Digital Data and Documentation
By Gerasimos Petratos, MD, MS, CEO, Hiteks Solutions Inc.
Now that there is digital data and documentation, the world of possibilities of using the data in clinical and administrative workflows is endless. Right?
Well, this is all dependent on the quality of the software which is used to produce the value-added information (aka analytics) that come from the raw clinical data and documentation.
As we all know, interpreting clinical data is difficult, so reading through the chart to understand the circumstances around the clinical care is often dependent on the documenting physician’s and nurse’s ability to describe their findings. So how can software be designed to aid in this process?
The key question to ask a software vendor is how their software is designed, not what it does now (although that is also important), but “what can it do given its underlying design?”
NLP Software is an good example because most vendors borrowed/licensed the technology from academia (legacy software), meaning it started as an established set of code that was able to accomplish the basic text parsing and assignment of clinical words and phrases to parts of speech. Then, they built on top of these basic parts the logic to match clinical nouns and other parts of speech to a standard terminology (i.e. ICD-10, SNOMED, RadLex, etc.).
The issue with using this design to help CDI is that although the legacy software can be made to recognize new terms and rules created with when to alert a user, the speed and configurability is a problem. Most systems are not fast enough for real-time processing and usually hard-code their logic, making it difficult to fit into multiple CDI workflows.
New Design for New Era
The timeliness of information seems to be paramount to a physician’s ability to respond consistently (high compliance). So the speed of a computer program producing the recommendations, either in fully automated fashion or CDI Specialist-validated, is paramount to getting the advice in front of the physician while they are still completing their documentation.
Further, the software cannot hinder the algorithms already being used by a manual process, otherwise it would take months to add in simple changes to mimic how CDI works, like when you want to begin adding “respiratory distress” to query for acute respiratory failure.
So the software must be “configurable”; for example, being able to add in local terms, abbreviations, section-based context, and time sequences of events since CDI workflow includes these needs. For instance, you should be able to ask for cases where there are Radiology reports with positive findings (e.g. lung atelectasis, opacification, etc.) to prompt for possible pneumonia and type.
Gerasimos Petratos, MD, MS is CEO, Hiteks Solutions Inc. His presentation: “Leveraging CDI as Domain Expertise in Automated Query Knowledge Engineering” will be given at the MaHIMA Western Mass Meeting on October 27, 2017. email@example.com