Medical Coding in Data Management: Manual vs Auto-Coding

Medical coding is an important step in data cleaning activities within data management. The main objective of performing medical coding is to have the medical terms interpreted uniformly and in a standardized format; hence medical coding is required to be completed by using standardized medical dictionaries.

medical coding data management

In CROS NT coding is usually performed within the clinical database system. However if the clinical database does not include a suitable coding functionality, the coding is completed outside of the system using a standard coding program in SAS which was developed internally.

Data which can be coded include adverse events, medical histories, concomitant medications and even physical examinations, if required. Standard dictionaries such as MedDRA and also WHODrug are typically used.

Typically all medical histories and adverse events are summarized by preferred term (PT) and by system organ class (SOC) whilst all concomitant medications are summarized by preferred term (PT), by anatomical main group (1st level of the ATC classification) and chemical subgroup (4th level of the ATC classification). All of this information are outputted in the data which are used for statistical analysis.

Medical Coding in the world of EDC

In the current world of EDC, the majority of terms should be auto-coded, therefore the coding efforts should be minimal. However this is not always the case and in order to get the term auto-coded the verbatim must exactly match with appropriate term available in the medical dictionary. When auto-coding fails, manual coding will take place.

There are many challenges while performing the manual coding of medical histories or adverse events such as multiple medical concepts recorded together which requires a correction prior to coding taking place. The multiple signs and symptoms might also be reported, in such cases the medical coder should query the actual diagnosis with the investigator site. Any unclear terms with insufficient details are also required to be queried with the site.  In the case where the actual surgery together with the condition is reported, the medical coder should check whether the condition is included in the separate record. If not, this can be split using self-evident corrections or queried with the site.

The use of self-evident corrections to update the spelling or expand any abbreviation can be used if the sponsor is willing to use it. If not, this is usually queried with the site to correct the spelling and expand any unclear abbreviations.

Once misspellings have been corrected, queries have been responded by sites, reviewed and closed by the Data Manager, the actual verbatim is coded to primary System Organ Class (SOC) as a standard.

The coding of concomitant medications brings its own challenges as well. As a standard, CROS NT is using WHO Drug Enhanced format B2 of the dictionary, however Enhanced format C can also be used if it is required by the sponsor.

In order to perform as accurate coding as possible, certain fields are required to be completed such as Indication, Route and also the Dosage in order to assign the correct ATC code.

The auto-coding is also used when the drug recorded in the CRF must match exactly with the appropriate term available in the WHO Drug dictionary. However if more than one ATC code is listed for a specific drug, the ATC selection will fail the auto-coding and will be used for manual coding instead. The medical coder will use the information recorded on the concomitant medication page to select the most appropriate ATC code.

auto coding data management WhoDD MedDRA

The actual coding is done to the direct drug match recorded in CRF, however the statistical summary is done on preferred term (PT).

At the end of the study, the coding is usually reviewed to ensure the consistency and accuracy of the selected codes. Only after the actual review will the coding will be sent to the client for the medical review and approval.

In summary, coding is a specialised activity and experience is built up over time in order to become familiar with the various coding dictionaries and tools. The amount of interpretation tends to be reduced through an automated process, allowing for a more standardised, accurate result. CROS NT has developed this expertise and offers coding activities as part of their data management services as well as any standalone requirements.