Tracking Specificity in ICD-10

2016 - 2 April - SCCM’s Quality Improvement Initiatives
David L. Carpenter, MPAS, PA-C
David L. Carpenter, MPAS, PA-C, discusses the nuances of ICD-10.

ICD-10 introduced significant complexity compared to ICD-9.(1) The number of codes increased from 14,000 to 69,000. Fracture codes alone increased from 747 to 17,099. In addition, there are more than 25,000 codes designating either right or left.(2) With these changes, the ICU presents a particular challenge. Diagnoses evolve over time, while patients’ conditions rapidly change. This adds additional levels of complexity to correct coding in the ICU. This also means increased impact to the ICU as ICD-10 changes require extra specificity in description and coding. At our institution, providers are responsible for entering ICD-10 information in both the electronic medical record and separate billing software. This article aims to provide education on coding two areas of specificity, coding to the full number of characters required and proper use of unspecified codes.
Coding for Proper Specificity
The first principle of specificity is that a code is invalid if not coded to the full number of characters required.(2) As noted in the previous article, ICD-10 contains a total of seven alphabetic or numeric characters. For certain codes, a specific number of characters are required to ensure proper specificity of the code.  ICD-10 coding manuals specify when additional characters (eg, 5th, 6th, or 7th characters) or placeholders are required. For example, ST elevation myocardial infraction (STEMI) and non-STEMI (NSTEMI) codes demonstrate how different levels of specificity have different character requirements, as follows:

I21 – STEMI and NSTEMI. Both require a fourth character.
I21.0 – STEMI of anterior wall. Requires a fifth character.
I21.01 – STEMI involving left main coronary artery. Requires no additional characters.

Codes I21 and I21.0 would be incorrect codes, since they are not coded to the full number of characters required. I21.01, on the other hand, does not require a 6th character since there is no designated 6th character for this code.
Another problem arises when providers assume that a greater number of digits equals greater specificity. This is compounded by the increase in characters from five to seven between ICD-9 and ICD-10.

A78 – Q fever. No further descriptor is required since this is a complete descriptor.
S59.919D – Unspecified injury of unspecified forearm. Despite having all seven digits, it is not specific to either laterality or injury.

While coding manuals show when a code needs additional characters, provider coding software may allow coding without the required number of characters (specificity). Organizations should review provider-submitted ICD-10 codes and request additional information and specificity before submitting billing documentation.

Appropriate and Inappropriate Use of Unspecified Codes
The second principle of specificity in coding revolves around the use of unspecified codes. Unspecified codes are for conditions for which there is insufficient information to code more specifically. The key to a diagnosis code is the information in the medical record. In general, unspecified codes should be used when(3):

• The unspecified code is the best choice to accurately reflect the healthcare encounter
• Sufficient clinical information to assign a specific code is unknown
• The code most accurately reflects what is known at the time of a particular encounter

The challenge comes when a provider has difficulty differentiating among various codes for a condition and instead chooses an unspecified code despite sufficient clinical evidence for a more specific code. In these cases, providers must pay careful attention to ICD-10 taxonomy to select the proper code. Conversely, it is incorrect to submit a more specific code simply to increase specificity if the code is not supported by the medical record. An unspecified code is most commonly used when additional testing is being done or when the condition is evolving.

Example: A patient is admitted to the ICU with pneumonia diagnosed by chest radiograph. A sputum sample is sent to the laboratory. The next day, the laboratory reports gram-negative rod bacteria in the sputum. The day after that, the sputum grows Pseudomonas aeruginosa. The codes for each successive day should be:

Day 1 – J18.9 Pneumonia, organism unspecified
Day 2 – J15.6 Pneumonia due to gram-negative rod (GNR) bacteria
Day 3 – J15.1 Pneumonia due to Pseudomonas aeruginosa

On day 1, it is appropriate to code the pneumonia as unspecified since there is no other information to specify the type of pneumonia. By day 2, the pneumonia can be specified as GNR bacterial pneumonia, but the organism is still unknown. By day 3, the encounter can be fully specified with the organism identified.
While it is appropriate to use unspecified codes when information is unavailable, in other cases their use is hard to defend.

Example: Patients with respiratory failure are common in the ICU. Three possible codes are:

J96.01 Acute respiratory failure with hypoxia
J96.11 Chronic respiratory failure with hypoxia
J96.90 Respiratory failure, unspecified, unspecified whether with hypoxia or hypercapnia

There is little justification for using the third code since the clinician should know whether the condition is acute or chronic as well as whether or not hypoxia is present.

A similar case would be a forearm fracture. There would little cause to use S52.90XB, unspecified open fracture of the lower forearm since, presumably, observation would lead a provider to determine whether the left or right arm is fractured.

Monitoring Specificity
While some leeway is expected during the initial introduction of ICD-10 codes, payers will expect increased specificity in coding as providers gain more facility with ICD-10. Organizations should measure this in two ways.

The first way is to review provider code input for codes lacking sufficient specificity in terms of characters. Feedback on codes that lack specificity as well as tracking the feedback for individual or group patterns can allow organizations to develop both targeted feedback and general education on proper coding.

The second way is to measure providers’ use of unspecified codes. As mentioned above, there are appropriate uses for unspecified codes, but excessive use should point to problematic coding. The process starts by obtaining a list of ICD-10 codes.(4) The list is then indexed for the word unspecified. The index is then compared to codes extracted from billing software. Finally, the extracted comparison is used to develop a scoring method. This can be expressed in a number of ways but the ratio of unspecified codes to total codes or the ratio of unspecified codes to other codes both provide valuable insight. These measurements can be tracked longitudinally and compared among providers. Outliers can be targeted for additional education while the aggregate information can be used for more general education.

ICD-10 represents a paradigm shift in encounter code selection. The increased number of codes as well as the attention to laterality requires increased specificity in code selection. Particular attention should be paid both to appropriate specificity and the use of unspecified codes in encounters. Ensuring proper specificity as well as appropriate use of unspecified codes decreases the chances of claim rejection. Finally, using feedback from specificity reviews as well as tracking the use of unspecified codes will allow an organization to anticipate coding problems and address them.


1. Carpenter DL. Introducing ICD-10. Mt. Prospect, IL: Society of Critical Care Medicine. Accessed April 20, 2016.
2. Centers for Medicare & Medicaid Services. Road to 10: The Small Physician Practice’s Route to ICD-10. Washington, DC: Centers for Medicare & Medicaid Services; 2015. Accessed April 20, 2016.
3. AAPC. 2016 ICD-10-CM Code Book. AAPC; 2016.
4. Centers for Medicare & Medicaid Services. 2015 ICD-10 CM and GEMs. Washington, DC: Centers for Medicare & Medicaid Services; 2014. Accessed April 20, 2016.