A few weeks ago on October 1st, the switch from ICD-9 diagnosis and procedure codes to ICD-10 became mandatory for all HIPAA-covered entities. ICD-10 is already used by most (if not all) other developed countries, and has been for years. The number of available codes under ICD-10 will increase by nearly 5-fold, and will provide more precision and detail regarding a patient’s condition and the health encounter. For example, diagnosis codes indicating an injury will typically provide more information about the location of the injury, its cause, as well as whether the health encounter is the initial encounter to treat the injury or a subsequent encounter. Most agree that ICD-10 will allow for a more accurate representation of patients’ conditions, and likely will be accompanied by fewer coding errors. Some are concerned about the impact the new coding system could have on quality assessments (and hence reimbursement), as well as the cost associated with its implementation and continued use, but almost everyone agrees the change is necessary.
For researchers who utilize claims-based administrative data, the switch to ICD-10 will present different types of opportunities and challenges than it does for healthcare providers. It is widely assumed that the increased detail and accuracy provided by ICD-10 will translate into more accurate and precise research, which is obviously good for researchers. At the same time, along with a series of validation studies that will be needed to demonstrate the sensitivity and specificity of the new codes, good crosswalks will also be needed to allow for consistency in methods across research that utilized ICD-9 to that which will utilize ICD-10.
Some crosswalks for individual codes have already been developed, including by CMS, the Agency for Healthcare Research & Quality, (AHRQ) and the American Academy of Professional Coders, (AAPC). Some of these have shown that while many codes have a direct mapping between ICD-9 and ICD-10, other codes do not have a clear mapping, and still others have no mapping at all. There may be no remedy for this. But, what is perhaps more important than a crosswalk for individual codes is to develop a crosswalk for estimates (of incidence, prevalence, rates of outcomes, etc.) produced by the individual codes, especially given that many diseases and health events are typically defined by a set or group of codes instead of by the presence of a single code.
It may not be possible to identify methods whereby rate estimates produced using ICD-10 codes are identical to those produced by ICD-9 codes, although the AHRQ Healthcare Cost an Utilization Project has produced a Clinical Classifications Software for ICD-10. However, perhaps duplicating rates should not be the goal: after all, maybe the increased detail allowed by ICD-10 codes will produce estimates that are different - but more accurate - than those that were possible using ICD-9 codes. However, it will be crucial that we understand the differences so that we can make valid comparisons between rates produced by ICD-9 codes versus those produced by ICD-10 codes, and can identify how much of an observed difference over time is due to the new codes, and how much is due to actual changes in the underlying health status of the population being studied.
For obvious reasons, longitudinal studies that span October 2015 will need to face this issue head-on, since these studies will track rates from before October 2015 (when ICD-9 was used) to after the implementation of ICD-10 codes. But cross-sectional analyses that utilize only data from October 2015 and later will also need to address the ICD-9 to ICD-10 switch. When these studies compare their results to those of previously published literature, much of those results (at least initially) will be based on ICD-9 codes, and therefore will require that the authors be able to accurately assess how much the differences they see may be due to the switch in coding methods.
For most researchers, the typical time-lag associated with claims-based data means that we will not have to deal with these issues directly for at least a year or more. However, now is the time when we should be thinking about how we will address these challenges so that we can take full advantage of the benefits provided to us by ICD-10. Additionally, as research utilizing ICD-10 codes begins to emerge here in the US, it will be important to keep these issues in mind when assessing and critiquing the quality of these studies.
Craig Solid, PhD, is owner and principal of Solid Research Group, LLC, an organization that helps shape healthcare delivery and national healthcare policy by quantifying and disseminating results of health research and quality improvement initiatives. For over 15 years, Dr. Solid has helped healthcare organizations discover effective treatments for disease, improve the quality of care delivery, and increase efficiency and effectiveness. Dr. Solid is a member of the American Medical Writers Association.