The mandated use of SNOMED seemed to sneak up on healthcare providers, ninja-style. Like a ninja, it has the opportunity to silently deliver deadly blows: confusing care coordination efforts, confusing patients, and contributing to adverse risk scoring and underwriting analysis with private payers.
While providers and health IT professionals are accommodating SNOMED into clinical workflows and converting existing relevant data points to SNOMED to comply with CMS mandates, no one seems to be talking about the potential impact on patient care.
The requirement of SNOMED began innocuously enough. To qualify for Meaningful Use Stage 1 Core measures, CMS declared that Eligible Hospitals, Critical Access Hospitals, and Eligible Providers would record patient “problems” as discrete data rather than narrative text (typically, acronyms).
The Medicare and Medicaid EHR Incentive Programs do not specify the use of 1CD-9 or SNOMED-CT® in meeting the measure for this objective. However, the Office of the National Coordinator for Health Information Technology (ONC) has adopted ICD-9 or SNOMED-CT® for the entry of structured data for this measure and made this a requirement for EHR technology to be certified. Therefore, EPs will need to maintainan up-to-date problem list of current and active diagnoses using ICD-9 or SNOMED-CT® as a basis for the entry of structured data into certified EHR technology in order to meet the measure for this objective.
So, in 2013, providers were told there are two standards for recording structured problem data. They can choose ICD-9 or SNOMED-CT®. Since the objective states “list of current and active diagnoses” and they use ICD-9 for diagnose coding, it should be fine to use ICD-9, right? What’s the difference between an ICD code and a SNOMED code for the same ailment?
In Achieve (Meaningful Use Stage 2) Compliance with SNOMED-CT, Brian Levy gave an excellent synopsis:
Problem lists have been around a long time. Historically, and in many outpatient offices and clinics today, physicians have maintained paper problem lists in the front of patient charts that are updated during each encounter. The problem list basically acts as a running record of the major or chronic conditions suffered by the patient.
By contrast, the diagnoses coded within a practice represent the conditions that prompted the services rendered during a visit. In other words, the diagnosis codes describe the ailments that justify billing for the procedures performed. After a visit, a physician typically marks the appropriate ICD-9 codes on a superbill and sends it to the billing staff.
Geraldine Wade of Clinical Informatics Consulting used this simple SNOMED vs. ICD-9 visualization as part of her presentation to Hong Kong’s eHealth Record Office:
It seems simple enough: providers should use SNOMED when recording clinical symptom and diagnosis data, which doesn’t impact billing and reimbursement processes. Given time and adequate clinician training on the differences between SNOMED and ICD codes required for the same diagnosis, that process adjustment could be made.
Fast forward to 2014, the CMS issued clarification that the patient problem lists must be recorded only in SNOMED. Also, any patient diagnosis data already recorded in the problem list must resolve to SNOMED with all active problems, whether newly introduced or historic chronic conditions. Eden Ware described the collective provider reaction to this news in SNOMED: What It Is And Why It Was Added To Stage 2 Meaningful Use:
I can hear my provider colleagues screaming now. ‘No! We are already swamped trying to figure out ICD-10! We just want to care for our patients!’” She goes on to assuage their fears, with visions of IT automation to the rescue:
The good news is that the industry hears you, and products are available in the healthcare IT market to facilitate the translation of your problem lists into reportable, standardized SNOMED-CT codes. It will be important to ask your EMR vendors how they are handling this Meaningful Use Stage 2 requirement. Many vendors are utilizing “maps” between ICD-10-CM and SNOMED-CT to ensure this goal can be met. The mapping from the ICD-10 code to SNOMED-CT occurs behind the scenes, and is easily retrievable. However, the provider’s time is not affected and the goal for meeting this core requirement of Meaningful Use Stage 2 is ensured.
Presto! The SNOMED conversion problem is solved. There are a number of mapping tools available to address the linkage between ICD-9 and SNOMED, including the crosswalk provided by the National Institute of Health’s National Library of Medicine, which was created specifically to address the translation required for patient problem lists. EHR vendors are also building conversion features into their products, grouping patients with the same ICD-9 diagnosis for instant translation to SNOMED, where a single click will instantly convert, for example, 40 patients’ ICD-9 code for CAD to the parent SNOMED code for coronary artherosclerosis.
One click, 40 patients’ records altered, all future C-CDA clinical documents compliant with Meaningful Use criteria.
But wait… We’re talking about a patient’s clinical diagnosis data.
IT geeks, in conjunction with EMR vendors and systems integrators, are going to be responsible for making an accurate determination of which ICD-9 diagnosis code maps to a SNOMED code? Is there any risk that these conversions might fail?
NLM blithely discloses the difficulty of achieving a perfect match (emphasis mine):
The Map tries to identify as many one-to-one maps as possible, however, due to the differences between the two coding systems, one-to-one maps cannot be found for some ICD-9-CM codes. This difference is usually due to one of two reasons. Firstly, in ICD-9-CM, some codes are “catch-all” codes that encompass heterogeneous diseases or conditions (e.g. pneumonia due to other specified bacteria). These codes, commonly known as “NEC codes” (not elsewhere classified codes), will not have one-to-one maps because of their nature. Secondly, since SNOMED CT is more granular than ICD-9-CM in most disease areas, some ICD-9-CM diseases or conditions are further refined as more specific concepts in SNOMED CT. For such cases, it is not possible to map to a more specific SNOMED CT concept without the input of additional information.
SNOMED is more granular than ICD-9 in most disease areas—not few, not some, but most.
Osteoarthrosis, the most frequently-occurring “unmapped” problem codes in the above image, has 20+ possible applicable SNOMED concepts related to the parent code 396275006 (Osteoarthrosis disorder), indicating more granular information such as site of osteoarthritis, or whether condition is chronic/endemic/degenerative, etc. Asthma has 25 possible derivations of SNOMED parent code 195967001 (Asthma disorder). Depression has close to 30 SNOMED codes describing parent code 35489007 (Depressive disorder), and the list goes on.
Here’s what the one-to-many mapping possibilities could mean to patients (including automated conversion tools):
The purpose of the SNOMED problem list is to inform all providers in the patient’s care continuum of any active or chronic conditions needing assessment and monitoring. With the single-click application of any given SNOMED code to an entire population of patients, it is highly likely that some, if not many, patients will be incorrectly assigned. It is unlikely that the patient will be clinically educated enough to identify, let alone explain, the difference to the network of providers participating in his or her care.
Is it reasonable to assume a patient suffering from severe asthma (370221004) would require a different care plan than a patient suffering from exercise-induced asthma (31387002)? How about the recently-discharged hospital patient who is suffering from asthma with irreversible airway obstruction (401000119107), referred to a primary care physician, whose Transition of Care problem list only states mild asthma (370218001)?
Sounds like a potential patient safety issue.
While the implications for care coordination are obvious and should be sobering, there are even more nefarious ramifications when you consider the EHR clinical data is also being shared with and consumed by insurance companies.
As more private payers are incorporating clinical document data into their member risk scoring and actuarial analytics, there are financial implications to the patient stemming from these automated (and, in some cases, arbitrary) mappings. The patient with mild depression (310495003) who is mistakenly coded as having major depressive disorder (370143000) may see medical insurance premiums increase, may experience a forced plan change with a reduced network of available doctors due to member attribution models applied based on clinical findings, and may have difficulty obtaining or renewing life insurance.
There are documented processes in place to rectify mistakes on consumer credit reports. This mandated SNOMED conversion introduces a new consumer protection need―how to address clinical data inaccuracies on a patient’s medical health record. A comprehensive process has yet to be introduced that would allow a patient to dispute a finding and have the information rectified at all points along the care continuum.
Is this what CMS had in mind when they mandated SNOMED for active and historical patient problems?
Welcome to the dark side of EHR interoperability.