Contact
Join
Login

The importance of data quality

Checking the accuracy of vital events records

Methods and tools to evaluate the quality of vital statistics

Tabulation and generation of vital statistics for national policy

Presentation, communication and dissemination of vital statistics

Learn how to use the Learning Centre

Move your way through the CRVS system or simply click on a topic to dive into a specific subject.

An account lets you:

  • Save resources from our Library
  • Track your progress through the Learning Centre
  • Sign-up for our free newsletter

Methods and tools to evaluate the quality of vital statistics

Importance of age, sex and cause of death in mortality statistics

The previous section outlined the minimum information countries should collect for the registration of deaths. Three of these characteristics – age, sex and cause of death – are particularly important regarding the quality of mortality statistics. 

Age may sometimes be reported inaccurately in death records. These errors are reflected in age-related mortality patterns when the death records are aggregated. When this occurs, one sees age heaping 1 in certain age groups. It is well documented that people, when unsure about the exact age of the decedent, tend to report ages that end in 0 or 5. For example, if a 68-year-old woman died, her family might report her as being about 70 if they were not sure of her exact age or date of birth. These errors need to be detected as they will bias the data, and should be corrected applying demographic techniques for smoothing age heaping.

In the example of age heaping below, the orange bars represent the number of deaths where the reported age of the decedent ends in 0 or 5.

Age heaping

Additionally, the percentage of records with unspecified age should be very low. If a significant number of records are missing age-related information, this is a sign of poor-quality data and these omissions should be corrected at the source.

Occasionally aggregating the mortality data by age and sex can also reveal implausible sex distributions from what would be expected, which might signal under registration of certain age groups – for example, older women. 

The percentage of death records without sex specified should be very low. If a significant number of records are missing sex data, this is a sign of poor quality data and these omissions should be corrected at the source. ( Deaths cause of death statistics )

By far the most difficult characteristic to check at the individual record level is the cause of death (COD) . Some plausibility checks may be possible at the registration level, especially with the aid of checks included in the software (for example, the software may highlight prostate cancer as a COD for a female as being implausible). However, it would not be possible for a lay person to validate the majority of CODs. The quality of COD statistics is best assessed at the aggregated level where cumulative errors or misuse of certain causes become evident. 

However, checking the quality of COD data only becomes possible when you have standards and benchmarks to check against; hence, the importance of using global standards. For mortality data, the International statistical classification of diseases and related health problems, revision 10 (ICD-10) from the World Health Organization is crucial for classifying COD. The core classification of ICD-10 is the three-character code, which is the basic minimal standard that all countries should aim for in their COD statistics. 

The percentage of death records without a COD specified should be very low. A significant number of records are missing COD information is a sign of poor-quality data, and these omissions should be corrected at the source ( Deaths cause of death statistics ).

1 Figure 2 in Adair T (2016). Lessons learned from recent experiences with the evaluation for the quality of vital statistics from civil registration in different settings. United Nations Expert Group meeting on the methodology and lessons learned to evaluate the completeness and quality of vital statistics data from civil registration. 



Read more

Age heaping, digit preference and techniques to smooth the data: US Census Bureau (1994).  Population analysis with microcomputers  volume I, presentation of techniques, Washington, pp. 18–25.

AbouZahr C et al. (2010).  Mortality statistics: a tool to improve understanding and quality, Working Paper no. 13, University of Queensland School of Population Health, Health Information Systems, Knowledge Hub, Brisbane, Australia. 


Guidance for assessing and interpreting the quality of mortality data using ANACONDA
Guidance for assessing and interpreting the quality of mortality data using ANACONDA

For users of ANACONDA (statisticians and/or analysts in health and statistics departments, researchers, or other experts working with mortality data).

Download resource (1.94MB)
More Info

Authors: Mikkelsen L, Lopez AD

Publication date: October 2017

Resource type: CRVS resources and tools

Related resources: Course prospectus: ANACONDA


© University of Melbourne 2018   For more information on copyright visit our website terms