Primary care physician supply and other key determinants of health care utilisation: the case of Switzerland
BMC Health Services Research
Primary care physician supply and other key determinants of health care utilisation: the case of Switzerland Andr Busato*1 and Beat Knzi2
0 Swisspep - Institute for Quality and Research in Healthcare , Postfach - CH 3073 Guemligen , Switzerland
1 Institute for Evaluative Research in Orthopaedic Surgery, University of Bern , Stauffacherstrasse 78, CH-3014, Bern , Switzerland
Background: The Swiss government decided to freeze new accreditations for physicians in private practice in Switzerland based on the assumption that demand-induced health care spending may be cut by limiting care offers. This legislation initiated an ongoing controversial public debate in Switzerland. The aim of this study is therefore the determination of socio-demographic and health system-related factors of per capita consultation rates with primary care physicians in the multicultural population of Switzerland. Methods: The data were derived from the complete claims data of Swiss health insurers for 2004 and included 21.4 million consultations provided by 6564 Swiss primary care physicians on a feefor-service basis. Socio-demographic data were obtained from the Swiss Federal Statistical Office. Utilisation-based health service areas were created and were used as observational units for statistical procedures. Multivariate and hierarchical models were applied to analyze the data. Results: Models within the study allowed the definition of 1018 primary care service areas with a median population of 3754 and an average per capita consultation rate of 2.95 per year. Statistical models yielded significant effects for various geographical, socio-demographic and cultural factors. The regional density of physicians in independent practice was also significantly associated with annual consultation rates and indicated an associated increase 0.10 for each additional primary care physician in a population of 10,000 inhabitants. Considerable differences across Swiss language regions were observed with reference to the supply of ambulatory health resources provided either by primary care physicians, specialists, or hospital-based ambulatory care. Conclusion: The study documents a large small-area variation in utilisation and provision of health care resources in Switzerland. Effects of physician density appeared to be strongly related to Swiss language regions and may be rooted in the different cultural backgrounds of the served populations.
Previous research on variations in the volume of
physician services in geographic areas has emphasized that an
additional use of services is not necessarily associated with
improved health in the corresponding populations.
After controlling for input prices and health status, it was
found that the volume of physician services is driven
partly by local practice patterns and partly by differences
in physician density and speciality. However, an
association between greater volume and demonstrable
improvement in outcomes was not found .
Such findings may have contributed to a governmental
decision to freeze new accreditations for physicians in
private practice in Switzerland. This legislation is based on
the assumption that demand-induced health care
spending may be cut by limiting the associated care offers, a
policy that initiated a controversial public debate in
Switzerland. Therefore, the purpose of this article is to
provide estimates of resource utilisation in Swiss primary
care and to determine relevant factors in this context.
The Swiss health system is based on principles of free
demand and supply and most services are reimbursed on
a fee-for-service system. The system provides
comprehensive coverage of high quality services, however, the
financial costs rank second (after the US) among OECD
countries and are still rising. Swiss health care is
characterized by federalist structures and is fragmented into 26
cantonal health systems[4,5]. National authorities have little
responsibility and nationwide reforms aimed at cutting
health care expenditures are therefore difficult to achieve.
Due to this highly decentralized governing of Swiss health
care, the specific goals of the study include the estimation
of spatial variability of availability and utilisation of
primary care resources, taking into account demographic,
socioeconomic, and cultural attributes of the population.
One of the problems associated with an estimation of the
regional availability of health care is that patients
frequently travel outside their residential areas to seek care,
especially those residing in remote and non-urban areas.
The formation of meaningful health service areas as a unit
of analysis is thus an important issue as it allows the
control of demand-induced utilisation phenomena,
including medical consumerism. We have defined health care
service areas for Swiss primary care based on the complete
claims data of compulsory health insurance obtained
from the data pool of all Swiss health insurers
(santesuisse) and used these areas as the units of analysis. We
therefore applied a new and alternative spatial model to
describe the utilisation and provision of primary health
care in Switzerland.
1018 ambulatory care service areas were created according
to Goodman et al. based on 2761 communities and the
community area codes of primary care physician practices
and the residential addresses of their patients. For each
service area, socio-demographic attributes were calculated
using community data available from the Swiss Federal
Statistical Office. The localisation index (LI) and the
market share index (MSI) were calculated as measures of the
health care utilisation and provision in each area. The LI
indicates the degree of localization of primary care
provided for the population in a given area and the MSI
indicates the degree of localization of primary care provision
from the provider's perspective .
Complete frequency data of all consultations reimbursed
by the compulsory basic health insurance were obtained
from the data pool of all Swiss health insurers
(santsuisse) for all Swiss ambulatory care physicians for the year
2004. These data consisted of two files. The first file
included a list of all Swiss ambulatory care providers
classified into 49 different medical specialties with board
certifications according to the Swiss Medical Association
(FMH), together with the area codes of the practice
locations. For this study, we used only consultations provided
by primary care physicians (FMH board certifications for
primary care, general internal medicine, general
practitioner without board certification, incorporated group
practice member). The second file included consultation
frequencies of each physician classified by gender, by 20
age groups, and by community of patient residence.
Patients' communities were therefore the smallest
observational unit of this study. We used only data from
consultations provided for illness and maternity.
Accidentrelated consultations were excluded due to different
health insurance coverage.
Descriptive analyses included the calculation of means
and medians, depending on the distribution of the data.
Univariate associations between variables were assessed
with Spearman correlation coefficients (rho). Statistical
models had to account for the two-level hierarchical
structure of the data; i.e. level I data at the level of 2762
community levels and level II data at the level of 1018 service
areas; consequently, multilevel models were used.
The main outcome variable of these models was the
annual per capita consultation rate with a primary care
physician for each community in the year 2004.
Preliminary analyses indicated a symmetrical and almost normal
distribution of this outcome variable. The following
explanatory variables were analysed:
Level I (community attributes)
- Average age of the population
- Female-to-male ratio
- Language code of the community (German, French,
- Proportion of non-Swiss citizens among the residents in
- A nine level classification of communities: Urban,
suburban, high income, peri-urban, tourist, industrial-tertiary,
rural-commuter, agricultural-mixed, agricultural rural.
This typology was developed by the Swiss Federal
Statistical Office in order to provide a meaningful representation
of communal characteristics for demographic and
socioeocomic investigations . The classification is mainly
based on the principle of hierarchic relationships between
urban centres and peripheral communities.
Level II (service area attributes)
- Area code of the service area (clustering variable)
- Number of primary care physicians per 10'000
inhabitants (GP density)
- Number of other, non-primary care physicians per
10'000 inhabitants (specialist density) in the same service
- Presence of a hospital providing ambulatory services in
the same service area (coded as a binary variable with 0 for
no hospital and 1 for at least one hospital).
Additionally, random effects variables for the intercept,
average age, and female-to-male ratio were included in the
model. The decision to include these random effects
variables was based on preliminary analyses of heterogeneity
of intercepts and slopes. The structure of the
variance-covariance matrix was specified as unstructured. Other level I
variables such as the income distribution within
communities and the ratio of > 65 vs. < 20 years of age also were
analysed initially. However, based on the extent of the
respective variance components and the criteria of model
fit, the set of level I variables mentioned above appeared
to account for most of the outcome variance and provided
the best fit between observed and expected data. A further
analysis indicated significant first-order interactions
between GP density and language region, between
nonSwiss population and language region, and finally
between female-to-male ratio and non-Swiss among the
resident population. The corresponding interaction terms
were thus also included in the model. Consequently, the
final model used in this study had the following structure
(fixed effects are denoted with ij, random effects with ij
and the redidual error is denoted with rij):
Yij = 00 + 01 GP densityj + 02 specialist densityj + 03 hospitalj
+ 10 agei + 20 FM-ratioi + 30 non-Swissi + 40 languagei + 50
community-typei + 11 GP densityj *languagei + 60 non-Swissi
*languagei + 70 FM-ratioi *non-Swissi + 0j + 1j agei + 2j
MF-ratioi + rij
Additional language-stratified analyses were performed
because of significant language-related first-order
interaction terms. Since there are only a few communities in
which Romansh is spoken, stratified analyses were
restricted to German-, French-, and Italian-speaking
communities only. The stratified models comprised the same
set of explanatory variables except language region and
the related interaction terms. Presence of hospital,
language region, and community type were treated as
classification variables, and continuous explanatory variables
were centred to facilitate parameter interpretation. Results
for classified data were interpreted as least-square means
(LS-Means) with 95% confidence intervals (CI95), and
the Bonferroni procedure was used to adjust for multiple
comparisons. Variance components of level II variables
were additionally calculated. Variance components
quantify the corresponding proportions of outcome
variation accounted for by specific variables and therefore
provide estimates about the relevance of regional
variables within Swiss primary care. Model fit was assessed
using residual analyses and the Akaike Information
Criterion (AIC) was used to compare several versions of the
model during model development. Residual analysis
showed no violation of basic assumption for linear
models. SAS 9.1 (SAS Institute Inc., Cary, NC, USA) and "proc
mixed" were used for all analyses and the level of
significance was set at 0.05 throughout the study.
The data obtained from the pool of Swiss health insurers
(santsuisse) included the complete records of the
21'413'299 consultations of all 6564 Swiss primary care
physicians with patients throughout Switzerland that
were reimbursed by compulsory health insurance in
2004. These primary care data accounted for 39.5% of all
ambulatory care physicians (i.e. primary care physicians
and specialists) and 56.5% of all ambulatory
consultations reimbursed by basic health insurance (consultations
provided by ambulatory departments of hospitals were
not included). The geographic distribution of physicians
and patient consultations across different community
types is given in Table 1.
Ambulatory care service areas were constructed from
community codes of physicians and patients[6,11] without
setting constraints on either population size or
localisation of utilizing or providing health services. 1018 service
areas emerged with at least one physician per area. Service
area-specific data on demographic and socioeconomic
attributes, and health care utilisation were calculated
based on the associated statistics of the included
communities (Table 2). These data indicate a wide range in how
primary care is provided and consumed across various
regions. Population data show a few urban areas with very
large populations that skew the corresponding mean to
Per capita consultation rate
Type of communitya
a Classification according to the Swiss Federal Statistical Office
the right. However, 75% of all service areas had less than
6750 inhabitants. Similar observations were also made
for the overall number of primary care physicians within
the areas (Table 2). 75% of all the areas had fewer than 5
physicians. Less skewed, but still highly variable, data
were observed for the measures of utilisation and
provision of primary care in terms of the consultation
frequencies LI and MSI. The most important variable in this
context was the annual per capita consultation rate with a
primary care physician, which showed an eleven-fold
variation across the regions. Consequently, this variable was
selected as the main outcome of the statistical procedures.
Positive and significant correlations were found between
regional population size and both LI (rho = 0.44) and MSI
(rho = 0.48), implying a higher localisation of utilisation
and provision of resources in highly populated areas.
The statistical procedures yielded significant effects for all
main explanatory variables except for the proportion of
non-Swiss citizens (Table 3). The model explained 36.4%
of the variation (pseudo-R2 statistics) of the annual per
capita consultation rates. The regional GP-density per
10'000 inhabitants was positively and significantly
associaProportion of non-Swiss citizens among residents
ated with the annual consultation rates in GP practices,
and increased by a factor of 0.10 for each additional GP in
a population of 10'000 inhabitants (Table 3). The density
of specialists in the same region was negatively associated
with the frequency of GP-consultations; i.e., the
GP-consultation rate decreased by a value of 0.01 for each
additional specialist in the population. A similarly significant
and "protective" effect was seen for the availability of
ambulatory services provided by hospitals (Table 4).
Residents in regions having ambulatory hospital departments
consulted their GP's 2.81 times per year, whereas residents
in "no hospital" regions had 2.99 visits. Higher average
age and the proportion of females in the population were
both positively and significantly associated with
GP-consultations, although no significant effect was seen for the
proportion of non-Swiss residents. However, the
interaction term between the female-to-male ratio and the
proportion of non-Swiss citizens was significant and
indicated high consultation rates in communities with a
high proportion of females and non-Swiss citizens.
Pair-wise comparisons of means indicated multiple
significant differences between language regions. Only the
consultation frequencies for German and Italian vs. Romansh
were not significantly different. Least square means
indicated furthermore that agricultural/rural and high-income
communities had the lowest and industrial/tertiary
communities had the highest frequencies (Table 4). All other
community types were closely scattered around the
overall Swiss mean of a 2.95 per capita consultation rate with
a primary care provider.
Significant first-order interaction terms indicated that the
effects of GP density and the proportion of non-Swiss
citizens were not constant across language regions.
Therefore, additional specific analyses stratified by language
(Tables 3 and 4) were performed. These data indicated
that language-specific effect estimates of the association
Table 3: Parameter estimates of continuous explanatory variables for annual per capita consultation rates with primary care
*significant parameter estimates (p < 0.05)
between GP density and consultation rates were lower
than the overall Swiss estimate. GP consultation rates
increase significantly by 0.06 for each additional GP in a
population of 10'000 inhabitants in the Swiss
Germanspeaking population, whereas considerably lower and
non-significant effects were observed in French- and
Italian-speaking regions (Table 3). Inverse patterns were seen
for the association between the GP-consultation rate and
the regional density of specialists. Specialists had a
negative effect on GP-consultation rates in all regions.
However, this effect was seen only in German-speaking regions
significantly associated with GP-consultations (Table 3).
The proportion of non-Swiss citizens appeared, in terms
of effect size, to be the most important factor among
continuous variables at the community-level for the Swiss
German population (Table 3). This implies that
GP-consultation rates increased significantly with higher
LSM 95% confidence limits
a Hospital providing ambulatory services in the same region.
Nonstratified, full model
Estimate 95% CI
Estimates of language-stratified models
German French Italian
tions of non-Swiss citizens by a factor of 2.08 for each
additional percent of non-Swiss citizens in a community.
Less prominent and non-significant effects were seen for
this factor in other Swiss language regions.
Variance component analyses showed that service areas
account for 39.4% of the total variance of
GP-consultations and imply considerable clustering of consultation
rates within areas (Table 5). Variance component analyses
at the the level of the service areas indicate that only 2.9%
of the explainable regional variation is accounted for by
the regional density of physicians and that 3.1% and
4.4%, respectively, are explained by specialist density and
the presence of a hospital providing ambulatory services
(Table 5). Language-specific analyses of variance
components show, however, that GP density is an important
factor in explaining GP consultation rates in Swiss German
populations, but not in French and Italian regions.
Inversely, in French-speaking regions, specialist density
and hospitals accounted for considerable amounts of
regional variation. No or no estimable effects were
observed in this context for Italian-speaking regions
There are several theoretical models in health systems
research that are used to explain the availability and
utilisation of health related resources [12-14]. Most of these
models distinguish between supply- and demand-related
factors, and this conceptual differentation was adapted,
therefore, for the following discussion.
Regulation of physician's supply
Swiss health care is characterised by a high degree of
decentralisation with strong regional and cantonal
influence and high cost . Coordinated and effective
interventions to maintain and improve cost-efficiency are
thus difficult to achieve. However, as an extraordinary
measure, federal legislation in the year 2000 gave the 26
cantonal authorities the authority to restrict the number
of ambulatory health care providers. This decision was
a intraclass correlation coefficient
b negative variance components
Proportion of regional variation
based mainly on an economic rationale regarding the
supply hypotheses of practice variation and should have
resulted in a more or less complete stop of new
accreditations of Swiss primary care providers in the following
years. Apparently, these governmental measures provoked
just the opposite effect as the number of newly accredited
physicians increased ! The association between
physician density and the intensity of utilisation of health care
resources thus remains a relevant factor in the current
debate about reducing costs in Swiss health care. In
contrast to previous analyses based on cantonal data, we
propose the use of health service areas as the unit of analysis
of potential determinants of health care utilisation. This
approach reduces the effect of patients seeking care
outside their residential area. As this paper demonstrates,
without this bias a much better representation of the
actual utilization and provision of resources in Swiss
primary care can be achieved.
Determinants of utilization of primary care resources
The analysis of per capita consultation rates across
language regions confirms well known cultural differences in
Switzerland that extend beyond health care ; but the
differences we observed between the French and other
language regions were larger than expected. Other
research in Switzerland showed that these differences are
neither related to variations in health status nor to
socioeconomic attributes of the respective populations and,
furthermore, the same research also indicated a lack of
theoretical specification and empirical models in this
context . We therefore can only assume that patients in
French-speaking communities directly seek care more
often from specialists without first consulting a primary
care provider. Similar mechanisms may also be at least
partially responsible for the observed disparities between
the different community types. However, traditionally
anchored behavioural patterns and effects of the political
and institutional environment including different
accessibility of resources can additionally influence the
consultation pattern of the underlying populations.
Additional research aimed at these topics is thus needed
to investigate the distribution of population-based
consultation patterns across language regions, community
types, medical specialties, and the corresponding
implications on cost for ambulatory care. Presently, we also see
strong political pressure to implement managed care, e.g.,
by establishing physician networks based on
gate-keeping. Our results may therefore help guide future
developments by setting population-based and regional priorities.
The data provide estimates of the importance of
sociodemographic factors relevant for the utilisation of primary
care resources in Switzerland. Effects of age and gender
confirm well known phenomena in this context.
However, particularly in the Swiss German population, the
effect of non-Swiss residents far exceeds corresponding
effects of other socio-demographic and supply-related
factors. These results thus confirm other research in this field,
showing a heterogeneity of emigrants across Swiss
language regions in terms of gender, ethnical background,
migration history and legal status of residency [19,20].
Possible explanations of these findings may be provided
within the conceptual framework of cultural
epidemiology where health and illness are understood as
socio-cultural categories that influence the perception of well being
. The findings support, however, ongoing efforts
facilitating information, prevention and health care for
migrants and improving the cultural competence of care
providers . In addition, the findings may also be
relevant to many other health care systems facing worldwide
refugee and migration phenomena.
Our data provide evidence that the number of physicians
accredited in a specific region is an important determinant
for the utilisation of Swiss primary care resources.
However, the associated effects are not consistent across Swiss
language regions and show an inverse relationship
between primary care providers and specialist care. The
findings not only imply considerable competition for
patients between specialists and primary care physicians
in the Swiss German regions, but also indicate that the
utilisation of primary care resources in these regions is
obviously more supplier-driven than in other regions. In
contrast, effects of hospital-based supply of ambulatory
services remain constant across language regions and
indicate consistent patterns of competition: the presence of
ambulatory hospitals reduces GP consultations by 0.18
per year and inhabitant.
The question remains, therefore, what are the major
driving forces for these findings? Is it supplier-induced
demand, clinical uncertainty, or patient needs; and what
are the roles of the 'zero risk society' or of medical
consumerism?  The length of the observation interval
may exclude simple fluctuating levels. It may be argued
that the available data provide no insight into the
decision-making process of individuals seeking and providing
care, and consequently that the data do not permit a
distinction between demand manipulation and clinical
ambiguity . However, our data cover the entire range
of non-accident-related primary care reimbursed by social
health insurance, including non-clinical incentives and
competition factors. These factors together may influence
professional activities of physicians working on a
fee-forservice basis. Nevertheless, supply-induced demand
cannot be excluded as an explanation of our findings,
particularly in Swiss German regions. However, variance
component analysis reflected not only the decentralized
structure of Swiss health care but also provided an
estimate of the potential importance of regulatory
interventions. From a national perspective, supply-related
components contribute only marginally to regional
variation in the use of primary care resources. Therefore, the
impact of a regulation of supply by, for example, freezing
the number of primary care physicians may have only a
very limited or even immeasurable effect on the
associated resource utilisation. However, language-stratified
analyses indicated that considerable amounts of variance
are accounted for by the GP-density in Swiss German
regions, and specialist and hospital supply in
Limitations and strengths
The results were derived from the claims database of all
Swiss health insurers and provide complete coverage of all
ambulatory or outpatient services reimbursed by
compulsory Swiss social health insurance in 2004. The spatial
model used in this study yielded service areas in which, on
average, 57% of the population used local health
resources, and/or physicians treated local patients in 61%
of all consultations. We therefore consider this approach
an appropriate spatial representation of utilisation and
provision of primary care in Switzerland. It can be
debated, however, whether a proportion of 36.4% of
explained variance of the outcome is enough to draw
relevant conclusions about the effect of physician density
and socioeconomic determinants on per capita
consultation rates. Our model was based on a set of only nine
explanatory variables and on an outcome with known
high random error. We consider it unlikely, given the type
and extent of data available, that the results are biased due
to measurement error. Furthermore, the explanatory
variables were derived from aggregated data at the community
and service area levels, and not from individuals the
interpretation of the results is thus limited to these two
distinct levels of aggregation and cannot be extended for
individual patients or physicians. Given these factors, we
believe the achieved amount of explained variance merits
publication of the data. Furthermore, the distinct
multicultural population mix in Switzerland offers excellent
conditions for studying health care use across different
cultures. A potentially significant imitation in interpreting
our data on resource use and volume in Swiss primary
care is related to the fact that the referral data are still
incomplete. A further limitation derives from the fact that
the activity levels of the physicians were not accounted for
in the analysis, i.e., full-time and part-time physicians
were equally weighted. Other limitations attach to the
cross-sectional nature of the study, since service areas are
not stable constructs over time because the availability
and utilisation of resources may change rapidly. The
methods applied in this work are not perfect, but they
may be helpful in describing, understanding, and
managing the underlying processes.
Our study confirms other reports documenting large
small-area variations in the utilisation and provision of
health care resources in the multicultural population of
Switzerland [15,23]. In contrast to other research, we
propose an alternative spatial model for analysing resource
utilisation in order to provide comparable,
utilisationbased service areas. We quantified the effects of health
system factors, including physician density, as key elements
of per capita consultation rates in Swiss primary care.
Physician density appeared to be a relevant factor associated
with resource utilisation. The effects of primary care
providers and specialists were, however, different across
language regions, and hence, cultural background. Only
limiting the number of care providers may not bring
significant changes in volume. Therefore, additional
measures to reduce unexplained variation in volume, such as
providing physicians with data on their resource use
compared with clinical reasoning based on evidence-based
guidelines and the practice patterns of their peers, may be
evaluated further. Early experiments using educational
forms of feedback proved successful in changing
physician behaviour and the associated costs in the long run
AB obtained the mandate to perform the study, he
performed all statistical analyses and wrote the first draft of
the manuscript, BK reviewed and completed the
manuscript with reference to all aspects of primary care.
The study was funded by the Swiss Health Observatory and, by
contractural agreement, the funding organization had no involvement in the
preparation of the present manuscript.
We acknowledge the work of Dr. H. Jaccard and M. Roth of the Swiss
Health Observatory and M. Bertschi of santesuisse for their help and
support in the project, and we thank Chris Ritter for his help preparing the
1. Wennberg JE , Fisher ES , Skinner JS : Geography and the debate over medicare reform . Health Aff (Millwood) 2004 : W96 - 114 .
2. Fisher ES , Wennberg DE , Stukel TA , Gottlieb DJ , Lucas FL , Pinder EL : The implications of regional variations in Medicare spending . Part 2: health outcomes and satisfaction with care . Ann Intern Med 2003 , 138 ( 4 ): 288 - 298 .
3. 832. 103 S : The Swiss Federal Council: Verordnung ber die Einschrnkung der Zulassung von Leistungserbringern zur Ttigkeit zu Lasten der obligatorischen Krankenpflegeversicherung . 2002 .
4. OECD/ WHO: OECD and WHO survey of Switzerland's health system . Organisation for Economic Co-operation and Development ; 2006 .
5. Herzlinger RE , Parsa-Parsi R : Consumer-driven health care: lessons from Switzerland . Jama 2004 , 292 ( 10 ): 1213 - 1220 .
6. Goodman DC , Mick SS , Bott D , Stukel T , Chang CH , Marth N , Poage J , Carretta HJ : Primary care service areas: a new tool for the evaluation of primary care services . Health Serv Res 2003 , 38 ( 1 Pt 1 ): 287 - 309 .
7. Klauss G , Staub L , Widmer M , Busato A : Hospital service areas - a new tool for health care planning in Switzerland . BMC Health Serv Res 2005 , 5 ( 1 ): 33 .
8. Singer J , Willett JB : Applied longitudinal analysis New York : Oxford University Press ; 2003 .
9. Schuler M , Dessermontet P , D J : Die Raumgliederungen der Schweiz: Swiss Federal Statistical Office ; 2005 .
10. Singer J : Using SAS PROC MIXED to fit multilevel models, hierarchical models, and individual growth models . Journal of Educational and Behavioral Statistics 1998 , 24 ( 4 ): 323 - 355 .
11. Wennberg J , Gittelsohn : Small area variations in health care delivery . Science 1973 , 182 ( 117 ): 1102 - 1108 .
12. Andersen RM : Revisiting the behavioral model and access to medical care: does it matter ? J Health Soc Behav 1995 , 36 ( 1 ): 1 - 10 .
13. Grossman M : The Human Capital Model of the Demand for Health : Newhouse; 2000 .
14. McGuire T : Handbook of Health Economics Physcian Agency , Chapt 9 Volume 1. Elsevier Science B.V; 2000 .
15. Crivelli L , Filippini M , Mosca I : Federalism and regional health care expenditures: an empirical analysis for the Swiss cantons . Health Econ 2006 , 15 ( 5 ): 535 - 541 .
16. Marty F : Praxiserffnungsinserate in der Schweizerischen rztezeitung von Mitte 1997 bis Ende 2004 . Teil 1. Primary Care 2005 , 5 : 455 - 458 .
17. Vatter A , Refli C : Do political factors matter for health care expenditures? A comparative study of Swiss cantons . Jnl Publ Pol 2003 , 23 ( 3 ): 301 - 323 .
18. Bisig B , Gutzwiler F : Gesundheitswesen Schweiz: Gibt es Unter-oder berversorgung? Band 1: Gesamtbersicht Volume 2 . Verlag Ruegger Zrich/Chur; 2004 .
19. FOPH: Forschung Migration und Gesundheit Swiss Federal Office of Public Health ; 2006 .
20. Gilgen D , Maeusezahl D , Salis Gross C , Battegay E , Flubacher P , Tanner M , Weiss MG , Hatz C : Impact of migration on illness experience and help-seeking strategies of patients from Turkey and Bosnia in primary health care in Basel . Health Place 2005 , 11 ( 3 ): 261 - 273 .
21. Fugelli P : The Zero-vision: potential side effects of communicating health perfection and zero risk . Patient Educ Couns 2006 , 60 ( 3 ): 267 - 271 .
22. Davis P , Gribben B , Scott A , Lay-Yee R : The "supply hypothesis" and medical practice variation in primary care: testing economic and clinical models of inter-practitioner variation . Soc Sci Med 2000 , 50 ( 3 ): 407 - 418 .
23. Nocera S , Wanzenried G : On the Dynamics of Physician Density; Theory and Empirical Evidence for Switzerland . Research Papers in Economics 2002 [http://ideas.repec.org/p/ube/dpvwib/ dp0208.html].
24. Winkens RA , Pop P , Grol RP , Kester AD , Knottnerus JA : Effect of feedback on test ordering behaviour of general practitioners . Bmj 1992 , 304 ( 6834 ): 1093 - 1096 .
25. Bradley C : Provision of cost information changes physician behaviour and reduces costs . Evidence-Based Medicine 1998 , 3 ( 2 ): 62 .