To interpret trends in mortality between ICD-9 and ICD-10, comparability ratios have been produced. These are simply the ratio of the number of deaths coded to a cause in ICD-10 to the number coded to the equivalent cause in ICD-9.
Comparability ratios reflect the net effect of the change. If the ratio is 1, the number of deaths coded to that cause is the same in both revisions. If the comparability ratio is 0.5, half as many deaths have been coded to that cause in ICD-10 as in ICD-9. A ratio of exactly 1 does not mean that assignment of deaths to that category is unchanged. It may mean that the same numbers of deaths have moved into and out of the category.
How to apply comparability ratios to examine trends over time
Comparability ratios should be applied to data coded in ICD-9 in order to examine trends over time. For a particular cause, the number of deaths coded to the equivalent cause in ICD-9 in the years being compared should be multiplied by the comparability ratio in order to give an 'expected' number of deaths which would have been coded to this cause in ICD-10. The ratio can also be applied directly to rates, to give an 'expected' rate. When applying comparability ratios, it should be borne in mind that they are based on 1999 data. Therefore when applying them to data for different years, the number of expected deaths in ICD-10 may not equal the total number of deaths in that year. This is because comparability ratios provide a simple approximation of the change in 1999. The effect will be very small in relation to the total number of deaths.
These ratios are only applicable to England and Wales mortality data from 1993 onwards, when ONS introduced automatic cause coding. Data for years prior to 1993 are not comparable with data for 1993 to 2000. The ratios can only be applied to causes defined by the exact group of ICD codes in both revisions that were used to calculate them. For example, the ratio for Ischaemic Heart Disease (ICD-9 410-414, ICD-10 I20-I25), which is 1.007 for females and 1.005 for males does not apply to Acute Myocardial Infarction (ICD-9 410, ICD-10 I21-I22), because of movements between these codes and others within the Ischaemic Heart Disease block. The ratio for Acute Myocardial Infarction is 0.926 for females and 0.937 for males.
When are comparability ratios not needed?
Where a comparability ratio is shown in the spreadsheet but its confidence interval includes 1, this means that the difference between the number of deaths allocated to the cause in ICD-9 and ICD-10 is not statistically significant. Adjustment will add little to comparisons over time.
When are comparability ratios needed?
Where a comparability ratio is shown in the spreadsheet and its confidence interval does not include 1, this ratio should be applied to data for 1993 to 2000 to adjust for the change to ICD-10 when making comparisons with data for 2001 and subsequent years. For causes with very large numbers of deaths, such as IHD, ratios close to 1 can still appear significant. In this case adjustment may have little effect on rates.
How should ratios be applied to specific age groups?
The comparability ratios in the spreadsheet can be applied to age-specific data as well as to data for all ages combined. Earlier analyses have shown that there is little difference in the age-standardised rates obtained if age-specific ratios are applied to data instead of an all-age ratio.
However, if the data being examined are to be used to examine trends in specific age groups, it may be sensible to apply age-specific comparability ratios. This is because there may be systematic differences by age in factors affecting the selection of the underlying cause of death. For example, the level of detail provided on a death certificate can differ by age. In addition, the likelihood of a person having chronic conditions varies by age. These conditions may be selected as the underlying cause of death instead of causes such as pneumonia.
Where there is a statistically significant pattern in comparability ratios by age, age-specific comparability ratios for the broad age groups under 75, 75-84 and 85+ are available.
How can trends be examined for causes of death which have ICD-10 codes but no equivalent in ICD-9?
Some causes listed in the table have no equivalent codes in ICD-9. For these causes of death, it is not possible to analyse trends before the introduction of ICD-10. However, by using the bridge coded data it is possible to identify the ICD-9 codes which were used to code deaths assigned to these causes in ICD-10. We plan to investigate the coding of mesothelioma in later work.
How can trends be examined for causes of death not covered in the comparability ratios available online?
Comparability ratios have been produced for 108 major causes of death. Diseases not covered by these groups can be examined using the bridge-coded data for 1999. This dataset contains counts of numbers of deaths by ICD-10 code, ICD-9 code, age and sex.