The Quality of Cause-Of-Death Statistics After the Introduction of the Electronic Coding System Iris/Musean

An Analysis of Mortality Data, 2005–2019

Cause-of-death statistics constitute an important basis for health reporting. Conclusions, especially from comparative evaluations, are valid only if the data quality has been taken into account. The data quality is assessed by the World Health Organization (WHO) also on the basis of the proportion of non-informative causes of death (“garbage codes”) among all causes of death. Non-informative causes of death are symptoms, unspecified/unknown causes of death, immediate causes of death (such as cardiac arrest), and intermediate causes of death triggered by underlying conditions (for example, heart failure, sepsis, pneumonia, renal failure) (1, 2). In Germany, the proportion of non-informative causes of death between 1998 and 2014 was between 1% and 14% (1) and 25% in cardiovascular deaths (3). Non-informative causes of death hamper effective prevention, as the underlying condition eventually leading to death remains unknown. Evaluating the quality of cause-of-death statistics is not the same as evaluating the correctness of a selected cause of death. This is possible only by means of regular postmortem examinations of a representative sample of deaths.

By 2018, seven federal states had introduced the electronic coding system IRIS/Muse in order to automate the plausibility check of the causal chain in the death certificate according to WHO rules (4) and the subsequent selection of the underlying condition (=underlying cause of death). In a simulation study using 67 660 death certificates, IRIS/Muse reduced the proportion of “unspecific codes” selected as underlying cause of death (5). A quality comparison of the cause of death statistics before and since the introduction of IRIS/Muse is yet to be carried out.

Methods

Data from the Federal Statistical Office of Germany (gbe-bund.de) for 2005–2019 for all deaths and deaths with common non-informative causes from a WHO short list (ICD-10 I10, I46, I50, I51.4–6, I51.9, I70.9, C76–C80 und R00–R99) (1) as well as further important intermediate causes of death (A41, J69, I26.9 und N17/N19) (2) were downloaded and analyzed by federal state for the age groups < 45 years, 45–64 years, 65–79 years, and ≥ 80 years. The years 2020–2022 were not included because of potential distortion by SARS-CoV-2 deaths.

The federal states were stratified according to when the IRIS/Muse software was introduced: early introduction 2011/12 (Baden-Württemberg, Saxony, Rhineland-Palatinate), late introduction 2016/17 (North Rhine-Westphalia, Berlin, Brandenburg, Saxony-Anhalt), and remaining states without introduction until 2018. For the year before IRIS/Muse was introduced (=index year) and the subsequent years, the proportion of non-informative causes of death were calculated.

Results

The quality of cause-of-death statistics in Germany has improved since 2005 (Figure). In those states where IRIS/Muse was introduced early, non-informative causes of death were consistently less frequent than in those states that adopted IRIS/Muse later or made no use of it at all (Figure). After the introduction in 2011/12, the proportion of non-informative causes of death in three states fell over the five following years from 13.0% to up to 10.2%. In the four states who adopted the IRIS/Muse system late, the mean proportion of 13.7% in these states fell only slightly subsequently (Figure, Table).

Figure

Changes in mean proportion of selected non-informative causes of death among all causes of death in 2005–2019 by federal state, stratified according to the time when IRIS/Muse was introduced

Table

Mean proportion (%) of selected non-informative deaths among all deaths in Germany in the year before IRIS/Muse was introduced

The non-informative causes of death, heart failure and unspecified or unknown causes of death (“Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified,” R00-R99) in Germany in 2019 were were selected as cause of death in roughly equal numbers, in a total of 72 000 deaths.. The proportion of heart failure decreased after the introduction of IRIS/Muse: in early adopting states from 5.4% in the index year to 4.1% after five years, and in late adopting states from 4.2% to 3.5% in the third year after the system had been introduced (Table). The proportion of causes of death coded R00-R99 decreased only in those states that adopted IRIS/Muse early and only in the initial four years (Table).

The quality of the cause-of-death statistics is age-dependent. The proportion of non-informative causes of death is highest in deaths <45 years of age (due to a high proportion of unknown causes of death) and in deaths ≥ 80 years of age (due to a high proportion of heart failure). After IRIS/Muse was introduced, the proportion of non-informative causes of death fell notably—and seemingly in a sustained manner—only in deaths ≥ 80 years. In deaths under the age of 45 years, however, the proportion of non-informative causes of death increased. The early introduction of IRIS./Muse was associated with a short term reduction only. The quality of cause-of-death statistics did not change in other age groups after IRIS/Muse had been introduced.

Discussion

Death certificates are the basis for any cause-of-death statistics. Only when at least one informative cause of death has been entered, an underlying condition can be selected as underlying cause of death that is compliant to the WHO rules—manually or automatically.

When taking into account non-informative causes of death, the quality of cause-of-death statistics in Germany has improved since 2005—especially thanks to a falling proportion of the heart failure as the cause of death. After IRIS/Muse was introduced, it seems that in deaths ≥ 80 years, other, informative causes of death that were additionally mentioned on death certificates were selected as underlying cause of death—for example, vascular dementia (ICD-10 F01), whose share has increased strongly over the years.

The fact that the quality of cause-of-death statistics in deaths <45 years has shown a worsening trend since 2005 is an unexpected finding.

Independent of whether IRIS/Muse was used, quality differences in cause-of-death-statistics between federal states are large. The proportion of selected non-informative causes of death in 2019 was between 9% in Saxony and ≥ 17% in Schleswig-Holstein, North Rhine-Westphalia, and Saarland.

Conclusions

An improvement in the quality of cause-of-death statistics after the introduction of IRIS/Muse seems to have occurred primarily in those states that introduced IRIS/Muse early on. Their cause-of-death federal statistics were on average of higher quality than those of the other states. The introduction of IRIS/Muse seems to result in a sustained reduction in the proportion of heart failure as a coded cause of death—particularly in deaths in older age.

An analysis of the differences in the quality of cause-of-death statistics between federal states might provide indications of quality improvement—for example, mandatory feedback possibilities for health officesin case information on a death certificate is incomplete or implausible. Those undertaking the postmortems should be familiar with the WHO definition of causes of death (4). Regular further training on this topic is desirable. Whether the planned introduction of electronic death certification will further improve the quality of cause-of-death statistics remains to be seen.

Susanne Stolpe, Bernd Kowall, Andreas Stang

Institute of Medical Informatics, Biometry and Epidemiology (IMIBE), University Hospital Essen, (Stolpe, Kowall, Stang), [e-mail]

Department of Epidemiology, School of Public Health, Boston University, Boston USA (Stang)

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

Manuscript received on 24 May 2023, revised version accepted on 8 August 2023.

Translated from the original Germany by Birte Twisselmann, PhD.

Cite this as:
Stolpe S, Kowall B, Stang A: The quality of cause-of-death statistics after the introduction of the electronic coding system IRIS/Muse—an analysis of mortality data, 2005–2019. Dtsch Arztebl Int 2023; 120: 793–4. DOI: 10.3238/arztebl.m2023.0190

References

1. Department of Information, Evidence and Research (2020): WHO methods and data sources for country-level causes of death 2000–2019. Global Health Estimates Technical Paper WHO/DDI/DNA/GHE/2020.2; WHO, Genf. https://cdn.who.int/media/docs/default-source/gho-documents/global-health-estimates/ghe2019_cod_methods.pdf (last accessed on 30 August 2023).

2. Johnson SC, Cunningham M, Dippenaar IN, et al.: Public health utility of cause of death data: applying empirical algorithms to improve data quality. BMC med Inform Decis Mak 2021; 21: 175

3. Stolpe S, Stang A: Nichtinformative Codierungen bei kardiovaskulären Todesursachen: Auswirkungen auf die Mortalitätsrate für ischämische Herzerkrankungen. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz. 2019; 62: 1458–67

4. World Health Organisation (WHO): International statistical classification of diseases and related health problems (ICD 10)—instruction manual Vol 2 (5th edition); Genf, 2016.

5. Eckert O: Verbesserte Qualität der nationalen und internationalen Todesursachenstatistik durch den Kodierkern MUSE. Statistisches Bundesamt (eds.): WISTA—Wirtschaft und Statistik 2017; 4: 118–32.