % |
5.0 % |
12.42 |
9.0 |
1 - 49 |
3.3 % |
| rural residence |
58.8 % |
3.38 |
2.0 |
1 - 20 |
5.4 % |
5.7 % |
10.80 |
5.5 |
1 - 41 |
2.5 % |
| normal weight |
48.0 % |
2.92 |
2.0 |
1 - 18 |
3.6 % |
4.9 % |
8.46 |
4.0 |
1 - 40 |
2.0 % |
| preobese |
50.2 % |
3.43 |
2.0 |
1 - 17 |
5.9 % |
4.6 % |
8.93 |
6.0 |
1 - 41 |
2.2 % |
| obese class 1 |
64.2 % |
3.24 |
2.0 |
1 - 20 |
3.9 % |
5.6 % |
12.23 |
7.0 |
1 - 35 |
3.0 % |
| obese classes 2 - 3 |
57.5 % |
4.21 |
3.0 |
1 - 14 |
10.0 % |
9.9 % |
20.62 |
17.0 |
5 - 49 |
8.6 % |
|
Notes: unadjusted data; m: mean, med: median, min-max: minimum-maximum, pctiles: percentiles.
1 Definition of BMI groups see text.
|
Regarding the frequency of visits among those who had reported at least one, patterns
are not entirely similar. While again no substantial difference is observed between
sexes, and while utilization again increases with age, no clear pattern is observable
across social classes. Also, differences both between the statutory vs. privately
insured and rural vs. urban dwellers seem to be of lesser magnitude. That is, those
insured in the “GKV” and those living in more rural areas tend to have any GP contact
with a higher probability than their private and urban counterparts, respectively.
Concurrently, those who did go obviously did not do so more frequently. Finally, there
is a clear contrast with regard to obesity: while those in classes 2 or 3 were only
second place in reporting any GP contact (57.5 %), those who did go did so more often
than those in obesity class 1 (mean 4.21, vs. 3.24 in the latter group). Also, looking
at high utilization the obese in class 2 - 3 have the second highest proportion of
people with at least eight GP visits over half a year (10.0 %), only outdone by those
respondents aged 65 - 74 years (10.6 %).
Hospital utilization as expected varies on a quite lower level than GP visits in terms
of the proportion of respondents reporting any utilization (for the total sample:
5.4 %). Distributions along the factors kind of sickness fund (more statutory) and
place of residence (more rural) are roughly comparable to those regarding GP visits.
For other factors, partially different patterns emerge. For instance, while only small
differences pertain to being hospitalized at all across social classes, those from
the lower echelon had clearly been inpatient longer than those from middle and upper
classes (mean = 16.60 days, vs. 10.9 and 8.25, respectively). Finally, regarding obesity,
results are clear-cut. First, the proportion of respondents in obesity classes 2 -
3 who had been hospitalized at all is approximately twice of those in the other groups
(9.9 %). Second, this group was hospitalized for extraordinarily large number of days
(mean = 20.62 days, vs. 12.23, 8.93 and 8.46 in the other BMI-groups, respectively).
Lastly, high utilization (corresponding to the upper 2.5 percentiles of the distribution
in the total sample) was much more prevalent again in the extremely obese compared
to the other groups: 8.6 % of the former spent seven or more days in hospital, while
even in the other obese group (class 1), this proportion was only 3 %, and ranged
down to 2.0 % in the normal weight group.
In order to hedge these results against chance variations and, more importantly, confounding
by other factors that may predispose to or enable utilization, inference statistical
modeling was conducted for both indicators of utilization (number of visits to GP,
number of days in hospital). This was accomplished by scrutinizing the probability
of respondents to report any relevant utilization at all (logistic models), frequent
utilization (among users: zero-truncated negative binomial models), and high utilization
(multinomial models, of which only the statistics for the ‘high vs. no’-contrast are
reported hereafter).
Table [3] shows the results of these models. Sex differences in utilization are negligible
and statistically insignificant in case of GP visits, and stronger but again insignificant
in case of days in hospital. Older respondents clearly reported more utilization in
terms of GP visits, most notably by 17.77 times higher odds of eight or more GP visits
in the oldest vs. the youngest group (p < 0.001). A similar assertion holds for days
in hospitals, however not reaching statistical significance other than for those aged
65 - 75 who report more days than the youngest group if hospitalized at all (IRR =
3.26, p < 0.05). Results for social class strongly resemble those in the descriptive
analysis reported before, but for the most part are not statistically significant.
Furthermore, noteworthy contrasts pertain to any GP utilization (vs. none) in that
those in statutory health insurance (vs. private) as well as rural (vs. urban) dwellers
report this more often. The statutorily insured also tend to have a comparatively
high chance for high GP utilization (OR = 3.09, ns). Regarding the considerable difference
between rural and urban dwellers, one explanation might be the lower density of medical
specialists in the districts adjacent to Augsburg City. Finally, turning to the focal
correlate of utilization scrutinized in the present study, the multivariable analyses
confirm the descriptive accounts of differences between the obese and normal weight
respondents. That is, while respondents in obesity class 1 differ significantly from
those normal weight in their tendency to report any visit to GP (OR = 1.84, p < 0.01),
among those who did go to the GP a similar assertion pertains only to those in obesity
class 2 or 3 (IRR = 1.63, p < 0.05). Also, the odds of high utilization are about
3.6 times higher in this latter group than in those in normal weight range (p < 0.05).
Finally, regarding utilization of hospitals, again the obesity classes 2 - 3 stand
out. Not only are the odds of at least one inpatient day 2.39 times higher among these
extremely obese than among the non-overweight group (ns), the former also report significantly
more days if hospitalized at all (IRR = 3.24, p < 0.05), and most notably have 5.4
times higher odds of high utilization (p < 0.01).
Table 3 Visits to general practitioners (GP) and inpatient days over half a year, by sex,
age, social class, health insurance, place of residence, and obesity (four BMI-groups[1], adjusted estimates, KORA-Survey S4 1999/2001, Sub-study “Costs of illness related
to obesity”)
|
number of visits to general practitioners |
number of days in hospital (inpatient) |
|
TWO-PART MODEL |
MULTINOMIAL MODEL |
TWO-PART MODEL |
MULTINOMIAL MODEL |
|
any utilization
vs. none |
if any:
how much? |
high utilization
vs. none |
any utilization
vs. none |
if any:
how much? |
high utilization
vs. none |
|
OR |
IRR |
OR |
OR |
IRR |
OR |
| women |
1 |
1 |
1 |
1 |
1 |
1 |
| men |
1.06 |
1.02 |
1.16 |
1.33 |
1.50 |
1.61 |
| 25 - 35 years of age |
1 |
1 |
1 |
1 |
1 |
1 |
| 35 - 45 years of age |
1.12 |
0.96 |
2.59 |
1.05 |
1.50 |
1.66 |
| 45 - 55 years of age |
1.10 |
1.31 |
4.53 |
0.60 |
1.51 |
1.36 |
| 55 - 65 years of age |
2.41*** |
1.57* |
7.64* |
1.04 |
1.66 |
1.94 |
| 65 - 75 years of age |
3.01*** |
1.82** |
17.77*** |
1.15 |
3.26* |
2.60 |
| upper social class |
1 |
1 |
1 |
1 |
1 |
1 |
| middle social class |
1.10 |
0.69* |
0.81 |
1.02 |
1.53 |
1.07 |
| lower social class |
1.42 |
0.78 |
0.97 |
0.81 |
2.66 |
0.89 |
| private health insurance |
1 |
1 |
1 |
1 |
1 |
1 |
| statutory health insurance |
2.46*** |
1.01 |
3.09 |
1.38 |
1.29 |
1.51 |
| urban place of residence |
1 |
1 |
1 |
1 |
1 |
1 |
| rural place of residence |
1.67*** |
1.06 |
1.59 |
1.15 |
0.90 |
0.72 |
| normal weight |
1 |
1 |
1 |
1 |
1 |
1 |
| preobese |
1.01 |
1.20 |
1.57 |
0.95 |
0.72 |
1.09 |
| obese class 1 |
1.84** |
1.13 |
1.49 |
1.14 |
1.72 |
1.49 |
| obese class 2 or 3 |
1.42 |
1.63* |
3.57* |
2.39 |
3.24* |
5.40** |
|
Notes: Three models each for GP visits and inpatient days (logistic for any utilization
at all, zero-truncated negative-binomial [conditional on utilization greater nil],
and multinomial for high utilization); OR: odds ratio; IRR: incident rate ratio; estimates
from multinomial models with reference group “no utilization” (high vs. no-contrast
shown only); * p < 0.05 ** p < 0.01 *** p < 0.001; 1 Definition of BMI groups see
text.
|
Discussion
Discussion
The present study set out to analyze obesity as a correlate of two selected indicators
of out- and inpatient health services utilization, based on data from a population-representative
health survey in the Augsburg region, Germany (KORA Survey S4 1999/2001). Results
can be summarized as follows. First, any visit to a GP to have happened at all was
significantly more probable among those in obesity class 1 but not among those in
classes 2 - 3, when compared to those in normal weight range. At the same time, among
those who did visit GP, those in obesity classes 2 - 3 did so significantly more often
than their normal weight counterparts, while this assertion does not hold for respondents
in obesity class 1. Second, an inpatient stay in hospital was significantly more probable
than in those normal weight only among the extremely obese (i. e. class 2 or 3). In
contrast, regarding the number of days in hospitals among users, both obese groups
spent longer periods of time in hospital than those in normal weight range. However,
this failed to reach statistical significance due to the smaller base rate of participants
who had been hospitalized at all. Third, regarding high utilization, in both the out-
and the inpatient sector only those in obesity classes 2 - 3 but not those in class
1 are significantly more likely to report “high” utilization than the normal weight
group.
Thus, the research questions stated at the outset of the present analysis can be answered
as follows. On one hand, obesity generally does go with a more pronounced utilization
of out- and inpatient health services. On the other hand, if such differences occur,
they for the most part hold only for those in obesity classes 2 - 3 for all outcomes
but visiting a GP at all. In other words, compared with normal weight participants,
any visit to a GP is more probable among those slightly but not among those extremly
obese; at the same time, a significantly higher frequency in visits to GP is found
only in the extremely obese. Similar assertions hold for numbers of GP visits tantamount
to high utilization, being hospitalized at least once, and being hospitalized for
at least seven days over half a year (if at all). In sum, and keeping in mind that
especially the difference between obesity classes 2 - 3 vs. 1 in regard to any GP
visit should not be over-interpreted, there seems to be a tendency for obesity to
be associated with excess health services utilization only if it is extreme.
Before drawing some conclusions based on these results, some limitations of the present
study have to be considered. First, utilization was assessed by self-reports, thus
falling short of the “gold standard” of insurance data in the context of visits to
physicians and inpatient days. However, the survey assessment strategy also holds
an advantage, namely to be able to compile a lot of other information on the individual
level such as psychosocial variables [17], which have been argued to be relevant to the issue of health care utilization as
well [15]. Second, the response rate of 80 % in a follow-up of participants of a health survey
that itself had a response rate of 67 % may only just be within methodical standards
in survey research. Third, the representativeness of the study sample from the Augsburg
region for the whole of Germany still has to be examined in future analysis (see below,
Future planning). Finally, and most importantly, stratified analysis as well as tests
for effect modifications was neither the focus of this study nor possible in a comprehensive
way due to methodical constraints (most notably sample size). Explorative analysis
of interaction terms between the BMI-factor and sex at any rate revealed that the
latter did not conspicuously modify the associations between obesity and utilization.
Nevertheless, the need to more deeply conduct subgroup analysis is acknowledged here,
both from the view of health services epidemiology and health economics [23]. At the same time, analyses such as those in the present study still have an added
value in health services research because they do shed light on excess utilization
attributable to certain health states such as obesity by taking into account that
utilization would have been incurred in the absence as well [24].
In sum, to our knowledge this is but the second population-representative study of
adults in Germany since 1990 that has assessed excess utilization attributable to
obesity. As Thode et al. [10]
[11] for the whole of Germany, our analyses indicate for a regional population in Southern
Germany that services by GP are generally utilized more by obese than normal weight
adults. Most notably, those among the extremely obese who had visited a GP at all
reported a higher frequency of visits. Going beyond the scope of the analysis by Thode
et al., our data suggest that both inpatient and high out- and inpatient utilization
is more pronounced as well, however only in those with obesity by classes 2 - 3.
In a nutshell, our results point to an excess utilization of out- and inpatient health
services especially by extremely obese adults. This also underlines the added value
of differentiating obesity classes 1 vs. 2 - 3 in health services utilization research.
Of course, this may depend on the kind of utilization. Meisinger et al., for instance,
in a recent study with female participants of the MONICA Augsburg Survey S3 (1994/95)
[25], among other things found delayed routine cancer screening use in obese vs. normal
weight women, but no excess utilization of inpatient services. In other words, for
cancer screening use (Meisinger et al.’s focal outcome variable) treating those with
a BMI ≥ 30 as one homogeneous group may be entirely appropriate. In contrast, the
present study demonstrates that distinguishing subgroups with different grades of
obesity may be quite important, e. g. when scrutinizing inpatient utilization.
Finally, this study underscores the need to treat and prevent (especially extreme)
obesity in order to appropriately manage out- and inpatient health care utilization.
This is a topical objective for health policy both in light of the high and increasing
prevalence of obesity (not least in the Augsburg region [26]), and considering the numerous options available for preventive policies in this
context [19]
[27].
Future planning
Future planning
While this paper has focused on describing and modeling the associations of obesity
with selected indicators of health services utilization, future plans of the GSF-Institute
of Health Economics and Health Care Management in cooperation with the Hannover Medical
School (Medical Psychology) within obesity research include the following:
Costing the problem
Using an adequate quantity of resource utilization, which will include a range of
parameters relevant both to direct and indirect costs, it is planned to assess the
health care costs attributable to obesity in the S4-sample described above. On this
basis, projections are planned to the Augsburg region, and possibly other geographic
and/or administrative units.
Exploring longitudinal associations
In follow-up surveys, it is planned to examine associations between duration of and
changes in obesity status, and the utilization and costs of health care.
Evaluating preventive interventions
Moving beyond costs of illness studies, it is projected to explore the economic impact
of policies to prevent obesity as a risk factor. In the past, such policies have proven
to have only limited success, in particular among socially disadvantaged groups. Against
this background, further research will pursue to identify promising approaches to
promote physical activity among such groups (e. g. women with a comparably low socio-economic
status), and especially their evaluability in terms of their financial and health
impact.
Acknowledgement
Acknowledgement
The investigation has been supported by the GSF-National Research Center for Enviroment
and Health. The authors wish to thank all present and former members of the KORA Study
Group, especially Christian Janßen PhD (now at University of Cologne, Institute and
Polyclinic of Occupational Medicine, Social Medicine and Public Health), Andreas Mielck
PhD and Walter Satzinger PhD (both at GSF-Institute for Health Economics and Health
Care Management), and Kerstin Wüstner PhD (now at University of Augsburg, Special
Chair for Applied Psychology), for their responsibilities taken in earlier phases
of the S4-subproject “Costs of illness related to obesity”, PD Rolf Holle PhD for
his guidance of what is feasible (and what is not), Angela Döring (GSF-Institute for
Epidemiology) for sharing her expertise in obesity research, and Hannelore Nagl and
Andrea Wulff (both at GSF-Institute for Health Economics and Health Care Management)
for their diligent assistance in medical documentation. Last but not least, we would
like to express our appreciation to Sebastian Baumeister (Ernst Moritz Arndt University
of Greifswald, Institute of Epidemiology and Social Medicine, and UCLA School of Public
Health, Department of Health Services) for his valuable co-operation both on substantive
and statistical issues relevant to the utilization of health care.
The article relates specifically to the following contributions of this special issue
of Das Gesundheitswesen: [26]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35].
