Background
In an ideal health care system all people should have equal access to medical care [
1]. Accessibility of medical services depends on both non-spatial factors (such as economic, cultural, and social issues or factors related to the organization of the health care system) and on spatial factors (such as geographical distance) [
2‐
4]. Back in 1866, Jarvis [
5] identified an inverse relationship between hospitalization rates and the distance between the patients’ residences and the location of the hospital, with people living closer to hospitals being more likely to use them. To date, such ‘distance decay effect’ has been replicated with remarkable persistence by numerous studies that suggest a universal pattern of reduced service utilization with increasing spatial and time-related distance between peoples’ residences and somatic [
2,
6‐
10] and psychiatric [
1,
11‐
22] service sites.
Importantly, distance decay effects were not only found for mental hospitals but also for outpatient clinics. Bürgy and Häfner-Ranabauer [
18], for example, explored the relationship between the utilization of a psychiatric emergency service in Mannheim, Germany, and the accessibility of that outpatient service for the inhabitants in its catchment area. When using the spatial distance (air distance) and the time-related distance (travel time required by public transportation) between patients’ residence and the location of the emergency service as a proxy of its geographical accessibility, they found an association between increasing home-to-service site distance and reduced contacts with the emergency service. Beyond geographical proximity, less favorable ecological conditions of a district (e.g., higher population density, worse housing conditions, or higher proportion of foreigners) and certain diagnoses (e.g. schizophrenia and substance use disorders) did likewise predict increased outpatient service utilization in that German study. Interestingly, however, there was no interaction effect between distance and diagnoses on service utilization; that is, service utilization decreased with increasing home-to-service site distance for all mental disorders to the same degree with no differences between diagnoses. The findings from that German study are highly interesting but they might not generalize to more comprehensive psychiatric outpatient services which also provide non-emergency care in less urban environments since the study was restricted to the utilization of an emergency facility during out of office hours in an urban area.
In a more recent study, Zulian et al. [
1] assessed the influenced of distance and of clinical and socioeconomic patient characteristics on the utilization of community-based mental health services in Verona, Italy. Spatial and statistical analyses unfolded a distance decay effect with different trends for the three types of community-based mental health services under examination: the strongest negative correlation between distance and the number of patients per inhabitant in a specific area (caseload) was observed for outpatient clinics, followed by community mental health centers offering day care and rehabilitation. For acute inpatient wards the correlation was the least pronounced but still statistically significant. When controlling for socioeconomic predictors of service utilization (such as age, gender, marital status, or occupational status), the aforementioned distance decay effects were even stronger. That is, the influence of distance on service utilization was underestimated in this service provision area if the influence of unequally distributed socioeconomic variables on service utilization was omitted. Despite examining the impact of diagnosis – and of other characteristics (such as age, gender, etc) – on service utilization, Zulian et al. desisted from analyzing whether distance decay effects vary depending on diagnoses. With respect to service planning, however, it may be important to know which patients “suffer” the most from a distance decay effect. Zulian et al. furthermore used the travel time by car as a proxy of distance. This might be a little precise indicator of spatial accessibility for those patients who do not have access to a car or who are not able to drive a car (which might particularly often apply to mentally ill people).
To the best of our knowledge, no past study on distance decay effects in mental health care settings unified the following characteristics: (a) spatial accessibility was assessed in terms of travelling time by public transportation (which is available to everybody); (b) inpatient and outpatient services were both considered allowing for comparisons between settings; (c) the sample included regular inpatient and outpatient services (i.e., it was not restricted to psychiatric emergency services); and (d) distance effects were examined depending on diagnostic groups. We therefore aimed at answering the following research questions based on data from one of the largest service provision areas in Switzerland:
(1)
Does distance (i.e., travel time by public transportation) between patients’ homes and mental health service facilities affect the utilization of inpatient and outpatient services, respectively?
(2)
Does the relationship between distance and service utilization depend on the primary diagnosis of the patients?
Discussion
Research on spatial factors affecting the accessibility of mental health services has repeatedly demonstrated a pattern of reduced service utilization with increasing spatial or time-related distance between peoples’ residences and service facilities [
1,
11‐
21]. Our results corroborate previous findings of such a distance decay effect in the outpatient setting [
1,
16‐
20,
22]. Importantly, we used the travel time by public transportation, which is available to (almost) everybody, as proxy for the distance between patients’ residence and the outpatient clinic. Furthermore, we controlled for the influence of available socioeconomic characteristics of the communities (e.g., the average tax amount per inhabitant) on service utilization. While a distance decay effect was evident in our outpatient setting irrespective of the type of mental disorder (i.e., for all of the most prevalent ICD-10 primary diagnoses), no such distance decay effect was found in our inpatient setting, except for patients with organic mental disorders (ICD-10: F0). A possible explanation would be that F0 diagnoses such as dementia occurred almost exclusively in elderly patients who are among the least mobile members of the society.
Differences in distance decay effects between the inpatient and outpatient treatment settings may be expected given that outpatients have to travel the home-to-service facility distance for every single visit in an outpatient clinic, whereas inpatients have to cover the home-to-hospital distance only twice (when entering the hospital and after discharge). In addition, the more severe and acute clinical condition of inpatients, who are often in need of emergency and/or involuntary admissions, may likewise mitigate the influence of distance on the utilization of inpatient wards [
27]. Even though the lack of a global distance decay effect in our mental hospital contrasts with some previous findings [
5,
12‐
15], other earlier studies did also not find a clear-cut association between the home-to-hospital distance and the probability of being hospitalized [
28,
29], or they at least reported less pronounced distance decay effects for inpatient services [
1]. The complete lack of a distance decay effect except for organic mental disorders in our data might be explained by the rural location of our hospital, with some larger cities in more remote areas of the hospital. As a result, in our service provision area any distance decay effect might have been outbalanced by higher caseloads in these remote cities. It is well known that urban areas typically have higher rates of mental disorders than rural areas [
30‐
35]. In fact, the first peak of the inpatient caseload in our service area, which occurred at a distance range of 10-15 min away from the hospital (Fig.
3b), may be explained by the high caseloads in the cities of Aarau and Wettingen (which are the most populous communities in our service area). Thus, in previous studies where mental hospitals were located close to the most populated urban centers, the finding of a distance decay effect might simply have reflected some decay in the prevalence of mental disorders when moving from urban to rural areas. Such “typical ecological distribution” or gradient of mental disorders is sometimes explained by the less favorable social and ecological conditions in urban areas [
31]. In line with this, our analyses, which included all available ecological characteristics of the communities as potential confounders, revealed that the proportion of immigrants – which is typically higher in urban communities and which may reflect less favorable social and ecological conditions – was predictive of inpatient service utilization. Such association between a high concentration of immigrants from foreign countries in a geographical area and a more intense utilization of inpatient [
32] and of outpatient [
18] services in that geographical area has already been reported by previous studies. It is important to note, however, that being an immigrant per se is not a risk factor for mental health service utilization. Immigrants do not more often use mental health services themselves. On the contrary, in our service provision system they were even underrepresented among inpatients (they accounted for 8.6% of the inpatient cases while they represented 22.8% of the Swiss population in 2011 [
23]). Thus, rather than immigrant status being an individual risk factor itself, it is the concentration of immigrants in a community that indicates a higher risk for inpatient utilization of inhabitants, probably because the concentration of immigrants represents less favorable social structure in that respective community which in turn increases the risk for inpatient utilization [
36].
In summary, our findings and their comparison with results from previous studies support the notion that the distance decay relationship in mental health services is not a simple and consistent one but rather results from a complex interaction between geographical proximity to services, socioeconomic conditions in local communities, the organization of mental health services, and the transportation infrastructure [
2,
3].
With regard to practical implications, some recommendations can be derived from our findings for the most effective location of mental health services. The distance decay effect in our outpatient clinics, which seemed to occur irrespective of the type of mental disorder, denotes the importance of decentralized outpatient clinics to meet the needs of the population as close as possible to where people live and to avoid people in remote areas being insufficiently supplied with mental health care. At a distance of 25 min traveling time by public transportation to the closest outpatient clinic, the proportion of inhabitants using our outpatient services (the caseload) was already reduced by more than 50%. Furthermore, the spatial accessibility of our outpatient clinics did not only affect the caseload (i.e., the proportion of inhabitants receiving any outpatient treatment at all) but it also affected the number of outpatient sessions that the patients received [
37]. This requires particular attention given that frequency and continuity of care are known to be important components for the effectiveness of outpatient treatment [
38]. If decentralized location of outpatient clinics is not possible for any reason, the provision of transportation services as a part of mental health care programs could be a promising way to enable equal access to mental health care even for those people located in remote areas [
39].
For inpatient care, distance decay effects seem to be much less pronounced than for outpatient clinics. In catchment areas with good public transportation systems the home-to-hospital distance might be even completely irrelevant for access to inpatient services. Thus, there seems to be much less empirical basis to decentralize mental hospitals to the same extent as outpatient clinics. However, even if the role of distance would be completely irrelevant for inpatient service utilization, this is not saying that mental hospitals should be located arbitrary in the service provision area. Instead, for patients’ and relatives’ convenience, mental hospitals should be located in or close to the largest communities where most people (and hence most people in need for inpatient treatment) live. This notion is supported by findings showing higher prevalence rates for mental disorders in urban areas [
30‐
35].
Limitations
Several limitations of this study have to be addressed. First, the ecological cross sectional design does not allow for causal inferences and bears the risk of ecological fallacy. We do not know to what extent the caseloads of the communities were affected by different prevalence rates of mental disorders or by other community-related ecological variables for which no data was available. As our outpatient clinics were located in the most urban areas of the catchment area, which typically have higher prevalence rates of mental disorders [
30‐
35], our results could simply show some decay in the prevalence of mental disorders from urban to rural areas. However, although decreasing prevalence rates of mental disorders or non-controllable ecological variables of the communities might have contributed to the observed distance decay effects, it is rather unlikely that they changed to the same degree with increasing distance from service facilities as the caseloads did.
Second, travel times between patients’ homes and service facilities were calculated using the main public transportation station in each community as starting point. The analyzed travel times were only an approximation of the real travel times of the individuals. However, even if individual door-to-door travel times would have been available from timetables for every inhabitant in the service area, such figures would not have been completely exact (traveling the same way twice almost never takes exactly the same of amount time, e.g. due to delays in public transportation).
Third, our analyses were restricted to those patients who were using the secondary mental health services of the PDAG. While the hospital of the PDAG provides the vast majority of inpatient treatments in the canton of Aargau [
24,
25], data on outpatient visits at general practitioners and psychiatrists in private practice was not available to this study. Nevertheless, it is unlikely that the distance decay effects in our outpatient clinics can be completely explained by the omission of outpatient cases and visits at physicians in private practice who run their medical practices in remote areas. Like our outpatient clinics, most private practices are located in urban communities of our service area suggesting that people in remote areas really are at risk of being insufficiently supplied with mental health care.
Fourth, as is usual in mental health services research, diagnoses were not assessed with structured clinical interviews such as the SCID-I [
40] and SCIDII [
41]. Their application would have been far too time-consuming for a routine clinical care setting. Concerning the reliability of our clinical diagnoses, however, miss-codings might be rare as we only analyzed broad diagnostic categories (ICD-10: F0, F1, F2 etc.). This was confirmed in a recent study in our mental hospital which showed good overall agreement of the commonly used clinical examination technique with SCID I assessments regarding primary diagnoses at the level of these broad diagnostic categories [
42].
Fifth, analyses on diagnostic subgroups were restricted to those primary diagnoses which were present in at least n = 200 cases, and the 95 (1.7%) cases with missing data on primary diagnosis were excluded from these subgroup analyses. A minimal subsample size of n = 200 cases might be considered arbitrary but it was intended to render reliable estimates of the distance decay effects within the most prevalent diagnostic subgroups.