Guest post: Teaching physicians to “think with ink”

by James S. Kennedy, MD, CCS, CDIP, CCDS

Several journals recently challenged our proficiency in diagnostic decision-making. A lead article in the September 12, 2017 Annals of Internal Medicine asserts that “several signs indicate that the focus on diagnostic reasoning in internal medicine training may be threatened.”

An article in JAMA published on the same day discussed diagnostic calibration and its impact on medical errors.

My experience in ICD-10-CM/PCS documentation and coding integrity is that many physicians know in their heads what is wrong with their patients; however, they have not been taught to “think with ink” in describing their patients' illness in the EHR using ICD-10’s language to ensure proper coding.

Why is this important? CMS and private payers hold physicians accountable for risk-adjusted outcomes and cost efficiency. CMS measures these metrics as a ratio of the observed metric to the expected metric, known as O to E.

The observed metric is what happens. It is the actual costs, lengths of stay, mortality rate, readmissions, and complications that our patients experience, all of which are easily measured once defined. Death is obvious, as is the length of stay. Cost can be determined using a cost accounting system or through multiplying various charges by a cost to charge ratio. Readmissions are not so easy since patients may be admitted to other facilities than our own. Complications are tricky; what is truly a complication versus usual or integral to the procedure? Even so, once defined, these tangible components are simple to measure.

The expected metric, however, is a more difficult measurement to arrive at. You and I may use a clinical algorithm based on various indicators, like APACHE, SOFA, CHA2DS2-VASc, ABCD2, or other scores, to measure disease severity. However, CMS, state governments, insurance companies, and many “quality measurement” companies use only ICD-10-CM/PCS diagnosis and procedure codes, and these codes are assigned only if I document the magic words in the record that a coder can properly abstract.

For example, let’s say that I am a spine surgeon planning a multi-level laminectomy for cervical spinal stenosis with spinal cord compression. My physical examination demonstrates 3/5 strength in the right arm and right leg, 4/5 strength in the left arm, and 3+/5 strength in the left leg with a Babinski sign. If I only state that these extremities are weak, the ICD-10-CM code for weakness adds no credit for the added risk my patient has.

On the other hand, if I describe that the patient has neurologic quadriparesis caused by the spinal cord compression, upon ICD-10-CM coding of this term, my case gets four to five additional days in expected length of stay, $12,000 in additional expected costs in MACRA/MIPS cost-efficiency modeling, $16,000–$20,000 in additional costs to finance the operating room and nursing care my patient will receive, and a higher expected readmission or complication rate.

In another example, let’s say that our risk-adjusted readmission rate is high, and we are on a hospital committee to bring it down. We can do one or both of the following:

• Spend the money to send a home health nurse out to the patient’s home every day and ensure the patient’s confounding factors are compliantly addressed. Yes, this is expensive and time-consuming, but it does reduce the actual readmission rate (the numerator in the O to E algorithm).
• Ensure all the comorbidities predicting this risk of readmission have been properly documented and coded in ICD-10-CM/PCS during the current admission and on an outpatient basis during the previous six to 12 months, all of which increase the expected readmission rate (the denominator in the O to E algorithm).

Note that in this workflow, I can keep my actual numerator metric (e.g., length of stay, cost, readmissions, mortality) the same but still favorably influence the risk-adjusted metric by increasing my denominator metric.

The trick, of course, is knowing what ICD-10-CM/PCS codes matter and how to ensure their proper documentation and capture to influence the denominator of the O to E ratio. You may wish to ask directors in your inpatient coding, quality, and case management departments how these methodologies work; explanations are not hard to find if you know where to look.

Editor’s note: This article was adapted from the original in JustCoding. Dr. Kennedy is a general internist and certified coder, specializing in clinical effectiveness, medical informatics, and clinical documentation and coding improvement strategies. Contact him at 615-479-7021 or at jkennedy@cdimd.com. Opinions expressed are that of the author and do not necessarily represent HCPro, ACDIS, or any of its subsidiaries.