District-Level Life Expectancy in Germany

Background: Identifying regions with low life expectancy is important to policy makers, in particular for allocating resources in the health system. Life expectancy estimates for small regions are, however, often unreliable and lead to statistical uncertainties when the underlying populations are relatively small.

Methods: We combine the most recent German data available (2015–2017) with a Bayesian model that includes several methodological advances. This allows us to estimate male and female life expectancy with good precision for all 402 German districts and to quantify the uncertainty of those estimates.

Results: Across districts, life expectancy varies between 75.8 and 81.2 years for men and from 81.8 to 85.7 years for women. The spatial pattern is similar for women and men. Rural districts in eastern Germany and some districts of the Ruhr region have relatively low life expectancy. Districts with relatively high life expectancies cluster in Baden-Wuerttemberg and southern Bavaria. Exploratory analysis shows that average income, population density, and number of physicians per 100 000 inhabitants are not strongly correlated with life expectancy at district level. In contrast, indicators that point to particularly disadvantaged segments of the population (unemployment rate, welfare benefits) are better predictors of life expectancy.

Conclusions: We do not find a consistent urban–rural gap in life expectancy. Our results suggest that policies that improve living standards for poorer segment of the population are the most likely to reduce the existing differences in life expectancy.

Life expectancy in Germany is about 83.3 years for women and 78.5 years for men, according to the most recent data from the Federal Statistical Office (1). Internationally, Germany ranks 30^th, trailing the leading countries for life expectancy by about 4 years for women and 3 years for men (2).

After only modest improvements since the 1970s, the former German Democratic Republic (referred to from here on as eastern Germany or the east) experienced a sudden increase in life expectancy after the reunification of Germany (3, 4). In contrast, life expectancy in the former Federal Republic of Germany (from here on, western Germany or the west) has increased steadily, at a rate comparable with other western industrialized countries. The current life expectancy for women no longer shows a difference between eastern Germany and western Germany, while there remains a gap of slightly over 1 year for men.

Comparisons of such large areas are certainly informative. Appropriate planning of health services, however, requires analysis at a finer geographic scale. Small-area estimates are therefore crucial in identifying marginalized regions. This is particularly important with regard to Art. 72 [2] of Germany’s constitution, which empowers the federal government to enact legislation to provide equivalent living conditions.

To improve the identification of such regions, we estimated—based on age-specific mortality rates—life expectancy at the district level for women and men in Germany.

Our article has two substantive goals. First, we wanted to provide reliable estimates for small areas by using methods that do not suffer from instability and high uncertainty when the underlying populations are small. Standard estimators are vulnerable to random statistical fluctuations when there are only a few deaths from which to estimate mortality in small populations. In addition, our statistical approach also produces estimates of the underlying uncertainty. Such estimates of uncertainty are valuable for assessing whether differences in life expectancy between two districts are real or merely the random outcome of small numbers of deaths in districts of modest size. To our knowledge, no such interval estimations of life expectancy at the district level in Germany have yet been published.

Second, we wanted to examine whether there are specific patterns of correlation between social and economic indicators at the district level and local life expectancy. It is well known that poor, disadvantaged, and less (formally) educated people have lower life expectancy than those with higher income, better economic prospects, or a university degree (5, 6, 7). This is likely due not only to income and resources per se, but also to the higher prevalence of unhealthy lifestyles such as poor diet, smoking, and excessive drinking or to occupation-specific hazards among people with lower income or lower level of education (8, 9). By producing district-level estimates for life expectancy in Germany and investigating their correlations with social and policy variables, we hope to provide useful data and a solid foundation for future, more detailed, analyses.

Methods

Data

For each combination of district, sex, and age group (0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, …, 80, 81, 82, 83, 84, 85+) we had access to the resident population on 31 December in 2014, 2015, 2016, and 2017 and to the numbers of deaths in the calendar years 2015, 2016, and 2017. These data—the most recent available for estimation of mortality—are made available by the federal and state statistical offices via www.regionalstatistik.de (10, 11). We calculated the person-years lived by age group and sex in each calendar year as the mean of two consecutive end-of-year populations (e.g., the number of person-years lived in 2015 by women aged 50–54 years is the average for the women of this age group alive on 31 December 2014 and on 31 December 2015). In order to reduce random fluctuations due to small populations, we pooled the data on deaths and the person-years lived from the three available years (2015, 2016, 2017).

With the exception of the merger of the districts Göttingen and Osterode am Harz in 2016, the district boundaries did not change in the period 2015–2017. Consequently, our estimates for Göttingen and Osterode am Harz pertain to 2015 only.

Population sizes in German districts span two orders of magnitude. On 31 December 2017 the smallest district was Zweibrücken, with about 33 900 inhabitants, while Berlin, with about 3.6 million inhabitants, was the largest. Median district population size is about 150 000, and 50% of districts have between 100 000 and 240 000 inhabitants.

Annual death counts span slightly less than two orders of magnitude. In 2017, for example, there were 449 deaths in Zweibrücken and 34 339 deaths in Berlin. Slightly more than half of all districts had between 1200 and 2700 deaths in 2017.

Relational Bayesian model

We use a new relational model to estimate mortality rates by age and sex in each district. Relational models, such as the Brass logit model (12), have been widely employed for populations with sparse data, and were also partly used for estimation of life tables in Germany’s component states after the register-based census in 2011 (13). The relational TOPALS model that we used was proposed by de Beer (14), and has good statistical properties even in very small populations (15, 16).

Our approach estimates age-specific mortality rates for all districts and both sexes with a Bayesian model that links parameters across districts. The model produces probabilistic mortality rates at the district level, from which we can calculate a probability distribution for life expectancy e₀ at the district level.

Based on the given death counts and person-years lived, this distribution assigns higher probability to e₀ values that are more likely according to the model. Its median serves as the point estimate of life expectancy at the district level. We represent the uncertainty of our estimates by means of the interval between the 10th and 90th percentiles, which spans 80% of possible values for local e0. Mathematical and statistical details can be found online at http://german-district-mortality.schmert.net.

Results

Figure 1 displays estimates of life expectancy and of 80% probability intervals for each district and both sexes. We highlight the 10 largest cities and the districts with the highest and lowest life expectancies for women and men. Based on the estimated mortality rates for the years 2015–2017, men in Bremerhaven had the shortest life expectancy, while men living in the district surrounding Munich (Landkreis) could expect to live around 5 years longer. We found the lowest life expectancy for women in the Salzland district of Saxony–Anhalt, the highest in the Starnberg district, southwest of Munich. The eTable provides a list of all districts, arranged by federal state, with point estimates and 80% interval estimates. The light blue and light red bars in Figure 1 indicate these 80% probability intervals. Because of the larger populations these bars are narrower for the largest cities than for less densely populated districts. The width of the interval estimates varies from 0.14 years for Berlin to 0.55 years for Ansbach and Osterode am Harz; the median is 0.36 years.

Figure 1

Life expectancy at birth for men (blue) and women (red)

Table

List of indicators used, their definitions, means and standard deviations

Figures 2 and 3 give a good impression of the distribution of life expectancy across districts in Germany. eFigures 1 and 2 show the same maps for men and women with the rankings of the individual districts. District ranks should be interpreted cautiously due to overlapping interval estimates, but they indicate the approximate position among the 402 rural districts. An alphabetical list of districts by federal state is given in the eTable. Paler shades, denoting districts with lower life expectancy, are found more frequently in the east than in the west, especially for males. However, also in western Germany there are districts with comparably low life expectancy, especially districts in the Ruhr region, such as Dortmund or Gelsenkirchen.

Figure 2

Estimates of life expectancy for men in all 402 districts of Germany expressed as whole numbers, i.e., 75 stands for 75.00–75.99 years, 76 for 76.00–76.99 years, etc.

Figure 3

Estimates of life expectancy for women in all 402 districts of Germany expressed as whole numbers, i.e., 81 stands for 81.00–81.99 years, 82 for 82.00–82.99 years, etc.

eFigure 1

Estimates of life expectancy for men in all 402 districts of Germany expressed as whole numbers, i.e., 75 stands for 75.00–75.99 years, 76 for 76.00–76.99 years, etc. The numbering of the districts shows the estimated ranking (1 = highest, 402 = lowest).

eFigure 2

Estimates of life expectancy for women in all 402 districts of Germany expressed as whole numbers, i.e., 81 stands for 81.00–81.99 years, 82 for 82.00–82.99 years, etc. The numbering of the districts shows the estimated ranking (1 = highest, 402 = lowest).

It is well known that economic development, local conditions, and the availability of medical services can play important roles in life expectancy (17, 18). We now go on to explore what correlations exist between district-level life expectancy and social and economic indicators from the INKAR database (INKAR, Indikatoren und Karten zur Raum- und Stadtentwicklung [Indicators and Maps for Spatial and Urban Planning]) (19).

For this purpose we used the following district-level indicators: population density (persons per km²), primary-care physicians per 100 000 inhabitants, gross domestic product (GDP) per capita, unemployment rate, child poverty (persons eligible under 15 years per 100 inhabitants under 15 years), housing subsidies (households that receive housing benefit per 1000 households), German Social Code II-based (“Hartz IV”) welfare benefits (persons eligible below age 65 years in %), and public assistance for the elderly (proportion of inhabitants ≥ 65 years in ‰) (“Grundsicherung im Alter”).

Figure 4 shows the strength of the relationships between life expectancies at the district level and each of the selected indicators. The bars in the figure represent standardized bivariate regression coefficients: the estimated changes in life expectancy with an increase of one standard deviation in the given indicator. For example, the standardized coefficient for the unemployment rate among men in the west is –0.6. The mean unemployment rate for men in the west was 5.40%, with a standard deviation of 2.47% (Table). Thus a 2.47% difference in unemployment rates between two western districts corresponds to a life expectancy 0.6 years lower than in the district with the higher life expectancy.

Figure 4

Expected change in life expectancy (years) when indicator increases by 1 standard deviation

Several patterns emerge clearly from Figure 4. The selected indicators are more strongly correlated with life expectancy for men (left panel) than for women (right panel). Here we examine only aggregated cross-sectional data, but individual-level longitudinal studies of mortality often show that social, demographic, and economic conditions affect men more strongly than women (20).

The correlations between population density, physician density, and life expectancy are relatively low. It is interesting, however, that these correlations point in different directions in the east and the west: in eastern districts life expectancy rises slightly with increasing population density and physician density, while in western districts it declines slightly. Therefore only in eastern Germany does living in an urban area or a more densely populated region confer an advantage with regard to life expectancy.

The data in Figure 4 also demonstrate that economic indicators are much stronger predictors of life expectancy than population density or the number of primary-care physicians per 100 000 residents. This contradicts speculation in the popular press (21) that differences in the proximity to medical care could explain regional differences in health and mortality.

The positive relationship between GDP and life expectancy across countries has been well known since the 1970s (17). The same pattern is visible in our analysis of districts: higher GDP per capita coincides with higher life expectancy.

Correlations between life expectancy and population density, physician density, and GDP are relatively low, however, compared with economic indicators that focus on the most disadvantaged residents of a district. As seen from the bars in the lower section of Figure 4, indicators such as unemployment rate, housing subsidies, and other measure of public assistance have notably higher correlations (all negative) with district-level life expectancy. It is also interesting to note that local unemployment and public subsidies are more strongly related to lower life expectancies in western Germany.

In our opinion the simple correlations illustrated in Figure 4 clearly show it is not the average economic conditions in a region that influence life expectancy, but rather the circumstances of persons at the lower end of the socioeconomic spectrum.

Discussion

We estimate that life expectancy varies across German districts by more than 5 years for men (5.3 years), and by almost 4 years for women (3.9 years). To put this into an international context, the lowest life expectancy for men (75.8 years) corresponds roughly to that in Oman (ranked 53^rd of 201 countries in UN estimates [2]), while the highest male life expectancy (81.2) is approximately equal to that in Australia (ranked 6^th). For women, the district with the lowest life expectancy is comparable with the life expectancy of Czech women (81.8, ranked 46^th). The highest female life expectancy in a German district is comparable with the life expectancy of women in South Korea (85.7, ranked 5th).

It is important to mention a few methodological limitations: Our estimates are based on the current population and death counts in each district, and could consequently be affected by selective migration. For instance, migration of particularly healthy people may reduce life expectancy in their home district and increase life expectancy in their new district. Neither our data nor our methods can control for this. In addition, models like ours that pool information across small units in order to lower variability tend to pull statistical outliers with small populations towards regional and national means. This is a classic bias–variance trade-off, and in some cases it may cause too much smoothing for a “real” outlier with a small population. Without a longer time series or additional information to identify such outliers, this is an unavoidable risk.

Although cross-sectional, aggregate data do not permit causal inferences, we explored bivariate correlations between district-level life expectancy and macrofactors often discussed in the media, such as population density or physician density. Our results indicate that these factors are notably less important than economic indicators that focus on poorer population segments, such as the unemployment rate or welfare payments.

Life expectancy summarizes the entire age profile of mortality in a single number. Thus, further analyses focusing specifically on infant and childhood mortality or mortality in higher age groups might yield different results. By providing all data, code, and results at http://german-district-mortality.schmert.net/ we encourage and enable other researchers to thoroughly understand our model, to reproduce our findings, to update them with new data, and to use them for independent analyses. Important examples of wider-reaching research initiatives would include more detailed examination of mortality by age and the integration of lifestyle factors such as smoking prevalence, alcohol consumption, or physical activity (22).

Conflict of interest statement
The authors declare that no conflict of interest exists.

Manuscript submitted on 9 October 2019, revised version accepted on 23 April 2020

Corresponding author
Prof. Dr. Roland Rau
Wirtschafts- und Sozialwissenschaftliche Fakultät
Universität Rostock
Ulmenstr. 69, 18057 Rostock, Germany

Cite this as
Rau R, Schmertmann CP: District-level life expectancy in Germany.
Dtsch Arztebl Int 2020; 117: 493–9. DOI: 10.3238/arztebl.2020.0493

►Supplementary material

eTable and eFigures:
www.aerzteblatt-international.de/20m0493

References

1. Statistisches Bundesamt (Destatis): Durchschnittliche Lebenserwartung (Periodensterbetafel); Tabelle 12621–0002; 2020. https://www-genesis.destatis.de/genesis//online?operation=table&code=12621–0002&levelindex=0&levelid=1588857487918 (last accessed on 3 March 2020).

2. United Nations Population Division: wpp2019: World Population Prospects 2019. R package version 1.1–0. www.CRAN.R-project.org/package=wpp2019 (last accessed on 3 March 2020).

3. Vogt TC: How many years of life did the fall of the Berlin Wall add? A projection of East German life expectancy. Gerontology 2013; 59: 276–82

4. Oeppen J, Vaupel JW: Broken limits to life expectancy. Science 2002; 296: 1029–31

5. Mackenbach JP, Bos V, Andersen O, et al.: Widening socioeconomic inequalities in mortality in six Western European countries. Int J Epidemiol 2003; 32: 830–7

6. Lampert T, Hoebel J, Kroll LE: Social differences in mortality and life expectancy in Germany. Current situation and trends. Journal of Health Monitoring 2019; 4: 3–14.

7. Wenau G, Grigoriev P, Shkolnikov V: Socioeconomic disparities in life expectancy gains among retired German men, 1997–2016. J Epidemiol Community Health 2019; 73: 605–11

8. Lantz PM, House JS, Lepkowski JM, Williams DR, Mero RP, Chen J: Socioeconomic factors, health behaviors, and mortality: results from a nationally representative prospective study of US adults. JAMA 1998; 279: 1703–8

9. Goldman, N: Mortality Differentials: Selection and Causation. In: Smelser NJ, Baltes PB (eds.): International encyclopedia of the social & behavioral sciences. Amsterdam: Elsevier 2001: 10068–70

10. Regionalstatistik. Die Statistischen Ämter des Bundes und der Länder: Bevölkerung nach Geschlecht, Nationalität und Altersgruppen; Tabelle 12411–03–03–4; 2019. www.regionalstatistik.de/genesis/online/data;sid=1ACC2309105E04B36B092B88D41E63B8.reg1?operation=begriffsRecherche&suchanweisung_language=de&suchanweisung=12411-03-03-4 (last accessed on 7 September 2019).

11. Regionalstatistik. Die Statistischen Ämter des Bundes und der Länder: Gestorbene nach Geschlecht, Nationalität und Altersgruppen - Jahressumme - regionale Ebenen; Tabelle 12613–02–02–4; 2019. www.regionalstatistik.de/genesis/online/data;sid=84C57B3423E5A0326EDC269B88930DF7.reg3?operation=begriffsRecherche&suchanweisung_language=de&suchanweisung=12613-02-02-4 (last accessed on 7 September 2019).

12. Brass W: On the scale of mortality. In Brass W (ed.): Biological aspects of demography. London: Taylor & Francis 1971: 69–110.

13. zur Nieden F, Rau R, Luy M: Allgemeine Sterbetafel 2010/12 - Neue Ansätze zur Glättung und Extrapolation der Sterbewahrscheinlichkeiten. Wirtschaft und Statistik 2016: 63–74.

14. de Beer JAA: Smoothing and projecting age – specific probabilities of death by TOPALS. Demogr Res 2012; 27: 543–92

15. Gonzaga MR, Schmertmann CP: Estimating age- and sex-specific mortality rates for small areas with TOPALS regression: an application to Brazil in 2010. Rev Bras Estud Popul 2016; 33: 629–52

16. Schmertmann CP, Gonzaga MR: Bayesian estimation of age-specific mortality and life expectancy for small areas with defective vital records. Demography 2018; 55: 1363–88

17. Preston SH: The changing relation between mortality and level of economic development. Pop Stud 1975; 29: 231–48

18. Cutler D, Deaton A, Lleras-Muney A: The determinants of mortality. J Econ Perspect 2006; 20: 97–120

19. Bundesinstitut für Bau-, Stadt- und Raumforschung (BBSR) im Bundesamt für Bauwesen und Raumordnung (BBR): Indikatoren und Karten zur Raum- und Stadtentwicklung (INKAR), 2019. www.inkar.de (last accessed on 14 January 2020).

20. Koskinen S, Martelin T: Why are socioeconomic mortality differences smaller among women than among men? Soc Sci Med 1994; 38: 1385–96

21. Weise K, Krüger L, Kleine R: Die bittere Wahrheit über das Leben auf dem Land. Bild Hamburg; 27 October 2017. www.bild.de/politik/inland/mangel/die-bittere-wahrheit-ueber-das-leben-auf-dem-land-53680998.bild.html (last accessed on 7 May 2020).

22. Atzendorf J, Apfelbacher C, de Matos EG, et al.: Do smoking, nutrition, alcohol use, and physical activity vary between regions in Germany? - Results of a cross-sectional study. BMC Public Health 2020; 20: 277

23. Bundesamt für Kartographie und Geodäsie: Verwaltungsgebiete 1 : 2 500000. VG2500. https://daten.gdz.bkg.bund.de/produkte/vg/vg2500/2014/vg2500_01-01.utm32w.shape.zip (last accessed on 7 May 2020).