OUP user menu

Hierarchical spatio-temporal mapping of premature mortality due to alcoholic liver disease in Hungary, 2005–2010

Csilla Nagy, Attila Juhász, Zoltán Papp, Linda Beale
DOI: http://dx.doi.org/10.1093/eurpub/ckt169 827-833 First published online: 11 November 2013


Background: Excess alcohol consumption is a major public health issue in Hungary, with high mortality rates from alcohol disease. In addition, excess alcohol consumption has been found to differ by socioeconomic and environmental factors. Methods: The spatial pattern of mortality from alcoholic liver disease across Hungary for the period 2005 and 2010, at municipality level, was investigated by sex using Bayesian modelling. The changes in mortality over this 6-year period were analysed using hierarchical spatio-temporal disease mapping. The association between the spatial distribution of mortality from alcoholic liver disease and cirrhosis and deprivation (by sex) was also assessed using a Hungarian-specific deprivation index. Results: A statistically significant association was found between mortality and deprivation status in males. Areas of significantly highest age-adjusted relative risks were found, for males, in the south-western part, and at the eastern border of the country. A number of areas showed area-specific trends that were significantly steeper than the national average mortality trend. Conclusion: Using our findings, population groups with increasing mortality trends were identified, which may help decision makers fund effective health promotion programmes to reduce mortality rates in the working-age population.


Alcohol-related health and social problems are major global concerns. Several reports have shown that alcohol consumption is the second largest risk factor for disease and injury burden in Central Europe.1,2 Hungarian indicators, even in 2010, concerning abuse of alcohol attributable to public health problems (including chronic liver diseases and cirrhosis mortality) were among the worst ones in Europe.3 Although chronic liver diseases and cirrhosis mortality shows a decreasing trend, in comparison with the average rate of the European Union members before May 2004 (EU-15),3 male mortality rates are still almost five times higher and female mortality rates are nearly four times higher. In recent years, the average per capita alcohol consumption also showed a decreasing trend among the adult population of Hungary; however, it is one of the highest (with 11.51 litres per capita in 2009) among the 27 European Union members states (EU-27).3 Furthermore, it is an unfortunate fact that the general health status of the Hungarian population is worse than is indicated by country-level socioeconomic status alone: Hungarian gross domestic product ranked 16th among the EU-27, whereas premature mortality ranked 22nd in 2010.3,4

Several studies have shown the importance of lifestyle and social-economic factors in the development of chronic liver disease.5–8 In Hungary, those with a high level of education have been shown to be less likely to develop chronic liver disease; conversely, a lack of social relationships and excessive alcohol consumption are risk factors.5,6 It has also been reported that more than 80% of mortality from chronic liver disease can be attributed to alcohol abuse in Hungary.9,10 Furthermore, beyond the quantity of alcohol, the consumption of poor-quality alcohol of dubious origin is also responsible for increased mortality.9–12

In the late 90s, considerable geographical differences in distribution of alcoholic liver disease mortality were clearly associated with alcohol consumption. For the working-age population, particularly males, high mortality risks have already been identified along the north-eastern-to-south-western axis of Hungary.13,14 The aim of this study was to analyse the spatial pattern of liver disease mortality at municipality level in Hungary and also to investigate its development over time using spatio-temporal disease mapping with routinely collected data.


This study focused on the development and geographic distribution of alcoholic liver disease mortality (BNO-10: K70) for ages 25–64 years in Hungary for the years 2005–2010. Maps comparable in space and time can give valuable information not only about geographic localization of alcoholic liver disease mortality but also on the development of alcohol abuse in general. The association between the spatial distribution of alcoholic liver disease mortality and deprivation by sex was also assessed using a Hungarian-specific deprivation index (DI).


Mortality data from the Hungarian Central Statistical Office was attained at the municipality level, by year, sex and 5-year age bands. Population data (also at the municipality level, by year, sex and 5-year age bands) were obtained from the Central Office for Administrative and Electronic Public Services and the National Institute of Environmental Health.

An area-based composite index was used to provide information about socioeconomic deprivation in Hungary.15 This DI was developed with data from 2001 census using principal component analysis based on seven elementary socioeconomic indicators (income, the level of education, the rate of unemployment, the rate of one-parent families, the rate of large families, density of housing and car ownership) at the municipality level and is described in detail.15

To produce more comparable statistics at the international level, in 2005, Hungary additionally implemented an Automated Coding System for medical causes of death that uses cause-of-death coding outlined by the World Health Organization.16

A comparison conducted by the Hungarian Central Statistical Office between data coded using the old and new methods showed an improvement in data quality and subsequently differences in the frequency of certain deaths. For example, using the automatically coded data, alcoholic liver disease mortality significantly decreased (up to 40%) in 2005.17 A Hungarian study of morbidity before 2005 found significant overestimation of chronic liver diseases, which may explain this decrease.13,17,18 This study, therefore, used the period between 2005 and 2010 to ensure that data were of reasonable quality and were comparable internationally.17

Risk analysis

The association between deprivation and alcoholic liver disease mortality (BNO 10:K70) was calculated using the risk analysis capability of the RIF software.19 Both directly standardized death rates (SDRs, standardized for the European standard population age distribution (1976)) and indirectly standardized mortality ratios (SMRs) were calculated for municipality bands grouped by deprivation quintiles. Homogeneity and linear trend tests were additionally carried out to check whether risk is statistically homogeneous across bands, and to test the global association of the DI and the relative risk of premature mortality due to alcoholic liver disease.15

Disease mapping

Disease mapping options in the RIF, along with the ability to export results directly to INLA, were used to investigate spatial patterns of alcoholic liver disease at the municipality level for the period 2005–2010.20 Hierarchical Bayesian modelling proposed by Besag et al.21 was used to smooth the results.

These smoothed SMRs were calculated using the observed cases of each municipality together with the expected cases based on the sex- and age-specific death rates for the Hungarian population. The posterior probability of each area being above average risk (i.e. SMR > 1) was computed as a measure of uncertainty.22

Spatio-temporal disease mapping

Disease mapping can reveal valuable information about geographical patterns of disease and mortality; however, it masks temporal trends, making it difficult to understand the development of these patterns. Recent advances in spatial epidemiology have extended pure spatial models to incorporate time trends and space–time interactions.

Using the spatio-temporal model in R with the INLA-specific package,23,24 we adopted the classical parametric model introduced by Bernardinelli et al.25 in which the log risk in a given area is assumed to be a linear function of time. Consequently, the linear trend of mortality for each area can be identified. The coefficients are area-specific and spatially structured. This formulation includes the same spatial components as the model in the diseases mapping section, a main linear trend (representing the global time effect, the main linear time-trend over all areas) and differential trends (identifying the interaction between time and space and representing the difference between the global trend and the area-specific trend). If differential trends are less than zero, the area-specific trend is less steep than the mean trend, whereas if differential trends are greater than zero, it implies that the area-specific trend is steeper than the mean trend.


The results of the risk analysis showed a significant linear association between the risk of premature mortality due to alcoholic liver disease and deprivation for both gender (males: χ2 homogeneity = 182.62, P = 0, χ2 linearity = 166.35, P = 0; females: χ2 homogeneity = 14.34, P = 0.01, χ2 linearity = 7.44, P = 0.01) for the years 2005–2010 in Hungary (table 1), which suggests that areas of highest deprivation showed a high risk of liver disease mortality among the Hungarian population.

View this table:
Table 1

Premature mortality due to alcoholic liver disease at the municipality level by DI quintile, Hungary, 2005–2010

DI quintileMalesFemales
Standardized death ratesa (per 100 000 person years) (95% CI)SMR (95% CI)Standardized death ratesa (per 100 000 person years) (95% CI)SMR (95% CI)
I. (least deprived)59.690.8318.850.92
V. (most deprived)91.041.2721.241.03
  • a Standard: age distribution of the European standard population (1976).

  • CI: confidence interval.

Standardized death rates of alcoholic liver disease mortality in the upper quintile band (i.e. the most deprived, SDRMales = 91.04/100000; confidence interval [CI] [85.59–96.84]; SDRFemales = 21.24/100000; CI [18.69–24.14]) exceeded the mortality in the lowest quintile band (i.e. the least deprived, SDRMales = 59.69/100000; CI [57.58–61.87; SDRFemales = 18.85/100000; CI [17.74–20.02]) by 53% in males and by 13% in females (table 1). However, mortality in the fourth quintile band (i.e. the deprived, SDRFemales = 22.54/100000; CI [20.66–24.59]) exceeded the mortality in the lowest quintile band by 21% in females (table 1).

Male mortality in the lowest deprivation quintile was 20% lower than the Hungarian average, whereas mortality in areas with the highest deprivation exceeded the national average mortality by 30%. These differences are unlikely to have occurred by chance (table 1). For females, in almost every quintile band, except for the lowest deprivation quintile (SMRFemales = 0.9; CI [0.87–0.98]), mortality fluctuated around the national average; however, these results were weak but still statistically significant (table 1).

Figure 1 shows the geographical distribution of smoothed SMRs for alcoholic liver disease mortality in Hungary for males and females. Differences in spatial distribution of alcoholic liver disease mortality are quite marked, except for some areas (to the east and south-east of the country) where both male and female mortality rates are significantly lower than other parts of the country. For males, areas of significantly highest age-adjusted relative risks were found in the south-western part and at the eastern border of Hungary (figure 1) (Supplement 1). For females, several areas of significantly low age-adjusted relative risks can be seen in the north-western part of Hungary. For males, the lowest age-adjusted relative risks are found in the capital city of Budapest and neighbouring areas; however, for females, these areas showed the significantly highest mortality (figure 1) (Supplement 1).

Figure 1

The spatial distribution of smoothed SMRs of premature mortality due to alcoholic liver disease in Hungary for males and females, 2005–2010 (Supplement 1)

The spatio-temporal analysis identified no national linear temporal trend in relative risks (RRNt) from 2005 to 2010 for females (RRNt = 1.004 [0.984–1.025]); however, a weak but significantly positive linear time-trend was observed for males (RRNt = 1.023 [1.011–1.036]).

Comparing local differences, by municipality, with the national trend over time did not reveal any significant differences for females; however, for males, some interesting patterns were observed (figure 2) (Supplement 2). Municipalities with a trend that significantly differed from the national time trend are shown together with posterior probabilities to show the associated uncertainty. Most of the municipalities with a more gradual trend compared with the national trend were mainly located in the western, south-western and in the south-eastern parts of Hungary. Areas with area-specific trend that was steeper than the national trend were identified in the north-western and north-eastern parts of the country.

Figure 2

The spatial distribution of differential trend in premature mortality due to alcoholic liver disease in Hungary for males, 2005–2010 (Supplement 2). Map: direction of time trend compared with the main (national) linear time-trend. Graphs: fitted relative risks for selected municipalities (Municipal trend), and the main linear time-trend (National trend) on a log scale obtained by spatial-temporal model

To illustrate the results of both the spatial and spatio-temporal approaches, the estimated time trend on a logarithmic scale, for selected municipalities, is also shown in figure 2 (graphs, municipal trend). Those areas with mortality levels lower than the national average were found in the eastern part of the country for males (figure 1, Supplement 1). Furthermore, these areas had time trends less steep than the national average (figure 2, Supplement 2). These findings suggest that predicted risk had decreased in these areas over the study period (see example of Berettyóújfalu) (figure 2) (Supplement 2). Areas were also identified in this part of the country with similar differential trends to that of the national trend, i.e. gently increasing (see example of Hajdúböszörmény and Nyíregyháza) (figure 2) (Supplement 2).

Areas with high mortality were detected in the south-western part of the country (figure 1) (Supplement 1); however, the differential trend was less steep than the national trend, indicating that the predicted risks had decreased in these areas for the investigation period (e.g. Nagykanizsa) (figure 2) (Supplement 2). Some areas were found where the municipality trend was steeper than the average trend for the 6-year study period (e.g. Kisbér or Ózd) (figure 2) (Supplement 2).


Extremely high mortality rates from chronic liver disease and cirrhosis (related to alcohol abuse) have been observed in several European countries, including Hungary.3,26-28 In Hungary, from the mid-90s, chronic liver disease mortality has been decreasing; however, in 2010, it remained almost five times higher than in the EU-15,3 with more than 80% potentially related to the alcohol consumption.9,10 These figures combined with findings from numerous international and national studies that show a significant reduction in life expectancy from birth due to alcohol abuse26,27 and demonstrate the importance of national health policies that address alcohol abuse.7–9,12,26–28

A World Health Organization study suggested that a significant part of the disease burden, from non-communicable diseases, was caused by lifestyle factors, with alcohol abuse being the second most important cause of disease burden within Central Europe, including Hungary.1

Previous national studies also showed alcohol-related mortality/morbidity to have different geographical patterns by sex.6,13,14 At the municipality level, a significant and positive association was found between alcoholic liver disease mortality and deprivation for males and females, but this association was weaker for females. Again, this analysis also found the sexes to exhibit distinct spatial patterns and explains the difference seen with deprivation. For males, areas with significant and high age-adjusted relative risks were found in the south-western, northern and north-eastern parts of Hungary. Areas with significant and low age-adjusted relative risks for males were located in Budapest and its neighbourhood. These regions were, interestingly, predominately those areas of high risks for females. Regular consumption of alcohol is significantly higher among younger and middle-aged women in Budapest compared with those living in the countryside.29 A past study reported distinct gender-related differences in alcohol consumption and alcohol-related problems (e.g. biological, psychological and social risk factors30) that give rise to different drinking frequencies and habits, as had previously been found.6,31 Among nations involved in the same study, the former communist countries in Central Europe (Czech Republic and Hungary) were described as having higher-than-average gender differences in drinking habits.31 According to national health survey reports, the chance of heavy drinking in Hungarian women was found to be lower than in men. Furthermore, a gender gap was identified regarding the association between alcohol abuse and certain socioeconomic factors (education, income status), e.g. higher level of education narrows the difference between the two genders.6

Drinking habits may not be the only underlying cause behind mortality indicators, as local alcohol policy interventions can also have an effect on spatial patterns of specific mortality. In Hungary, two drinking cultures can be found: one formed of frequent drinkers (Mediterranean ‘model’) and another characterized by drinkers who consume large quantities (Nordic ‘model’).6 The results of this study showed that wine consumption was more common in the western part of Hungary, whereas hard spirits tended to be drunk in the central and northern part of the country.6 Despite these findings, to date, neither local spatial patterns of drinking habits nor regional alcohol policies or strategies have been mapped for Hungary.

Understanding the spatial distribution is, however, only part of the story. For public health planning, specifying priorities and allocating limited resources requires not only knowledge of the spatial variation of alcohol-related mortality but also information as to its development over time. A weak but significantly positive national linear time-trend was observed for males. The spatio-temporal analysis in this study identified a number of areas in Hungary that are priority areas, with local municipality trends that are steeper than the average trend for the 6-year study period for males. These trends identify these areas as areas with increasing risk, irrespective of the overall level of predicted risk. The spatio-temporal analysis highlighted a number of areas with significant low mortality rates combined with significant increases in risk over time. These areas deserve particular attention, as early, targeted intervention can reverse a deteriorating health trend. It is also interesting to note that no such temporal variations were detected for females. The results of this study demonstrate that the spatio-temporal modelling of data has additional benefits over a purely spatial analysis.

These findings demonstrate the use of a spatio-temporal approach that not only maps levels of mortality, but can also determine the direction of any changes, highlighting areas with decreasing or increasing mortality. The main advantage of the adopted spatio-temporal disease mapping method is, therefore, that the information is shared in both space and time and the parametric formulation is straightforward. However, the parametric model has some limitations. Assuming that each area has a linear time-trend is restrictive and inappropriate for longer periods.25,32 Although the smooth temporal evolution models introduced by Knorr-Held,33 Martínez-Beneito et al.34 or Abellan et al.35 may offer a more appropriate fit to the data in some cases,36 the parametric model used in this analysis is significantly easier to implement and interpret and nevertheless provides valuable information for public health professionals in everyday practice.

A number of other limitations must also be considered. First, routinely collected mortality data are prone to certain biases. Overestimation of cases with alcohol-related liver disease was a known issue in Hungary,13,17,18 and although the implementation of the Automated Coding Systems in 2005 caused a reduction in cases, and data after 2005 should correspond to the true situation, the raw data are still being revised. Second, the DI in this study is based on variables from the 2001 census and, therefore, as with all such indices, the index will become less accurate over time.17 The most recent census (2011) data are not yet available. The agreement of the initial analysis with past Hungarian research looking at chronic liver disease suggests that results are reasonable.6 Finally, this study looked at alcoholic liver disease mortality that was directly related to alcohol consumption; however, it should be noted that alcohol consumption in a wider sense has many other harmful effects on health. It is a contributor in cardiovascular disease mortality, malignant neoplasms, accidents and poisonings.17,37

Programmes designed to improve the overall health of the Hungarian population by reducing alcohol abuse tend to focus on either the qualitative or quantitative modification of consumption habits.38–40 The results of this research have identified a number of areas where the trend differs from the national picture and the population shows deteriorating mortality rates from alcohol-related liver disease. Identifying those areas that would benefit from health promotion programmes can help decision makers to reduce the overall burden of mortality from alcohol-related diseases.

The results of such spatio-temporal analysis also have wider implications for the analysis of alcohol-related disease across the European Union. Deaths related to alcohol consumption show a pattern of decline; however, it remains a significant health burden and indeed there may be areas that, as with the case of Hungary, do not exhibit the overall pattern of decline we expect. Identifying these areas can ensure that resources are effectively targeted.

Conflicts of interest: None declared.

Key points

What was already known?

  • Chronic liver disease and cirrhosis within European countries is showing a decline, but remains a significant health burden particularly in Hungary.

  • The effects of alcohol consumption and different socioeconomic factors on premature mortality are well known.

What this paper adds?

  • This study found significant association between premature mortality due to alcoholic liver disease and deprivation at the municipality level in Hungary.

  • Our results revealed the spatial pattern of premature mortality due to alcoholic liver diseases at the municipality level in Hungary.

  • Spatio-temporal disease mapping technique was implemented using routinely collected data.

  • The spatio-temporal analysis revealed areas with increased risks for males in Hungary for the investigation period.


The authors thank Professor Róza Ádány of the University of Debrecen Medical and Health Science Centre for the encouragement and many useful discussions. They also thank Dr Anna Páldy of the National Institute of Environmental Health for the opportunity of participation in the EUROHEIS2 project (2007–2010).


View Abstract