August 27, 2013

Some Errors in Codes Likely Come From Inadequate Interest By Physicians – at the Beginning


Once upon a time, even though Medicare was still the most reliable payer for health services at acute-care hospitals, assignment of ICD-9 codes was not considered to be of any great importance there.


Hospitals were paid in a variety of ways, from per-diem payments to portions of charges to allowable fees (and percentages of them). Then the earth cooled and the dinosaurs disappeared and some mathematicians from Yale University created a diagnosis-related group (DRG) methodology for predicting hospital costs. The federal government latched onto this predictable system and created the Inpatient Prospective Payment System, implementing it in 1983. The Medicare payment methodology changed from an open-ended, fee-for-service model to a fixed fee model.  And this was all based on ICD-9 code assignment.

Medical records librarians, as they were known then, were given the responsibility for assigning ICD codes from physician documentation in hospital charts, and rules were created to control assignment of these codes, sequencing of these codes, and grouping to DRGs. And some of the physicians at the hospitals couldn’t care less.

Well, why is it necessary to understand all of this? At the inception of the DRG system, to some degree it was only the medical records librarians who became hospital coders who cared. And they were given the responsibility to interpret what a physician wrote in a chart in order to be able to assign ICD codes for the DRG assignments so that the hospital could get paid in this new system – yet nobody gave them any help. Well, as some physicians couldn’t care less, let’s just say any help provided was occasionally less than stellar. Mistakes were made, not because the folks who came up with the concepts or advice did something wrong; they just didn’t have some of the clinical insights needed to arrive at the proper conclusions (and, with their history of being rebuked by some physicians, they made do with what they had). Who could blame them?

What became significant about the errors is twofold – most of them resulted in higher DRG payments than the case deserved, for one thing. Also, international statistics derived from data from the United States shows that we were bad at patient care. And neither of these outcomes should have been acceptable.

Anemia of Pregnancy, Delivered with Complication

The first issue I became aware of emerged back during the time of Y2K. I was doing an audit at a hospital that had been questioned about the apparent overuse of 648.22 and 285.1 in obstetrics cases that led to higher reimbursements than the DRG payor thought were valid. The circumstance seemed to be that virtually every child delivery was associated with a lower-than-normal hemoglobin level for the newly non-pregnant female, and, when the patient was treated with iron, blood transfusion or repeated CBC, both codes were assigned. I looked at the index and read the definition of the ICD codes in the 640 series. All of them were conditions that existed during pregnancy or at the time of delivery. Then I got to the 648 series, which cites “the listed conditions when complicating the pregnant state.” I looked at all of the entries in 648, which included diabetes and thyroid disease and drug dependence and maternal congenital cardiac disorders and so forth – all conditions that had to have been present, based on the grouping, throughout the entire pregnancy. Then I looked at the classification of fifth digits and recognized that the complication could have occurred during the pregnancy, during the delivery, or after the delivery. But the primary condition had to exist during the entire pregnancy. And one was supposed to assign a code for the complication separately –if one occurred.

So I wrote to Coding Clinic and it came up with its advice delivered during the first quarter of 2002. Specifically, the advice indicated that even if the 648 condition in question didn’t occur until after the delivery, one should assign the code and also code the complication. Basically, they said that anemia of pregnancy, delivered with complication, caused anemia when the patient did not have anemia during the pregnancy. They based that on the conclusion that “pregnancy,” according to the definition they created in the early 1980s, included the time period from inception until two weeks after delivery (the entire “puerperium.”)

I couldn’t accept the logic. I called the physician in charge of the World Health Organization for the northern half of the western hemisphere: Roberto Becker, MD. He was totally taken aback and dismayed by the coding guidance interpretation of the intent of the 648 series and the advice on use of the code for a condition that did not exist until after the delivery. He said that the 648 series is intended to be assigned to a condition that existed throughout the mother’s entire pregnancy, if not a significant portion of that pregnancy, in order to be eligible for use of that code – and how anyone could interpret any different way was beyond his comprehension.

So the statistics in the U.S. show that we probably don’t have pregnancies with thyroid disease or pregnancies with diabetes or pregnancies with congenital heart disease any more than the rest of the world, but we have an awful lot of anemic ladies getting pregnant – even getting pregnant with anemia due to acute blood loss.

In ICD-10, this issue seems to be corrected, as no anemia that did not exist during the pregnancy is reportable with the O99.0 series – marking a total reversal from ICD-9.

Cardiorenal Syndrome


Cardiorenal syndrome is assigned 404.90, stating “without heart failure” and “with CKD, stages 1–4 or unspecified.” Obviously, from the classification, one has to presume that cardiorenal syndrome always occurs due to hypertension. But that couldn’t be further from the truth. The heart disease associated with cardiorenal syndrome can occur due to ischemic heart disease, alcoholic heart disease, amyloid heart disease, or any origin of heart disease at all. And the renal disease of cardiorenal syndrome can occur due to diabetes, obstruction, rhabdomyolysis, lupus nephritis, or any origin of renal disease at all.

Certainly, both or either of these conditions can be caused by hypertension, but to presume that they are all a result of hypertension is wrong. We know from coding rules that renal disease is linked to hypertension automatically without additional specification, but heart disease is never automatically linked to hypertension. And the fact that acute decompensation of heart disease may have an adverse effect on renal function (plus the fact that acute decompensation of renal disease may have an adverse effect on heart function) is totally ignored within the current ICD classification. Cardiorenal syndrome needs the identification of any chronic heart disease, its cause and any chronic kidney disease, and the identification of any acute decompensation of either or both, plus their cause(s), to describe the disease fully – and 404 has no business being associated with coding of cardiorenal syndrome unless the physician specifically indicates that hypertension is involved.

In ICD-10, I13 codes, with or without heart failure and with CKD stages 1 through 4 and stage 5 or ESRD are exactly the same as they are in ICD-9. It basically indicates that no cardiorenal syndrome can exist without both elements being due to hypertension (and that is incorrect).

Hypostatic Pneumonia

Let’s take a look at ICD code 514. I have written tomes about this. It’s a sore point in the involvement of all members of the cooperating parties because they all have involvement in the interpretation of definitions, misguided advice, and inappropriate classification of the condition originally called “apoplexy of the lung.” In order to appreciate the intended meaning fully, one has to go back to ICD Pre-version 1. This was a condition found at autopsy for people who were admitted for something else, yet who happened to die. Many of these people were severely malnourished and had been lying in basically one position for weeks to months – likely with tuberculosis, cancer or some other terminal disease – whose lung tissue became so congested with blood because of the lack of positional change that it developed the feel of liver to the pathologist’s hand. Any local inflammation was due to the solidified blood in the bronchi and alveoli and not due to an infectious source, unless one followed the original condition. It variously was called congestion of the lung, dropsy of the lung, active congestion of the lung, hypostatic pneumonia (with no consideration that this was an infectious process, as are other pneumonias), or other terms. It was one of the original 100 causes of death (No. 94) in the Bertillon Classification of the 1890s. It was defined as a terminal condition of the lungs.

But without proper direction by medical historians or an overabundance of physicians who cared, the people who developed coding instructions, interpretations and guidance only had the words to go by. So they took the various terms used through history for this condition and said that any time a physician uses one of the terms, you assign 514. It started with Coding Clinic’s publication from the third quarter of 1988, when it told coders to assign this code when it was addressed as non-cardiac chronic pulmonary edema. Then, during the second quarter of 1998, a subsequent publication included information about assigning the code for documentation of hypostatic pneumonia (which led a lot of people to assume that, if a patient develops pneumonia after having atelectasis, the physician should call it hypostatic pneumonia). They disregarded the fact that patients were treated with antibiotics at the time, and it obviously was thought to be a bacterial pneumonia. Despite the fact that a Centers for Medicare & Medicaid (CMS) authority said this circumstance should be assigned ICD code 486, current advice persists even today. Online sources also have bought into the bad definitions, because they’re the main ones you can find – so they help perpetuate the myth that it is a real pneumonia.

Additionally, issues arise because of physician documentation of “pulmonary edema” when acute pulmonary edema is implied, but the record does not include the word “acute.” Guidance leads the coder to 514. If the physician actually implies that the patient has chronic pulmonary (hypostatic) edema because of prolonged lying in one place, noting that he does not mean congestive heart failure or pulmonary edema in end-stage renal disease (or that the condition is related to some acute incident or trauma), then the treatment should involve simple mobilization (and the patient was likely in the hospital for a totally different reason). 514 is only 514 when that’s what it is. It should not be based on inadequate documentation. That’s inadequate advice to coding professionals.

In ICD-10, we find that J18.2 carries exactly the same implications that 514 has had with ICD-9, namely that the condition represents pneumonia and that it can be either broncopneumonia or lobar pneumonia. And it’s still wrong.

SIRS + Infection = Sepsis

SIRS (systemic inflammatory response syndrome) is a syndrome of changes in vital signs and white blood cell counts as the systemic response to inflammation, with the inflammation in particular being key. Whether it’s infectious or not, it’s still a response to inflammation. Yet Coding Clinic doesn’t care. In its issue from the first quarter of 2010, with an example given of a physiological response to a medication (not an inflammatory issue at all, not an allergy, just the effects of the drug on a patient’s circulatory system), coders yet again are advised to assign an ICD code based on the documentation (but who really knows what the physician documented without taking a look at the chart?)

Obviously, the authors of that piece must think coding professionals are not smart enough to think for themselves – to realize that, given epinephrine, a person will experience a rise in pulse rate and respiratory rate; running down the block, a person will have a rise in pulse rate and respiratory rate; given a shot of steroids, a person will develop an elevation in white blood cell count and probably pulse rate; in the face of atrial fibrillation with rapid ventricular response, there will be tachycardia, and often tachypnea. These things are not inflammation, and just because the physician says that the patient has two of the four signs of SIRS, you should not automatically assign anything from the 995.9x series without first getting clarification that the abnormalities are being caused by an inflammatory process. When I sent Coding Clinic a letter discussing this situation and offering this opinion, their response was, “we already said this is the way it is; we’re not clinicians, so just do it the way we say.” Oh, come on.

Then there’s sepsis. Sufficed to say, SIRS plus infection is not sepsis – unless it is. These statistics are made up, but very illustrative. Probably 99.9999 percent of people with a cold or bronchitis or an ingrown toenail will have two of the four criteria identified in 1991 by a group of well-intentioned physicians as indicative of “SIRS + infection = sepsis,” and most of these people never get to see a doctor. Probably 95 percent of people with these signs and symptoms who do see a medical professional get something for treatment and go home. Of the very, very few people who even get to an emergency room, probably over 90 percent are treated and released. Of the few who finally are admitted to the hospital with two of the four criteria and an infection, maybe 30 percent actually have sepsis. Yet our coding guidelines tell us that SIRS (or, at least, two of the four criteria) in the face of infection is to be coded as sepsis 100 percent of the time. And that’s ludicrous.

This is all derived from two issues. First, all of the studies done on patients whose evaluation led to the identification of these criteria were already on critical care units. So they were already in the smallest possible group of people. Some had an infectious source and some didn’t, but still others had an inflammatory source of a different nature. Naturally, those patients with the criteria and an infection had sepsis, but all the rest of the population of the world had been weeded out. Secondly, subsequent studies done on a wider population of patients than only ICU patients demonstrated clearly that these criteria were inadequate to predict the presence of sepsis, even in face of infection, without identifying that the patient was critically ill first. Jean Louis Vincent, MD, editor-in-chief of Critical Care, had written in 1997, “dear SIRS, I don’t like you any more,” and just wrote in an editorial in his journal,

“The confusion related to sepsis definitions and terminology was amplified some 20 years ago when participants at a North American consensus conference confused signs of infection, such as fever and altered white blood cell count, with signs of sepsis.”

As a result, the world knows that the United States is inundated with sepsis – we can’t keep anyone healthy. That our sepsis population rose from 310,000 in 2004 to more than 750,000 by 2009. That most of our patients get sent to the ICU for sepsis from the emergency room, whereas most of the rest of the world has people getting admitted to the hospital in general (and, if and when they get sick, then they get transferred to the ICU). That our length of stay for sepsis patients and costs per patient is lower than it is anywhere else in the world. Could it maybe be related to the fact that half of the patients coded as sepsis don’t have it? But the coding rules say we have to code it that way – all of the coding rules, advice and guidelines. And hospitals are being paid about $5 billion extra a year because of this error alone. And the RACs are taking a bunch of it back. And we’re paying for that from both sides.

ICD-10 may offer an improvement in that it eliminates the code for SIRS due to infection completely and only permits assigning codes in the R65.1 range for SIRS due to noninfectious sources (such as burns or acute pancreatitis). We will still have R65.2 for severe sepsis, which is the way it should be. As of the writing of this article, ICD-10 official coding guidelines do not refer to SIRS at all, except as it relates to noninfectious sources. If this doesn’t change, we will be considerably better off. Sepsis is sepsis, and that’s how it should be.

Complications in the Neonate

Newborns are a different breed. There’s not a lot of attention being paid to newborn codes, or advice, even though all of us were at one time newborns. It’s because virtually no newborns are Medicare beneficiaries, so most folks don’t pay them a lot of attention unless they’re their own. Newborns may come into this world with congenital conditions  however, based on maternal environment or genetic problems (or complicated by the process of being born). Neonates are considered by the coding world as “newborns” through the first 28 days of life. Now this leads us to some issues.

There is a series of codes that is designed to identify problems with being born, whether they are genetic problems or other congenital problems or issues that happen to occur on the way down the birth canal. The identification of these things makes the delivery higher-risk merely because of their presence. The intent of the code sets is to name those conditions that make being born a potential problem. They are not intended to be assigned for conditions that have absolutely nothing to do with the process of being born, from the sperm and egg into the world. It doesn’t matter when in life these things are identified; the code may be assigned for them. However, because of the well-intentioned coders who were given the enormous task of dealing with code sets without clinical insight or resources to ask, many came to the conclusion that anything that happened to a person within the first 28 days of life was eligible to be assigned one of these codes, even if they had nothing at all to do with the process of being born.

So we have a lot of complicated births identified long after discharge from the hospital (within 28 days of life) being added to the birth statistics – and, naturally, we have a higher incidence of complication of births than any country in the world. In addition, the DRG reimbursement for every one of these cases is three to five times what it would have been if the child were 28 days and one second old. That’s because the proper code only was supposed to be assigned based on the original intent, and the way the rest of the world treats these codes. We’re the only ones messing it up. And it’s because of the basic tenets all coders in the U.S. were taught – that a child is a neonate through the first 28 days, for all coding purposes. And that’s wrong.

The ICD-10 official coding guidelines address this issue briefly, only saying that, if it’s not clear that a condition was acquired by the birth process or was community-acquired, the default would be the code used for conditions occurring due to the birth process. If coding professionals are informed about this difference, they likely will come to the correct conclusion in almost all cases. However, some will be used for assigning “complication of delivery” codes regardless, and this will not help our statistics.

Robert S. Gold, MD

(1942-2016) The late Robert S. Gold, MD, was a nationally known physician, responsible for having championed clinical documentation with a peer-to-peer educational approach in hospital organizations. Dr. Gold was a cofounder and the CEO for DCBA, Inc., a consulting firm that concentrates on development of Clinical Documentation Improvement (CDI) programs that aid in proper data streams, proper communication within the medical records and proper reimbursement. Dr. Gold served on the ICD10monitor editorial board from 2011 through January 2016.