- Healthcare organizations are constantly producing and collecting medical data to help treat patients more effectively, and are also using the data for analytics purposes such as population health. Organizations need a computer assisted coding system (CACS) to ensure data is consistent throughout the organization to successfully sort and code that data.
CACS uses natural language processing to pick out important terms and phrases from ICD-9 CM and ICD-10 CM, as well as the AMA’s Current Procedural Terminology. These code sets are consistently used throughout organizations to record different diagnoses and procedures.
The system also provides context for clinician notes to distinguish between different uses. For example, the CACS can determine the difference between a diagnosis of heart disease as opposed to a family history of heart disease. This adds structure to potentially unstructured documentation data so it can be made actionable.
Traditionally this coding was done manually by coders, but the influx of data over the past several years has driven the need to automate this process. While coders are still needed for more complex projects and troubleshooting, the majority of day to day coding is performed by CACS, according to an official HIMSS blog post.
“Since its early beginnings in outpatient specialty areas, CAC systems have been adopted to improve medical coding workflows, increase medical coding accuracy, and balance medical coding resources to focus on more volume and complex cases,” wrote HIMSS contributor Deborah Kohn. “The recent and significant increase in the adoption of CAC systems in inpatient environments is linked to the same, compelling, justification benefits.”
CACS make traditional data coding simpler, but organizations also need to make sure that the tool will fit into their overall health IT infrastructure strategy so it can help address new and upcoming challenges. Kohn advised that automatic medical code generation or suggestions and automatic data abstraction are two key areas of focus to address future challenges.
Automatic medical code generation will generate suggested medical codes based on the information the clinician enters so the clinician can view all options associated with certain symptoms or situations. Automatic data abstraction will automatically extract coded data elements for user-defined purposes such as reporting or clinical registries.
“CAC systems use either structured data input or NLP engines, or both,” said Kohn. “NLP engines use artificial intelligence to identify concepts in the unstructured text data and to associate medical codes from controlled vocabularies to relevant phrases in the text. NLP engines must be able to interpret and combine concepts in terms of morphology, syntax, semantics, and real-world knowledge.”
Some organizations have evaluated the best way to approach integrating CACS into their infrastructure and are developing a strategy to ensure providers are using their automated and manual coding resources to the best of their ability.
An AHIMA study conducted at Cleveland Clinic found that pairing CACS with manual coders does reduce coding time, reporting that the combination increased coder productivity by more than 20 percent. However, the study also revealed that CACS paired with manual coders does not increase accuracy. Cleveland Clinic was able to increase its productivity without compromising accuracy, but implementing just the CACS without a manual coder resulted in a lower recall and precision rate.
The AHMIA study concluded that CACS needs to be “tuned” over time.
“The study tested the precision and recall performance for the CAC alone at implementation and then six months later,” the report authors concluded. “Since the NLP engine learns-known as tuning-over time, the CAC recall rate improved for coding both diagnoses and procedures as a result.”
“Going forward, Cleveland Clinic could run 25 cases through the CAC at regular intervals to evaluate improvements in precision and recall over time.”
Organizations can’t implement a CACS and expect it to seamlessly integrate with existing IT infrastructure or perform with perfect accuracy. While the technology automates certain processes and makes it easier to comb through more data, manual coders are still needed to teach the CACS and use it to enhance their performance.
“CAC systems, like all software solutions, are complex information systems requiring a substantial investment in time, dollars, and resources,” Kohn concluded. “Therefore, it is essential for any healthcare organization to develop a strong knowledge-base of key CAC solution features and functions. Only after obtaining such knowledge will healthcare organizations be able to contribute to the newer challenges of today’s healthcare industry.”