Background: School non-completion and early work disability is a great public health challenge in Norway, as in most western countries. This study aims to investigate how medically based disability pension (DP) among young adults varies geographically and how municipal socioeconomic conditions interact with non-completion of secondary education in determining DP risk. Methods: The study includes a nationally representative sample of 30% of all Norwegians (N = 350,699) aged 21–40 in 2010 from Statistic Norway’s population registries. Multilevel models incorporating factors at the individual, neighbourhood and municipal levels were applied to estimate the neighbourhood and municipality general contextual effects in DP receipt, and detect possible differences in the impact of municipal socioeconomic conditions on DP risk between completers and non-completers of secondary education. Results: A pattern of spatial clustering at the neighbourhood (ICC = 0.124) and municipality (ICC = 0.021) levels are clearly evident, indicating that the underlying causes of DP receipt have a systematic neighbourhood and municipality variation in Norway. Non-completion of secondary education is strongly correlated with DP receipt among those younger than 40. Socioeconomic characteristics of the municipality are also significantly correlated with DP risk, but these associations are conditioned by the completion of secondary education. Living in a socioeconomically advantageous municipality (i.e. high income, high education levels and low unemployment and social security payment rates) is associated with a higher risk of DP, but only among those who do not complete their secondary education. Although the proportion of DPs was equal in rural and urban areas, it is evident that young people living in urban settings are more at risk of early DP than their counterparts living in rural parts of the country when controlling for other risk factors. Conclusion: The association between school non-completion and DP risk varies between municipalities and local socioeconomic environments. The interplay between personal characteristics and the local community is important in DP risk among young adults, implying that preventive measures should be directed not only at the individual level, but also include the educational system and the local community. Keywords: Norway, Secondary education, Disability pension, Social inequality, Community characteristics, Geographic information systems, Multilevel * Correspondence: firstname.lastname@example.org Department of Neuromedicine and Movement Science, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology, Trondheim, Norway Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Myhr et al. BMC Public Health (2018) 18:682 Page 2 of 15 Background associated with fewer health-promoting behaviours The proportion of young adults prematurely leaving the [33, 39], higher morbidity [33, 40, 41] and all-cause labour market due to disability pensions (DPs) has mortality . A number of studies have examined increased significantly during the last decade . Recent the effect of local socioeconomic conditions on the statistics have shown that 1.8% of the Norwegian popu- incidence of DP receipt, but less attention has been lation between the ages of 18 and 29 receive DP, which paid to the variation in contextual risk across sub- is almost double since 2007 . The leading reason for groups of the population. It is plausible that the so- DP receipt among individuals below 40 years of age in cioeconomic context of the area may not equally Norway is mental illness . Brage and Thune  attrib- affect health for all people and certain personal char- uted this increase in part to more precise diagnostics, acteristics and features of the social environment may changes in health status and growing requirements in act as moderators [34, 43]. In other words, there may the labour market. Early work-life exit among young be statistical interactions between personal character- people is a great public health challenge and a threat to istics, features of the residential context and the the Nordic welfare state model, which depends on high health outcomes studied. According to the relative employment rates . Young adults who come to rely deprivation hypothesis, individuals who are disadvan- on social insurance benefits for most of their life course taged, relative to others in a certain neighbourhood, place a high socioeconomic burden on their society . will experience stress-inducing social comparisons, They also experience substantial lifetime consequences which may have adverse consequences for individual in terms of health and socioeconomic marginalisation health [34, 44]. [6, 7]. Increasing DP rates will, therefore, ultimately lead This study investigates how medically based DP among to a society with larger socioeconomic and health dispar- young adults varies geographically, and how municipal ities. DP receipt is also an important area for scientific socioeconomic conditions interacts with non-completion inquiry because DP is an indicator of society’s health sta- of secondary education in determining DP risk. tus as a whole, given that DP eligibility criteria are The specific aims of the study were: strictly medical . Social factors present at different levels of society, at (i). to explore geographic distributions of non- the individual, family, community and national levels, completion of secondary education and DP among strongly influence the health of young people . young adults in Norway; Individual factors related to early DP have been exten- (ii). to assess how neighbourhood and municipality sively investigated, showing a clear educational gradient differences relate to DP risk in young adulthood; and. with heavy clustering of DP among non-completers of (iii). to examine whether municipal socioeconomic secondary education [10, 11]. Low education achieve- conditions interact with the association between ment is associated with lower work participation, higher school non-completion and risk of DP in young risk of long-term socioeconomic marginalisation [12–15], adulthood. unemployment  and mental and physical health issues [9, 17, 18]. School non-completers are also far more likely Methods to receive DP [10, 11, 19] or depend on other medical and Study population non-medical public benefits early in life [10, 11, 20]. This study builds on a 30% random sample, stratified by Moreover, numerous studies have shown that childhood age, gender and municipality of residence of the entire adversities, such as parental disability and low socioeco- Norwegian population aged 21–40 years in 2010 (N = nomic status, are associated with physical and mental 395,514), extracted from Statistic Norway’s event data- health problems [21–27], low educational achievement base, FD-trygd . These data are linked to the [28, 29] and work disability [10, 30, 31] later in life. National Education Database (NUDB) through a unique A large body of research have linked area characteris- 11-digit personal identification number assigned to all tics, both physical and social, to a range of health behav- Norwegian citizens. Entitlement of DP was observed at iours and health related outcomes [32–35]. Official the end of 2010, when respondents were between the statistics demonstrate, for instance, large geographical ages 21 and 40. The main focus of this study centred on variations in DP recipient rates in Norway, and that the mechanisms for exclusion from working life. Hence, certain structural (contextual) factors may partly explain individuals entitled to a DP due to cognitive abnormal- this variation . Nordic population studies have shown ities (N = 527) (mainly those with extensive cognitive that the prevalence of DP correlates with municipal disabilities), most of whom never achieve ordinary paid socioeconomic conditions, such as economic develop- work, were excluded from the study. See Fig. 1 for inclu- ment, unemployment rate and education level [36–38]. sion and exclusion criteria for the present study. The Moreover, socioeconomically disadvantaged areas are final sample size was 350,699 individuals. The unique Myhr et al. BMC Public Health (2018) 18:682 Page 3 of 15 Fig. 1 Flow chart of the participants in the present study who where included in the analysis. The proportion of eligible subjects with complete data is 88.7% personal identification number allowed us to identify following criteria: (1) be between the ages 16 and 67, (2) information about the individual’s registered parents (or have been a member of the national insurance program caregivers). We merged the dataset with census informa- for at least 3 years (all residents of the country are mem- tion on individuals’ home municipalities, using macro bers) and (3) have undergone appropriate medical treat- statistics on demography, employment and economic ment and rehabilitation that might improve their development from the Norwegian Social Science Data earning ability. (NSD) regional database. Explanatory variables Measures The outcome variable Individual level For each individual, we sourced infor- Our dependent variable was whether the individual was mation on age, gender, employment record and parental registered as a DP recipient in the National Insurance DP from the FD-trygd database. Parental DP is known Administration (FD-trygd database) at the end of the to be associated with both low educational achievement follow-up period in 2010. In Norway, the eligibility and early DP [10, 30] and was, therefore, included as a criteria for granting a DP is strictly medical, based on an covariate in the analysis. NUDB provided secondary assessment that a person’s earning ability is permanently education data on non-completion, defined as having reduced by at least 50% due to illness, injury or disabil- not obtained a secondary education degree by age 21. ity. In addition, the applicants need to meet the The variable was used as both an explanatory and Myhr et al. BMC Public Health (2018) 18:682 Page 4 of 15 moderator variable in the final analysis. Official statistics variation due to unmeasured factors. Hence, we fitted a  have shown that men dominate the 20–29 group of three-level random intercept model by using maximum DP recipients, while women are overrepresented in the likelihood estimation [48–50] to distinguish the individ- 30–39 age group. This study accounted for the bias this ual, neighbourhood and municipality sources of vari- disparity could present by interacting age with gender in ation in DP receipt. The multilevel framework allows us the analysis. In other words, the model reflected the to simultaneously examine the effects of group-level and effects of gender at different ages. individual-level predictors while also accounting for non-independence of observations (clustering) within Neighbourhood level The FD-trygd database provides higher-level units. We modelled the prediction of DP re- information on neighbourhood of residence, which con- ceipt in young adulthood in 10 steps. First, we estimated stituted our second level of analysis. We used the indi- an “empty” model, only including a random intercept, vidual’s recorded census enumeration district in 2010, which represents the variation in DP between the three which is the lowest geographical level for Norwegian initial levels. This allowed us to determine the impact of population statistics, to identify their neighbourhoods the neighbourhood and municipality context in DP . The binary variable “rural” identifies the neighbour- receipt . Models 2–4 in Table 2 included all the hood of residence as rural or urban. Urban settlements individual level variables. In Table 3, we extended the have clusters of homes where at least 200 people live random intercept logit model for the relationship within a distance of 50 m or less, while the rural areas between school non-completion and DP risk to allow have a lower population density than this threshold . non-completion effect to vary across municipalities. Multilevel models with many random components are Municipality level The third unit of analysis comprised computationally demanding and, given our large dataset, all the 430 Norwegian municipalities in 2010. Norway’s such models became intractable. Thus, to keep the municipalities are subject to several common national model simple, a two-level random slope model (i.e. laws and regulations, which means that they represent individuals nested within municipalities) was fitted in relatively homogeneous and, therefore, comparable order to examine whether the relationship between units. Our model included a spatial lag variable and a school non-completion and DP risk varies between set of municipality characteristics that describe the municipalities. A likelihood ratio test (LR test) was socioeconomic conditions. The spatial lag variable is the used to compare the random intercept and the ran- mean of the age-adjusted DP rates in neighbouring mu- dom slope model’s goodness of fit. In the final steps, nicipalities and is included to account for spatial we included all the neighbourhood and municipality dependencies that may exist in the larger regional con- variables and adjusted for age, gender and parental text. Education level, defined as the percentage of per- DP receipt (Table 4). Models 2–5 added the inter- sons aged 20–39 who completed secondary education, action terms of non-completion of secondary educa- and income (the average gross income for all municipal tion with the municipality variables: education level residents aged 17 years and above) were used to evaluate (Model 2), gross household income (Model 3), the importance of the municipal socioeconomic environ- unemployment rate (Model 4) and the rate of social ment. The analysis also included the percentage of security payments (Model 5). inhabitants aged 20–39 years receiving unemployment To quantify the influence of neighbourhood and benefits and social security benefits, which reflect the municipality of residence in DP receipt, we computed socioeconomic environment more indirectly. The muni- the median odds ratio (MOR)  and the intraclass cipality characteristics enter the model as continuous correlation coefficients (ICCs) . The MOR trans- (grand mean centred), 2010 census variables (except the lates the area level variance on the log-odds scale income variable, which was only available for 2009), and and may, in our case, in a simplified way, be inter- are sourced from NSD’s regional database. preted as the increased median odds for receiving a DP if an individual lived in another neighbourhood Statistical approach (or municipality) with a higher risk [52, 53]. Thus, We investigated the relevance of the residential context the higher the MOR, the greater the contextual ef- as well as the association between municipal socioeco- fects. The ICC expresses the correlation in the out- nomic conditions and DP receipt in young adulthood come (i.e. DP receipt) between two individuals taken and tested the hypothetical interactions using logistic randomly from the same neighbourhood (or munici- multilevel models [48–50]. Individuals (level 1) are pality). By using the latent variable method [50, 54], nested within neighbourhoods (level 2, N = 12,894), which considers the variance from a standard logistic which are nested within municipalities (level 3, N = 430). distribution (π /3 = 3.29), we calculated the ICC with Each of these contexts may condition individual level the following formula: Myhr et al. BMC Public Health (2018) 18:682 Page 5 of 15 Results Var V ICC ¼ 2 Descriptive statistics Var V þ f Table 1 presents descriptive information for the indi- vidual and contextual variables among receivers and non-receivers of DP. At the end of 2010, a total of The percentage of proportional change in variance 7065 (2.0%) individuals were receiving DP, of which (PCV) at the respective levels quantifies the percentage 83.2% (N = 5876) had not completed secondary educa- of the variance of the empty model explainable by pre- tion. Of those who received DP, 37% (N = 2615) had dictor variables inserted into the more complex models no previous employment records registered in the . The model parameters were estimated by a mixed FD-trygd database, compared to 2.7% (N = 9191) in effects method using Stata/MP software (version 13). the non-receiving population. We used geographic information system (GIS) [56, 57] to explore and visualise the geographical patterns of Spatial pattern of secondary education non-completion school non-completion and DP receipt among young and DP rates among young adults adults. Measures of global spatial correlation were calcu- The geographical distribution of secondary education lated using the Global Moran’s I statistic and local indi- non-completion (Fig. 2) and DP rates (Fig. 3) differs cators of spatial association (LISA), which evaluates greatly among the Norwegian municipalities. The drop- whether the pattern expressed (i.e. school non-completion out rates in the 430 municipalities in Norway have a and DP receipt) is clustered, dispersed or random [58, 59]. clear geographical pattern (Fig. 2), with high dropout Finally, we used ArcGIS 10 for Desktop for the spatial rates in the northern and south-eastern regions, and low analysis. dropout rates in western Norway. The prevalence of DP Table 1 Individual and community characteristics of young receivers and non-receivers of disability pension (DP) Variable DP receivers (N = 7065) Non-receivers (N = 343,634) N% N % Individual level variables Male 3673 52.0 174,696 50.8 Mean Age (SD) 33.4 5.50 30.7 5.92 Secondary education non-completion 5876 83.2 110,719 32.2 Birth cohort 1970–1974 3204 45.4 94,279 27.4 1975–1979 1798 25.5 81,946 23.9 1980–1984 1180 16.7 80,537 23.4 1985–1989 883 12.5 86,872 25.3 No previous employment records 2615 37.0 9191 2,7 Maternal DP 1544 21.9 32,551 9.5 Mother’s identity unknown 282 4.0 23,570 6.9 Paternal DP 1218 17.2 27,781 8.1 Father’s identity unknown 114 1.6 18,669 5.4 Neighbourhood level variable Rural place of residence 1304 18.5 63,435 18.5 Municipality level variables Socioeconomic variables Mean SD Mean SD Secondary or tertiary education % 66.0 4.64 66.7 4.46 Unemployment % 3.3 0.76 3.2 0.71 Social security benefits % 6.1 2.07 5.7 1.86 Disability benefits % 9.7 2.91 8.7 2.76 Gross household income (thousands) 339.4 382.2 349.8 398.5 a b Significant difference (p-value≤0.05) between groups tested by chi square test Significant mean difference (p-value≤0.05) between groups tested by independent sample t-test Myhr et al. BMC Public Health (2018) 18:682 Page 6 of 15 Fig. 2 The geographic distribution of non-completers of upper secondary education (percentages) among individuals aged 21–40 years (born between 1970 and 1989 in Norway, 2010 among young adults varies from zero to 8.3%, with an autocorrelation (LISA) for both non-completion and DP average of 2.0% for the total country. rates confirmed these patterns, as does Fig. 4, which Measuring the global spatial correlation with the shows the results from both analyses. Specifically, high Moran’s I estimator revealed significant clustering in non-completion rates cluster in much of northern both school non-completion and DP rates, with correla- Norway, while clustering of DP rates is very limited here. tions of .23 and .12 and z-scores of 13.8 and 7.6, respect- High DP rates cluster in the southern region, but high ively. However, comparing Fig. 2 and Fig. 3 revealed non-completion rates do not. Finally, the western region distinct geographical distributions of the two groups. shows substantial overlap in low-rate clustering for both The spatial patterns, especially in the northern region, variables. showed a clear clustering of school leavers, but far less clustering of DP rates. The south showed concentrations Parametric estimation of municipalities with high DP rates without high drop- The prevalence of early DP at the neighbourhood and out rates, while western Norway showed low dropout municipality levels differs. In the first step, we estimated rates and low DP rates. An analysis of local an “empty” model. With only the second and third Myhr et al. BMC Public Health (2018) 18:682 Page 7 of 15 Fig. 3 The geographic distributions of disability pensions (percentages) among young adults aged 21–40 years (born between 1970 and 1989 in Norway, 2010 random intercepts in our model, we found that the ICC DP receipt after the age of 21–25 for the 1985–89-cohort, are 0.124 and 0.021. In other words, model 1 (in Table which complicates the comparison between the cohorts. 2) suggests that 12.4 and 2.1% of the variation in DP risk Being male, older than average or having parents (mother can be attributed to differences between neighbourhoods and/or father) who receive DP are all correlated with and municipalities, respectively. Table 2 shows the higher DP risk before age 40. The interaction term with individual and parental covariates of DP receipt in young age and gender are negative and statistical significant, adulthood. Non-completion of secondary education is indicating that the positive association between males and positively associated with DP receipt, and this association DP decreases over time. A complete lack of employment seems to have strengthened itself over the last two de- history is associated with the largest DP risk. cades. The association between school non-completion In Table 3, we extended the random intercept logit and DP receipt is stronger for individuals born in the model for the relationship between the probability of period 1985–1989 compared to their counterparts born receiving DP and non-completion of secondary educa- between 1970 and 1974. However, our data do not capture tion to allow the impact of non-completion to vary Myhr et al. BMC Public Health (2018) 18:682 Page 8 of 15 Fig. 4 Local indicators of spatial association (LISA) for secondary education non-completion rates and disability pension rates across municipalities. The two-level random intercept Fig. 3. Living in municipalities where neighbouring model, which is nested in the random slope model, is municipalities have high DP rates correlates with higher rejected at the 5% significance level (using a likelihood ratio DP risk, even when adjusting for individual and munici- test), suggesting that the impact of school non-completion pal socioeconomic variables. Models 3–5 (in Table 4) does vary between municipalities. suggest that there is co-variation between the municipal Turning to the neighbourhood and municipality socioeconomic environment and the individual DP risk. variables (Table 4), we found that rural settlement is as- However, these associations are conditioned by the sociated with lower risk of DP in young adulthood. This completion of secondary education. In other words, the corresponds well with the patterns observed in Fig. 3, effect of the municipal socioeconomic environment changes where clusters of some of the country’s highest DP rates dependent on whether or not the individual has completed are found in the densely populated areas of eastern and secondary education. Among non-completers, advantageous southern Norway. The spatial lag variable is positive and municipal socioeconomic conditions, such as high income significant, indicating that early DP has a regional clus- and education levels and low unemployment and social se- tering effect. This confirms the clustering mapped in curity payment rates, are all associated with higher DP risk. Myhr et al. BMC Public Health (2018) 18:682 Page 9 of 15 Table 2 The impact of non-completion of secondary education and its interaction with period of birth on the probability of receiving disability pension (DP) Model 1 Model 2 Model 3 Model 4 Coef SE Coef SE Coef SE Coef SE Fixed effects Secondary education non-completion 1.8604*** 0.0449 1.8063*** 0.0452 1.5831*** 0.0464 Cohort 1970–74 ref ref ref 1975–79 −0.3847*** 0.0819 −0.4013*** 0.0822 −0.1267 0.0851 1980–84 −0.6735*** 0.1243 −0.7075*** 0.1246 −0.2871* 0.1322 1985–89 −1.6452*** 0.1982 −1.6817*** 0.1987 −1.6253*** 0.2109 Non-completion*Cohort 1* 1970–74 ref ref ref 1*1975–79 0.5471*** 0.0775 0.5619*** 0.0777 0.4528*** 0.0799 1* 1980–84 0.7691*** 0.0958 0.7825*** 0.0960 0.4657*** 0.0995 1* 1985–89 1.7037*** 0.1528 1.7275*** 0.1529 1.2582*** 0.1551 Male 0.5092** 01526 0.4742** 0.1532 0.7858*** 0.1659 Age 0.0734*** 0.0092 0.0737*** 0.0092 0.1365*** 0.0101 Age*male −0.0177*** 0.0045 −0.0169*** 0.0045 −0.0222*** 0.0049 Maternal DP No ref ref Yes 0.6020*** 0.0318 0.5009*** 0.0354 Mother’s identity unknown 0.2322** 0.0808 −0.1565 0.0914 Paternal DP No ref ref Yes 0.4298*** 0.0347 0.2996*** 0.0388 Father’s identity unknown −1.6170*** 0.1218 −2.5678*** 0.1334 No previous employment records 3.7955*** 0.0405 Random effects Neighbourhood variance (95% CI) 0.4775 0.0289 0.4297 0.0285 0.4319 0.0288 0.3022 0.0283 PCV −10.0% −9.6% −36.7% ICC(%) 0.1240 0.0069 0.1133 0.0070 0.1142 0.0070 0.0827 0.0074 MOR 1.93 1.87 1.87 1.69 Municipality variance (95% CI) 0.0823 0.0129 0.0741 0.0127 0.0619 0.0115 0.0622 0.0119 PCV −10.0% −24.8% −24.4% ICC (%) 0.0214 0.0033 0.0195 0.0033 0.0164 0.0030 0.0170 0.0032 MOR 1.32 1.30 1.27 1.27 -2loglikeliehood 67,825.348 59,195.306 58,211.102 48,854.622 The proportional change in variance expresses the change in variance at the particular level from the empty model ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05 Keeping other variables constant, the predicted effects percentages with secondary or tertiary education). Doing of the municipal socioeconomic variables, such as edu- this reveals that the probability of receiving a DP among cation level, can be evaluated by adding together the school non-completers increases with increasing level municipal education level (percentages of inhabitants of municipal education level, whereas among school with secondary or tertiary education) and school completers the probability is more or less constant non-completion term and their interactions after filling (< 0.5%) regardless of education level (Fig. 5). Among in for different levels of municipal education (i.e. non-completers residing in a municipality with 60% Myhr et al. BMC Public Health (2018) 18:682 Page 10 of 15 Table 3 Parameter estimates and log-likelihood values for the random intercept and random slope logistic regression models Random intercept Random slope (coefficient) Parameter Coef SE Coef SE Individual level Intercept −5.1818*** 0.0322 −5.1694*** 0.0376 School non-completion 2.3159*** 0.0354 2.2977*** 0.0364 Municipality level random part Residual variance intercept 0.0914 0.0131 0.0941 0.0245 Residual variance slope 0.0337 0.0237 Intercept-slope covariance −0.0148 0.0208 -2Log likelihood 61,010.002 60,994.002 BIC 61,048.3 61,045.07 AIC 61,016 61,002 ***p < 0.001 Likelihood ratio test: LR chi2 = 16.65, p-value = 0.0002 of inhabitants with secondary or tertiary education Interactions with community peers and adults shape the the risk is 3.5%, and in a municipality with 75% the norms, values, aspirations and, ultimately, the behav- risk has increased to about 5%. iours of the residents. Hence, advantaged neighbour- hoods, where most adults have attained advanced formal Discussion education and steady jobs, will foster behaviours and This study examined how medically based DP among attitudes within the next generations that are conducive young adults varies geographically, and how municipal to success in both education and work. In less advan- socioeconomic conditions interact with non-completion taged communities, where the share of the population of secondary education in determining DP risk. Findings participating in the labour force is low and the depend- from the current study reinforce the relevance of the ency on welfare benefits is high, the positive attitudes residential context in DP risk among young adults. In toward education and work career may be less common. support of previous studies, we found that Rege, et al.  suggest that being disability may have non-completers of secondary education are more likely “contagion” effect in the community, meaning that the to receive early DP than completers. Our parametric es- propensity to receive DP increases when many people timation, however, suggest that the association between around you also depend on DP. Community characteris- school non-completion and DP receipt varies across tics represent more than the sum of their parts. Socio- municipalities. The key contributions of this study are economic factors at the level of individuals may fail to related to the exploration of how different municipal protect even the health of well-off people if they live in socioeconomic conditions interact with non-completion socioeconomically disadvantaged neighbourhoods , of secondary education, to alter DP risk in young adult- and socioeconomically privileged neighbourhoods may hood. Municipalities with high socioeconomic profiles impose an added health risk to the marginalised. are, in general, associated with both a lower risk of Not only did we find that municipal factors were non-completion and DP receipt [36–38, 60]. But this correlated with young inhabitants’ DP risk, but we also association does not hold for all groups. We found that uncovered regional effects. The DP rates in neighbour- non-completion has a stronger association with DP in ing municipalities are significantly associated with DP socioeconomically advantaged municipalities. In other risk, suggesting inter-municipal processes. Most munici- words, living in a high-status municipality (i.e. high palities are embedded within larger regional contexts, income, high education levels and low unemployment and common historical, political, economic, and cultural and social security payment rates) is associated with factors shape them. Todd  suggests that the inher- higher risk of DP among those who do not complete ited regional differences in social structure affect our their secondary education. practices and values, which, in turn, will dispose us to Spatial clustering of DPs, which is evident at the muni- think and act institutionally. In western Norway, a cipality level, can be interpreted in light of Wilson’s region with traditional Christian orthodoxy , people assumption that neighbourhood characteristics influence strongly value education and express this value by collective socialisation processes by shaping the type of attaining higher formal education. Official statistics show role models youth are exposed to outside their homes. that the population of western Norway, in general, has Myhr et al. BMC Public Health (2018) 18:682 Page 11 of 15 Table 4 The impact of non-completion of secondary education, municipal socioeconomic factors and their interactions on the probability of receiving disability pension (DP) Model 1 Model 2 Model 3 Model 4 Model 5 Coef SE Coef SE Coef SE Coef SE Coef SE Fixed effects Individual level Non-completion 2.2416*** 0.0329 1.0613* 0.4800 1.3491*** 0.2916 2.6458*** 0.1449 2.5850*** 0.1049 Neighbourhood level Rural settlement −0.3896*** 0.0404 −0.3894*** 0.0404 −0.3889*** 0.0404 −0.3896*** 0.0404 −0.3887*** 0.0404 Municipality level Socioeconomic factors Education 0.0226*** 0.0052 0.0074 0.0080 0.0221*** 0.0052 0.0225*** 0.0052 0.0224*** 0.0052 Household income −4.14e- 8.95e- −4.11e- 8.95e- −6.20e- 1.12e- −4.13e- 8.96e- −4.09e- 8.94e- 06*** 07 06*** 07 06*** 06 06*** 07 06*** 07 Unemployment −0.0286 0.0309 −0.0285 0.0308 −0.0283 0.0308 0.0760 0.0475 −0.0285 0.0308 Social security benefits 0.0127 0.0107 0.0129 0.0107 0.0130 0.0107 0.0129 0.0108 0.0611*** 0.0175 DP spatlag 0.0302** 0.0117 0.0303** 0.0117 0.0306** 0.0117 0.0300* 0.0117 0.0302** 0.0117 Interactions non-completion and municipality factors 1*Education 0.0177* 0.0072 1*household income 2.60e-06** 8.47e- 1*Unemployment −0.1248** 0.0433 1* Social security −0.0572*** 0.0164 benefits General contextual effects Neighbourhood variance 0.4214 0.0288 0.4206 0.0288 0.4201 0.0288 0.4213 0.0288 0.4206 0.0288 PCV −11.8% −11.9% −12.0% −11.8% −11.9% ICC (%) 0.1120 0.0070 0.1118 0.0070 0.1117 0.0070 0.1120 0.0070 0.1118 0.0070 MOR 1.86 1.86 1.86 1.86 1.86 Municipality variance 0.0503 0.0104 0.0503 0.0104 0.0503 0.0104 0.0503 0.0104 0.0500 0.0103 PCV −38.9% −38.9% −38.9% −38.9% −39.2% ICC (%) 0.0134 0.0027 0.0134 0.0027 0.0134 0.0027 0.0134 0.0027 0.0133 0.0027 MOR 1.24 1.24 1.24 1.24 1.24 -2loglikelihood 58,304,708 58,298,74 58,295,14 58,296,46 58,292,728 a b adjusted for age, gender and parental DP receipt The proportional change in variance expresses the change in variance at the particular level from the empty model ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05 better health and has the highest life expectancy in the association may emerge from risk factors not directly country . Thus, our finding of high secondary educa- modelled. Poor health in adolescence is, for example, tion completion rates and low DP prevalence in western strongly associated with school non-completion . Norway is not surprising. Like Markussen, et al. , we Such health problems may, indeed, lower the chances of found that northern Norway has a higher non-completion finding a job and increase the probability of receiving rate than any other region. Yet, this region has no system- DP in young adulthood. Our analysis, however, show atic clustering of high DP prevalence. Even outside of that the relationship between school non-completion our focal age group, the education level here is relatively and DP risk varies between municipalities and that the low and many employers have traditionally not required a municipal socioeconomic environment has a substantial secondary education. impact on this relationship. Previous Nordic population In line with previous Norwegian population studies studies have demonstrated a relationship between [10, 11], we found a strong association between school municipal socioeconomic conditions and DP prevalence non-completion and DP risk in young adulthood. This [36–38]. What stands out in our study is our finding Myhr et al. BMC Public Health (2018) 18:682 Page 12 of 15 Fig. 5 Predictive margins of school completers and non-completers predicting probability (Pr) of receiving DP by percentages of municipal residents with secondary or tertiary education that living in municipalities with high education and in- to Reime and Claussen , we found that rural settle- come levels and low unemployment and social security ment was associated with a lower risk of DP among benefit rates was associated with higher DP risk among young school leavers. The education level in rural non-completers. Advantageous socioeconomic condi- parts of Norway is generally lower compared to more tions are generally associated with increased individual urban areas, with easier access to jobs not requiring a probability of both completing secondary education and formal education. successfully entering the labour market [32, 38]. Never- Based on the relatively high portion of young people theless, population and community characteristics can, receiving DP in Norway, one might question how DPs indeed, interact with individual characteristics  and, are granted. Disability benefits granted to young people have differing impacts across population groups. are a substitute for lost income due to disability. In Non-completers in areas with a lower education level order to be entitled as a young disabled person, one may, according to the relative deprivation hypothesis must be under 26 years old upon becoming seriously [34, 44], be less at risk than non-completers in more and permanently ill, and said illness must be clearly socioeconomically advanced areas. Mishra and Carleton documented by a medical doctor/specialist. The causes  demonstrated that subjective feelings of relative of the increasing proportion of disabled young people in deprivation are linked with poorer physical and mental Norway are, however, highly complex and unclear . It health. Moreover, social distance, distrust and lack of is primarily mental illness that causes young people to cohesion between population groups often characterise become disabled. One important explanation is likely communities with high material and social inequalities that it has become more difficult for young people with . This may lead to higher stress levels, especially for mental illnesses to obtain and retain employment . those at the bottom of the social ladder, resulting in Another explanation of this growth is tied to changes in higher prevalence of stress-related morbidity and health social security schemes and expectations related to risk behaviours [71, 72]. Moreover, young adults without receiving valuable welfare schemes. As time-limited a secondary school degree may face greater difficulties in disability benefits were replaced with work-disbursement the labour market in societies with ample access to benefits in 2010, many were transferred to this new highly qualified applicants. Disparities between workers’ benefit. Today, about 70% of those who received tem- resources and structural features of the job market may porary disability benefits have been granted DP . negatively impact their health . School non-completion A major strength of our study is the use of large, may, in other words, be more detrimental and contribute nationally representative registry data with multiple to stronger health selection in socioeconomically advanced explanatory factors at the individual, family and neigh- areas. Hence, municipalities with seemingly strong socio- bourhood levels linked to population-based municipal economic profiles pose added risks for disadvantaged socioeconomic factors. The use of high quality, official young adults without a secondary school degree. Similar longitudinal registry data covering almost the entire Myhr et al. BMC Public Health (2018) 18:682 Page 13 of 15 Norwegian population greatly minimises the risk of between risk factors and early DP. Advantageous munici- selection bias and random errors in our analyses. How- pal socioeconomic conditions will, in general, increase ever, using such a large dataset introduces the risk of both the individual probability of completing secondary identifying significant, but inconsequential effects . education and successfully entering the labour market. Although studies based on large sample size have many However, among non-completers, these municipal condi- advantages, marginally significant effects observed in tions are associated with a higher risk of receiving DP in such studies typically mean that the predictive effect of young adulthood. Furthermore, living in rural communi- the exposure is quite modest . Moreover, a large ties lowers the risk of early DP. The mostly rural northern dataset with multiple explanatory variables introduces Norway has the highest non-completion rates in the coun- the risk of over-adjusting for inter-level confounding, in try without particularly high levels of DP. These commu- that contextual factors might determine some of the nities offer relatively well-paid jobs in the maricultural and individual level variables . We limited the risk of fishing industry that do not require high formal education. over-adjustment by including only a small number of Young adults with no previous employment records have individual and family level predictors that previous the highest risk of receiving DP. Young people who do not research has shown to affect DP risk. Our study has sev- finish secondary education are more marginalised in eral limitations. First, there are many methodological societies that place a higher weight on formal education. challenges in the analysis of neighbourhood contextual As more students complete their education, the potential effects, such as identification of the appropriate bound- marginalisation and barriers into the job market for those aries , endogeneity , structural confounding  who drop out increases. As our results suggest, environ- and excessive extrapolation in multilevel modelling . mental factors are important determinants of risk, and An obvious challenge related to the estimation of neigh- measures aimed at lowering DP rates will probably fail to bourhood effects in Norway is the major differences in reach their potential without an understanding of the risks population density between the different regions of the posed by the local environment. Future efforts to promote country. A 30% random and stratified sample of the social equality and successful transitions to adulthood population leads to a small number of study participants with regard to work and health should focus on the inter- in a significant proportion of neighbourhoods. Moreover, play between the local community and individual factors. the issue with selective residential mobility poses inter- Funding pretational challenges. For instance, advantageous socio- AM is funded by the Research Council of Norway and Frisknett AS (Industrial economic circumstances and healthier individuals tend PhD Scheme, Grant number: 208173/O30). to move to or remain in less deprived neighbourhoods Availability of data and materials . Finally, an ideal study should include longitudinal Due to the legislation governing scientific ethics, the data that support the explanatory data at multiple appropriate levels and allow findings of this study are only available on request in accordance with the the levels (i.e. the context) to change over time. How- agreement with the owner of the data, Statistic Norway, and the approver of the study, the Regional Committees for Medical and Health Research Ethics ever, our data do not allow us to control for this and, (REC) in Mid-Norway. Please see http://www.ssb.no/en/omssb/tjenester-og- thus, prevent the adoption of this analytic framework. verktoy/data-til-forskning for the procedure and requirements to obtain microdata from Statistic Norway. Conclusions Authors’ contributions This study underlines the importance of completing sec- AM, TH and TH designed and planned the study. AM and TH (i.e. T. Halvorsen) ondary education in the prevention of medically based structured and analysed the data, and AM, ML, TH and TH interpreted the data DP among young adults in modern society. However, and wrote the manuscript. All authors take responsibility for the integrity and accuracy of the data analysis and the decision to submit this paper for the study also demonstrates the significance of the resi- publication. All authors read and approved the final manuscript. dential context and local socioeconomic environment in individual variation in DP receipt. Low educational Ethics approval and consent to participate achievement and DP receipt have several central deter- The present study is based on retrospective analysis of registry data. The Regional Committees for Medical and Health Research Ethics (REK) of Mid-Norway approved minants in common, but comparing the geographical the study and the data linkage procedures (permission 2011/783). The ethic distributions of non-completion and DP rates reveal re- committee REK formally waived the need for consent. The exemptions were given gional divergence. The risk factors manifest themselves because our study used data registries where the information was collected from sources other than the persons themselves. at different structural levels, and a risk measured at the individual level may have a different effect when evalu- Competing interests ated at the municipal level. This creates divergence in The authors declare that they have no competing interests. the geographical distribution of non-completion and DP rates, and anything but a multilevel analysis would likely Publisher’sNote conflate these results. Moreover, this study suggests that Springer Nature remains neutral with regard to jurisdictional claims in the population under study largely defines the relationship published maps and institutional affiliations. Myhr et al. BMC Public Health (2018) 18:682 Page 14 of 15 Author details Epidemiol Community Health. 2011;65(1):57–63. https://doi.org/10.1136/jech. Department of Neuromedicine and Movement Science, Faculty of Medicine 2009.092569. and Health Sciences, Norwegian University of Science and Technology, 22. Boe T, Sivertsen B, Heiervang E, Goodman R, Lundervold AJ, Hysing M. Trondheim, Norway. Faculty of Nursing and Health Sciences, Nord Socioeconomic status and child mental health: the role of parental University, Levanger, Norway. Department of Health Research, SINTEF emotional well-being and parenting practices. J Abnorm Child Psychol. Technology and Society, Trondheim, Norway. 2013;42(5): 705–15. https://doi.org/10.1007/s10802-013-9818-9. 23. Boe T, Overland S, Lundervold AJ, Hysing M. Socioeconomic status and Received: 7 August 2017 Accepted: 7 May 2018 children's mental health: results from the Bergen child study. Soc Psychiatry Psychiatr Epidemiol. 2012;47(10):1557–66. https://doi.org/10. 1007/s00127-011-0462-9. 24. Galobardes B, Lynch JW, Smith GD. Is the association between childhood References socioeconomic circumstances and cause-specific mortality established? 1. OECD. Sickness, disability and work: breaking the barriers: a synthesis of Update of a systematic review. J Epidemiol Community Health. 2008;62(5): findings across OECD countries. Paris: OECD Publishing; 2010. https://doi. 387–90. https://doi.org/10.1136/jech.2007.065508. org/10.1787/9789264088856-en. 25. Chen E, Martin AD, Matthews KA. Socioeconomic status and health: 2. Ellingsen J. Utviklingen i uføretrygd1 per 31. Mars 2017 [in Norwegian]. Oslo: do gradients differ within childhood and adolescence? Soc Sci Med. Norwegian Labour and Welfare Administration (NAV); 2017. https://www. 2006;62(9):2161–70. nav.no/no/NAV+og+samfunn/Statistikk/AAP+nedsatt+arbeidsevne+og 26. Repetti RL, Taylor SE, Seeman TE. Risky families: family social environments and +uforetrygd+-+statistikk/Uforetrygd/Uforetrygd+-+Statistikknotater the mental and physical health of offspring. Psychol Bull. 2002;128(2):330. 3. Brage S, Thune O. Ung uførhet og psykisk sykdom [In Norwegian]. Oslo: Arbeid 27. Poulton R, Caspi A, Milne BJ, et al. Association between children's og Velferd: Norwegian Labour and Welfare Administration (NAV), 2015:37–49. experience of socioeconomic disadvantage and adult health: a life-course 4. Dølvik J, Fløtten T, Hippe J, Jordfald B. The Nordic model towards 2030. A study. Lancet. 2002;360(9346):1640–5. https://doi.org/10.1016/s0140- new chapter? NordMod2030: Final report. Fafo-report, 2015. 6736(02)11602-3. 5. OECD. Mental health and work. Paeris: OECD Publishing; 2013. https://doi. 28. Davis-Kean PE. The influence of parent education and family income on org/10.1787/9789264178984-en child achievement: the indirect role of parental expectations and the home 6. Avendano M, Berkman LF. Labor markets, employment policies, and health. environment. J Fam Psychol. 2005;19(2):294. In: Berkman L, Kawachi I,Glymour M, eds. Social epidemiology Second ed. 29. Myhr A, Lillefjell M, Espnes GA, Halvorsen T. Do family and New York: Oxford University Press. 2014:182–233. neighbourhood matter in secondary school completion? A multilevel 7. Bartley M, Ferrie J, Montgomery SM. Health and labour market study of determinants and their interactions in a life-course perspective. disadvantage: unemployment, non-employment, and job insecurity. Social PLoS One. 2017;12(2):e0172281. determinants of health. New York: Oxford University Press; 2009. p. 78–96. 30. Pape K, Bjorngaard JH, De Ridder KA, Westin S, Holmen TL, Krokstad S. 8. Claussen B. Restricting the influx of disability beneficiaries by means of law: Medical benefits in young Norwegians and their parents, and the experiences in Norway. Scand J Public Health. 1998;26(1):1–7. contribution of family health and socioeconomic status. The HUNT study, 9. Viner RM, Ozer EM, Denny S, et al. Adolescence and the social determinants Norway. Scand J Public Health. 2013;41(5):455–62. https://doi.org/10.1177/ of health. Lancet. 2012;379(9826):1641–1652. https://doi.org/10.1016/S0140- 6736(12)60149-4. 31. Harkonmaki K, Korkeila K, Vahtera J, et al. Childhood adversities as a 10. Gravseth HM, Bjerkedal T, Irgens LM, Aalen OO, Selmer R, Kristensen P. Life predictor of disability retirement. J Epidemiol Community Health. 2007;61(6): course determinants for early disability pension: a follow-up of Norwegian 479–84. https://doi.org/10.1136/jech.2006.052670. men and women born 1967-1976. Eur J Epidemiol. 2007;22(8):533–43. 32. Nieuwenhuis J, Hooimeijer P, Meeus W. Neighbourhood effects on https://doi.org/10.1007/s10654-007-9139-9. educational attainment of adolescents, buffered by personality and 11. De Ridder KA, Pape K, Johnsen R, Westin S, Holmen TL, Bjorngaard JH. educational commitment. Soc Sci Res. 2015;50:100–9. School dropout: a major public health challenge: a 10-year prospective 33. Reijneveld SA. Neighbourhood socioeconomic context and self reported study on medical and non-medical social insurance benefits in young health and smoking: a secondary analysis of data on seven cities. J adulthood, the young-HUNT 1 study (Norway). J Epidemiol Community Epidemiol Community Health. 2002;56(12):935–42. https://doi.org/10.1136/ Health. 2012;66(11):995–1000. https://doi.org/10.1136/jech-2011-200047. jech.56.12.935. 12. Bäckman O, Jakobsen V, Lorentzen T, Österbacka E, Dahl E. Early school 34. Stafford M, Marmot M. Neighbourhood deprivation and health: does it leaving in Scandinavia: extent and labour market effects. J Eur Soc Policy. affect us all equally? Int J Epidemiol. 2003;32(3):357–66. https://doi.org/10. 2015;25(3):253–69. 1093/ije/dyg084. 13. Bäckman O, Jakobsen V, Lorentzen T, Österbacka E, Dahl E. Dropping out in 35. Stafford M, McCarthy M. Neighbourhoods, housing, and health. In: Marmot Scandinavia social exclusion and labour market attachment among upper MG, Wilkinson RG, editors. Social determinants of health. 2nd ed. Oxford: secondary school dropouts in Denmark, Finland, Norway and Sweden: Oxford University Press; 2006. p. 297–317. Institute for Futures Studies; 2011. 36. Laaksonen M, Gould R. The effect of municipality characteristics on disability 14. OECD. Education at a glance 2014: OECD Indicators: OECD Publishing. p. 2014. retirement. Eur J Public Health. 2013;24(1):116–21. https://doi.org/10.1787/eag-2017-en 37. Reime LJ, Claussen B. Municipal unemployment and municipal typologies 15. Bäckman O, Nilsson A. Pathways to social exclusion—a life-course study. Eur as predictors of disability pensioning in Norway: a multilevel analysis. Scand Sociol Rev. 2010;27(1):107–23. https://doi.org/10.1093/esr/jcp064. J Public health. 2013;41(2):158–65. 16. Caspi A, Wright BRE, Moffitt TE, Silva PA. Early failure in the labor 38. Krokstad S, Magnus P, Skrondal A, Westin S. The importance of social market: childhood and adolescent predictors of unemployment in the characteristics of communities for the medically based disability pension. Eur J transition to adulthood. Am Sociol Rev. 1998;63(3):424–51. https://doi. Pub Health. 2004;14(4):406–12. https://doi.org/10.1093/eurpub/14.4.406. org/10.2307/2657557. 39. Kavanagh AM, Goller JL, King T, Jolley D, Crawford D, Turrell G. Urban area 17. Marmot MG, Bell R. Fair society, healthy lives. Public Health. 2012;126(Suppl disadvantage and physical activity: a multilevel study in Melbourne, 1):S4–10. https://doi.org/10.1016/j.puhe.2012.05.014. Australia. J Epidemiol Community Health. 2005;59(11):934–40. https://doi. 18. Marmot MG, Wilkinson RG. Social determinants of health. Oxford: Oxford org/10.1136/jech.2005.035931. University Press; 2006. 40. Subramanian SV, Kawachi I, Kennedy BP. Does the state you live in make a 19. Myhr A, Haugan T, Espnes GA, Lillefjell M. Disability pensions among young difference? Multilevel analysis of self-rated health in the US. Soc Sci Med. adults in vocational rehabilitation. J Occup Rehabil. 2015;26(1): 95–102. 2001;53(1):9–19. https://doi.org/10.1016/S0277-9536(00)00309-9. https://doi.org/10.1007/s10926-015-9590-5. 20. OECD. Society at a Glance 2016: OECD Social Indicators. Paris: OECD 41. Browning CR, Cagney KA. Neighborhood structural disadvantage, collective Publishing; 2016. https://doi.org/10.1787/9789264261488-en efficacy, and self-rated physical health in an urban setting. J Health Soc 21. Amone-P’olak K, Burger H, Huisman M, Oldehinkel AJ, Ormel J. Parental Behav. 2002;43(4):383–99. psychopathology and socioeconomic position predict adolescent offspring's 42. Marinacci C, Spadea T, Biggeri A, Demaria M, Caiazzo A, Costa G. The mental health independently and do not interact: the TRAILS study. J role of individual and contextual socioeconomic circumstances on Myhr et al. BMC Public Health (2018) 18:682 Page 15 of 15 mortality: analysis of time variations in a city of north West Italy. J 67. De Ridder KA, Pape K, Johnsen R, Holmen TL, Westin S, Bjorngaard JH. Epidemiol Community Health. 2004;58(3):199–207. https://doi.org/10. Adolescent health and high school dropout: a prospective cohort study of 1136/jech.2003.014928. 9000 Norwegian adolescents (the young-HUNT). PLoS One. 2013;8(9): 43. Wilkinson RG. Ourselves and others - for better or worse: social vulnerability e74954. https://doi.org/10.1371/journal.pone.0074954. and inequality. In: Marmot MG, Wilkinson RG, editors. Social determinants of 68. Shouls S, Congdon P, Curtis S. Modelling inequality in reported long term health. Oxford: Oxford University Press; 2006. p. 341–57. illness in the UK: combining individual and area characteristics. J Epidemiol Community Health. 1996;50(3):366–76. 44. Wilkinson RG. Ourselves and others–for better or worse: social vulnerability 69. Mishra S, Carleton RN. Subjective relative deprivation is associated with and inequality. In: Marmot MG, Wilkinson RG, eds. Social Determinants of poorer physical and mental health. Soc Sci Med. 2015;147:144–9. Health. Oxford Oxford University Press. 2006:341–57. 70. Kawachi I, Kennedy BP, Lochner K, Prothrow-Stith D. Social capital, income 45. Akselsen A, Lien S, Siverstøl Ø. FD-Trygd, list of variables. Oslo/Kongsvinger: inequality, and mortality. Am J Public Health. 1997;87(9):1491–8. Statistics Norway/Department of Social Statistics/Division for Social Welfare 71. Wilkinson RG. Health, hierarchy, and social anxiety. Ann N Y Acad Sci. 1999; Statistics; 2007. 896(1):48–63. 46. Ellingsen J. Utviklingen i uføretrygd per 31. mars 2018 [in Norwegian]: The 72. Wilkinson RG. Income inequality, social cohesion, and health: clarifying the Norwegian Labour and Welfare Administration; 2018. https://www.nav.no/ theory—a reply to Muntaner and lynch. Int J Health Serv. 1999;29(3):525–43. no/NAV+og+samfunn/Statistikk/AAP+nedsatt+arbeidsevne+og+uforetrygd 73. Dahl E, Elstad JI. Recent changes in social structure and health inequalities +-+statistikk/Uforetrygd/ in Norway. Scand J Public Health Suppl. 2001;55:7–17. 47. Statistics Norway. Population and area in urban settlements 2016 [In 74. Kann IC, Kristoffersen P. Arbeidsavklaringspenger – et venterom for Norwegian]. [Official statistics ]. https://www.ssb.no/befolkning/statistikker/ uførepensjon? [In Norwegian]. Arbeid og velferd: Norwegian Labour Welfare beftett/aar. Administration (NAV). 2014;2:101–15. 48. Goldstein H. Multilevel statistical models. London: Arnold; 1995. 75. Kaplan RM, Chambers DA, Glasgow RE. Big data and large sample size: a 49. Rabe-Hesketh S, Skrondal A., ed. Multilevel and longitudinal modeling using cautionary note on the potential for Bias. Clin Transl Sci. 2014;7(4):342–6. Stata, Volume II: Categorical responses, counts, and survival. Third ed. Texas: https://doi.org/10.1111/cts.12178. STATA press; 2012. 76. Lindström M, Merlo J, Östergren P-O. Social capital and sense of insecurity 50. Snijders TAB, Bosker RJ. Multilevel analysis: an introduction to basic and in the neighbourhood: a population-based multilevel analysis in Malmö, advanced multilevel modeling, second ed. London: Sage Publishers; 2012. Sweden. Soc Sci Med. 2003;56(5):1111–20. 51. Merlo J, Chaix B, Yang M, Lynch J, Råstam L. A brief conceptual tutorial of 77. Merlo J, Ohlsson H, Lynch KF, Chaix B, Subramanian S. Individual and multilevel analysis in social epidemiology: linking the statistical concept of collective bodies: using measures of variance and association in contextual clustering to the idea of contextual phenomenon. J Epidemiol Community epidemiology. J Epidemiol Community Health. 2009;63(12):1043–8. Health. 2005;59(6):443–9. 78. Oakes JM. The (mis) estimation of neighborhood effects: causal inference 52. Larsen K, Merlo J. Appropriate assessment of neighborhood effects on for a practicable social epidemiology. Soc Sci Med. 2004;58(10):1929–52. individual health: integrating random and fixed effects in multilevel logistic 79. Messer LC, Oakes JM, Mason S. Effects of socioeconomic and racial regression. Am J Epidemiol. 2005;161(1):81–8. residential segregation on preterm birth: a cautionary tale of structural 53. Merlo J, Viciana-Fernández FJ, Ramiro-Fariñas D. Population RGotLDotA. confounding. Am J Epidemiol. 2010;171(6):664–73. Bringing the individual back to small-area variation studies: a multilevel analysis 80. Norman P, Boyle P, Rees P. Selective migration, health and deprivation: of all-cause mortality in Andalusia, Spain. Soc Sci Med. 2012;75(8):1477–87. a longitudinal analysis. Soc Sci Med. 2005;60(12):2755–71. 54. Browne WJ, Subramanian SV, Jones K, Goldstein H. Variance partitioning in multilevel logistic models that exhibit overdispersion. J R Stati Soc: A (Stat Soc). 2005;168(3):599–613. 55. Merlo J, Chaix B, Ohlsson H, et al. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health. 2006;60(4):290–7. 56. Chaix B, Leyland AH, Sabel CE, et al. Spatial clustering of mental disorders and associated characteristics of the neighbourhood context in Malmö, Sweden, in 2001. J Epidemiol Community Health. 2006;60(5):427–35. 57. Chaix B, Merlo J, Subramanian S, Lynch J, Chauvin P. Comparison of a spatial perspective with the multilevel analytical approach in neighborhood studies: the case of mental and behavioral disorders due to psychoactive substance use in Malmö, Sweden, 2001. Am J Epidemiol. 2005;162(2):171–82. 58. Anselin L. Local indicators of spatial association—LISA. Geogr Anal. 1995; 27(2):93–115. 59. Anselin L, Getis A. Spatial statistical analysis and geographic information systems. Ann Reg Sci. 1992;26(1):19–33. 60. Ainsworth JW. Why does it take a village? The mediation of neighborhood effects on educational achievement. Soc Forces. 2002;81(1):117–52. https:// doi.org/10.1353/sof.2002.0038. 61. Wilson WJ. When work disappears: The world of the new urban poor: vintage, 2011. 62. Rege M, Telle K, Votruba M. Social interaction effects in disability pension participation: evidence from plant downsizing*. Scand J Econ. 2012;114(4):1208–39. 63. Todd E. The causes of progress. Culture, authority and change. Oxford: Blackwell Pub; 1987. 64. Rokkan S. Geography, religion and social class: crosscutting cleavages in Norwegian politics. In: Lipset SM, Rokkan S, editors. Party systems and voter alignments: cross-national perspectives. New York: Free press; 1967. 65. Norwegian Institute of Public Health. Oslo: Public health report: Life expectancy in Norway; 2016. https://www.fhi.no/en/op/hin/befolkning-og- levealder/levealderen-i-norge/ 66. Markussen E, Lødding B, Holen S. De’ hær e’kke nokka for mæ [in Norwegian]. Oslo: The Nordic Institute for Studies in Innovation, Research and Education (NIFU); 2012.
BMC Public Health – Springer Journals
Published: May 31, 2018