Bilevel Parameter Learning for Higher-Order Total Variation Regularisation Models.

We consider a bilevel optimisation approach for parameter learning in higher-order total variation image reconstruction models. Apart from the least squares cost functional, naturally used in bilevel learning, we propose and analyse an alternative cost based on a Huber-regularised TV seminorm. Differentiability properties of the solution operator are verified and a first-order optimality system is derived. Based on the adjoint information, a combined quasi-Newton/semismooth Newton algorithm is proposed for the numerical solution of the bilevel problems. Numerical experiments are carried out to show the suitability of our approach and the improved performance of the new cost functional. Thanks to the bilevel optimisation framework, also a detailed comparison between TGV 2 and ICTV is carried out, showing the advantages and shortcomings of both regularisers, depending on the structure of the processed images and their noise level.

Socioeconomic status and ischaemic heart disease mortality in 10 western European populations during the 1990s.

OBJECTIVE: To assess the association between socioeconomic status and ischaemic heart disease (IHD) mortality in 10 western European populations during the 1990s. DESIGN: Longitudinal study. SETTING: 10 European populations (95,009,822 person years). METHODS: Longitudinal data on IHD mortality by educational level were obtained from registries in Finland, Norway, Denmark, England/Wales, Belgium, Switzerland, Austria, Turin (Italy), Barcelona (Spain), and Madrid (Spain). Age standardised rates and rate ratios (RRs) of IHD mortality by educational level were calculated by using Poisson regression. RESULTS: IHD mortality was higher in those with a lower socioeconomic status than in those with a higher socioeconomic status among men aged 30-59 (RR 1.55, 95% confidence interval (CI) 1.51 to 1.60) and 60 years and over (RR 1.22, 95% CI 1.21 to 1.24), and among women aged 30-59 (RR 2.13, 95% CI 1.98 to 2.29) and 60 years and over (RR 1.36, 95% CI 1.33 to 1.38). Socioeconomic disparities in IHD mortality were larger in the Scandinavian countries and England/Wales, of moderate size in Belgium, Switzerland, and Austria, and smaller in southern European populations among men and younger women (p < 0.0001). For elderly women the north-south gradient was smaller and there was less variation between populations. No socioeconomic disparities in IHD mortality existed among elderly men in southern Europe. CONCLUSIONS: Socioeconomic disparities in IHD mortality were larger in northern than in southern European populations during the 1990s. This partly reflects the pattern of socioeconomic disparities in cardiovascular risk factors in Europe. Population wide strategies to reduce risk factor prevalence combined with interventions targeted at the lower socioeconomic groups can contribute to reduce IHD mortality in Europe.

Education level inequalities and transportation injury mortality in the middle aged and elderly in European settings.

OBJECTIVE: To study the differential distribution of transportation injury mortality by educational level in nine European settings, among people older than 30 years, during the 1990s. METHODS: Deaths of men and women older than 30 years from transportation injuries were studied. Rate differences and rate ratios (RR) between high and low educational level rates were obtained. RESULTS: Among men, those of low educational level had higher death rates in all settings, a pattern that was maintained in the different settings; no inequalities were found among women. Among men, in all the settings, the RR was higher in the 30-49 age group (RR 1.46, 95% CI 1.32 to 1.61) than in the age groups 50-69 and > or = 70 years, a pattern that was maintained in the different settings. For women for all the settings together, no differences were found among educational levels in the three age groups. In the different settings, only three had a high RR in the youngest age group, Finland (RR 1.33, 95% CI 1.01 to 1.74), Belgium (RR 1.38; 95% CI 1.13 to 1.67), and Austria (RR 1.49, 95% CI 0.75 to 2.96). CONCLUSION: This study provides new evidence on the importance of socioeconomic inequalities in transportation injury mortality across Europe. This applies to men, but not to women. Greater attention should be placed on opportunities to select intervention strategies tailored to tackle socioeconomic inequalities in transportation injuries.

Neighbourhood unemployment and all cause mortality: a comparison of six countries.

STUDY OBJECTIVE: Studies have shown that living in more deprived neighbourhoods is related to higher mortality rates, independent of individual socioeconomic characteristics. One approach that contributes to understanding the processes underlying this association is to examine whether the relation is modified by the country context. In this study, the size of the association between neighbourhood unemployment rates and all cause mortality was compared across samples from six countries (United States, Netherlands, England, Finland, Italy, and Spain). DESIGN: Data from three prospective cohort studies (ARIC (US), GLOBE (Netherlands), and Whitehall II (England)) and three population based register studies (Helsinki, Turin, Madrid) were analysed. In each study, neighbourhood unemployment rates were derived from census, register based data. Cox proportional hazard models, taking into account the possible correlation of outcomes among people of the same neighbourhood, were used to assess the associations between neighbourhood unemployment and all cause mortality, adjusted for education and occupation at the individual level. RESULTS: In men, after adjustment for age, education, and occupation, living in the quartile of neighbourhoods with the highest compared with the lowest unemployment rates was associated with increased hazards of mortality (14%-46%), although for the Whitehall II study associations were not statistically significant. Similar patterns were found in women, but associations were not statistically significant in two of the five studies that included women. CONCLUSIONS: Living in more deprived neighbourhoods is associated with increased all cause mortality in the US and five European countries, independent of individual socioeconomic characteristics. There is no evidence that country substantially modified this association.

Trends in socioeconomic disparities in stroke mortality in six european countries between 1981-1985 and 1991-1995.

This study assesses whether stroke mortality trends have been less favorable among lower than among higher socioeconomic groups. Longitudinal data on mortality by socioeconomic status were obtained for Finland, Norway, Denmark, Sweden, England/Wales, and Turin, Italy. Data covered the entire population or a representative sample. Stroke mortality rates were calculated for the period 1981-1995. Changes in stroke mortality rate ratios were analyzed using Poisson regression and compared with rate ratios in ischemic heat disease mortality. Trends in stroke mortality were generally as favorable among lower as among higher socioeconomic groups, such that socioeconomic disparities in stroke mortality persisted and remained of a similar magnitude in the 1990s as in the 1980s. In Norway, however, occupational disparities in stroke mortality significantly widened, and a nonsignificant increase was observed in some countries. In contrast, disparities in ischemic heart disease mortality widened throughout this period in most populations. Improvements in hypertension prevalence and treatment may have contributed to similar stroke mortality declines in all socioeconomic groups in most countries. Socioeconomic disparities in stroke mortality generally persisted and may have widened in some populations, which fact underlines the need to improve preventive and secondary care for stroke among the lower socioeconomic groups.

Socioeconomic inequalities in mortality among elderly people in 11 European populations.

STUDY OBJECTIVE: To describe mortality inequalities related to education and housing tenure in 11 European populations and to describe the age pattern of relative and absolute socioeconomic inequalities in mortality in the elderly European population. DESIGN AND METHODS: Data from mortality registries linked with population census data of 11 countries and regions of Europe were acquired for the beginning of the 1990s. Indicators of socioeconomic status were educational level and housing tenure. The study determined mortality rate ratios, relative indices of inequality (RII), and mortality rate differences. The age range was 30 to 90+ years. Analyses were performed on the pooled European data, including all populations, and on the data of populations separately. Data were included from Finland, Norway, Denmark, England and Wales, Belgium, France, Austria, Switzerland, Barcelona, Madrid, and Turin. MAIN RESULTS: In Europe (populations pooled) relative inequalities in mortality decreased with increasing age, but persisted. Absolute educational mortality differences increased until the ages 90+. In some of the populations, relative inequalities among older women were as large as those among middle aged women. The decline of relative educational inequalities was largest in Norway (men and women) and Austria (men). Relative educational inequalities did not decrease, or hardly decreased with age in England and Wales (men), Belgium, Switzerland, Austria, and Turin (women). CONCLUSIONS: Socioeconomic inequalities in mortality among older men and women were found to persist in each country, sometimes of similar magnitude as those among the middle aged. Mortality inequalities among older populations are an important public health problem in Europe.

Effects of the characteristics of neighbourhoods and the characteristics of people on cause specific mortality: a register based follow up study of 252,000 men.

STUDY OBJECTIVE: To assess the strength of the associations between neighbourhood characteristics and mortality, after adjusting for individual characteristics. DESIGN AND SETTING: 1990 census records of over 25 year old men in the Helsinki Metropolitan area linked to death records in 1991-1995; almost 1.22 million person years and 15 000 deaths. Individual characteristics were education, occupation based social class, housing tenure, housing density, and living arrangements. Proportion of manual workers, proportion of over 60 year olds, and social cohesion were measured for 55 small areas, and SAS Glimmix was used to fit multilevel models. MAIN RESULTS: Men in areas with high proportion of manual workers and low social cohesion have high mortality, particularly among 25-64 year olds. About 70% of this excess mortality is explained by compositional differences of people living in these areas. Accidents and violence, circulatory diseases, and alcohol related causes contribute most to these area effects. Area characteristics do not consistently modify or mediate the effects of individual socioeconomic characteristics on mortality. CONCLUSIONS: As compared with individual characteristics neighbourhood characteristics have modest independent effects on male mortality. Furthermore, individual socioeconomic characteristics are associated with mortality independently of area characteristics. Rather than the characteristics of areas, other social contexts, such as peer groups and family settings may be more fruitful targets for further research and policy on contextual effects on mortality.

Change in male and female life expectancy by social class: decomposition by age and cause of death in Finland 1971-95.

STUDY OBJECTIVE: To quantify the contribution of different causes of death and age groups for trends in life expectancy for two major social classes. DESIGN AND SETTING: Prospective study of mortality in Finland among all over 35 year old men and women. Baseline social class (manual/non-manual) was from the 1970, 1975, 1980, 1985 and 1990 census records, and follow up was by computerised record linkage to death certificates for 1971-1995. MAIN RESULTS: From the early 1970s to the early 1990s life expectancy at age 35 increased by about five and four years among Finnish men and women respectively, with largest gains among 55-74 year old men and 65-84 year old women. Life expectancy increase was 5.1 years among non-manual and 3.8 years among manual men; corresponding figures for women were 3.6 and 3.0 years. In the 1980s, when differences in life expectancy increased most rapidly, decline in cardiovascular disease mortality was more rapid in the non-manual than the manual class. Furthermore, increasing mortality for alcohol associated causes, "other diseases", and accidents and violence were most prominent in the manual class. CONCLUSIONS: Explanations of increasing social inequalities in mortality that are based on one underlying factor are difficult to reconcile with the variability in the cause specific trends in social inequalities in mortality. The contribution of older ages to social inequalities in mortality should be more widely recognised.

Income differences in mortality: a register-based follow-up study of three million men and women.

OBJECTIVES: We analyse whether the relationship between net household income and mortality form a continuous linear gradient or is curvilinear, assess the attenuation of this association after adjusting for confounding and reverse causality, and assess the strength of the association by age and cause of death. DESIGN AND SETTING: Prospective study of mortality in Finland among all men and women over 30 years old. Information on household income and sociodemographic factors was from the records of the Finnish tax authorities and the 1990 census. Income data were available for more than 95% of the cohort. Follow-up was by record linkage to death certificates in 1991-1996; altogether about 261 000 deaths. RESULTS: The all-cause mortality ratio between the lowest and the highest household income decile is 2.37 (95% CI : 2.30-2.44) among men and 1.73 (95% CI : 1.67-1.80) among women. Adjusting for household structure, spouse's economic activity, social class, education and own economic activity attenuates the relationship by 61% among men and 52% among women. The association between income and mortality is mainly linear before and after adjusting for confounding, and the association is strong for all 5-year age groups below 60-64 years, after which it declines rapidly in strength. CONCLUSIONS: The mainly linear nature of the relationship and the strong attenuation after adjustment for other socioeconomic factors and economic activity status, and the age pattern of the relationship indicate that a large part of the relationship is unlikely to be due to direct causal effects of poverty and material hardship. Rather, income seems to be related to accumulation of factors that increase mortality over the whole range of incomes.

Mortality differences by parental social class from childhood to adulthood.

STUDY OBJECTIVE: To examine mortality differences by parental social class and cause of death from age 5 to age 34. DESIGN: Register-based follow up study based on census records for 1985 and 1990 linked with death records for the period 1987-95. SETTING AND SUBJECTS: The study covers all males and females in non-manual and manual classes in Finland aged 5-34 years in 1987-95 (8135 deaths). Parental social class is defined on the basis of the occupation of the head of household at the time the child was 0-14 years. MAIN OUTCOME MEASURES: All cause mortality, mortality from diseases, mortality from accidents and violence, and alcohol related mortality during the period 1987-95. MAIN RESULTS: At ages 5-14 there is no systematic gradient in mortality by parental social class. Both absolute and relative differences increase with age. The relative rate of male all cause mortality among manual class descendants at ages 25-29 compared with that of upper non-manual class descendants is 1.60 (95% CI 1.37, 1.86). At ages 30-34 the relative rate among males is 1.95 (95 % CI 1.58, 2.42) and among females 1.47 (95% CI 1. 03, 2.10). Among males alcohol related causes of death account for 70% of the excess mortality of sons of manual class parents compared with sons of upper non-manual class parents at ages 25-34. At ages 25-34, both among females and males, the contribution of diseases to the mortality difference increases. CONCLUSIONS: Parental social class has an impact on mortality after childhood mainly through health related behaviours and lifestyles up to age 34.

Diminishing educational differences in breast cancer mortality among Finnish women: a register-based 25-year follow-up.

OBJECTIVES: This study examined trends in breast cancer mortality by education, age, and birth cohort. METHODS: Census records of Finnish women 35 years and older were linked with death records for 1971 through 1995. RESULTS: Excess breast cancer mortality of more-educated women has declined rapidly, mainly because of increasing mortality among less-educated women and stable or decreasing mortality among more-educated 35- to 64-year-old women. During the 1990s, mortality among more-educated 50- to 64-year-old women declined particularly fast. CONCLUSIONS: The causes of declining differences by education in breast cancer mortality are difficult to verify, but they may be due in part to narrowing differences in reproductive behavior among the younger birth cohorts and to a period effect possibly associated with the introduction of breast cancer screening in the late 1980s.

Socioeconomic inequalities in mortality among women and among men: an international study.

OBJECTIVES: This study compared differences in total and cause-specific mortality by educational level among women with those among men in 7 countries: the United States, Finland, Norway, Italy, the Czech Republic, Hungary, and Estonia. METHODS: National data were obtained for the period ca. 1980 to ca. 1990. Age-adjusted rate ratios comparing a broad lower-educational group with a broad upper-educational group were calculated with Poisson regression analysis. RESULTS: Total mortality rate ratios among women ranged from 1.09 in the Czech Republic to 1.31 in the United States and Estonia. Higher mortality rates among lower-educated women were found for most causes of death, but not for neoplasms. Relative inequalities in total mortality tended to be smaller among women than among men. In the United States and Western Europe, but not in Central and Eastern Europe, this sex difference was largely due to differences between women and men in cause-of-death pattern. For specific causes of death, inequalities are usually larger among men. CONCLUSIONS: Further study of the interaction between socioeconomic factors, sex, and mortality may provide important clues to the explanation of inequalities in health.

Bias related to the exclusion of the economically inactive in studies on social class differences in mortality.

BACKGROUND: To assess how the exclusion of the economically inactive affects levels and trends in social class differences in mortality among men and women at different durations of follow-up. METHODS: Records of the 1970, 1975, 1980 and 1985 censuses on Finnish men and women aged 35-64 linked with records of all deaths during 1971-1990. RESULTS: Exclusion of the economically inactive population underestimates the class differences in the total population by about 25% among men and 60% among women. The bias does not disappear if the first 5 years of follow-up are excluded and the bias can lead to erroneous conclusions about the trends in social class differences in mortality. CONCLUSIONS: Analyses based on the economically active population may lead to significant underestimation of social class differences in mortality, introduce biases in international comparison and may only partially capture the causal mechanisms underlying these mortality differences. Our results further show that although the bias diminishes during the follow-up, it is by no means eliminated after the first 5 years. The underestimation of social class differences in mortality created by the exclusion of the inactive should be more widely recognized and more accurate data on previous occupations should be collected.

Occupational class and ischemic heart disease mortality in the United States and 11 European countries.

OBJECTIVES: Twelve countries were compared with respect to occupational class differences in ischemic heart disease mortality in order to identify factors that are associated with smaller or larger mortality differences. METHODS: Data on mortality by occupational class among men aged 30 to 64 years were obtained from national longitudinal or cross-sectional studies for the 1980s. A common occupational class scheme was applied to most countries. Potential effects of the main data problems were evaluated quantitatively. RESULTS: A north-south contrast existed within Europe. In England and Wales, Ireland, and Nordic countries, manual classes had higher mortality rates than nonmanual classes. In France, Switzerland, and Mediterranean countries, manual classes had mortality rates as low as, or lower than, those among nonmanual classes. Compared with Northern Europe, mortality differences in the United States were smaller (among men aged 30-44 years) or about as large (among men aged 45-64 years). CONCLUSIONS: The results underline the highly variable nature of socioeconomic inequalities in ischemic heart disease mortality. These inequalities appear to be highly sensitive to social gradients in behavioral risk factors. These risk factor gradients are determined by cultural as well as socioeconomic developments.

The effects of differential unemployment rate increases of occupation groups on changes in mortality.

OBJECTIVES: This study estimated the effects of changes in unemployment rates of occupation groups on changes in mortality in a period of increasing unemployment. METHODS: Census records for all 20- to 64-year-old economically active Finnish men in 1985 were linked to information on unemployment and deaths in 1987 through 1993. RESULTS: Change in mortality was similar in occupation groups in which unemployment rates increased at a different pace. These relationships were similar for all age groups and for mortality from diseases as well as accidents and violence. CONCLUSIONS: Unemployment does not seem to cause mortality in the short term. Excess mortality rates among unemployed individuals observed in previous studies may have been due in part to selection.

Do education and income buffer the effects of death of spouse on mortality?

In this paper, we estimate the effects of the death of a spouse on the mortality of the survivor in different education and income groups. These socioeconomic resources may buffer the harmful effects of the stressful life event of the loss of one's spouse. The data come from a prospective study of mortality among all 35- to 74-year-old married Finnish persons. Follow-up was established by record linkage to death certificate registers for 1986-1991 (about 86,000 deaths, of which almost 5,500 occurred among the bereaved). The relative mortality after the death of one's spouse was broadly similar in different education and income groups. Absolute differences in mortality rates between bereaved and nonbereaved persons were larger in the lower end of the social spectrum, however. This pattern held for four broad categories of death: both sexes and two age groups (35-64 years and 65-74 years). The degree to which socioeconomic resources buffer the effects of death of spouse depends on whether it is assessed in terms of rate differences or rate ratios. Nevertheless, regardless of measurement choice, the effects of bereavement exist in all socioeconomic groups analyzed in this study. Furthermore, because of the high absolute level of mortality, the burden of excess mortality experienced after the death of one's spouse is heavier in the lower social strata.

Socioeconomic inequalities in health expectancy in Finland and Norway in the late 1980s.

Studies on health inequalities have usually focused either on mortality or on morbidity. This concerns national studies as well as international comparisons of health inequalities. This paper seeks to bridge the gap by applying health expectancy as a synthetic overall measure of health. The purpose of the study is to compare socioeconomic inequalities in health expectancy in Finland and Norway in the late 1980s. Additionally, the major methodological issues in the use of health expectancy in the study of health inequalities are identified. Data on mortality by level of education derive from linked national follow-up studies (1986-1990) of population censuses. Data on the prevalence of morbidity by level of education derive from nationally representative surveys of the noninstitutionalised adult population in 1985/87. Persons aged 25-74 years were included. Four measures of morbidity were used: limiting long-standing illness, extremely limiting long-standing illness, functional disabilities and perceived less than good health. The association between mortality/morbidity and level of education in each 5-year age/sex group was determined by a regression-based method. Partial life expectancies and partial health expectancies for ages 25-74 were then calculated by using the mortality quotients and morbidity prevalences predicted by the regression model for those at the top and the bottom of the educational hierarchy in each 5-year age group, using an application of the method first presented by Sullivan. Although various measures of health expectancy were used, the result were consistent. In absolute terms the size of socioeconomic inequalities in health expectancy in Finland and Norway is on the same level. In relative terms, however, the size of inequalities in health expectancy is greater in Norway. If one considers premature mortality to be more severe than any indicator of morbidity, the mortality morbidity mix of the health inequalities is less favourable to Finland, since the size of absolute inequalities in mortality is greater in Finland. Health expectancy measures provide a promising measure for assessing and comparing the pattern and the size of health inequalities.

Mortality in a large population-based cohort of patients with drug-treated diabetes mellitus.

OBJECTIVES: This paper presents detailed cause-specific data about excess mortality among diabetic persons in Finland, by age and sex. METHODS: Five-year follow-up data on the Finnish population aged 30 through 74 years were analyzed. During these 5 years, 11,215 persons with diabetes and 102,843 persons without diabetes died. The diabetic population was defined as people who were entitled to free medication for diabetes at the beginning of the follow-up period, that is, at the end of 1980. RESULTS: The relative mortality of persons with drug-treated diabetes compared with nondiabetic persons was higher among women (3.4) than among men (2.4). Almost three quarters of the mortality excess was due to circulatory diseases. For most other causes of death, too, diabetic persons had higher than average mortality. The exceptions were lung cancer, chronic obstructive pulmonary disease, and alcohol poisoning. CONCLUSIONS: Diabetes is a general risk factor for untimely death and makes a significant contribution to overall national death rates, particularly for circulatory diseases. Lower than average mortality from smoking-related diseases and alcohol poisoning, however, warrant optimism about the effects of health education among diabetic persons.

Sociodemographic mortality differences among the oldest old in Finland.

This study examined mortality differences and trends by several sociodemographic characteristics among the Finnish elderly aged 80 years or over during the period of 1971-90. The analyses were based on comprehensive data sets compiled by means of linking individual death records and census records for the entire population of Finland. Poisson regression was applied as the main statistical tool. For both sexes, life expectancy at age 80 was about 1 year longer among those with a higher education than among those with basic education. A similar difference was found between former upper nonmanual workers and manual workers. Slightly lower than average mortality was observed among the married, among those living in Western Finland, and among the Swedish-speaking population. Mortality declined during the study period in all subgroups, with no consistent signs of either convergence or divergence of mortality levels. The results suggest that at least some further decline of mortality even among the oldest old is possible.