Several recent posts (this one among them) have talked about how the switch to ICD-10-CM and ICD-10-PCS will affect medical coding. Chief among the concerns is that coding productivity will go down because the new US ICD-10 coding schemes simply have more codes. This was even a brief topic of discussion on the #HITsm chat last night.
If coding continues as it has, being mostly a human based operation, this will likely be the case. However, I see this as an opportunity for healthcare IT vendors. I spent several years working in Natural Language Processing. I know that automated coding systems can approach and potentially exceed human skill levels using NLP techniques. This would especially be the case if machine learning and feedback techniques are applied.
The challenge is that the NLP system has to be designed in rather special ways. It will need to be able to provide its reasons for coding one way or another -- something that humans can do naturally, but which software does not do well. The importance of this stems from the need to deal with audits on coded information. I think the natural first step is to look at providing "coding assistance" to professional coders, rather than trying to replace them, as this takes advantage of existing processes.
In addition to throughput, the benefit of automated coding could be higher accuracy and greater consistency. It looks like an interesting opportunity in the health IT space.