Geografisk Tidsskrift, Bind 97 (1997)The Geographical Distribution of Mental Disorder in Denmark 1971-1990: A Study of First Admission RatesPeter Astrup T. Madsen Side 132
AbstractIn the period 1971-90 the first admission rate to Danish psychiatric hospitals and wards decreased by approximately 45%. This might reflect the similar decrease in number of available psychiatric beds. Generally, the decrease of first admission rates was larger in the metropolitan area than in the rest of the country. Consequently, the disparity in admission rates between town and country was reduced. Number of available beds has gained increasing importance as a predictor of variations between counties and municipalities whereas the importance of distance to hospital and degree of urbanization has declined. As regards socio-demographic variables special importance should be attached to variables for social aid and the employment rate in the social and health sector. However, it is generally erallydifficult to determine whether regional variations are due to differences in treatment facilities or morbidity differences. From a disaggregate analysis it will appear that the diagnoses schizophrenia and alcoholism are especially related to low status areas whereas a connection with social status regarding neuroses and manic-depressive psychoses is less obvious. KeywordsPsychiatry,
first admissions, socio-demographic variables,
bedcapacity, Peter Astrup
Madsen: Slotsvej 14, 8543 Hornslet, Denmark.
This paper examines the regional variations of first admission rates of psychiatric patients. Up until 1976 psychiatry was organized by the Danish state with treatment at the psychiatric hospitals ofthat time. In 1976 psychiatry became a county matter, however, in the early 19905, a further decentralization has taken place with parts of both the financial and organizational responsibility being transferred to the municipalities. The increasing interest during recent years in analyses of the regional variations of psychiatric morbidity should be perceived on the basis of this decentralization (Andersen 1987). In Denmark, studies of regional variations have mostly been carried out within the somatic area with special focus on regional variations in medical treatment during the 1980s (Andersen & Loft, 1987) and in this connection of the nature of different treatment facilities. The use of such facilities is probably related in part to their availability and proximity. Already in the middle of the last century Jarvis (1850) demonstrated that persons living near psychiatric hospitals were the ones most frequently admitted. This effect should be avoided by using data on first admissions as in this study. Another approach of regionalization is in the identification of specific socio-demographic indicators that co-vary spatially with incidence rates of morbidity, the socalled ecological approach (Smith, 1976). The characteristics of first admissions are expected to be more in accordance with the population they represent than is the case for the total number of admissions which also includes patients with repeated contacts with the psychiatric hospital service. The overall aim is to study the extent to which regional variations of first admissions can be attributed to regional variations of morbidity (demand) or whether such differences reflect regional differences of treatment facilities (supply). Geographical variations of first admissions will be described in relation to town size and illustrated by a number of variables for socio-demographic conditions in the municipalities. Furthermore, it is assumed that first admissions are related to the availability of treatment facilities and the existing hospital bed capacity. In general, the pattern is described for all diagnoses. At the end of the article a disagreggate analysis is made with a detailed analysis in connection with the pattern of selected diagnosis groups. Side 133
Materials and MethodsThe object of the study is all first adult admissions to psychiatric wards and hospitals as well as neurosis clinics and alcohol clinics in the period 1971-1990, inclusive. As dependent variable the directly standardized rate of incidence is calculated for men and women together, with the age structure in 1971 for the total Danish population as standard. The incidence, ie the number of new cases appearing during a certain observation time, eg a year, is related to the population figures in municipalities and expressed in terms of per thousand. The aim of the standardization is to make the rates independent of possible differences in age structure between the municipalities. The admission data derive from Det psykiatriske centralregister (The Danish Psychiatric Case Register) (Munk-Jørgensen et al, 1993). The rate is calculated as an annual average of the cumulative rate of incidence for each of the following five year periods: 1971-1975,1976-1980,1981-1985 and 1986-1990. These
calculations have been madefor: 1. 275
municipalities in Denmark including Copenhagen
and
Frederiksberg 2. 12
urbanization groups (Danmarks Statistik, 1988)
In connection with the municipality specific analysis, linear regression analysis is used for the estimation of variations of the dependent variable: the logarithm 10-rate as a function of the independent variables. The independent variables can be classified as follows: 1. Variables
for treatment facilities: hospital bed capacity,
2. Socio-demographic variables for demand relations: a total of 34 different variables are included in the background report for demographic, socio-economic and social conditions (Madsen, 1997). The contents and extent of these categories of variables cover what literature refers to as adequate and relevant for an ecological approach in connection with studies of regional variations of mental disorder (Burgess et al., 1992). In connection with a cluster analysis of municipalities, a principal component analysis has been used to reduce the 34 variables to 5 underlying dimensions. On the basis of various selection methods in connection with a factor analysis of the studied variables and a stepwise regression analysis only two significant socio-demographic variables have been included in the final model as well as the measure of hospital bed capacity in order to account for the regional variations. The relations between the independent variables and first admissions are determined on the basis of aggregated data information for the individual municipality and are made in principle as a cross sectional analysis. As the degree of accuracy on which the admission rate is determined for the individual municipality especially depends on the municipality size, a weighting procedure is used based on the reciprocal variance of the rate estimates (Podcock et al., 1981). All DiagnosesIn the period
1971-1990 there was a total of 193,767 first
The decrease in
rate of first admissions from 3.04 to 1.63 ![]() Table 1: The development of first admissions as well as measures for directly standardized admission rate and hospital bed capacity 1971-1990 the whole of Denmark Side 134
thousand inhabitants of 46.6 per cent. A larger drop of the admission rate for women than for men may be due to the fact that a number of disorders with the highest admission frequency for women, like neuroses, has been decreasing during the period in question (Madsen, 1997). The Development Classified according to Urbanization GroupsThe decrease of the directly standardized rate on a national basis will appear from similar calculations for urbanization groups. This is investigated by means of the comparative ratio, relating the directly standardized rate for each urbanization group to the rate for the whole of Denmark from Table 1 for each period. A value higher than 1.0 indicates for each period a level above the average for the whole country and vice versa for values below 1.0. ![]() Figure 1: The comparative ratio for 12 urbanization groups (basis=l.o). Figure 1 shows
the comparative ratios for 12 urbanization It appears that the rate - compared to the country average - is related to the degree of urbanization: the more people who live in built-up areas the higher the rate. During the whole period the three municipalities of Copenhagen, Frederiksberg and Gentofte have had the highest admission rate however, with decreasing values. Rural district municipalities, ie municipalities where the largest town counts less than 2,000 inhabitants, have had the lowest value through the whole period. However, there has been a tendency of relatively increasing values of the studied ratio in the least urbanized areas from 1971 to 1991. This shows a clear convergence regarding a more even distribution: a decline for metropolis areas and an increase for other groups. Municipality Specific AnalysesThe directly standardized rate for the country as a whole, ie the nation-wide average, has been calculated at 1.63 first admissions per 1,000 inhabitants over 14 years, cf. Table 1, whereas the municipality average is 1.34, regardless of municipality size, cf. Figure 2. Distributed on the individual municipalities, this briefly illustrates the distribution of first admissions in the period 1986-1990. In general, the rates seem to concentrate with high values for municipalities in the counties of West Sealand as well as in definite peripheral areas like Bornholm and the northern part of Viborg county in Jutland. Municipalities with a psychiatric hospital or ward also seem to be represented with relatively high rates in these areas. This applies ia to the municipalities of Copenhagen, Frederiksberg, Hillerød, Holbæk, Dianalund, Svendborg, Ribe and Thisted. Correlation and Regression AnalysisThe connection between admission rates for municipalities and various explanatory variables is illustrated by means of a linear regression analysis for the period 1986-1990. The choice of independent variables depends on a principal component analysis in connection with studies of the intercorrelation of the variables, ie test for multicollinarity Side 135
![]() Figure 2: First admissions 1986-90. of directly standardized rate in municipalities per 1,000 inhabitants > 14 years. (Belsley et al., 1980). Furthermore, the simple correlation of certain variables with the dependent variable has also been studied to reach a meaningful interpretation of the included variables. Of the studied demand variables the connection with the first admission rates is highest for "social security benefits" in municipalities which corresponds with a similar analysis carried out by Juel & Schousboe (1992). On the basis of the Side 136
![]() Table 2: Multiple regression analysis of logarithm rate 86_90 with demand variables (D) and structural variables (S) for the period 1986-1990. With the inclusion of only three explanatory variables it has been possible to describe more than 40 per cent of the total variation of first admissions between municipalities. The significant variables should not be assigned too much predictive value. In this connection they merely illustrate examples of a certain connection with the dependent variableand not any causal relations. For the variable "share of employees in the social and health sector" it could be presumed that this would be influenced by the existence of psychiatric treatment centres in the municipality, which is why the employment in this sector could be perceived as a variable for proximity to psychiatric institutions. The Development 1971-1990Below a similar regression analysis is made with structural variables, also for the three preceding periods (Table 3). A rather uniform pattern appears both as regards the sign characters, size and changes over time of the independent variables. The variable "bed capacity" appears to be of increasing importance as a predictor for the admission rate. At the same time it appears that location of treatment facilities in the municipalities is of decreasing importance. This is also the case for degree of urbanization which, for the first three periods, has the greatest explanatory value. This confirms the development according to Figure 1 as regards the classification into 12 urbanization groups. It should be added that the variable for urbanization in this connection should be regarded as an "umbrella variable" for a number of socio-demographic relations for which ![]() Table 3: Multiple regression analyses of logarithm rates for periods with structural variables. Side 137
variable
information has been collected only for the period
Diagnosis Specific AnalysisThe above analysis uses the rate for all first admissions as a dependent variable for geographical variations of the admission pattern which, of course, covers considerable variations of this measure because of the different diagnoses involved in the material. An ICD-8 breakdown constitutes 15 main diagnosis groups (Sundhedsstyrelsen, 1991). As such a disaggregation will increase the uncertainty for the diagnosis specific rates the 275 municipalities have been combined by means of a cluster analysis to constitute 15 municipality clusters. This has been based on the 34 socio-demographic variables for municipalities. A principal component analysis has been used to reduce the variation in the data to 5 underlying dimensions, cf. Table 4. This shows that 77.2 per cent of the total variance for the 34 socio-demographic variables can be reduced to 5 underlying dimensions with the content which has been named according to the most important variables correlated with the component in question. This is illustrated in Table 4 showing a variable with the highest loading (correlation). A similar method has been used in connection with an analysis of psychiatric admissions in the city of Nottingham, England (Giggs, 1986). The cluster analysis is performed as a so-called "quick cluster analysis" (Norusis, 1990). The clusters vary in size from representing 1 municipality to comprising 60 municipalities which, of course, gives rise to some methodological problems regarding variance and homogeneity, etc. Subsequently an attempt has been made to name the individual clusters in relation to a characteristic "representative", ie a municipality located "near" the gravity centre of the cluster in question, ie the average component score which is calculated for each of the 5 components by the analysis. On the next page, Table 5 characterises the 15 clusters on the basis of the component analysis. The clusters are grouped into four sectors and are further ranged according to urban density. The highest density is found in Frederiksberg, the municipality in the centre of Greater Copenhagen. At the other end of the scale "Møn" is characterized as the most rural cluster. At the same time the table reflects the comparative ratio for standardized rates for 4 selected diagnosis groups and all diagnoses for each cluster. A value higher than 1.0 also in this case indicates a level above the average for the whole country. It appears that schizophrenia and alcoholism are especiallyrepresented in metropolitan areas whereas manic-depressive, ie affective psychoses to a higher extent are found in provincial towns and rural areas. See Figure 3 for the two selected diagnoses. The distribution of neuroses seems to be more incidental. The strength of the similarity between the geographical distribution of the 5 ![]() Table 4: Principal component analysis of 34 socio-demographic variables with indication of explained variance and the most important variable according to a varimax rotation. Side 138
![]() Table 5: Comparative ratio of standardized rate for diagnoses seen in relation to homogeneous socio-ecological areas (clusters). ![]() Table 6: Product moment corre, .. , A . , lationfor 4 diagnosis groups and .... Side 139
categories is
tested by means of the Pearson product All the intercorrelation produces positive values which indicate a certain accordance in the spatial incidence pattern, however, for values below 0.6 the connection is not significant at the 0.01 level which is why there is only a certain accordance between the pattern for schizophrenia and alcoholism. This tendency can
also be proved for the 3 previous Multiple Regression AnalysisBy means of a multiple regression analysis it is examined whether variations between clusters for directly standardized logarithm rates of first admissions can be described by characteristics for these clusters. As independent variables for this purpose the component scores are used for the 5 components which appeared indirectly from the cluster analysis and Table 5. Table 7 gives the result of such an analysis for all diagnoses compared with first admissions for the diagnosis alcoholism. Furthermore, the table reflects the components included in the multiple linear model with indication of explained variance after each step. A similar method has been used previously by Giggs (1986). The components
show a direct connection with the It appears that 63.4% of the variation of all first admissions for these clusters can be explained statistically by the defined components. The relatively high degree of explanation seen in relation to Table 2 is due to the high ![]() Figure 3: Comparative ratio for selected diagnoses in relation to clusters (basis = 1.0). ![]() Table 7: Summary results ofmul. , . , .. tiple regression analyses oflogar. , , ithm rates and component score (demand variables) for the period 1986-90. aggregation level
for analysis units which blurs much of The equivalent regression analysis for the diagnosis categoryalcoholism gives a statistically better explanation of geographical variations (almost 90%) and includes more explanatory components. This is, of course, due to the fact that all diagnoses form an inhomogeneous group in which possible connections with underlying socio-demographic Side 140
variables cannot always be assigned the same value. Thus it often occurs that a diagnosis like manic-depressive psychoses is relatively strongly represented in high status areas while the reverse is the case for the diagnosis of schizophrenia (Giggs, 1986), cf. also Table 5. Figure 4 illustrates how well the regression model fits the data. The deviations from the expected value (horizontal line = 1.0) are, of course, biggest for all diagnoses where the statistical explanation ability was smaller than for alcoholism. For all diagnoses it appears that the model underestimates the municipalities of Copenhagen and Frederiksberg and overestimates the values for other clusters in the metropolitan area. To some extent the opposite is the case for alcoholism. As will appear from Tables 5 and 6 the diagnoses of schizophrenia and alcoholism show a somewhat similar distribution pattern. Even though it is impossible to say anything about the causal relations, there is no doubt that in areas with considerable social problems and social comedowns, ie component 5 which has a close connection with "permanent social security", there are more cases of alcoholism than in affluent areas. According to Table 7 there is a tendency towards many first admitted alcoholics in affluent areas. This may be due to methodological relations in connection with so-called ecological fallacy (Rosen et al., 1985) as alcohol problems more generally occur in big city areas than in the country, cf. inclusion of component 3 (single and big city conditions) in step 4. At the same time, social status is generally higher in urban areas compared to rural districts which may result in the abovementioned ConclusionThe analysis in terms of time of the development in psychiatric admissions shows that a significant drop in the admission rate can be attributed especially to a corresponding drop in number of beds, ie a confirmation of "Roemer's principle" to the effect that the number of beds offered creates its own demand (Roemer, 1991). The importance of bed capacity has increased for regional variations of first admissions which has thus become a limiting resource that can be attributed to political decisions to reduce the bed capacity at psychiatric hospitals. Concurrently with the decrease in number of beds availableit appears that differences between town and country have been effaced. This regional equalization could be due to the fact that there is a more widespread acceptance of mental disorders in all classes of the community or that the balance of social or psychic problems over time has shifted ![]() Figure 4: Ratio for the observed and expected directly standardized rate in clusters 1986-90 (basis= 1.0). Side 141
to exist also in rural areas. There may be several reasons for this development, however, part of the explanation could be the rural exodus in connection with the change from an agricultural society to a modern industrial society with a more marked labour division than before. In the beginning of the 1970s the regional variations could possibly be attributed to conditions related to the centrally organized state hospitals of that time. In those days distance from the hospital played a more significant role than today with the increased mobility and decentralization of the psychiatric treatment (Sundhedsstyrelsen, 1991). In general, geography today (here expressed in terms of variables for urbanization and access to psychiatric treatment) seems to be of less importance than before. However, in municipalities with a psychiatric institution, the admission rate is larger than expected. Measures for possible variations of psychiatric morbidity seem to explain the major part of the geographical variation more than differences in the existence of treatment facilities, cf. Table 2 which, for the period 1986-1990, shows that 36.3 per cent of the total variation between municipalities is due to socio-demographic conditions whereas this share lies at a considerably lower level for the structural variables in the same period, cf. Table 3. Of all studied variables for the period 1986-1990, the closest connection with the rate in municipalities for first admissions exists for "paid social security benefits". This variable, however, should be considered as a variable for a number of social and economic conditions in municipalities; however, no information has been obtained about these conditions. The connection with socio-demographic variables has only been studied for the period 1986-1990 where these show a relatively weak connection with all diagnoses. This connection is considerably strengthened with a disaggregation into diagnosis groups, however, with an aggregation to municipality clusters. This part of the analysis shows that the established clusters are well suited for studies of social events, especially if they can be related to conditions linked with the defined components, like in this case first admissions among the mentally disordered. It should be noticed that the principal component analysis has been carried out independently of the admission pattern. Furthermore, the clusters used and their characteristics could be used in connection with a description of other social events, like crime, suicide and drug abuse. In general, the analysis has shown significant variations in the geographical distribution of mental disorders, however, a complete appreciation of these variations requires a knowledge of both supply and demand variables. A number of indicators have been used in this analysis for such conditions which have proved to be associated with the selected diagnosis groups. Thus, efforts must be made for a diagnosis specific approach. The ideal approach would be to link background information directly related to the patients admitted; however, the present aggregate analysis could contribute to put forward hypotheses regarding possible connections between socio-ecological variables and psychiatric admissions with a view to more detailed analyses at the individual level. The method can also be used to analyse all admissions to estimate the total consumption of resources used in connection with psychiatric treatment. In this connection it is necessary when interpreting the geographical distribution to consider the so-called drifting tendencies because psychiatric patients tend to move over time to certain areas. This happens for voluntary or compulsory reasons, however, to various degrees according to the mental disorder in question following selective mechanisms in connection with migration (Giggs, 1984). With the ongoing decentralization of the district psychiatry it will be the task for Danish municipalities in the future to assess the need for psychiatric treatment, in cooperation with the counties. Consequently, a current evaluation and distribution scale is to be desired in order to plan the professional efforts in the treatment of the mentally afflicted, perhaps on the basis of some of the methods described in this paper. AcknowledgementsThe article is
based on a paper presented at the 35th European
Translation:
Dorthe Fuglede, The Aarhus School of Funds made available by Statens Samfundsvidenskabelige Forskningsråd in Denmark (The Danish Research Council of Social Science). Excerpt of the paper previously published in Ugeskrift for Læger (Danish Medical Weekly), (Madsen, 1995). ReferencesAndersen, TF.
(1987): Regionale variationer i sundhedsvæsenets
Andersen,
T.F. & Madsen, M. & Loft, A. (1987): Regionale
variationer Belsley, D.A.
& Kuh, E. & Welsch, R.E. (1980): Regression
Diagnostics: Burgess, P.M. & Joyce, CM. & Pattison, P.E. & Finch, SJ. (1992): Social indicators and the prediction of psychiatric inpatient service utilisation. Soc. Psychiatry Psychiatr. Epidemiol. 27: 83-94. Danmarks
Statistik (1988): Kommunalstatistisk databank. Giggs, J.A.
(1984): Residential mobility and mental health. In:
Giggs, J.A.
(1986): Mental disorder and ecological structure in
Jarvis, E.
(1850): The influence of distance from proximity to an
Juel, K.
& Pedersen, J.E. & Schousboe, M. (1992):
Københavns Madsen, P.A.
(1995): Geographical variations in first-admission
Læger; 157:
1853-1857. Madsen, P.A.
(1997): Regionale variationer i førstegangsindlæggelser
Munk-Jørgensen, P. &
Kastrup, M. & Mortensen, P.B. (1993): Munk-Jørgensen, P. & Lu'tzhøft, J. & Jensen, J. & Strømgren, E. (1992): Trends in psychiatric hospitalization in Denmark. A 10-year register-based investigation. Acta Psychiatr. Scand. 86: 79-83. Norusis, M.
(1990): SPSS/PC+ Advanced Statistics. Chicago. Pocock, S.J.
& Cook, D.G. & Beresford, A.A. (1981):
Regression Roemer, M.I.
(1961): Bed supply and hospital utilization. Hospitals
Rosen, M.
& Nystrøm, & Woll, S. (1985): Guidelines for
regional Smith, C.J.
(1976): Geography and Mental Health. Resource Paper
Sundhedsstyrelsen (The National Board of Health) (1986): Klassifikation af sygdomme. Systematisk 5 cifret fortegnelse (Classification of diseases. Systematic 5-digit register. Copenhagen). Sundhedsstyrelsen (1991):
Omstilling på psykiatriområdet. |