Analysis of patient flows for orthopedic procedures using small area analysis in Switzerland

Background In general cantons regulate and control the Swiss health service system; patient flows within and between cantons are thereby partially disregarded. This paper develops an alternative spatial model, based upon the construction of orthopedic hospital service areas (HSAOs), and introduces indices for the analysis of patient streams in order to identify areas, irrespective of canton, with diverse characteristics, importance, needs, or demands. Methods HSAOs were constructed using orthopedic discharge data. Patient streams between the HSAOs were analysed by calculating three indices: the localization index (% local residents discharged locally), the netindex (the ratio of discharges of nonlocal incoming residents to outgoing local residents), and the market share index (% of local resident discharges of all discharges in local hospitals). Results The 85 orthopedic HSAOs show a median localization index of 60.8%, a market share index of 75.1%, and 30% of HSAOs have a positive netindex. Insurance class of bed, admission type, and patient age are partially but significantly associated with those indicators. A trend to more centrally provided health services can be observed not only in large urban HSAOs such as Geneva, Bern, Basel, and Zurich, but also in HSAOs in mountain sport areas such as Sion, Davos, or St.Moritz. Furthermore, elderly and emergency patients are more frequently treated locally than younger people or those having elective procedures. Conclusion The division of Switzerland into HSAOs provides an alternative spatial model for analysing and describing patient streams for health service utilization. Because this small area model allows more in-depth analysis of patient streams both within and between cantons, it may improve support and planning of resource allocation of in-patient care in the Swiss healthcare system.


Background
Since January 1997, all Swiss hospital discharges are collected yearly in the Swiss Federal Statistical Office's medical statistics of stationary institutions. Each discharge record is labelled with a residence code called medstat, which is an aggregate of several postal code areas. Each medstat region has 3'500 -10'000 inhabitants and is created according to socio-economic and geographic coherence criteria. Switzerland is divided into 612 medstat regions, of which 240 contain at least one hospital [1].
After a start-up period, from 2000 on the data collection may be considered complete. By means of these data, an exact inventory of the status of the Swiss health care supply and hospital usage can be established. In addition to a traditional analysis based on cantons, studies based on hospital service areas (HSAs) can be performed. HSAs are aggregates of medstat regions in which at least one medstat region with at least one hospital is represented. Their definition is based on the small area analysis methodology described for health service research [2,3].
The segmentation of Switzerland into HSAs offers a meaningful spatial model that enables more detailed examination of stationary hospital services used by HSA residents and nonresidents. Affording insight into the geographical distribution of hospital usage [4][5][6][7][8], HSAs enable the description of variability in patient flows and measurement of the extent of local and nonlocal (to an HSA) treatments. Several indices that describe patient streams can help identify areas that attract and treat local or nonlocal residents, and HSAs allow more precise analysis of potential health supply shortages or overcapacities. Also, the focus of HSA studies can be sharpened to single medical disciplines (internal medicine, surgery, etc.), individual diagnoses (ICD10 [9]), specific treatments (CHOP-codes, a translation and adaptation of the US classification ICD-9-CM volume 3, [10]), one type of hospital (acute, rehabilitation), or applied to insurance-based accommodation type (private, semi-private, or general). Instead of using the previously defined general hospital service areas by Klauss et al, this paper defines orthopedic hospital service areas (HSA O s) that use Swiss orthopedic discharge data from 2000-2002. There were several reasons for defining orthopedic HSA. First the main focus of the research in our institute is on orthopedics. Secondary it is well known now that the federal discharge data from 1998 until 2000 were not as complete as the later data. The second data set ordered from the Swiss Federal Office of statistics obtained only orthopedic procedures for the years 2000-2002, but with much more patient information as the first data set. Third, because within the different time periods of the data sets, in which hospitals were closed, pooled together or newly opened based on a new health insurance law in Switzerland, it was decided to build new and for this study orthopedic specific HSAs.
The study describes patient flows within the country using a model based on small areas instead of larger administrative, cantonal areas. This focus on patient flows rather than on utilization was chosen as literature describing the use of small area analysis for analysing patient streams is hardly available. The study shows how certain partially newly defined patient stream indices can identify diverse, important characteristics of areas, and may assist in the formulation of hypotheses to explain variability in patient flows.

Definition of hospital service areas for orthopedic procedures
To construct hospital service areas for orthopedic procedures (HSA O s), a medstat of a person's residence (home medstat) and medstat where treatment took place (treatment medstat) must be available for each hospital discharge (as described in detail by Klauss [3] and Goodman [2]). Briefly, HSA O definition consists of 3 steps: ▪ For each home medstat region, the treatment medstats in which the first, second, and third highest number of discharges have taken place are calculated. For medstat regions with at least one hospital, home medstat and treatment medstat can be the same. Thereafter all medstat regions are allocated to the treatment medstat in which the highest number of discharges has taken place (primary HSA O ). Theoretically for Switzerland, there now may be 240 HSA O s as Switzerland has 240 medstat regions with at least 1 hospital.
▪ The medstat regions are displayed in the Swiss medstat map (ArcMap) according to the assigned treatment medstat with the highest number of discharges. HSA O s are then examined for violations of the plurality rule [3,11]. This means that if the sum of discharges in the secondand third-place medstats is much higher than the one for the assigned medstat, and the second and third-place medstat belong to the same HSA O "X", the medstat in question is also assigned to this HSA O "X". In addition, if medstats are geographically disconnected from their primary HSA O , they also may be assigned to an adjacent HSA O if this HSA O contains the second-or third-place medstat (continuity).  Additional variables in the dataset were age, admission type (emergency or planned procedures), and class of occupied bed. The class or accommodation type of the occupied bed is defined by the Swiss insurance system and means that a bed is covered by the respective insurance type. Both basic and semiprivate beds are provided in public hospitals, whereas private clinics only offer private insured beds. Insurance class of a bed does not necessarily relate to a patient's insurance coverage, for a patient may temporarily upgrade his/her insurance coverage to semi-private for hospitalization in a semiprivate bed, which offers higher comfort with fewer beds per room (max two beds per room for semiprivate, one bed per room for private insured beds). The data set does not contain information on the insurance coverage of the patient; therefore the class of occupied bed is used as a surrogate marker to describe the effect of insurance coverage on the indices.
At the time the study was undertaken only data from 2000-2002 were available. As the Swiss federal institute for statistics is very restrictive in making data available newer data for this study (although possibly available by now) where not ordered later on. In addition, the Swiss healthcare system and hospital politics underwent larger re-organisations during the past 10 years, including hospital closures, pooling or new openings, and therefore data including larger time periods may distort results and lead to false conclusions.

Statistical analysis
Statistical analyses where performed with SAS 9.1 ® (SAS Institute Inc., Cary, NC, USA), geographic representations of data using ArcGis (ArcView8.2 ® , ESRI, Redlands CA, USA). Preliminary analyses indicated that most outcome variables were derived from symmetrical distributions; therefore univariate, linear models were used to analyse these data and the Bonferroni procedure was applied for pairwise comparisons in case of significant overall F tests. Residual analyses were applied to assess the fit between observed and modelled data, and R 2 -values were used to estimate the amount of variance of outcomes accounted for by the models. Correlations of continuous indicators with skewed distribution were assessed with Spearman's rho. The significance level was set at p < 0.05 throughout the study.
For 6 HSA O s, including the large urban HSA O s Bern, Zurich, Aarau, and Fribourg, the localization index for pri-  Netindices are partially associated with the insurance class of the occupied bed ( Whereas no significant differences in LI, MSI, and netindex are observed between children and adults, as indicated by a higher average LI and MSI elderly patients seem in general to be served more locally than younger persons (< 65 years old) -an indication that elderly people tend to be less mobile than younger people.
The insurance class of occupied beds is clearly associated with the LI and netindex, with private beds showing the lowest average LI and highest negative net index. This may be due to the fact that patients demanding a private bed are more willing to travel outside their HSA O . When such a bed is not available in their own HSA O , they expect better treatment in another HSA O , or they seek treatment in a special clinic or demand specific treatments not provided in every hospital.
Netindices represent some typical Swiss geographic and socio-cultural settings. As a nation with large winter sport areas, HSA O s with a positive net index for emergency procedures are seen in the mountain areas in the Bernese Alps, Valais, and Graubünden. This may reflect the many leisure-related accidents involving non-HSA O residents that occur in these areas. In contrast, for planned procedures a positive net index is observed for the main large urban HSA O s of Lausanne, Basel, Bern, Zurich, and Lugano, as well as in the mountain HSA O s Sion and Davos.

Discussion
Small area analysis is a technique that uses large administrative data bases to study regional variation of healthcare resources and utilization. A persistent finding in health services research is that health care delivery and hospital utilization in the Western world vary widely between countries and between areas within countries [14,15]. This phenomenon is also observed in Switzerland. The reasons for geographic variations may be many -including the type of healthcare offered, socio-economic factors, and geographic influences -and have been discussed in detail by others [16][17][18]. Wennberg [18] concludes that the most plausible sources of variation in hospital utilization in small areas are organizational factors, including the distribution of resources, availability of beds, and balance between primary and secondary care. How relevant these sources of variation are to Switzerland cannot be determined from the descriptive data.
Although Switzerland's health care system is cantonally structured, patient streams may be influenced by multiple regulatory agreements among and between hospitals, communes, regions, and cantons. Discussions on the financing of hospitals currently are of great importance. For example, the "Neue Zürcher Zeitung" reported (2. Juni 2005) that the city of Zurich is demanding a reconfiguration of hospital regions because many non-city residents come to Zurich for treatment without their communities sharing the costs of health care. Small area analysis of hospital service areas can reveal new information and offer a new spatial model for hospital financing and planning.
Localization Index for all, emergency and planned orthopedic procedures Figure 4 Localization Index for all, emergency and planned orthopedic procedures. Legend: Geographic distribution of the LI, A = all procedures, B = emergency procedures, C = planned procedures The 3 introduced indices for patient streams allow the identification of areas with different characteristics -attractiveness, importance-of the provided hospital care and may guide health care planners in decision making processes. For example, although the method of creating HSAs implicates a tendency to regions with high LI, still HSAs may occur with LI < 50%. According to the method for building HSAs such a region has to be defined as a unique HSA, but still more than 50% of its residents are treated elsewhere in many different (surrounding?) HSAs. The implication of such a finding might be that the offered orthopedic supply in such an area should be questioned. Is it no longer needed or should the quality be revised? Furthermore, high positive netindices, not only draw the attention to HSAs with a centralized function (university or canton hospitals), but also -as in the case of Switzerland-to areas with special needs according to their geographic function (like regions within large ski areas). Netindex for all, emergency and planned orthopedic procedures Figure 5 Netindex for all, emergency and planned orthopedic procedures. Legend: Geographic distribution of the netindex, A = all procedures, B = emergency procedures, C = planned procedures flows. In addition, the data cannot discriminate hospital type (acute, rehabilitation, general, private, etc.). Rehabilitation clinics, for example, may attract many patients from outside their HSA O s and therefore should be excluded from the analysis when examining utilization of inpatient care. Furthermore, the underlying area model of medstat regions for building HSA O s was created using exclusively federal health statistics based on aggregated postal code areas; the aim was to achieve comparability of socio-demographic and geographic factors. How well such an aggregated area model represents the real world of patient streams remains unclear. Although postal codes might have yielded more precise estimates of the influence of patient utilization upon the size and shape of HSA O s, their use was precluded due to data confidentiality laws. Yet despite its aggregated nature, the medstat region may be considered as a reasonably valid area model for creating HSA O s. One last problem not considered in this paper is that HSA O definition should also include a criterion for minimum population size or number of discharges, because areas with a low population size may produce outlier values. However, it is unclear where to set a cut-off value, and the risk of combining HSA O s that do not belong together for geographic or demographic reasons might exist. The entire procedure of creating HSA O s would then become arbitrary and difficult to reproduce. Although individual HSA O values are treated as independent, in reality they are not; as Tobler's first law of geography states, "Everything is related to everything else, but near things are more related than distant things" [19]. HSA O s with central functions may automatically create low LI and negative netindices for their neighbouring HSA O s.

Implications
Geographic information systems increasingly may allow communities, health care services researchers, and health care policy makers to link data sources, thus highlighting areas where public health interventions can be applied. The definition of health service areas for orthopedic procedures and the display of patient flows between such service areas can be used to characterize areas for their attractiveness for the local and nonlocal population. HSA O s may thereby reveal possible needs and demands Market share index Figure 6 Market share index. Legend: Geographic distribution of the market share index which can help to assist hospital planning and policymaking.
Although these areas do not follow conventional municipal or cantonal borders, they may more closely represent the real world for hospital planning. Further investigations using overall discharge data for all procedures are on their way.

Conclusion
Switzerland can be divided into 85 orthopedic hospital service areas. These areas show large variations with regard to patient streams and a trend to centralized health care services seen in the large urban areas, some central mountain areas, and isolated border regions. Older patients in general seem to be served more locally than children and adults; so too are patients who occupy a general bed. Large mountain sport area HSA O s serve more non-HSA O residents for emergencies than areas not located in the mountains. Patient-stream analyses on a small area level can help to identify areas with higher and lower attractiveness and importance for local and nonlocal residents, and may aid health care policy makers in hospital planning procedures.