Regional aspects of diagnosis and orthopedic use for back pain

>> What has so far not been sufficiently taken into account are regional differences in socio-demographic influencing factors: For example, there are no systematic studies at the individual level as to whether socio-demographic influencing factors such as age, gender, unemployment and citizenship have the same strong influence on the probability of diagnosis everywhere, or whether it is in the city -Country differences exist. In general, relatively little is known about the factors influencing orthopedic use when back pain is diagnosed, apart from studies with a comparatively small or very specific study population (Borys et al. 2013). Knowledge of risk and user groups, broken down in as much detail as possible, would be of great importance for the efficient planning of target group-specific treatment offers such as the disease management program “Chronic Back Pain”, which came into force on October 1, 2019, and for the potential avoidance of surgical interventions that are not absolutely necessary ( Casser et al. 2016).
While in many countries such as the USA, Great Britain and Canada, geographic information systems (GIS) and spatial-statistical methods are now used as standard in the planning of regional healthcare facilities (Cromley and McLafferty 2011), the use of GIS in the German healthcare system has only increased in recent years has become more important (Thißen et al. 2017). Much of the information contained in statutory health insurance (GKV) data, such as age, disease burden, type of insured person and availability of service providers, depends very much on the insured person's place of residence (Kauhl et al. 2016). The regionality of this data is a crucial feature: using regional statistical methods, it is possible to calculate in relative detail which group of insured persons in which regions have an increased diagnosis risk and an increased probability of utilization. In principle, treatment offers can be better adapted to the relevant target groups in the regions, thereby saving considerable costs.
This study examines the small-scale distribution of back pain and the associated orthopedic use and examines in more detail the regionally specific sociodemographic characteristics that are associated with the diagnosis and use of orthopedic services for back pain.
Data basis and methods
All AOK Nordost policyholders who were insured with AOK Nordost as of July 1, 2017 were included in this study. AOK Nordost insures 1.66 million people in the federal states of Berlin, Brandenburg and Mecklenburg-Western Pomerania. Back pain was defined as the presence of a confirmed diagnosis with the ICD code M54. A total of 478 thousand insured people had a confirmed diagnosis of back pain in 2017. To calculate the treatment rate, all insured people who received orthopedic treatment for back pain in 2017 were used (147 thousand).
For the analysis of possible influencing factors, various variables were included at the individual and aggregate level, which based on previous results are considered to be important influencing factors for back pain and orthopedic use. In addition to age and gender, many results indicate that back pain is more common in people with low socioeconomic status (Kuntz et al. 2017). Therefore, the insured types of unemployed and long-term unemployed were considered together as an indicator of low socioeconomic status (Kuntz et al. 2017). However, studies on the impact of citizenship on the presence of back pain are not clear (Bermejo et al. 2012; Razum et al. 2008). Against this background, citizenship was included in the study as a possible influencing factor. Some studies come to the conclusion that, in addition to sociodemographic factors, obesity also plays a decisive role in the development of back pain (Neuhauser, Ellert and Ziese 2005). For this reason, the presence of a confirmed diagnosis of obesity in at least 2 quarters was considered as an additional possible risk factor.
In addition to the presence of a low socioeconomic status at the individual level, the structural disadvantage (deprivation) of the living quarters was also taken into account, since it has been shown for many chronic diseases that the structural disadvantage of the place of residence can also represent an important influencing factor for the presence of chronic diseases (Kauhl et al. 2018b, 2018c; Maier 2017). For some - but not all - diseases, there is a connection between the availability of (specialist) doctors and a higher probability of diagnosis
observe (Kauhl et al. 2019). Therefore, for this study, the orthopedist density was included in sliding catchment areas of 30 minutes car travel time in accordance with the G-BA gravity model (own calculation) (Sundmacher et al. 2018). Building on this, car density as an indicator for mobility in rural areas was also considered as a possible influencing factor. The population data for calculating the gravity model and the data on car density were available from Nexiga at the level of the 8,850 residential areas in northeast Germany.
Cartographic representation of the prevalence
and treatment rates
The standard population of Germany was used to map the age- and gender-standardized prevalence and treatment rate. Since the number of residents and insured people also differs between the residential areas, a Bayesian smoothing procedure was used, which always weights the respective rates towards the average of the neighboring residential areas. This means that the respective rates are stabilized by information from neighboring residential areas and random fluctuations based on small numbers of cases are balanced out. To put it simply, the smoothing process can be used to represent rates on a very small scale with the statistical reliability of higher aggregation levels (Besag, York and Mollie). Only through this process does a cartographic representation of small-scale prevalence and treatment rates become meaningful. The results of the smoothing process were exported and mapped in ESRI ArcGIS 10.6. To calculate the diagnosis prevalence, the ratio of age- and gender-standardized sick insured persons to all insured persons was calculated. To represent the treatment rate, the ratio of age- and gender-standardized insured persons who received orthopedic services from an orthopedist to all insured persons with a confirmed diagnosis of back pain was used.
Statistical analysis
Since regionally very detailed data was available for this analysis at both the individual and aggregate levels, two regression models were chosen that explicitly take regionality into account: The first model is a Bayesian global logistic regression model, which determines a regression coefficient per variable for northeast Germany as a whole. In the global model, regional effects were taken into account through the neighborhood relationships of the residential areas. This means that, according to Tobler's first law of geography, living areas that are closer are more similar to each other than living areas that are further away (Sui 2004) and this fact must be explicitly taken into account in a regression model. The second regression model - a Bayesian spatially varying coefficient model - creates a local regression coefficient for each region. These regression coefficients can be mapped. What is of interest to this study is, for example, whether unemployment is a risk factor for the diagnosis of back pain everywhere, or whether regional differences exist. In order to be able to analyze age as an influencing factor regionally, the age of the insured was divided into two categories: Older than 50 and younger than 50, since the average age of AOK Nordost insured people for 2017 is around 50 years. The dependent variable was the presence of a confirmed diagnosis of back pain (yes/no) or whether orthopedic treatment for back pain had been received from an orthopedist (yes/no). The calculation of the smoothing procedure and the regression models was carried out using the R package INLA (Rue, Martino and Chopin 2009). The results were presented using ESRI ArcGIS 10.6.
Results
Regional distribution of the prevalence and treatment rate of back pain
The age- and gender-standardized prevalence of diagnosed back pain was 26.2% overall in 2017 and is therefore within the range of previous results (Henn et al. 2014). The highest prevalence is in West Berlin and some communities in Mecklenburg-Western Pomerania. In Brandenburg the prevalence is predominantly below average. According to standardization for age and gender, the proportion of insured people receiving orthopedic treatment for back pain out of all insured people diagnosed with back pain was 19.9%. The highest treatment rates are in West Berlin and the districts of Uckermark, Oder-Spree and Spree-Neiße.

Factors influencing the diagnosis of back pain
With each additional year of life, the risk of diagnosis increases by 2.5%. Women have a 37% higher risk of diagnosis than men. Insured persons with foreign citizenship have a 31% higher risk than insured persons with German citizenship. Our results make it clear that unemployed insured people have a 68% higher risk of diagnosis than non-unemployed insured people. The presence of a confirmed diagnosis of obesity (M2Q) is also significantly associated with the diagnosis of back pain: Insured people who have been diagnosed with obesity have a 74% higher risk of being diagnosed with back pain than insured people without diagnosed obesity. We were unable to determine any significant connection to the structural disadvantage of the place of residence (deprivation) for the diagnosis of back pain. Although our study shows a significant connection between the diagnosis of back pain and the density of orthopedic surgeons, the effect is relatively small. If the density of orthopedists between regions increases by one orthopedist per 100,000 inhabitants, the risk of diagnosis increases by 0.9%.
Factors influencing the use of orthopedics for back pain
Overall, only a significant connection can be found between orthopedic use and the characteristics (older) age, female gender, diagnosed obesity and density of orthopedists. With each additional year of age, the probability of utilization increases by 0.9%. Women are 27% more likely to see an orthopedist when diagnosed with back pain than men. If obesity is diagnosed, these insured people are 21% more likely to make use of it than insured people without obesity. The influence of the density of orthopedists on utilization is relatively pronounced: with each more orthopedist per 100,000 inhabitants between the regions, the probability of utilization increases by 8.9%.

Region-specific factors influencing the diagnosis of
Back pain
All observed variables associated with back pain show regional differences in the strength of the association. Insured people with foreign citizenship only have an increased probability of diagnosis in Berlin. In Berlin-Tempelhof and Friedrichshain-Kreuzberg in particular, insured people with foreign citizenship have a more than 30% higher probability of diagnosis than insured people with German citizenship. The influence of individual unemployment is positive almost everywhere. However, unemployed insured people in Berlin and Schwerin have a 40% higher risk of diagnosis than non-unemployed insured people, while unemployed insured people in the Uckermark, for example, only have a 10% higher risk. Overall, insured people over 50 have an increased risk of diagnosis compared to insured people under 50. However, insured people over 50 years old in Berlin and parts of the Prignitz district have a diagnosis risk that is more than 3 times higher, while insured people over 50 years old in the Ostprignitz-Ruppin district only have a risk of up to 1.8 times as high. Across northeastern Germany, women have a higher risk of being diagnosed than male insured people. However, pronounced regional differences can also be identified here: In the Oder-Spree district, women have a diagnosis risk that is more than 56% higher than men, while women in the Mecklenburger-Seenplatte district, for example, only have a diagnosis risk that is up to 40% higher. The presence of diagnosed obesity is also a common risk factor for the diagnosis of back pain. Berlin particularly stands out here: insured people with diagnosed obesity have a diagnosis risk that is more than twice as high as insured people without diagnosed obesity. The influence of orthopedist density is not positive everywhere; Only in the more rural regions of Brandenburg and Mecklenburg-Western Pomerania does the individual diagnosis risk increase by 3-4% with one more orthopedist per 100,000 inhabitants. In Berlin and parts of Brandenburg, a slightly negative connection can even be observed.

Regionally specific factors influencing the use of orthopedics for back pain
The variables age: over 50, gender: female, diagnosed obesity and orthopedic density show very different regional relationships to orthopedic use. Although the connection between the age group over 50 and the use of orthopedics is positive almost across the board, the strength of the connection shows very strong regional fluctuations. While in the north-east of Mecklenburg-Western Pomerania and in the south and south-east of Brandenburg, insured people over 50 years of age with diagnosed back pain are more than twice as likely to be used compared to those under 50, the same group of insured people in some regions such as the Oberhavel district only up to 25% higher probability of use compared to those under 50 years old. In almost all regions, female insured people are more likely to make a claim than male insured people. However, the strength of the connection is subject to strong regional fluctuations: in the north-east of Mecklenburg-Western Pomerania and in the south and south-east of Brandenburg, female insured people are more than 36% more likely to make a claim than male insured people. In other regions, such as the southern part of the Mecklenburger-Seenplatte district, female insured persons, for example, are only up to 20% more likely to see an orthopedist if they are diagnosed with back pain. The presence of diagnosed obesity is also a widespread influencing factor for the use of an orthopedist. However, there are again strong regional differences: While insured people with diagnosed obesity in the districts of Vorpommern-Greifswald, Schwerin, Ludwigslust-Parchim and Teltow-Fläming have a more than 40% higher probability of utilization than insured people without diagnosed obesity, the same group of insured people in others Regions such as in some parts of Berlin
and the southern part of the Oberhavel district only have a 10% higher probability of utilization. Although the density of orthopedists is - as expected - positively associated with the use of orthopedics in a large part of the regions, there are again regional fluctuations: While in Berlin, the southern and northeastern part of Brandenburg, the probability of use increases by up to one more orthopedist per 100,000 inhabitants increased to 9%, a slightly negative relationship can even be observed in some regions of Northern Brandenburg and Mecklenburg-Western Pomerania.
discussion
In contrast to some chronic diseases such as
for example hypertension (Kauhl
et al.), the results indicate very small-scale differences in the prevalence and use of orthopedics for diagnosed back pain.
With regard to the diagnosis of back pain, the results essentially confirm the results of previous studies: older age, female gender, unemployment, foreign citizenship and the presence of diagnosed obesity are important influencing factors of back pain (Raspe 2012; Kuntz et al. 2017). Although a connection between back pain and low socioeconomic status has been documented in the scientific literature, based on the results we can only document a connection with individual unemployment. Other factors such as level of education and income at the insured person level may also be relevant, but this information is not available. While the influence of structural disadvantage in the place of residence (deprivation) can be considered certain for a large number of chronic diseases (Kauhl et al. 2018a; Kauhl et al. 2018c), this cannot be proven based on our results for back pain. At the same time, the study makes it clear that the presence of a diagnosed obesity has a stronger influence on the diagnosis of back pain than the presence of individual unemployment - although it must be emphasized at this point that the presence of a diagnosis of obesity does not reflect the prevalence of obesity in the population : While it is assumed that 23.3% of men and 23.9% of women in the general population of 18 to 79 year olds are obese (Mensink et al. 2013), the age- and gender-standardized prevalence of diagnosed obesity is in the billing data of AOK Nordost at 12.4%.
With regard to the use of orthopedic services for diagnosed back pain, it can be seen that at the level of individual insured person characteristics, only (higher) age, gender and the presence of diagnosed obesity are significant predictors of the use of orthopedic services for back pain. At an aggregate level, only a connection could be determined between the density of orthopedists and the use of orthopedic services.
With the help of detailed regional statistical methods, it was made clear that the socio-demographic characteristics of insured persons that are associated with the diagnosis and use of orthopedics for back pain depend heavily on the respective region. Especially when it comes to the indicators of foreign citizenship, individual unemployment and diagnosed obesity, it is clear that insured people with these characteristics have a significantly increased risk of being diagnosed, especially in Berlin. Our study makes it clear that a target group-specific treatment offer such as the DMP Back Pain in Berlin should potentially address a different insured structure than in the rural regions of Brandenburg and Mecklenburg-Western Pomerania. It would be interesting to ask at this point whether insured people with these characteristics in Berlin actually have an increased risk of back pain from an epidemiological perspective (for example due to heavy physical activity), or whether the regionally different coefficients are just the result of different levels of utilization behavior. However, the regionally varying strength of the connection to the density of orthopedists must be viewed against the background of the lack of information in the doctor master data: information on the average waiting time, equipment and specialization of the orthopedist is not available. Consequently, these regional differences may need to be understood as a consequence of this disregarded information. For example, in regions with a negative connection to the density of orthopedic surgeons, waiting times could be correspondingly higher, or in these regions there is a specialization in body regions other than the back. At the same time, the results reveal a strength of the methodology used in this study: only through the use of regionalized regression models could regionally different utilization patterns for the diagnosis and orthopedic treatment of back pain be made visible. This means that the results can be used to plan treatment offers in northeast Germany that are tailored to local needs.
Supply-induced demand has long been assumed in Germany, which is held responsible for a shortage and oversupply. For example, in regions with a high density of orthopedic surgeons, fewer surgical procedures are carried out for muscoskeletal diseases (Pollmanns,
Wesermann and Drösler 2018; Schäfer, Pritzkuleit, Jeszenszky et al. 2013). Accordingly, it can be assumed that orthopedic practices tend to treat patients conservatively over a longer period of time before an operation is recommended (Schäfer, Pritzkuleit, Hannemann et al. 2013). Our analysis underlines this finding in that we were able to demonstrate a positive relationship between orthopedist density and orthopedic utilization for a large part of the regions based on our data. At this point, it would be useful to compare the data from the inpatient sector on the frequency of operations for muscoskeletal diseases in order to find out in more detail whether, conversely, operations are performed more frequently in regions with a low density of orthopedic surgeons.
With regard to the requirements planning of the orthopedists and surgeons, who have been merged into a group of specialists in the requirements planning guidelines since January 16, 2019, the question arises as to whether this merger may lead to a reduction in the range of conservative measures in favor of surgical - and therefore more expensive - measures ( Medicine 2019), although surgical measures do not necessarily lead to a better result than conservative measures for certain indications (Schäfer, Pritzkuleit, Hannemann et al. 2013). However, this question cannot currently be answered using the regional regression models used here.
Conclusion
This study illustrates the very small-scale differences in the prevalence and treatment rate of back pain in northeast Germany. The results of the regionalized regression models also show that the risk of diagnosis and the probability of using orthopedics for back pain vary significantly regionally within the risk groups. Our study therefore provides evidence of regionally different utilization patterns for the diagnosis and orthopedic utilization of back pain. Our analysis thus provides a foundation for adapting orthopedic care even more closely to the specific risk and user groups in the regions. Our results also make it clear that the prevalence of back pain and the resulting use of orthopedists is partly a supply-induced prevalence and demand. However, it is not possible to make a statement about possible incorrect or under-supplied care based on our results. For this purpose, a comparative analysis with the data from the inpatient sector would be necessary in order to be able to comprehensively assess the care situation. <<