urban small area variation in adolescents' health behaviour

Pergamon Soc. Sci. Med. Vol. 45, No. 7, pp. 1089-1098, 1997 © 1997 Elsevier Science Ltd. All rights reserved

PII: S0277-9536(97)00036-1 Printed in Great Britain 0277-9536/97 $17.00 + 0.00

URBAN SMALL AREA VARIATION IN ADOLESCENTS' HEALTH BEHAVIOUR

S A K A R I K A R V O N E N ~* and A R J A H. R I M P E L A 2

'Department of Public Health, P.O. Box 41 (Mannerheimintie 172) FIN-00014, University of Helsinki, Helsinki, Finland and -"I'ampere School of Public Health, Tampere University, Tampere, Finland

A~tract--Our previous study indicated that region plays a relatively small role in adolescents' health behaviour. Here, the regional patterning of health behaviour is studied further by shifting the focus to small areas. First, we test whether small area socioeconomic, demographic and housing characteristics correlate with health behaviour. The analysis then turns to the relationship between these characteristics and their individual level correlates. We wish to ascertain if behaviour is related to small area characteristics similarly for both genders and for adolescents' socioeconomic characteristics. The Adolescent Health and Lifestyle Survey data from 1989-1995 (16- and 18-year-olds, n = 1048, response rate 71%) were linked with data describing 33 subareas of Helsinki, the capital of Finland. Smoking, alcohol use, abstention from dietary fat and physical activity were used as lifestyle indicators. Gender apparently influences the extent to which the area plays a role. Logistic regression demonstrated that prolonged unemployment predicted low prevalence of abstention from dietary fat (traditional dietary patterns) among girls and heavy drinking among boys. High total rate of unemployment predicted lower physical activity among girls. Also owner-occupied housing correlated positively with girls' physical activity. Although the individual level socioeconomic characteristics were not as strongly related to health behaviour as the small area factors, a low level of education predicted smoking and alcohol use and, among girls, decreased physical activity. We conclude that small area characteristics, especially the level of unemployment of the area, may be even more strongly related to health behaviour than individual socioeconomic characteristics. © 1997 Elsevier Science Ltd

Key words- -smal l areas, adolescents, health behaviours, gender

INTRODUCTION

Previous studies have indicated that the larger regional context of adolescents' residence plays a fairly limited role in influencing their health behaviour (Karvonen and Rimpel/i, in press; Klein and Pittman, 1993). Additionally, there appears to be less observable regional variation and it is largely reducible to social structure (Duncan e t al. , 1993; Karvonen, 1995). Here, we continue analysing the regional patterning of health behaviour by shifting the focus to small areas, or local units such as the districts of a city.

Earlier, health behaviour was found to be related to small areas. Within Dublin, health-damaging behaviour was more prevalent in high mortality small areas than low mortality areas (Johnson et

al. , 1991). In Britain, health behaviour was more clearly patterned by small area than by region. For example, the patterning of alcohol use and diet by social class and gender differed according to small area context (Blaxter, 1990, pp. 117-124; Elm6n and Sundh, 1994).

The applicability and definition of small areas may, however, differ interculturally. For example, the British small area context greatly differs from

*Author for correspondence.

the Finnish one (Blaxter, 1990, pp. 78-79). In Britain variations in health or health behaviour are measured at the local authority level. In Finland such variation is measured at the district level. Also, there may be differences in the relative extent to which small areas differ from each other. In Great Britain the socioeconomic segregation of population into small areas is likely to be quite pro- nounced, thus resulting in a rather homogeneous population structure within areas (Townsend et al. ,

1988). However, although the population may be selected according to their residence area, in the Finnish context the population is likely to be socioeconomically mixed even in high status areas (Lankinen, 1994).

In a sum, small area variation can tentatively be expected to comprise three groups of factors, all of which are potentially relevant to adolescents' health behaviour. First, small area variation reflects differences in the sociodemographic composition of the total adolescent population. This group includes those contextual socioeconomic characteristics of the small areas that cannot be derived from the characteristics of all people living in the areas. Second, small areas may possess some cultural fea- tures that affect all people, or in the case of subcul- tures, a subgroup of people living in the area. A strong local tradition or religious culture are the

1089

1090 Sakari Karvonen and Arja H. Rimpel~i

Fig. 1. The small areas

most obvious examples here. The third group of factors has to do with the health-related and beha- vio,lral consequences of urbanisation. As small areas differ in their level of urbanisation there is a broad range of factors that may produce variation. These factors range from the direct consequences of urbanisation itself, such as pollution, to factors resulting from the density of the population, such as the availability of services, and its cultural consequences, such as the relative anonymity in social exchange within metropolitan areas.

In this study, small areas are defined on the basis of the social and demographic characteristics of districts within Helsinki, the capital of Finland (Fig. 1). The city of Helsinki has an approximate population of 500,000 inhabitants and is flanked by two other major cities (Hels inki alueittain, 1988, p. 17). The greater Helsinki area thus includes a total of ap- proximately one million inhabitants. This study concentrates only on the city of Helsinki itself which is the most urbanised area of this region.

By concentrating on one particular city we aim to simplify the analysis by ruling out the third group of factors mentioned earlier: within Helsinki, small area variation is likely to be produced by sociodemographic factors or local cultures. The level of urbanisation does not vary between areas to the degree that it would be a potential determinant of health behaviour.

For the present purposes the city of Helsinki is quite suitable as it is marked by large internal variation in social structure, housing conditions and demography (Hels inki alueittain, 1988, pp. 417 424). A distinct modern lifestyle can also be found there (Keskinen, 1991). Among adolescents this is marked by the wide use of alcohol together with health pro- moting dietary choices, such as abstaining from the

of Helsinki, Finland.

use of butter and high fat mi lk--both products traditionally present in the Finnish diet (PrfittS~l~ et al., 1986; Siurala, 1991).

The small area characteristics fall into three categories: demographics, housing and socioeconomic characteristics. Tentative explanations have been offered for the correlation between these characteristics and health behaviour. First, the demographic composition of the population may operate through social control (KandeL 1991). Thus, in areas where the average age is high, adolescents' behaviour may be subject to more social control than in areas with high adolescent population. Second, housing is a potential factor influencing adolescents' health behaviour. One conceivable pathway is through migration, as families moving to urban areas are likely to enter areas with a lot of temporary, rented apartments. Individual migration correlates with smoking, which suggests that smoking is used as a means to adjust to the new environment (Isohanni et al., 1991). Through greater visibility of smoking the individual level relationship may result in social acceptance of smoking in these contexts. Third, the socioeconomic composition of the population, particularly employment status, potentially influences health behaviour. At an individual level unemployment associates with the fiequency of drinking pro- blems (Hammarstr6m, 1994; Lahelma et al., 1995) and, among adolescents, with increased overall drinking levels (Janlert and Hammarstr6m, 1992). Thus, in small area contexts where unemployment is high particularly drinking may enjoy social acceptance, at least in subgroups prone to use alcohol.

The purpose of our study is to explore these relationships. We aim, first, to find out whether the characteristics of small areas correlate with adolescents' health behaviour. Second, we aim to analyse

Small area variation in health behaviour

Table 1. Response rates (%) and number of respondents by age and gender

1091

AGE AND GENDER 16 yrs 18 yrs

Girls Boys Girls Boys Total

Response rate (%) 84.3 66.2 79.0 57.6 70.7 n 292 257 278 221 1048

the relationship between the small area characteristics and their individual level correlates in order to discover whether the small area behaviour pattern characteristics are similar for both genders and other socioeconomic characteristics of adolescents.

MATERIAL AND METHODS

Data for the study come from two sources. The individual level data concerning adolescents' health behaviour and their socioeconomic characteristics come from the (Finnish) Adolescent Health and Lifestyle Survey (AHLS) (Nuorten terveystapatutkimus, 1979; Rimpel/i et al., 1988). They have been linked with data describing the characteristics of the small areas of Helsinki. These data came from the Statistical Office of the city of Helsinki.

In the AHLS nationally representative data were collected by mailed questionnaires in 1989, 1991, 1993 and 1995. The sample that was obtained from the Central Population Registry was based on consecutive days of birth so that all Finns aged 16 and 18 and born on the sample days were included (in 1989 six and in 1991-1995 12 consecutive dates in a period between May and September).

Based on municipality codes those adolescents aged 16 and 18 that were residents of Helsinki were selected for this study. All together this comprised a total of 1482 adolescents, of which 71% satisfac- torily completed the questionnaire (Table 1). The mean ages of respondents were 16.5 years and 18.5 years. Response rates were lower among boys than girls and among 18- than 16-year-olds.

Data from each survey were pooled to statisti- cally justify subdividing the sample according to the individual socioeconomic and small area variables. This pooling was possible as the timing of the study, core questions and the study populations were practically identical in all four surveys.

Five health behaviour measures were used. These were: daily smoking, weekly use of alcohol, monthly intoxication, abstaining from the use of milk and fat-containing spreads, and physical activity twice a week or more often. Non-respondents to single questions on health behaviour (1.0-1.6% of the total data) were not included in the respective analyses.

Smoking behaviour was measured on the basis of the rate of daily smoking (34% of respondents). Daily smokers were defined as all those that reported smoking daily or more often, having smoked a cigarette a week or less than a week ago

and having smoked more than 50 cigarettes during their lifetime. Thus even those previous smokers who had stopped smoking less than a week ago were considered to be daily smokers.

Alcohol use was investigated on the basis of two structured questions. Subjects were asked to indi- cate the general frequency of their alcohol use on a nine-point scale ranging from "daily" to "I do not use alcohol". Alcohol use at least once a week (33% of respondents) was analysed. The second question concerned heavy alcohol use: "How often do you drink so that you end up being heavily intoxicated?" Those that reported having been intoxicated at least once a month were regarded as heavy drinkers (26% of respondents).

Dietary behaviour was studied with similar questions: "What kind of milk do you drink usually?" and "What kind of spread do you use on bread?" with six alternatives in both questions describing different spreads/milk (examples of product names of spread/milk were mentioned in parentheses). As the "unhealthy" options were extremely rare among the urban population, the "healthier" option was chosen as the object of study. Abstainers from dietary fat were defined as all those who used no fat- containing spreads and used no milk or only non- fat milk (52% of respondents).

Two questions about physical activity were asked: "How often do you engage in physical ac- tivities or take part in sports during your leisure time (1) in exercises, competitions or games organ- ised by sports clubs; and (2) in other types of physical activity?" Subjects responded from seven alternatives ranging from "never" to "approxi- mately every day". The frequency of twice a week or more often was accepted for analysis (60% of respondents).

In addition to age and gender the individual sociodemographic characteristics included social status of head of the household (three categories; see Table 2), adolescents' own level of education (three categories), family type (two categories) and the location of the place of residence. These measures were based on adolescents' self-reports.

The data describing the small areas of Helsinki included a district code used in the administrative division of the city (Helsinki alueittain, 1988, p. 3). The AHLS data included municipality and post codes, and these were used to link the data with a particular small area. The following groups of small area characteristics were used to describe each of the 33 small areas of Helsinki: (1) the socioeco-


Table 2. The characteristics by small area (n = 33), their means and standard deviations

Small area Name 1 2 3 4 5 6 7 8 code

101 Vironniemi 33.5 11.3 13.8 61.1 50.8 7.7 17.6 35.4 102 Utlanlinna 32.4 17.2 15.9 62.3 51.0 6.5 13.9 36.3

Kampin- 32.8 11.6 16.6 59.6 48.5 5.8 13.5 36.2 103 malmi 104 Taka-T6616 29.3 12.4 14.8 58.5 47.1 6.5 11.4 36.7 105 Lauttasaari 33.8 16.1 13.0 67.9 61.5 6.5 16.6 35.0 201 Reijola 28.0 8.3 17.4 55.7 48.9 7.6 16.5 36.4

Munkki- 33.4 14.3 13.3 69.5 57. I 7.5 17.4 33.2 202 niemi 203 Haaga 26.2 8.5 14.8 63.4 53.1 5.0 16.5 38.6 204 Pit/ij/inm/iki 20.3 6.6 14.8 65.7 51.5 8.1 21.0 36.4 205 Kaarela 20.0 5.0 17.6 60.4 47.8 4.4 27.8 33.8 301 Kallio 21.4 4.1 20.9 48.9 53.5 5.7 7.30 40.1 302 Alppiharju 18.8 3.6 22.5 47.9 57.7 4.0 6.80 42.7 303 Vallila 17.2 3.8 23.2 47.6 43.8 6.0 9.80 40.1 304 Pasila 19.0 3.4 17.6 52.8 39.6 8.6 23.3 34.4

Vanhakau- 21.9 5.7 17.7 57.7 46.7 4.3 20.1 36.1 305 punki 401 Maunula 19.0 6.5 21.3 62.3 41.0 2.5 16.9 41.3 402 L/insi-Pakila 35.3 4.5 11.7 84.2 84.2 2.2 23.9 34.9

Tuomarin- 34.5 4.0 10.8 79.4 84.2 t.3 34.9 30.5 403 kyl/i 404 Oulunkyl/i 26.8 5.1 14.0 68.3 61.9 6.8 21.2 34.4 405 lt/i-Pakila 28.2 4.5 14.3 79.2 75.7 4.8 27.6 38.9 501 Latokartano 14.7 3.3 20.7 63.7 48.2 2.9 21.9 36.9 502 Pukinm/iki 19.9 2.9 19.0 60.4 50.8 3.5 24.9 31.9 503 Malmi 19.6 3.1 18.7 65.1 55.9 4.3 28.3 34.5 504 Suutarila 17.0 3.0 16.9 69.5 66.6 1.3 31.6 32.3 505 Puistola 19.3 2.4 14.4 69.9 61.0 5.1 30.7 33.5 506 Jakom~iki 6.8 2.6 25.8 58.8 32.3 1.9 23.0 41.3 601 Kulosaari 34.5 20.2 13.1 73.6 58.6 5.3 18.6 41.9 602 Herttoniemi 22.7 6.4 20.3 62.2 46.7 3.4 19.1 38.5 603 Laajasalo 28.6 7.6 14.0 71.7 62.6 4.4 26.3 36.3 701 Vartiokyl/i 20.8 7.6 16.6 67.7 61.1 2.9 22.2 37.8 702 Myllypuro 15.5 6.7 22.6 66.5 50.4 2.7 20.8 45.1 703 Mellunm/iki 13.3 3.8 23.6 61.4 49.7 3.4 25.4 39.2 704 Vuosaari 19.0 7.3 20.0 61.7 50.6 4.0 24.7 36.6 Mean 23.7 7. I 17.3 63.8 54.6 4.8 20.7 36.5 Standard 7.4 4.6 3.8 8.5 11.3 2.0 6.8 2.8 deviation

1 = Upper white collar workers (%); 2 = Swedish population (%); 3 = unemployment (%); 4 = married couples (%); 5 = owner-occupied housing (%); 6 = dwellings provided by the employer (%); 7 = adolescents and children (%); 8 = prolonged unemployment (proportion of the total of the unemployed) (%).

nomic character is t ics (average o f earnings) , p ro-

po r t ion o f popu la t i on wi th a s tudy degree (i.e.

middle level educat ion) , u n e m p l o y m e n t rate ( two

variables), the percentage o f blue collar workers ,

the percentage o f uppe r white collar worke r s and

the percentage o f non-nuc lea r families; (2) the

demograph ic characterist ics (the p r o p o r t i o n o f

popu la t i on over 65), the percentage o f you th and

chi ldren ( f rom 0 to 19 years), the percentage o f

those w h o were native speakers o f Swedish*; and

(3) character is t ics o f hous ing and families o f the

small area (the percentage o f owner -occupied

houses , the percentage o f rented hous ing and the

*Swedish is the other official language of Finland. At the beginning of the century the Swedish-speaking population formed the basis of the aristocracy and the administrative structure of the area, and it is still over- represented in the academic, managerial and art pro- fessions.

percentage o f t e m p o r a r y residences, the percentage

o f non -mar r i ed and marr ied couples). The measures

were c o n t i n u o u s variables.

METHODS OF ANALYSIS

In the first instance variables describing the small

area character is t ics were reduced by correlat ing

them with heal th behav iour variables. All those

characterist ics were included that showed a statisti-

cally significant corre la t ion (range f rom 0.09 to

0.15) wi th at least one heal th behav iour variable.

The n u m b e r o f small area character is t ics was

fur ther reduced by correlat ing them with each

other . Wi th in each o f the three g r oups the small

area character is t ics were highly correlated. At this

po in t those characterist ics that showed the highest

i n t r ag roup corre la t ion and lowest in te rg roup corre-

lat ions were selected to describe the small areas.

Small area variation in health behaviour 1093

Thus, for example, of the original seven measures describing housing only the percentage of owner- occupied housing and the percentage of dwellings provided by the employer were included. Owner- occupied housing correlated highly negatively with rented housing and showed lower correlations with the variables from the other groups than the rest of the housing characteristics. Similarly, as dwellings provided by the employer correlated only with the proportion of these dwellings of rented apartments and the latter measure showed higher correlations with variables from the other groups, the overall percentage of dwellings was selected.

In the final analysis, eight variables were selected into multivariate analyses. (1) The socioeconomic characteristics were measured by three variables: the overall percentage of unemployment, the proportion of prolonged unemployment among the unemployed, and the percentage of upper white collar workers. (2) The demographic characteristics included the percentage of youth and children (population under 19 years) and the percentage of the Swedish-speaking population. (3) The characteristics of housing and families included, in addition to the two measures mentioned above (owner-occupied housing and dwellings provided by the employer), the percentage of married couples.

To locate the characteristics geographically their distribution of by small area are shown in Table 2. Socioeconomically poorest areas in terms of high unemployment and low rate of upper white collar workers were situated in the eastern parts of the Central Helsinki (small areas 301-3). They also included some northeastern areas such as Jakom/iki (506), Maunula (401) and Latokartano (501) as well as the eastern suburbs (702 3) (Fig. 1). Prolonged unemployment was high somewhat surprisingly also in Kulosaari (601) that is a relatively well-to-do area judging by other socioeconomic criteria. The very low overall unemployment rate in the area suggests that unemployment is structurally somewhat different from the one observed in the other areas with high rate of prolonged unemployment.

The demographic distribution of the population was illustrated with the proportion of Swedish- speaking population and adolescents and children. Native Swedish speakers were most prevalent in traditional good reputation areas: central (101-4) and western (105) parts of Helsinki and Kulosaari. Lowest rates were found in the northeastern suburbs (501-6). High percentage of adolescents and children as well as of married couples were most typical in northern (403, 405) and northeastern suburbs (503-5) while in the centre both rates were low (103-4, 301-3). Married couples were also prevalent in Kulosaari (601) and Laajasalo (603).

Finally, the composition of housing was exempli- fied with the proportion of owner-occupied housing and dwellings provided by the emloyer. Owner- occupied housing was most typical form of housing

in Northern Helsinki (401-2, 405, 504) where employer-provided housing was rare. The proportion of owner-occupied housing was lowest in Jakom/iki (506).

At this stage also the respondents' participation or failure to respond to the questionnaire by the small area variables was ascertained by correlating the response status with the small area variables. None of the variables was significantly related to the response status (data not shown).

The independent effect of small area characteristics as well as of adolescents' own sociodemographic characteristics was assessed by means of logistic regression analysis. It was performed step- wise using the statistical package SPSS for Windows (version 6.1) (Bryman and Cramer, 1994). In multivariate tables age was controlled first. After that all significant small area variables were fitted into the model and finally the individual socioeconomic variables were added.

Having determined the main effects model for each of the behaviours, the significance of the interactions between those individual level variables that were in the model and the small area variables (including those not in the model) were tested by fitting them into the main effects model. This was done to study whether the individual level socioeconomic differences varied by small area characteristics or not and to ensure that there were no higher order interactions between variables that were not themselves significant.

The models were fitted separately for girls and boys, as the preliminary (correlation) analyses suggested that the patterning of behaviour was different. Odds ratios (OR) for health behaviours were calculated. The significance of adding a variable in the model was tested by the Pearson chi- square test (P < 0.05).

RESULTS

First, the relationship between the small area characteristics was explored by correlating them with each other (Table 3). Bivariate correlations suggested that area characteristics of affluence tend to cluster together as the proportion of upper white collar workers correlated strongly with the proportion of Swedish population and low rate of unemployment. Those characteristics that describe relatively stable residence areas also combined. Here, these were low unemployment rates, high rates of married couples, owner-occupied housing and youth and children. Employer-provided dwellings was relatively independent of the other characteristics as shown by low to moderate correlations with the other characteristics. Such was the case with the rate of prolonged unemployment, although, not surprisingly, it correlated with overall unemployment rate.


Table 3. Pearson's correlation coeflicents between small area (n = 33) characteristics (P-values)

Small area characteristic I 2 3 4 5 6 7 8

1. Upper white collar workers 1.00 2. Swedish population 0.68

(0.000) 1.00 3. Unemployment -0.84 -0.45 1,00

(0.000) (0.009) 4. Married couples 0.45 0.14 -0.67 1.00

(0.008) (0.43) (0.000) 5. Owner-occupied housing 0.54 -0.02 -0.67 0.77 1.00

(0.001) (0.92) (0.000) (0.000) 6. Dwellings provided by the 0.37 0.44 -0.31 -0.32 -0.26 employer (0.03) (0.01) (0.08) (0.07) (0.15) 7. Youth and children -0.14 -0.40 -0.25 0.64 0.42

(0.45) (0.02) (0.16) (0.000) (0.02) 8. Prolonged unemployment -0.34 0.08 0.59 -0.32 -0.37

(0.05) (0.65) (0.000) (0.07) (0.04)

1.00

0.44 1.00 (0,01)

-0.13 0.52 (0.48) (0.02)

1.00

The distribution of respondents by their own sociodemographic characteristics is shown in Table 4. Boys and 18-year-olds were slightly under- represented among the respondents. In terms of the other sociodemographic characteristics most of the respondents had come from nuclear upper white collar families and had studied at upper secondary school. Obviously, as the adolescents comprised an urban population, nine out of ten of the respondents lived in urban central areas.

Most of these characteristics correlated significantly with the small area characteristics (data not shown). Their correlations can be summarised to suggest that nuclear upper white collar family background and studying at upper secondary school associates, not surprisingly, with living in areas with larger proportions of upper white collar families and employer-provided dwellings as well as a low proportion of adolescents and children. Also, a high proportion of Swedish population and a low proportion of married couples were related to some of these characteristics. The correlations were significant and ranged from 0.07 to 0.12.

Smoking

In multivariate tables daily smoking was associated with small area characteristics among girls only (Table 5). For girls, the percentage of adolescents in the area interacted with individual level of education although the main effect of the small area variable was not significant. Although significant, the interaction produced no marked effect on smoking patterns by level of education. According to the main effects model the odds for smoking increased with decreasing level of education, so that female school leavers had the highest odds of daily smoking.

Smoking was also highest among male school leavers. It was significantly higher in other schools as well than in the reference group. For boys from non-nuclear families smoking was significantly higher than in nuclear families.

Alcohol use

Alcohol use was not related to the small area characteristics among girls (Table 5). Among boys frequent heavy drinking increased with increasing rate of prolonged unemployment of the area

Table 4. The distribution of respondents by sociodemographic characteristics

Sociodemographic characteristics n %

Age 16 yrs 551 53 18 yrs 497 47

Gender Girls 570 54 Boys 478 46

Social status Upper white collar 426 42 Lower white collar 324 32

Manual workers 257 26 Level of education

Upper secondary school 673 64 Other school 242 23 School leaver 133 13

Family type Standard 696 67

Non-standard (mainly one-parent 342 33 families)

Location of residence Centre 922 90

Outside centres 102 10

Smal l a r ea va r i a t i on in hea l t h b e h a v i o u r 1095

Table 5. Final models according to logistic regression analyses predicting adolescents" health behaviour; 16 and 18 yrs

Health behaviour Gender Final model Model ;(2 df

Daily smoking Boys EDUC + FAM 49.0 3 Girls ADOL b + EDUC 10.2 3

Weekly alcohol use Boys AGE 45.8 1 Girls AGE + EDUC 9.6 2

Monthly drunkenness Boys AGE + P R O L N G + EDUC 14.2 2 Girls EDUC 20.3 2

Frequent physical Boys ~ activity

Girls O W N E R + U N E M P + PLACE + EDUC¢ 15.7 3 Abstaining dietary fat Boys ~' "

Girls P R O L N G 4.4 I

Individual level: EDUC = level of education, FAM = family type, PLACE = location of residence; small area level ADOL = rate of youth and children, P R O L N G = prolonged unemployment, O W N E R = owner-occupied U N E M P = unemployment;

aonly the constant; bincludes significant interaction: EDUC*ADOL; Cincludes significant interactions: E D U C * U N E M P , P L A C E * O W N E R

(italic): housing,

(Table 6). Thus, the odds for heavy drinking were over threefold in areas where prolonged unemployment was most prevalent compared to areas with lowest rate. Heavy drinking was also significantly related to boys' individual level of education, so that school leavers but not boys from other schools drank more heavily than the reference group. The

variation of heavy drinking by education was, how-

ever, smaller than that related to small area charac-

teristics.

Among girls, individual level of education was

significantly related to both the frequency of alco-

hol use and heavy drinking. The odds for alcohol

Table 6. Summary of the main effects logistic regression models showing associations between health behaviour, small areas and individual socioeconomic characteristics, expressed as relative odds; 16 and 18 yrs

Health behaviour Gender Small area/individual characteristic OR

Daily smoking Boys Level of education: upper secondary 1.00 other 2.7T'

school leaver 4.92 ~ Family type: nuclear 1.00

non-nuclear 1.87" Girls The rate of adolescents and children

low (6.8%) 1.00 high (35.3%) 0.93

Level of education: upper secondary 1.00 other 3.29 ~'

school leaver 5.93" Weekly alcohol use Boys b

Girls Level of education: upper secondary 1.00 other 1.20

school leaver 2.53 ~ Monthly drunkenness Boys Prolonged unemployment rate

low (31.0%) 1.00 high (45.0%) 3.26 ~

Level of education: upper secondary 1.00 other 1.34

school leaver 2.48 ~ Girls Level of education: upper secondary 1.00

other 2.22 ~ school leaver 3.35 ~'

Physical activity Boys Girls Owner-occupied housing

low (32.3%) 1.00 high (84.2%) 4.5P ~

Unemployment rate low (11.0%) 1.00 high (26.0%) 0.96

Level of education: upper secondary I.O0 other 0.63"

school leaver 0.34 Residence: centre 1.00

outside centre 2.90" Abstaining dietary fat Boys

Girls Prolonged unemployment rate low (31.0%) 1.00 high (45.0%) 0.33"

"95% significance level excludes 1.00. honly the constant.

1096 Sakari Karvonen and Arja H. Rimpelfi

use was highest among school leavers according to both drinking measures.

Physical activi O'

Among boys physical activity was not related to the small area characteristics. On the other hand a rather complex relationship between small area characteristics and physical activity was found for girls. The final model included four variables and two interactions (Table 5). The two interactions produced, however, only minor changes in the relationships estimated by the main the effects model. According to this model, physically active girls were more than four times more prevalent in areas with a high percentage of owner-occupied housing than in areas with the lowest percentage of this type of housing.

The individual level characteristics showed lower association with girls' physical activity than did small area characteristics. Still, physical activity fell significantly with lowering level of education. For school leavers the odds for physical activity were three times lower than for the reference category. For those living outside the centre of the city, physical activity was increased.

Abstaining from dietary fat

Abstaining from the use of dietary fat among girls was related to the rate of prolonged unemployment of the area. The highest percentage of dietary fat abstainers was found in districts with low rate of prolonged unemployment (Table 6). In highly unemployed areas the relative odds for abstention were as low as 0.33. Among boys, on the other hand, none of the small area factors were related to abstention from dietary fat.

DISCUSSION

The purpose of this study was to explore the patterns of health behaviour in small areas among young people. The study focused on one particular region, the city of Helsinki, to see to what extent small area characteristics correlate with modern health behaviour.

Indeed, clear small area effects were found in some of the behaviours. Consistently, these correlated higher with health behaviour than with individual level factors. Especially, unemployment in the area proved to be related to health behaviour as two significant relationships and a third suggestive one emerged in the study. Heavy drinking among boys and traditional dietary patterns (signified by less abstinence from dietary fat) among girls increased in areas with increased rate of prolonged unemployment. In addition to these, regular physical activity among girls was slightly lower in areas where the total unemployment rate was higher.

These results could, of course, derive from pro- blems in the data. There are three main areas that

could introduce this type of bias: selection of the respondents, the validity of the measures and link- ing data from different sources. The survey data may suffer from the selection of respondents from a small area. However, the response rates were not correlated with factors that described small areas, so it not likely that the selection of respondents took place systematically to the degree that the results were seriously biased.

Generally, the validity of self-reported behaviour measures has been assessed to be quite good (Akers et al., 1983). A recent study on dietary fat use by Roos (1995) has, however, shown that the validity of two measures used in this study--bread spread and milk--has decreased. Further, the study suggests that the extent to which these measures represent the overall level of dietary fat obtained from food is related to social status. Still, in Roos' study the two dietary measures appeared among the best indicators of dietary fat.

The small area statistics that were linked to the survey data were gathered at a different time than the individual level data. For example, the unemployment rates for small areas were obtained from population statistics at the end of 1994 while the median year for the pooled survey data was 1992. During this time a deep economic recession hit Finland (Santam~iki-Vuori and Sauramo, 1994). The average unemployment rate almost doubled within one year. According to a study by Lankinen (1994), however, unemployment increased in all small areas and the worst unemployment rates still can be found in those small areas with lowest socioeconomic status.

The consequences of this discrepancy in time between data sets were estimated by dividing the data into two subsets, one for 1989-1991, the other for 1993-1995, and multivariate analyses were run for both subsets. Overall, the results were similar in both subsets to the analyses using the whole data set. Still, as small area data describing the earlier study years were not available, changes over time in the small area effects remain possible.

To explain the relationship between unemployment, prolonged or overall, of an area and adolescents' health behaviour, a number of potential theories would have to be assessed ranging from broad contextual small area effects (Colby et al., 1994) to factors that enable or hinder lifestyle options (Abel, 1991). This is why only tentative in- terpretations can be offered here.

The increased heavy drinking among boys in areas with high level of prolonged unemployment is likely to result from a combination of cultural factors and availability of alcoholic beverages. Although studies of the relationship between unemployment and the general frequency alcohol use in individuals have shown somewhat inconsistent findings (Hammarstr6m, 1994; Lahelma et al., 1995), heavy drinking seems to be increasing, at least for

Small area variation in health behaviour 1097

some of the unemployed. Through the visibility of their drinking patterns the cultural acceptance of drunkenness may, then, be encouraged. At the same time, the company of frequent alcohol users may give adolescent boys increased access to alcohol. For girls, such company may not seem as attractive, given that it mainly consists of drunken males (Connolly et al., 1992).

The relationship between abstaining from dietary fat and prolonged unemployment can probably be interpreted more properly from factors that are associated with the unemployment rate of the area than the unemployment itself. This is why we suggest that the socioeconomic status of the area is reflected in dietary patterns. The fact that the individual measures of social class were not related to abstaining from dietary fat suggests that within the urban population there are influences that operate through the broader health culture. Health cultural influences facilitate behavioural options among adolescents of the whole area which in turn relates to the area structure of Helsinki. In Finland, unlike Britain for instance, there is less socioeconomic segregation among the population. Even the "better- off" areas include people from the lower social classes, and vice versa.

A different societal process probably lies behind the relationship between physical activity and owner-occupied housing. Factors related to better housing, not the quality of housing itself, may explain why girls' physical activity tended to increase with an increased proportion of owner- occupied housing. We suggest that the relationship derives from two groups of factors: better availability and access to high quality facilities combined with an upper class culture that highly favours fit- ness (Karisto, 1988; Karisto et al., 1993).

This interpretation is supported by our findings that showed that geographically the small area characteristics tend to cluster. Several of the measures of affluence of the area associated with each other. These "affluent" areas were also situated in the core areas of the city, such as the western and southern central areas of Helsinki, that traditionally have a better reputation. Respectively, Lankinen (1994) showed that the worst unemployment areas are still the areas with the lowest proportion of owner-occupied housing and highest rates of rented housing. According to Lankinen the absolute differences between small areas have been stable while the relative socioeconomic differences have increased. Thus, the recession seems not to have influenced variation between small areas. Although due to limited numbers in the data the geographical area patterns in health behaviour could not be analysed, our findings suggest that the socioeconomically poor areas may also host poor health behaviour among adolescents.

Our analysis suggests that the small area characteristics are patterned by gender. It appears that

girls' lifestyle is related to more complex factors, which may operate also small area level, than boys. Among boys, only heavy drinking varied by small area characteristics. However, to provide a full discussion of the matter of gender and health behaviours goes beyond the scope of this particular study. Thus, this study appears to agree with earlier studies (Pr~itt~il~i, 1988; Kelleher et al., 1992: Shopland et al., 1992) showing different patterns and covariates of health behaviour by gender. We conclude that gender significantly shapes both the personal and societal resources that one has for choosing between options of behaviours as well as the range of the options themselves. In other words, not only may resources be limited for a particular gender, but also gender operates at the level of opportunity structures; i.e. the choices that lead to adopting one behaviour may be perceived differ- ently by the two genders, respectively.

We explored whether the characteristics of small areas are related to health behaviour in Finnish adolescents. Some clear relationships between small area characteristics and health behaviour+ especially of girls, were found. On the other hand, at the small area level the other individual socioeconomic characteristics of adolescents were weakly correlated to health behaviour in urban environment. In most cases the small area characteristics have more powerful influence on health behaviour than an adolescent's own socioeconomic background. As the regional level apparently has lost its ability to serve as an indication of health behaviour, the local context may in future become more influential from the point of view of measuring behavioural variation.

REFERENCES

Abel, T. (1991) Measuring health lifestyles in a compara- tive analysis: theoretical issues and empirical findings. Social Science & Medicine 32, 899.

Akers, R. L., Massey, J., Clarke, W. and Lauer, R. M. (1983) Are self-reports of adolescent deviance valid? Biochemical measures, randomized response, and the bogus pipeline in smoking behaviour. Social Forces 61, 234.

Blaxter, M. (1990) Health and Lifestyles. Routledge, London.

Bryman, A. and Cramer, D. (1994) Quantitative Data Analysis for Social Scientists. Routledge, London.

Colby, J. P. Jr, Linsky, A. S. and Straus, M. A. (1994) Social stress and state-to-state differences in smoking and smoking related mortality in the United States. Social Science & Medicine 38, 373.

Connolly, G. M., Casswell, S., Stewart, J. and Silva, P. A. (1992) Drinking context and other influences on the drinking of 15-year-old New Zealanders. British Journal of Addiction 87, 1029.

Duncan, C., Jones, K. and Moon, G. (1993) Do places matter? A multi-level analysis of regional variations in health-related behaviour in Britain. Social Science & Medicine 37, 725.

1098 Sakari Karvonen and Aria H. Rimpel/i

EIm6n, H. and Sundh, V. (1994) Mortality in childhood, youth and early adulthood. European Journal o f Public Health 4, 274.

Hammarstr6m, A. (1994) Health consequences of youth unemployment--review from a gender perspective. Social Science & Medicine 38, 699.

Helsinki alueittain (1988) The City of Helsinki, Statistical Office, Helsinki.

lsohanni, M., Moilanen, I. and Rantakallio, P. (1991) Determinants of teenage smoking, with special reference to non-standard family background. British Journal of Addiction 86, 391.

Janlert, U. and Hammarstr6m, A. (1992) Alcohol consumption among unemployed youths: results from a prospective study. British Journal of Addiction 87, 703.

Johnson, Z., Jennings, S., Fogarty, J., Johnson, H., Lyons, R., Doorley, P. and Hynes, M. (1991) Behavioural risk factors among young adults in small areas with high mortality versus those in low-mortality areas. International Journal o f Epidemiology 20, 989.

Kandel, D. (1991) The social demography of drug use. Milbank Quarterly 69, 365.

Karisto, A. (1988) Liikunta ja el/imfintyylit. In Uuteen Liikuntakulttuuriin, ed. E. Sironen, p. 43. Vastapaino, Tampere.

Karisto, A., Pr/itt~il/i, R. and Berg, M-A. (1993) The good, the bad, and the ugly. Differences and changes in health related lifestyles. In Regulating Markets Regulating People. On Food and Nutrition Polio', eds U. Kjaernes et al., p. 185. Novus forlag, Oslo.

Karvonen, S. (1995) Regional differences in drinking among Finnish adolescents. Addiction 90, 57.

Karvonen, S. and Rimpel~, A. (in press) Regional diver- sity in smoking among Finnish adolescents a comparison of conventional administrative and cultural categorizations. Health and Place.

Kelleher, K., Rickert, V. I., Hardin, B. H., Pope, S. K. Farmer, F. I. (1992) Rurality and gender. Effects on early adolescent alcohol use. American Journal of Diseases of Children 146, 317.

Keskinen, V. (1991) Urbaania nuorta etsim/iss~i. In Nuoret ja muutos, eds J. L/ihteenmaa and L. Siurala, p. 175. Central Statistical Office of Finland Studies, Helsinki.

Klein, H. and Pittman, D. J. (1993) Regional differences in alcohol consumption and drinkers' attitudes toward drinking. American Journal of Drug and Alcohol Abuse 19, 523.

Lahelma, E., Kangas, R. and Manderbacka, K. (1995) Drinking and unemployment: contrasting patterns among men and women. Drug and Alcohol Dependence 37, 71.

Lankinen, M. ( 1 9 9 4 ) Taantuvatko ldhiOt? Pddkaupunkiseudun kerrostaloldhidt sosiaalisen segregaa- tion valossa. Helsingin kaupungin tietokeskus 3/1994, Helsinki.

Nuorten terveystapatutkimus (1979) Tutkimusohjelman Idh- tdkohdat ja toteutus. M48/79, Kansanterveystieteen julk- aisuja, Tampere.

Pr/itt~il~i, R. (1988) Socio-demographic differences in fat and sugar consumption patterns among Finnish adolescents. Ecology of Food and Nutrition 22, 53.

Pr~itt/il/i, R., Rahkonen, O. and Rimpel~i, M. (1986) Consumption patterns of critical fat sources among adolescents in 1977-1985. Nutrition Research 6, 485.

Rimpel/i, M., Rimpel~i, A., Rahkonen, O. and Teperi, J. (1988) The evolution of the Juvenile Health Habit Study 1977 1985. In Health Behaviour Research and Health Promotion, eds R. Anderson et al. Oxford University Press, Oxford,

Roos, E. (1995) Validity and comparison of three satu- rated fat indices. Scandinavian Journal of Nutrition 39, 55.

Santam~iki-Vuori, T. and Sauramo, P. (1994) The economic depression and youth unemployment in Finland. Young 2, 52.

Shopland, D. R., Niemcryk, S. J. and Marconi, K. M. (1992) Geographic and gender variations in total tobacco use. American Journal of Public Health 82, 103.

Siurala, L. (1991) Urbaanin nuoren vapaa-aika. Helsingin kaupungin tietokeskuksen tutkimuksia 7, Helsinki.

Townsend, P., Phillimore, P. and Beattie, A. (1988) Health and Deprivation. Inequality and the North. Croom Helm, Kent.

urban small area variation in adolescents' health behaviour

Documents