Discussion 7.2 Comparing dump codes in ICD9 and ICD10

Please note that the pages on this website do not reflect the methods that were used to generate the final set of results for the GBD-2010 study, which was published in December 2012.

-- If you are interested in the methods or results of GBD-2010 and subsequent revisions, please visit the official website of IHME's GBD project (http://www.healthdata.org/gbd). The materials described on the website are now obsolete but these webpages have been retained as an internet archive of the work of the group.

-- Please visit the website www.globalburdenofinjuries.org to find out more about other closely related collaborations of our group members.

Last Update: July 15th 2009

Status of discussion: Closed

This paper is a sub-discussion of the broader discussion on unspecified categories (Discussion Topic 7). It focuses on a comparison of the dump codes in ICD9 vs ICD10 and the implications for GBD-2005.

Where we stand and the way forward

• Preliminary comparisons of deaths coded to ICD9, ICD9 Basic Tabulation List, and ICD10 in multiple countries has been conducted.
• A draft manuscript is being prepared for circulation in the expert group for feedback; results will also be discussed with the expert group in the GBD-Injury symposium in Melbourne July 29 2009; we expect to submit the manuscript for publication mid-August 2009.

Results

[Note: This figure has multiple sheets. To view other sheets press the >> button on the figure]

Figure 1: Comparison of injury mortality estimates using ICD-9 and ICD-9 BTLcodes for selected country-years for which detailed ICD-9 data was available.

Figure 1 shows that the use of ICD-9 BTL codes leads to a distortion in the estimates of unintentional injuries. Falls are over-estimated at the expense of other unintentional injuries, especially road injuries. The discrepancies exceed 10% in most countries and exceed 20% in some. The discrepancy is smallest for Slovenia. As examples,

• In France, falls are overestimated by 70% and road injuries are underestimated by 18%.
• In Italy, falls are overestimated by 44% and road injuries are underestimated by 26%.

Table 2: Distribution of deaths assigned to key partially-specified categories in the ICD-10 and ICD-9 eras.

Table 2 describes the distribution of deaths assigned to key partially-specified cause of death categories in these countries during the ICD-9 and ICD-10 eras. Comparing Figure 1 and Table 2 reveals that the degree of discrepancy associated with ICD-9 BTL codes is closely related to the number of deaths ascribed to unspecified unintentional injuries. Thus, in Slovenia and N. Zealand, where the number of deaths assigned to these dump codes is relatively small, Figure 1 shows that ICD-9 BTL based estimates match closely with the ICD-9 based estimates. On the other hand, for countries where Figure 1 shows a big discrepancy (e.g. France, Italy, Spain), Table 2 shows that a relatively large fraction of unintentional injuries are assigned to unspecified causes.

[Note: This figure has multiple sheets. To view other sheets press the >> button on the figure]

Figure 2: The flow of the key partially-specified codes within the unintentional injuries from ICD-9 to ICD-10.

Figure 2 explores the flow of ICD codes that comprise the unspecified unintentional injury category from ICD-9 to ICD-10. During the ICD-9 era, of the three dump codes (E887, E9289 and E9299) that make up the this category, the bulk of deaths are coded to E887 (Fracture - cause unspecified). The total of deaths assigned to E887, E9289 and E9299, matches closely the number of deaths assigned to the ICD10 dump code for unintentional injuries, X59 (Exposure to unspecified factor).

Figure 3 (a)

[Note: This figure has multiple sheets. To view other sheets press the >> button on the figure]

Figure 3 (b)

[Note: This figure has multiple sheets. To view other sheets press the >> button on the figure]

Figure 3: Discontinuities in the time histories of injuries before and after the switch in coding from ICD-9 to ICD-10. Results are shown grouped by, (a), external cause, and (b) grouped by country.

Taken together the results in, Figures 1–3 and Table 2, suggest a set of characteristics that occur commonly in several countries, which we refer to as the “typical pattern”. These characteristics are:

1. Falls+ & Road Injuries-: When we compare deaths coded to the ICD-9 BTL Falls and Road Injury categories with estimates using the detailed ICD-9 data in conjunction with the default GBD methods of age- sex- proportional redistribution of partially specified categories applied to the detailed ICD-9 coded, we find that the ICD-9 BTL overestimates Falls and underestimates Road Injuries. In general, all other unintentional injuries (besides falls) are also underestimated.
2. Of the three ICD-9 codes that make up the unspecified unintentional injury category, E887 captures the bulk of the deaths.
3. The time trend of the sum of the three ICD-9 codes that make up the unspecified unintentional injury category (E887+E9289+E9299) is continuous with the trend for the ICD-10 code X59.
4. This “typical pattern” is most pronounced for countries where a large fraction of deaths are coded to the unspecified unintentional injury category (Table XX).
5. Inter-personal and self-inflicted deaths are unaffected – i.e. estimates of these deaths using ICD-9 BTL data match those using ICD-9 detailed data.

Table 3: The extent to which different countries follow the typical pattern and notable deviations from it.

In addition to the typical-pattern noted above and the country-specific effects described in Table 3, there are a few other notable patterns evident in the time histories displayed in Figure 3:

• Mechanical plus sharp objects: This category has a large drop in most countries across the transition from ICD9 to ICD10. These countries are Australia, Austria, Belgium, Canada, N. Zealand, Spain, USA, and, could also include Cuba, and Slovenia. In Australia, there is evidence from two years of dual coding to ICD 9 and ICD 10 (REF) that shows large case flow from E919 to a residual transport group that includes special vehicles primarily used for agriculture, on industrial premises, and in construction. This analysis also showed a flow of deaths to W49, which is the dump code for exposure to the inanimate mechanical forces block. [Question: Would this problem be reduced by assigning all V83-V85 cases to the machinery category]
  • Adverse effects of medical treatment: Different countries show remarkable differences in the number of deaths coded to this category. See, for instance, the difference between deaths assigned to this category in Belgium and France (Figure 1).