The burden of diabetes mortality in Finland 1988-2007 - A brief report

Increasing incidence of especially type 2 diabetes has been reported in many countries worldwide [1]. At the same time studies have reported higher rates of premature morbidity and mortality among diabetics than in the general population irrespective of the country, study period or age-group studied [see e.g. [2–7]]. For example, in a meta-analysis including data from 37 studies worldwide the excess risk for fatal coronary heart disease was estimated to be 3.5 among women and 2.06 among men [3]. The disease burden related to diabetes has been reported to be distributed unevenly across the population. Earlier studies have reported higher diabetes prevalence [8], higher rates of diabetes related complications and excess mortality [9–12] among those with lower as compared to those with higher socioeconomic position. For example, a study from ten European countries reported an overall prevalence ratio of 1.6 of diabetes among men with lower secondary education compared to men with tertiary education, among women, the ratio was 2.2. The risk ratio for mortality varied from 1.3 to 5.2 among men and from 2.0 to 10.7 among women in the 1990s and early 2000s [8].

Earlier research on the burden of diabetes mortality has mainly examined mortality among patients with a diagnosis of diabetes and examined diabetes as the underlying cause of death along with other main causes. Whereas different strategies have been used to identify the total diabetes population (e.g., clinical examination, register based data, survey data), part of the diabetes population is likely to be always missed (e.g., undiagnosed cases, persons treated by diet only). Another method used for analyzing the burden of diabetes mortality has been to estimate the number of deaths attributable to diabetes in the general population. Traditionally only the underlying cause of death has been taken into account, but there are some relatively recent studies that also take into account contributory causes of death [13–15]. Few of these studies have examined mortality differences in different population groups, e.g., regionally [14] or by socioeconomic position [13]. However, Miech and colleagues [13] found widening educational differences in diabetes mortality in the U.S. between 1989 and 2005.

The aim of this study is to examine the total burden of diabetes mortality in the general Finnish population by age, gender and socioeconomic position and trends in it from 1988 to 2007. We further aim to analyze socioeconomic differences in potential years of life lost due to diabetes. We assess diabetes both as the underlying and contributory cause of death for the total population aged 35-80 years. In order to evaluate the total burden of diabetes, we analyzed both underlying and contributory causes of death since diabetes associates with several potentially fatal chronic complications, including cardiovascular and renal disease.

The data used in the study are based on information from different administrative registers linked with death records by Statistics Finland. These individual level data consist of an 11 percent random sample of the population residing in Finland in 1987-2007 and an 80 percent oversample of persons who died during those years. We used weights in the analyses to take account of this sampling design. Immigrants were dropped out and emigrants censored (1.2% of the sample). In all our analyses diabetes related deaths refer to diabetes as an underlying or a contributory cause, classified by Statistics Finland according to the International Classification of Diseases (ICD). The 9th revision of the classification was used for years 1988-1995 (code 250) and the 10th revision for years 1996-2007 (codes E10-E14).

Information on occupation-based socioeconomic position was available for every 5^th year: 1985, 1990, ..., 2005. These were used for the mortality analyses in the years 1988-1990, 1991-1995, ..., 2006-2007 respectively. Four groups were used: (1) upper non-manual, (2) lower non-manual and (3) manual workers and (4) farmers or other entrepreneurs. Those working at home were classified according to the occupation of the head of the household [16]; in the 2000s more than 80% of women with children aged 3+ were employed. The classification we use is also retrospective i.e. information on occupation for the economically inactive population (e.g. pensioners and unemployed) was searched from earlier years. This classification has been used previously for studies of social inequalities in mortality [17]. A small group of students and persons for whom socioeconomic position could not be determined (2% of men, 3% of women) was excluded from the analyses concerning socioeconomic position. Confirmatory analyses based on a three category education variable replicate the main diabetes mortality trends observed for occupation-based socioeconomic position.

For the trend analysis between years 1988-2007 we calculated age-specific diabetes death rates by gender and age-standardized diabetes death rates by gender and socioeconomic position. Standardization was done by using the direct method and using the sum of all person-years in the total population in 5-year age groups as the standard.

The analysis of potential years of life lost (PYLL) between the ages 35 and 80 due to diabetes related deaths by socioeconomic position was performed for the period 2004-2007. PYLL was calculated as the sum of deaths weighted by the difference between age at death and age 80[18]. We calculated 95% confidence intervals for PYLL on the basis of the Taylor linearized variance estimator with Stata version 11.1 [19].

This study examined the total burden of diabetes mortality in the Finnish population in 1988-2007. Both underlying and contributory causes were included. A systematic age pattern was detected in diabetes related deaths for both genders: the older the age-group the higher the mortality. Diabetes related mortality declined during the follow-up period especially among women, and the decline was especially large in the oldest age-groups. More research is needed to evaluate the causes of these trends. In line with earlier studies of total mortality among persons with diabetes and studies of diabetes related mortality [9, 10, 13, 20] a clear socioeconomic pattern was detected among the three employee categories (upper non-manual, lower non-manual, manual) among both men and women: the higher the socioeconomic position the lower the mortality. The contribution of diabetes related mortality to potential years of life lost was eight percent among men and six percent among women. Among men, the contribution of diabetes to potential years of life lost was similar in all socioeconomic groups suggesting that diabetes related mortality does not contribute to socioeconomic differences in total mortality. However, among women the contribution of diabetes to potential years of life lost was larger in lower groups.

The longitudinal nature of our individual level data enabled us to analyze trends in diabetes mortality. The Finnish Causes of death register is considered valid and reliable by international standards [21]. Furthermore, the rate of confirmation of the diagnosis by autopsy in Finland is high compared to many other countries (30% for all deaths and about 60% of those of working age) [22]. A limitation of our study is that we use ICD9 for defining the conditions analyzed in 1988-1995 and ICD10 from 1996 onwards. The update in ICD nosology is likely to have at least some contribution to the change in the level of mortality between 1995 and 1996. However, it is not likely to influence gender, socioeconomic or age differences in diabetes related mortality.

It is possible however that our data may still underestimate the number of diabetes deaths. We thus compared the diabetes deaths in our data to those in earlier studies examining mortality among diabetes patients [20, 23]. We observed no significant difference between the two approaches in the level of diabetes related mortality, which we consider a validation to both approaches of analyzing diabetes as a cause of death. However, both approaches are still likely to be underestimates to the extent that there are persons suffering diabetes without knowing it and diabetes may be under-recorded in death certificates [24].

According to our results diabetes related mortality declined especially among older age groups in Finland suggesting improvements in diabetes care in the 1990s and 2000s particularly among women. Since diabetes incidence has increased in Finland and elsewhere [1], improvements in diabetes care have been instrumental in offsetting this change. In order to further reduce the burden of diabetes a better treatment balance of diabetes in order to prevent complications would significantly decrease the burden of diabetes mortality. Use of both underlying and contributory causes of death certification may be useful in monitoring trends and population sub-group differences in the burden of diabetes in order to take into account also the contribution of diabetes to mortality due to its complications.