DIY ICD-10 conversion Part 2

November 15th, 2013 / By Ron Mills, PhD

If you are a provider, you probably put ICD-9 codes on claims in order to get paid. If you are a payer, you receive claims with ICD-9 codes on them. (The “you” in those sentences has to be taken very broadly – it could mean, for example, “the software used by the coders employed by the company you outsource your revenue management to – or your claims adjudication to.”) In any case, I’m not going to talk just now about the ICD-9 codes on your claims.

So where else do you have ICD-9 codes? You might be surprised. They can turn up anywhere someone wants to identify a set of patients clinically and isn’t satisfied just using English and trusting that everyone will interpret it consistently. A case manager may want to identify all the “diabetic” patients in her care. An analyst may want to see if the number of ER visits for patients with “multiple significant trauma” is increasing. A computer program that pays claims may need to identify patients whose “medical condition” justifies paying for “breast augmentation.” A contract between payer and provider might specify a carve-out rate for “gamma-knife procedures.”

Any time the specification of a medical condition or procedure needs to be made precise, you might see ICD-9 codes. I can hear physicians snorting in derision over my concatenation of “clinical,” “precise,” and “ICD.” I agree. ICD – any version including 10 – is a pretty blunt instrument for representing the clinical aspects of a patient encounter. Rarely do physicians use it for their purposes, which has led them to invent more detailed systems like SNOMED. Perhaps when computers get even more powerful and can be fed a few years worth of rich EHR data to learn from, they can deduce broad categories from fine detail. But for now, blunt is what the rest of us who are concerned with payment and analysis prefer.

For want of a better term, I’ve been calling a list of ICD-9 codes that specify a set of patients a “policy.” This series is addressed to people whose job it is to rewrite their policies using lists of ICD-10 codes so that they specify the same set of patients. There are two ways of doing that: recoding and translation.

Before we get into the details (which any DIY converter must have), we digress to clear up some common confusions – ones I’ve frequently had to address in practice. I apologize in advance to the 80% of you who already know the material in the next few paragraphs, but my responsibility is to the other 20%.

Common confusion 1: What is a code? This discussion is concerned with four coding systems:

  • ICD-9-CM diagnoses
  • ICD-10-CM diagnoses
  • ICD-9-CM procedures
  • ICD-10-PCS procedures

The first three of these are organized hierarchically – in essentially an outline form. Here’s an example from ICD-9-CM diagnoses:

250 Diabetes mellitus

250.0 Diabetes mellitus without mention of complication

250.00 Diabetes mellitus without mention of complication, type II or unspecified type, not stated as uncontrolled

250.01 Diabetes mellitus without mention of complication, type I [juvenile type], not stated as uncontrolled

250.02 Diabetes mellitus without mention of complication, type II or unspecified type, uncontrolled

250.03 Diabetes mellitus without mention of complication, type I [juvenile type], uncontrolled

250.1 Diabetes with ketoacidosis (4 codes)

250.2 Diabetes with hyperosmolarity (4 codes)

and so on through 250.9.

Here 250 is further divided into 250.0 through 250.9 and most of those are further divided – for example, the division of 250.0 into 250.00 through 250.03.

All of these are commonly called “codes.” Technically, only those that are not further divided – only 250.00, 250.01, 250.02, and 250.03 in the example above – are “codes.” They are the only ones you can put on a claim if you want to get paid. The others go by various names, but we’re going to call them “headers.”

What’s the confusion? You will find many policies written in terms of headers: “… all patients with diabetes, ICD-9-CM code 250.” What does that mean? Some computer programs just label 250 an “invalid code.” Others add zeroes or nines until they find a real code – in this case 250.00, which almost certainly does not mean what the policy writer meant. What we do is interpret a header as all the codes underneath it. You may say 250, but you mean 250.00 or 250.01 or … or 250.93 – 40 codes in all.

This information is critical when you try to take the meaning over into ICD-10. A header in ICD-9 often does not correspond to any single header in ICD-10. To capture the same set of diabetic patients in ICD-10 that the codes under 250 in ICD-9 capture, you have to use 140 ICD-10 codes, which reside under the headers E10, E11 and E13.

So in what follows, when we say “code” we mean the codes that are not further divided. When you encounter a header, you will have to decide from the context what it means. For most of the techniques we are going to discuss, your best bet is to expand it into all the undivided codes in the outline it heads.

How do you do that? Here is my opportunity to recommend that if you are going to play this game, get a copy of the rule book. Specifically, you need the ICD-9-CM and ICD-10-CM/PCS coding books. In the old days, you had to buy an ICD-9-CM code book (technically the “ICD-9-CM Tabular and Index”). Now if you Google “ICD-9-CM code” a bunch of things that purport to be free on-line equivalents pop up, but I didn’t find any of them that gave the feel of the coding system the way a book does. For ICD-10-CM/PCS, the situation is better. Google “ICD-10-CM” and pick the first link to That should lead you to a free PDF of the entire ICD-10-CM diagnosis code set, in outline form, very much like the books of old. Similarly, if you go to the site and choose “2014 ICD-10 PCS and GEMs” from the left side, you can get free PDFs of the ICD-10-PCS procedures.

I’ve hit my word limit. I have three more common confusions to cover and then we can get down with the techniques for converting policies. More soon.

Ron Mills is a Software Architect for the Clinical & Economic Research department of 3M Health Information Systems.