Regional differences in reasons for consultation and general practitioners’ spectrum of services – Results of a cross-sectional observational study

Among other factors, the patients’ consultation reasons and the GPs’ spectrum of services determine process and outcome of the medical treatment. So far, however, there has been little information on differences in reasons for consultations and performed services between urban and rural areas. Thus, the goal of the study was to investigate these factors in relation to the regional location of the GP practices. We conducted a cross-sectional observational study based on standardised GP interviews in a quota sampling design. All counties and independent cities within a radius of 120km around Hamburg were divided in three regional categories (urban, rural, environs) and stratified proportionally to the population size. Differences in the number of reasons for consultation and services were analysed by multivariate linear regressions in mixed models adjusted for random effects on the levels of the German federal states and administrative districts. Differences in individual consultation reasons and services were identified by logistic regression models via stepwise forward and backward selection.

Keywords: General practice, Reasons for consultation, General practitioners' spectrum of services, Regional comparison, Urban-rural differences

Background
In developed health care systems, there is a rational health care planning based on available data. But usually not all relevant data are used for this process. For example, in Germany, the regional associations of statutory health insurance physicians and the regional federations of the statutory health insurances are obliged to perform a mutual needs-related planning of the number of accredited physicians within the statutory health insurance system. In order to prevent over-and undersupply with physicians, ratios of the number of physicians in different specialties and the population size in specific planning areas are defined. These ratios can be adapted to special features of the region or the regional infrastructure, especially if other planning areas contribute to the health care in the respective area. Other criteria for deviating from these ratios are regional data on morbidity, sociodemographic parameters or socioeconomic status of the population [1].
On the one hand, according to these criteria, there are substantial regional disparities. For example, diabetes and obesity are more prevalent in rural areas [2], while depressive disorders play a slightly larger role in core cities [3]. Age groups differ largely between eastern and western Germany and between urban and rural areas [4]. Additionally, from a socio-economic perspective, there are inconsistent regional differences between urban and rural areas. For example, while highly qualified jobs are concentrated in the centres of settlement areas [5], there is at the same time markedly more spending capacity-adjusted income poverty in urban areas [6].
On the other hand, not all factors relevant for the utilisation of health care are covered in these criteria.
For example, barriers and enablers for access to health care differ between the regions. The preliminary qualitative study of the analysis presented here reported that GPs and patients in rural areas rather discussed long distances to the doctor's office, long waiting times and the GPs' large workload. In contrast, high patient fluctuation and fierce competition among the GPs were frequent topics of discussions in urban areas [7].
There are also differences between the regions in patient behaviour and the GPs' understanding and perception of their role. Patients in rural areas seemed to consult their GP relatively late in urgent consultation matters, while a certain fraction of the patients in cities already consulted their GP for banalities [7]. Furthermore, GPs in rural areas were more likely to identify themselves as family doctors and persons of trust for their patients while their role as medical service provider and referrer to specialists was more emphasised in urban areas [8].
Presumably, these aspects are linked to regional variations of the patients' reasons for consultation and the GPs' spectrum of services, which also determine process and outcome of the medical treatment. A study by Steinhäuser et al. showed clear differences between rural and urban GPs regarding 25 of the 42 most frequently carried out procedures (such as "ambulatory blood pressure monitoring" or "adolescent examinations") [9]. So far, however, there has been relatively little information on the difference in reasons for consultations and services between urban and rural areas.
Thus, the goal of the study presented here was to investigate reasons for consultations and services of GPs in relation to the regional location of their practices in Northern Germany.

Methods
The project "Outpatient Healthcare Research North (Ambulante Versorgungsforschung Nord -AVFN)" was designed as a cross-sectional observational study. Methods, research hypotheses and statistical models had been described in the study register ClinicalTrials.gov (NCT02558322) before starting the survey and in the published study protocol [10].
Three categories had been defined for the regional comparison according to the so-called "structural settlement of district types" of the German Federal Institute for Research on Building, Urban Affairs and Spatial Development [11]. The category "urban areas" included urban municipalities, the category "environs" urbanized districts and rural district with signs of agglomeration, and the category "rural areas" sparsely populated rural districts.
A quota sampling design was chosen to represent as many regionally different healthcare situations in the study area as possible. The purpose of this design was to raise the probability of also including underserved regions into the study where usually many GPs are unwilling to participate in a study due to their heavy workload. The goal was to recruit at least 80 GPs per regional category for the projectcategorised by individual administrative districts (counties and independent cities) and proportionally to the respective population size in the districts [10].
At first and for determining the survey area, all administrative districts were included in the study where at least 20% of the land area was located within a radius of 100 km (ca. 62 miles) linear distance around the study centre (University Medical Centre Hamburg-Eppendorf). As initially, despite regular follow-up actions, the desired sample size of 240 GPs could not be achieved, the radius around the study centre was expanded to 120 km (ca. 75 miles) according to the approach described in the study protocol [10]. The thus chosen administration districts for the study were derived from the German Federal States of Bremen, Hamburg, Mecklenburg-Western Pomerania, Lower Saxony, Saxony-Anhalt and Schleswig-Holstein.
GPs' recruitment was based on a database of the Department of General Practice at the University Medical Centre Hamburg-Eppendorf as well as on the databases of the respective regional associations of statutory health insurance physicians in the selected areas. GPs were eligible for the study if they had an accreditation as statutory health insurance physician in the respective administrative district and used an EDP system facilitating to draw up a list of all patients treated over the preceding quarter (3-month accounting period).
Eligible GPs were contacted in writing by mail and asked to participate in the study. Participating GPs were visited by a staff member of the project and personally interviewed. In rare cases, it had been impossible to completely finish the GP's interview during the office visit for lack of time. Regarding these cases, the interview was later completed by telephone.
The interviews were based on standardised questionnaires and contained information regarding the GPs' age, gender, workload, and postgraduate training as well as data on the practice. Additionally, 99 different reasons for consultations from 17 areas / organ systems and healthcare services involving 38 different procedures were documented by means of a standardised instrument developed on the basis of the International Classification of Primary Care (ICPC-2) [10; [12][13]. These data are described as the proportion of the respective consultation reason or the respective service of all consultation reasons or services documented at the respective practice. The GPs' survey was conducted from 12 June 2015 to 27 April 2017.
Regional differences were initially described using the chi-square test and t-test. The number of different consultation reasons and the GPs' service spectrum were subsequently analysed by means of multivariate linear regressions in mixed models adjusted for random effects on the levels of the German federal states and administrative districts within the federal states. In these analyses, the control variables were stepwise included in three statistical models. A potential improvement of the model fit was determined by the likelihood ratio test.
To be able to interpret these results better, significant differences in reasons for consultations and services between urban areas, environs, and rural areas were identified by logistic regression analyses via stepwise forward and backward selection with p ≤ 0.05 as inclusion and p ≤ 0.10 as exclusion criterion. Variables showing characteristic values of a lower frequency than 1% had been excluded in advance. Thus, 14 ICPC 2 areas / organ systems, 29 different reasons for consultations and 22 different services remained in the statistical models for selecting variables.
An alpha level of 5% (p ≤ 0.05) was defined as statistically significant for all analyses of the inferential statistics. Stata 15.1 was used for data preparation and data analysis.

Results
In our quota sampling design, we were able to adhere to the stratification rules defined in the study protocol. Figure 1 shows a map of the respective regions. Primary care practices in 34 of the 37 selected administrative districts (91.9%) could be represented in the dataset. No study participants could be recruited from three districts of the environs (Delmenhorst, Diepholz and Osterholz). Initially, we recruited 280 GPs from the selected regions. However, no interview could be carried out with 69 study participants due to time-related or organisational reasons (e.g., sick primary care partners, problems with the patient management software). In the end, we were able to personally interview 211 GPs.
Tables A1 to A3 in the additional files describe the recruitment process of the stratification groups. In total, 12% of the GPs in urban areas, 13% of the GPs in the environs and 25% of the GPs in the rural areas had to be recruited from other stratification groups of the respective regional category than originally planned.
The sociodemographic data, the number of treated patients and the professional qualification of the interviewed GPs are shown in Table 1. The GPs' average age was 54.5 years and 65.4% of GPs were male. On average, GPs saw 1,032 patients per quarter with a slightly higher number of patients in rural areas than in urban areas. In Germany, GPs can work individually in private practices (individual practice) or together with other physicians -either in private practices in which each physician submits his own claims (group practice) or in private practices in which the claims of all physicians are joined and submitted as one bill (joint practice). In our study, GPs mostly worked in individual or joint practices and less frequently in group practices or medical care centres ("Medizinische Versorgungszentren", MVZ). Individual practices were noticeable more frequently found in rural areas than in urban areas and in our sample, group practices including MVZ had been solely located in urban areas and the environs.
77.3% of GPs had postgraduate training as specialist in general practice and 23.2% had postgraduate training as specialist in internal medicine. The postgraduate training "specialist in general practice" was reported more often in rural areas than in the environs and the postgraduate training "specialist in internal medicine" was reported less often in rural areas than in the environs. 6.6% of GPs reported that they had not completed a postgraduate specialist training. On average, GPs reported 1.5 additional postgraduate trainings. The most frequent were in emergency care and palliative medicine. 31.8% of GPs had no completed additional postgraduate trainings. Table 2 shows the average relative rate of individual consultation reasons of all reported reasons for consultations. Most frequently, GPs described reasons for consultations from the ICPC 2 areas / organ systems "cardiovascular system", "musculoskeletal system" and "endocrine/metabolic and nutritional". In total and on average, GPs reported 59.1 ± 15.4 different reasons for consultations that had been addressed by their patients during consultations at least once per month.
A multivariate analysis adjusted for age and gender showed no statistically significant regional variation in the number of different consultation reasons (see Table A4 in the additional files).
However, the stepwise variable selection showed that reasons for consultations from the areas "ear" (odds ratio (OR) 1.42; p=0.046), "skin" (OR 1.38; p=0.013) and "musculoskeletal system" (OR 1.13; p=0.003) as well as the consultation reason "lymph/lymphoedema in the leg, chronic wounds" (OR 1.78; p=0.006) were more often reported in rural areas than in urban areas. On the other hand, consultation reasons from the area "social problems" (1/OR 1.19; p=0.006) were less frequently reported in rural areas than in urban areas. Additionally, "shoulder problems" (OR 1.50; p=0.005) were reported more often in rural areas than in the environs and "acute infections of the respiratory tract" (1/OR 1.10; p=0.024) were less frequently reported in rural areas than in the environs. Table 3 shows the average relative rate of the individual services of all reported services. The most frequently reported services were "blood test", "referral to a specialist" and "continuous care for the chronically ill". The average number of different services provided was 30.3 + 3.9. Compared to urban areas, a somewhat broader spectrum of services was reported in rural areas (31.1 ± 3.7 vs. 29.4 ± 3.9; p=0.010).
This difference was, at first, confirmed in the multivariate multilevel analysis controlled for age and gender (see Table 4). However, after including postgraduate training, the regional difference disappeared and the model fit improved significantly (p=0.021). In contrast, the additional inclusion of practice type and number of patients treated by quarter produced neither significant changes in the estimates of the other independent variables nor a further improvement of the model fit (p=0.502).
The stepwise selection of variables showed that the service "local injection / infiltration" (OR 1.26; p=0.046) was more frequently reported in rural areas than in urban areas and the services "check-up 35" (1/OR 1.47; p=0.013), "psychosomatic primary care" (1/OR 1.20; p=0.003) as well as "lifestyle counselling, social counselling" (1/OR 1.11; p=0.003) were less frequently reported in rural areas than in urban areas. Additionally, "urine analysis" (OR=1.12; p=0.028) was more frequently reported in rural areas than in the environs.

Main findings
In our study, GPs from regions of different degrees of urbanisation reported how often their patients had presented specific reasons for consultations and how often GPs had responded with providing different services. In total, the number of different reasons for consulting GPs was similar in urban and rural areas, however, the frequency of individual reasons for consultations had been different. In contrast, GPs from rural areas performed a larger number of different services than GPs from urban areas. Similar to the analysis of consultation reasons, the frequency of the individual services also differed between the regional categories.
On the one hand, our study suggests that GPs in rural areas spent less time on communicative, preventive, and consultant medicine than GPs in urban areas. On the other hand, GPs in rural areas more often reported to perform services that are usually provided by ambulatory specialist care in urban areas. The GPs' postgraduate training seemed to play an important role for the service spectrum offered by GPs. The multivariate models showed that the broader spectrum of services reported by rural GPs was statistically explained by a higher level of postgraduate training.

Strengths and limitations of the survey
GPs' recruitment was not carried out by random selection but by quota sampling. By adhering to the 25% maximum deviation from the recruitment plan defined in the study protocol [10] -we were able to include 91.9% of the administrative districts into our study and also to include GPs from lessfavoured areas difficult to reach by means of transport, such as Kalbe (Milde) or Helgoland. The downside of the quota sampling was, that in some districts many GPs had to be contacted to adhere to the stratification plan while in other districts the recruitment process had to end early without follow-up and not all GPs willing to participate could be included in the study. This led to the high number of 4,956 contacted GPs and the comparatively low participation rate of 4.3% interviewed GPs.
Despite the fact that -by definition -the participation rate is not important in quota sampling, it could affect the representativeness of the GP population. However, it should be noted that the response rate is not the only indicator for a selection bias. We believe that, in our study, a quota sampling is a better procedure to represent the regionally different healthcare situations than random sampling, because we suppose that the variation between underserved and overserved districts is greater than the variation between GPs within the districts.
Additionally, although random selection is a better known and clearly more frequently used method, studies have shown quota sampling not necessarily to be linked to a larger selection bias than random sampling [14]. To gather some information about a possible selection bias we performed a comparison of the data of study participants in the included regions with the statistics of the German national association of statutory health insurance physicians. This analysis showed, with one exception, only relatively minor deviations. Thus, GPs participating in our study had only been slightly older (urban areas: +0.9 years; environs: +0.4 years; rural areas: +0.6 years) and slightly more often males than the basic study population of the selected districts (urban areas: +3.6%; rural areas: +3.6%). A more distinct selection bias was found regarding the GPs' gender in the environs where the proportion of men compared to the basic study population had increased by 15.6 percentage points [15].
Besides the uneven gender distribution, there is one more aspect of the regional category "environs" that needs to be discussed. The "environs" by definition include urbanized districts and rural district with signs of agglomeration. However, our data suggest that, depending on the analysed item, features of primary care in the environs is either similar to "urban areas" or similar to "rural areas" (e.g. the rate of the consultation reason "social problems" resembles rural areas while the rate of the service "lifestyle counselling, social counselling" resembles urban areas). Therefore the "environs" should be considered as mixed category and findings regarding this regional category need to be interpreted with care.
The statistical methods provided additional strength by considering potential confounders and the cluster structure of the dataset. Aside from the two multivariate analyses on the range of reasons for consultations and services used for analysing the main question of this study, the stepwise variable selection also identified regional differences in the individual reasons for consultations and services. It should be noted that this approach reacts sensitively to differences in the distribution of the variables.
The identified model thus might not necessarily be the best set of variables to describe the regional differences in individual services and consultation reasons. Since the explained variance had been relatively low (the pseudo-R 2 averaged 0.091), there might also be other important differences between the regions than shown in the identified predictors.
As in all studies based on interviews, the physicians' answers might also have been influenced by memory gaps, errors or social desirability. However, using interview data facilitated a series of analyses that would not have been possible on the basis of routine data. For example, services paid by flat rate could be depicted individually in our study and reasons for consultations, such as "social problems" which are difficult to code in ICD-10, could be recorded. Another limitation is the fact that the recruitment area has been restricted to northern Germany and the results therefore do not necessarily represent the remainder of Germany.

Comparison with the literature and discussion of results
Our study suggests that the frequency of the individual reasons for consultation differ between urban and rural areas. Compared to urban areas, fewer consultations for "social problems" were reported in rural areas indicating that GPs in urban areas probably had to provide more psychosocial support for their patients than GPs in rural areas. This could be the result of regionally different housing situations.
One of the results of the EU contribution to the World Mental Health Surveys Initiative was that life in urban areas seemed to increase the risk of affective disorders [16]. An analysis of statutory health insurance physicians' billing data also reported on the increased prevalence of depressive disorders in German core cities [3].
"Acute infections of the respiratory tract" had also been reported less frequently in rural areas than in the environs. This may be an indication that social medical tasks, such as issuing a statement of illness, are given higher priority in districts with a higher degree of urbanisation than rural areas. This point had already been stated by some participants of our preliminary qualitative study by saying that GPs in urban areas are often only perceived as medical service providers or referrers to other specialists [8].
In Germany, patients are permitted to consult ambulatory specialists without the referral of a GP. In our study, GPs in rural areas more often reported reasons for consultations for which urban patients probably would have directly used the services of a specialist, primarily concerning orthopaedics, dermatology and otorhinolaryngology. This finding might be explained by the different regional availability of physicians in outpatient care. On average, rural areas included in this study only showed a lower density of ca. 10% of physicians in primary care but a 40% to 50% lower density of ophthalmologists, ENT specialists and orthopaedists in comparison to urban areas or the environs [15].
GPs in rural areas reported to treat a larger number of patients and to offer a broader spectrum of services than GPs in urban areas. Other studies showed the same results. Steinhäuser et al., for example, reported that GPs in rural areas carried out a broader range of procedures, particularly those usually playing a role in specialist care [9]. As early as 1998, Boerma et al. investigated regional differences in the spectrum of services provided by GPs from 30 European countries and described a more comprehensive range of services by primary care practices in rural areas and indicated a smaller role as gatekeeper of the inner city primary care practices [17]. Mehring et al. also described a higher ratio of coordinated care by GPs of patients from rural areas [18]. Additionally, a study on the treatment of bowel cancer highlighted a greater involvement of GPs from rural areas in coordinating care as well as clinical and psychosocial support for patients in rural areas [19].
Our study also showed differences in the type of individual services. The more frequent mentioning of "local injection / infiltration" or "urine analysis" in rural than in urban areas or environs might be interpreted as an indication that classic somatic primary care services had been carried out more frequently in rural areas than in urban areas or environs. House calls had also been reported more frequently in rural areas than in urban areas. Pochert et al. also described this regional difference in their study on the house call workload of GPs in Germany. Their study showed an average number of 16.5 house calls per week in rural areas whereas only 12.5 house calls per week were carried out in urban areas [20].
Regarding the level of services, GPs in urban areas placed greater emphasis on psychosocial support.
A study by Görig et al. showed that GPs in rural areas offered fewer lifestyle consultations for preventing cardiovascular disorders than GPs in cities [21]. As already mentioned, there is a higher prevalence of psychological disorders, such as depression or anxiety disorders in urban than in rural areas [22][23]. This may lead to the conclusion that GPs from urban regions have to provide psychosocial services for the care of these patients more frequently.
The multivariate analyses showed an association between gender and the provided service spectrum.
This finding might possibly be explained by the fact that women referred to other specialists more frequently or earlier than men. A series of studies examining the influence of physicians' gender on the treatment of patients revealed that female physicians were more likely to adhere to clinical practice guidelines than their male colleagues [24][25].

Conclusion
Our study pointed out a number of regional differences that could be relevant for health care planning, especially regarding the services provided in rural and urban areas. For example, the GPs in urban areas spent less time on preventive medicine than GPs in rural areas . This might indicate a shortage of preventive services in rural areas. Likewise, the supply with psychotherapy and psychosocial support in urban areas is generally better than in rural areas. In Germany almost half of the psychotherapists are based in the cities where just under 25% of the population lives [26]. However, GPs from urban areas spent more time on psychosocial support than GPs from rural areas. This might point to an underassessment of psychosocial problems in rural areas.
Our study suggests that GPs in rural areas more often provide services that are usually part of the specialized ambulatory care. This could be the result of a low availability of specialists in rural areas, which has been widely discussed in literature [27,15]. There are many strategies, that could help motivating young physicians to work in the country, e.g. expanding rural infrastructure. But as there generally are too few young physicians, other strategies should also be taken into consideration, e.g. telemedicine, mobile practices or delegation of medical services. In our study, GPs in rural areas mostly worked in individual practices while GPs in regions with a higher degree of urbanisation more often worked in group and joint practices or medical care centres. This could be the result of the lower mean age of GPs working in urban areas as practices with a larger staff of physicians usually provide better and more family-friendly working conditions, which could be attractive to young physicians.
Another finding that could be relevant is this context was that the broader service spectrum by rural GPs was statistically explained by a higher level of postgraduate training. On the one hand, this result could imply that the GPs from rural areas compensate for specific needs of the regional patient population by completing relevant (additional) postgraduate trainings. On the other hand, this could also imply that predominately GPs with a higher degree of postgraduate training are drawn to rural areas, e.g. because they feel better prepared for the local conditions. Further research should clarify the role of postgraduate training for interventions addressing the shortage of rural GPs.

List of abbreviations
EDP: electronic data processing; ENT: ear, nose, and throat; GP: General practitioner; ICPC-2: International Classification of Primary Care

Declarations
Ethics approval and consent to participate The study was approved by the Ethics Commission of the Hamburg Medical Association on 12 August 2013 (file number PV 4535). All study participant gave written informed consent to study participation.

Consent for publication
Not applicable.

Availability of data and material
The datasets analysed during the current study are not publicly available, because data sharing with other researchers was not part of patients' informed consent.

Competing interests
The authors declare that they have no competing interests.

Funding
The study was supported by the Hamburg and Schleswig-Holstein Association of Statutory Health Insurance Physicians. The funding body had no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.
Authors' contributions MS, IS and HH conceived and designed the study. DL, HOW, TR and AK significantly contributed to the study design and/or discussion of the study results. IS analysed the data. IS and HH drafted the manuscript. All the authors commented on the draft and read and approved the final version of the manuscript.      Legend: light blue areas -administration district of the category "rural areas"; medium blue areas -administration district of the category "environs"; dark blue areas -administration district of the category "urban areas"; grey and white shaded areas -not represented administration district.