Mapping Relationships Do Not Predict Impact of ICD-10-CM on Patient Safety Indicator Reporting
In a new research study published in the Journal of the American Medical Informatics Association (“Challenges and remediation for Patient Safety Indicators in the transition to ICD-10-CM”), the researchers claim to have found that the transition to ICD-10-CM will lead to underreporting or overreporting of some Agency for Healthcare Research and Quality (AHRQ) Patient Safety Indicators (PSIs). However, no conclusions regarding the impact of ICD-10-CM on PSIs should be made as a result of this study, as the approach used in this study does not accurately predict the impact of ICD-10-CM on patient safety indicator data.
The researchers’ sweeping claim about the future is based on a mapping of a mapping. The study takes the General Equivalence Mappings (GEMs) and maps them to one of three “complexity types” (“convoluted” is one) defined by the researchers using criteria of their own creation. These complexity types are then used as indicators of whether a PSI will be accurately reported or not. Interpretations that pile a mapping on top of a mapping get us farther from the real world business of coding medical records and recording patient safety issues using PSIs, not closer.
The researchers fail to acknowledge that code assignment for PSI reporting is not accomplished by mapping ICD-9-CM codes to ICD-10-CM codes. Rather, ICD-10-CM codes will be assigned directly by human review of medical record documentation and determination of the most appropriate code(s) based on coding rules/guidelines and knowledge of the ICD-10-CM coding system gained through comprehensive coding education. A medical record for an encounter after the ICD-10-CM transition will not be coded from the GEMs any more than coders today code records in ICD-9-CM by looking at a mapping from ICD-8.
Here is how a record is coded in the real world, regardless whether the codes used are ICD-9-CM or ICD-10-CM:
- The coder reviews the medical record for a specific encounter (hospital stay, doctor visit), looking for evidence that a condition or procedure should be coded according to the guidelines for that sector of the healthcare delivery system (e.g., hospital inpatient, outpatient radiology, physician office).
- The coder looks up the key terminology in a book, electronic version of ICD-9-CM/ICD-10-CM, or specialized coding software and follows the instructional notes or guidelines for that type of disease (diabetes) or type of procedure (biopsy) to find the code that most closely matches the information in the medical record.
- The coder records the code, and it is saved as part of the patient’s permanent medical record for that encounter.
- Once the record has been completely coded it is submitted by the healthcare provider to the payer and also stored in the provider’s data warehouse. The process is exactly the same for ICD-9-CM as it is for ICD-10-CM. There are no mappings involved.
The developers of the AHRQ PSIs did not rely solely on the GEMs for the conversion of the PSIs to ICD-10-CM. They appropriately used the GEMs as a starting point for identifying potential candidate codes for inclusion in the PSIs and then refined the codes to meet the intent of the respective PSI. The PSI developers’ goal was to identify those codes in the applicable code set (whether ICD-9-CM or ICD-10-CM) that will best capture the patients that should be included in a given PSI—patients who have a pressure ulcer, for example. The PSI developers don’t want to know whether the relationship between the ICD-9 pressure ulcer code and the ICD-10 pressure ulcer code is ‘convoluted’ or not. They want to know if a patient has a pressure ulcer. As it happens, there are excellent ICD-10 codes for capturing patients with pressure ulcers, and also excellent (though slightly less detailed) ICD-9 codes for specifying pressure ulcers. Does this mean that because there are more detailed ICD-10 codes, more pressure ulcer patients will show up in the ICD-10 data? Of course not. If one patient has a pressure ulcer of the left heel and another patient has a pressure ulcer of the right heel, both medical records are coded using the same, less specific, code in ICD-9 –CM, whereas, in ICD-10-CM, there are separate codes to identify left and right. Thus, in ICD-10-CM, there is more detail about the pressure ulcer than in ICD-9-CM, but regardless of whether the encounter is coded in ICD-9-CM or ICD-10-CM, both of these patients would be classified as having a pressure ulcer of the heel.
Because not all ICD-10-CM codes included in a mapping from an ICD-9-CM code necessarily meet the intent of a PSI, not all mapped codes are included in the ICD-10 PSIs, despite the researchers’ suggestion to the contrary. For example, the article states that postoperative respiratory failure (PSI-11) could be underreported with the transition to ICD-10-CM because of forward GEMs mapping to both procedure and non-procedure ICD-10-CM codes. However, AHRQ only included the procedure-related ICD-10-CM codes in PSI 11, so this is not a valid issue (see the proposed changes with the ICD–10–CM/PCS conversion of AHRQ Quality Indicators™ at http://www.qualityindicators.ahrq.gov/icd10/).
Some of the researchers’ conclusions clearly reflect a lack of understanding of the ICD-10-CM code set. For example, it is stated that the failure to include ICD-10-CM codes for “subsequent encounter” in the PSIs will result in underreporting. However, this is not a valid conclusion, as the ICD-10-CM Official Guidelines for Coding and Reporting state that “initial encounter” should continue to be coded as long as the patient is receiving active treatment for the condition, which is comparable to how the related ICD-9-CM codes are used today.
As CMS has often stated, in public meetings and in documents posted on their web site, the GEMs are not to be used for coding medical records. The GEMs were created as a tool for the conversion of data from ICD-9-CM to ICD-10-CM and ICD-10-PCS and vice versa. They serve as a temporary transition tool and are essentially “ICD-10 training wheels.” As organizations begin to dual code medical records (in both ICD-9-CM and ICD-10-CM/PCS) and thus have access to substantial amounts of real ICD-10-CM/PCS coded data, they are relying less on the GEMs. Actual coded data is a much better source of information than mapped data. So any results that use a highly abstracted superstructure built on top of the GEMs in order to make predictions about coding behavior should be taken with a very large grain of salt. This study may be an interesting informatics study, but the approach used in this study does not accurately predict the impact of ICD-10-CM on patient safety indicator data.
Finally, the authors of this research study miss the primary point of why it is essential the US move to ICD-10. Healthcare data continues to deteriorate as long as we continue to use a completely antiquated coding system. Numerous benefits of the transition to ICD-10 have been cited, including the provision of better information for:
- Measurement of quality of care, outcomes, patient safety, disease management
- Support for new healthcare delivery innovations — medical homes, value-based purchasing, etc.
- Improved public health through real-time monitoring and identification of potential health threats.