ICD Code Crosswalks: No Substitute for ICD-10 Compliance
ICD Code Crosswalks: No Substitutefor ICD-10 ComplianceWhile crosswalk solutions may appear compelling, their usefulness issignificantly limited by implementation complexity and expense, aswell as reliance on manual labor.Executive SummaryWhen it comes to converting claims betweenICD-9 and ICD-10 code sets, ICD crosswalks areoften presented as a sensible, even long-term,solution to achieving ICD-10 compliance. However,a crosswalk-based approach — which involves theautomated application of maps to convert claimsbetween ICD-9 and ICD-10 — is fraught with com-plexity and added expense and can even increasethe need for manual review of data and claims.The challenge is that mapping ICD crosswalks isinexact by definition, as the mapping betweenICD-9 and ICD-10 codes is nondeterministic.That is, while applying a forward map to anICD-9 code generates an ICD-10 code, applying abackward map to the resulting ICD-10 code doesnot always result in the original ICD-9 code. Infact, the American Medical Association (AMA)indicates that one-to-one matches exist for fewerthan 25% of ICD-9 to ICD-10 codes.1These typesof mismatches negatively impact the business,resulting in reduced auto-adjudication rates, lossof information and increased customer servicecall volumes, as well as other potential issues.Crosswalking in either direction, then, is not asubstitute for fully implementing ICD-10 andmust be seen as a transitional option of lastresort. Where crosswalks seem unavoidable,careful attention must be paid to the true costs ofchoosing this solution over migration to full ICD-10compliance. These costs, plus related IT and oper-ational issues, are highlighted in the commoncrosswalk implementations discussed below.The Crosswalk ConundrumCrosswalks are the automated application ofmaps to convert between two code sets — in thiscase, ICD-9 and ICD-10. There are two types ofICD crosswalks that can be used as a transitionalapproach to achieving ICD-10 compliance:• Backward crosswalk: This type is used toconvert ICD-10-based data to ICD-9-based data,to support processing or analytic and reportingneeds.• Forward crosswalk: This type is used to convertincoming ICD-9-based claims received on orafter the compliance date of Oct. 1, 2014, toICD-10-based claims so they can be processednatively as ICD-10 claims in systems that arefully ICD-10 compliant.The challenge that healthcare organizations facewith either type of map is that ICD-9 and ICD-10• Cognizant 20-20 Insightscognizant 20-20 insights | may 2013
2codes do not always map one-to-one exactly.The AMA findings indicate that there is a greatdeal of ambiguity to address when attempting tomatch ICD-10 to ICD-9 and vice versa (see Figure1). Multiple options arise when creating the mapson which crosswalks are based, resulting in opera-tional issues for all of the most common transi-tional crosswalk solutions.2Backward CrosswalkBackward crosswalks (ICD-10 to ICD-9) aregenerally performed to address one of two needs:• System insulation: In this case, a core systemis scheduled to be decommissioned and cannotbe completely migrated to an ICD-10 compliantsystem before the ICD-10 compliance date. It isimportant to note that when ICD-10 codes arebackward-crosswalked to ICD-9 for processingby non-ICD-10 compliant systems, businesspartners and downstream applications — bothcustom and commercial-off-the-shelf (COTS) products —will still require the originalICD-10 codes because theywill be operating with ICD-10codes. An organization canrealistically insulate onlya very small part of its ITlandscape from full ICD-10compliance.• Longitudinal analytics and reporting: Inthis case, a backward crosswalk can supportanalytic and reporting needs that span theICD-10 compliance date for when ICD-9 codes(for claims with a date of service before Oct. 1,2014) need to be compared with ICD-10 codes(for claims with a date of service on or after Oct.1, 2014). Examples include case management,utilization management or research.Backward crosswalks can support systeminsulation, longitudinal analytics and reportingfor short-term pain relief, but these also introducesignificant disadvantages.System Insulation Crosswalk IssuesWhen organizations need to perform systeminsulation, they use crosswalks to replace theincoming ICD-10 codes with ICD-9 codes, basedon a backward map. The insulated system com-municates with downstream applications andtrading partners with the original ICD-10 codes.The ICD-9 codes used in the insulated systemmust be replaced with the original (not mapped)ICD-10 codes.Specifically, a crosswalk intercepts incoming(upstream) claims before they reach the systemrather than making the system ICD-10 compliant.An adapter intercepts downstream interfacescontaining ICD-9 codes and replaces (not aforward crosswalk) the crosswalked ICD-9 claimswith the original ICD-10 code.The advantage of this approach is that it tem-porarily eliminates the need to remediate thesystems identified for insulation. However, it alsoposes two disadvantages:• An external viewer application is required toenable internal and external entities — includingcustomer service and appeals/grievances users— to view the original ICD-10 and ICD-9 codesside-by-side. Extensive supportive functionsin this external view are often required. Forexample, a customer service rep will see theoriginal ICD-9 code and the mapped ICD-10code. Depending on who is making the inquiry,callers will reference either code (originalICD-9 or mapped ICD-10). The service rep willneed to be able to work with both codes andunderstand the map to resolve issues.cognizant 20-20 insightsFigure 1ICD-9 to ICD-10 Percentage of MatchesMapping Categories ICD-10 to ICD-9 ICD-9 to ICD-10No Match 1.2% 3.0%1-to1 exact match 5.0% 24.2%1-to-1 approximate match with one choice 82.6% 49.1%1-to-1 approximate match with multiple choices 4.3% 18.7%1-to-many match with one scenario 6.6% 2.1%1-to-many match with multiple scenarios 0.2% 2.9%An organizationcan realisticallyinsulate only a verysmall part of its ITlandscape from fullICD-10 compliance.
• The ability to edit the ICD code (e.g., downcodingor bundling) is more complex to implementbecause of the need to simultaneously edit theoriginal ICD-10 code and the crosswalked ICD-9code, since each have a different system ofrecord.Figure 2 details the complexity of an effectivearchitecture for crosswalking a noncompliantclaims processing system.1. Incoming ICD-10 coded claims are crosswalkedto ICD-9 using either custom software or aCOTS tool.2. The original ICD codes for the crosswalkedincoming claims are persisted to a data store,to be used to feed systems downstream fromthe core claims system. The data store willneed to persist the following:>> Incoming ICD-10 codes.>> Unique identifiers for the claim because theclaim ID is typically generated inside theclaims processing system.3. A custom adapter intercepts the claimsprocessing system downstream, accessinginterfaces containing crosswalked ICD-9 codes.It then replaces the crosswalked ICD-9 codeswith the original ICD-10 code before passing itto the downstream recipient, such as an appli-cation or business partner, or before sendingdata back to the claims processing system (inthe case of downstream inbound interfaces). Insome cases, such as data stores for reportingpurposes that require both original and cross-walked claims to be accessible, the ICD-10 codecan be added to the interface; this enables boththe crosswalked ICD-9 code and original ICD-10code to be passed to the downstream recipient.The adapter has two distinct parts:>> A shell that hosts all of the interface-specif-ic adapters and provides required supportand common functionality.>> A collection of adapters, each supportingone of the ICD interfaces and hosted withinthe adapter shell.4. Modifications to the claims processing system(configuration and/or code), downstream appli-cations and trading partners (as needed) todirect ICD-affected interfaces to the adapterrather than the ultimate destination.3cognizant 20-20 insightsFigure 2Backward Crosswalk for System Insulation: A Conceptual Architecture15432Internal core systemsICD-9-basedadjudication rulesCore claims systemand surround code(ICD-9 claimsprocessing)RedirectICD-impactedinterfacesto adapterApplication supporting:• Investigative services,including providerservice and A&G• External entitiesviewing claimsDownstreamapplicationsand tradingpartnersCore claimsdata (ICD-9)Claim datafor viewerRetrievedICD-10 codewith versionOriginal claimICD-10 codesand uniqueidentifiersOriginal ICD-10 codesICD-9 codesNo ICD codesAdapterICD-9ICD-9ICD-10ICD-10EDIICD-10claimCrosswalk
cognizant 20-20 insights 45. An application external to the claims processingsystem is required to support:>> Investigative needs of stakeholders, bydisplaying both original ICD-10 codes andcrosswalked ICD-9 codes to support pro-vider service representatives and appeals/grievances users. The application will queryboth the data store persisting the incomingICD-10 (item 2) and the claims processingsystem data store to display both the origi-nal and crosswalked codes to meet investi-gative efforts and other needs.>> External entities viewing claims processingsystem claims that require access to origi-nal incoming ICD-10 code.>> Requirements for synchronizing edits toICD-9 and ICD-10 codes, if relying on indi-viduals’ editing codes is not adequate. Anextension to the applications is required toachieve this function.Longitudinal Analytics and Reporting CrosswalkChallengesHealthcare entities frequently have longitudinalanalytic and reporting needs in areas such ascase management, utilization management andresearch that draw on ICD codes accumulatedover a period of time. Filling these needs is chal-lenging, particularly when the time period extendsfrom ICD-9 usage to beyond the ICD-10 compliancedate because these analytics and reports willhave both ICD-9 and ICD-10 codes that cannotbe directly compared. A possible solution to thelongitudinal problem is to implement a crosswalkthat maps ICD codes in the subset of data neededfor longitudinal reporting to a common basis—initially ICD-9 and subsequently ICD-10, once thevolume of ICD-10 data is sufficient. This ICD-stan-dardized subset of data is stored in a data martand used as the basis for longitudinal analytic andreporting needs.The issues with inexact matching of codes onICD maps limit the usefulness of this approach. Acareful, case-by-case assessment should be madeto determine whether the crosswalk approachwill meet the specific business needs. When acrosswalk is deployed, it is very possible thatadditional effort will be required for a lengthyperiod of time to manually interpret reportscontaining a mix of ICD-9 and ICD-10 data. For lon-gitudinal reporting, this crosswalk does not alwayswork; in fact, it may be necessary to manuallycontend with the mix of ICD-9 and ICD-10 codesfor longitudinal reporting in these cases.Figure 3 shows a conceptual architecture of acrosswalk supporting longitudinal analytics andreporting spanning the ICD-10 compliance date.The elements of the solution are highlighted andnumbered to convey workflow sequence.1. ICD-10 based data is selected from the datawarehouse (or other data stores) and cross-walked to ICD-9 using either custom softwareor a COTS tool.Figure 3Backward Crosswalk to Support Longitudinal Analyticsand Reporting: A Conceptual ArchitectureICD-10-based(date of serviceon or after10/1/2014)ICD-9-based(date of serviceprior to10/1/2014)Data martstandardized onICD-9 to supportlongitudinalneedsICD-9-basedlongitudinal analyticsand reportingBackwardCrosswalkICD-10ICD-9ICD-9ICD-9
5cognizant 20-20 insights2. ICD-9 based data is loaded into a data mart thatsupports longitudinal analytic and reportingneeds from the crosswalk and directly from thedata warehouse.3. Analytics and reports are developed to use theICD-9 standardized data mart to meet longitu-dinal analytic and reporting needs.While this solution shows the backward crosswalkinitially used after the compliance date, atransition to an ICD-10-based data mart and ICD-10-based analytics and reports would be madewhen the volume of native ICD-10 data is signifi-cant compared with ICD-9 history.Forward Crosswalk ConcernsForward crosswalks (ICD-9 to ICD-10) are generallycreated to address two needs:• Noncompliant partners: The organizationwants to accept and process transactions frompartners that are noncompliant as of Oct. 1,2014.• Longitudinal analytics and reporting: Theorganization needs to support analytic andreporting activities that span the ICD-10compliance date, where ICD-9 codes (forclaims with a date of service before Oct. 1,2014) need to be compared with ICD-10 codes(for claims with a date of service on or afterOct. 1, 2014) This is typical in such areas ascase management, utilization management orresearch.The following subsection explores the challengesof using forward crosswalks to manage noncom-pliant claims received after the ICD-10 deadline.As describe earlier, the longitudinal analyticsand reporting approach with a forward crosswalkis similar to using a backward crosswalk, with atransition made once the volume of ICD-10-baseddata is significant compared with the volume ofICD-9-based data.Noncompliant PartnersOrganizations are developing strategies fordealing with noncompliant partners, primarilyproviders that will not be ICD-10 compliant onOct. 1, 2014. Noncompliance is a particular issuewith smaller providers in rural areas because ofthe cost of compliance and lack of resources. Fourbroad approaches exist for dealing with noncom-pliant providers:• Deny the claim: This is often not an optionbecause of the need to preserve relations withproviders and the potential impact to networkscaused by losing providers.• Provider outreach: This involves workingwith providers to help them become ICD-10compliant, using mechanisms such as superbillsto make compliance easier.• Third-party recoding: Organizations canidentify and partner with other organizationsthat can recode provider claims in ICD-10 ontheir behalf. Leveraging volume across all non-compliant providers will help reduce costs.• Crosswalk incoming claims from ICD-9 toICD-10: Forward crosswalks are used to upcodeICD-9-based claims to ICD-10.Two crosswalk approaches are explored. Thefirst scenario shown in the conceptual architec-ture depicted in Figure 4 crosswalks the claimssubmitted by noncompliant providers as soon asthe payer receives them.1. Noncompliant provider submits a claim on orafter Oct. 1, 2014, using ICD-9 codes.2. Claims are crosswalked, adding ICD-10 codesbased on the ICD-9 codes.3. The core claims system processes the upcodedclaim as if it were a native ICD-10 claim.4. Downstream systems and partners work withthe upcoded ICD-10 claim in the same way asany other ICD-10 based claim.Figure 4ForwardCrosswalkCore claims systemand surround code(ICD-10 claimsprocessing)ICD-10 ICD-10ICD-9ICD-9claim ICD-912 34Pre-Processing Forward Crosswalk to Support Noncompliant Providers
cognizant 20-20 insights 6In addition to the complexity and cost of imple-menting this solution, the following significantissues limit its application.• The inexact nature of the map might be suitablefor claims adjudication; however, it is likely tobe problematic for some internal and externaldownstream uses.• Legal implications may exist because thisapproach essentially changes the claim.• Payers face the possibility of future litigation ifproviders feel at some point that the payer-con-trolled map is used for the financial benefit ofthe payer. Even if the litigation is groundless, itwill consume resources and generate negativepublicity.• Policies, rules and contracts with ICD codeswill have to be maintained and configuredwithin systems for both ICD-9 and ICD-10 afterthe compliance date. By forgoing a forwardcrosswalk and instead using provider outreach,superbills and third-party re-coders to addressnoncompliant provider claims, the ICD-9 basedpolicies, rules and contracts are no longerneeded beyond the compliance date becauseall ICD-9-based claims will have a date ofservice before then.The scenario shown in Figure 5 depicts noncom-pliant claims processed as ICD-9 claims and thencrosswalks them, post-processing, before sendingthem to downstream systems and businesspartners.1. Noncompliant provider submits a claim on orafter Oct. 1, 2014, using ICD-9 codes.2. The core claims system processes the claimnatively, using ICD-9 codes.3. Processed claims are crosswalked, addingICD-10 codes based on the ICD-9 codes.4. Downstream systems and partners work withthe upcoded ICD-10 claim the same way as anyother ICD-10 based claim.In addition to the issues mentioned for the pre-processing crosswalk approach (as shown inFigure 4), this alternative has two additionalchallenges:• Crosswalk implementation becomes morecomplex because numerous downstreaminterfaces with systems and partners thatare ICD-10-compatible need to be addressed.Multiple crosswalks might need to be imple-mented.• Claims adjustment from downstream systems,stakeholders and partners becomes increas-ingly complex. The adjustment will be made inICD-10, but it must then be applied to a claimprocessed natively using ICD-9.Reconsidering the Value of CrosswalksForward and backward ICD crosswalks are not asubstitute for fully implementing ICD-10. Whilecrosswalk solutions may appear compelling, theirtrue usefulness is significantly limited by themany problems with implementation complexityand expense. Further, crosswalk solutions all toooften involve additional manual labor.Organizations need to carefully analyze thespecific situation and evaluate the alternativesto determine whether a crosswalk solution isthe optimal and effective way to meet businessobjectives. Often, implementing native ICD-10compliance is the far more effective solution,resulting in cleaner, more comprehensive dataand streamlined processes for managing thetransition to ICD-10.Figure 5ForwardCrosswalkCore claims systemand surround code(ICD-9 claimsprocessing)ICD-9 ICD-10ICD-9ICD-9claim ICD-912 34Post-Processing Forward Crosswalk to Support Noncompliant Providers