(1023) NLP- A Tool to Address Documentation Gaps and Improve Revenue

To determine if the use of Natural Language Processing (NLP), an Artificial Intelligence (AI) platform, aids in detection of poor documentation of anemia in pregnancy and in turn, potentially improve hospital reimbursement.

Study Design:

A retrospective cohort review of 14,418 records from January 1^st 2019 to December 31^st 2020.  By use of Machine Learning (ML), we used an NLP algorithm (SPAI ® Platform developed by Gynisus Inc.) to identify cases of anemia in pregnancy, set by predefined criteria. We tracked those that were lacking documentation in structured ‘problem-lists’. Since absence of structured documentation led to suboptimal capturing of the diagnosis in the hospital’s billing platform, we then translated our findings to potential loss of revenue by use of common diagnosis group related (DRG).

Results:

Of the 14,418 charts reviewed, 4,499 (31.2%) patients were found to be anemic by definition of Hemoglobin < 11 g/dL and/or Hematocrit < 33%. Of the 4,499 patients with anemia in pregnancy, the NLP based algorithm was able to capture 3,685 with documentation gaps, of which 188 had structured field diagnosis present that was entered late, and 3,497 that were completely lacking a structured field diagnosis. Only 626 patients had documentation of anemia in structured fields entered in a timely fashion. A manual validation sample of patients with documentation gaps was performed. Figure 1 presents the average timeline of those with late documentation. Figure 2 summarizes NLP-based findings. We found a potential two-year loss range (DRG dependent) of approximately 30 million dollars to 65 million dollars.

Conclusion:

NLP and ML-based systems have the ability to process free text and clinical data and translate it to usable information. NLP can be used to optimize revenue cycle management by uncovering financial loss due to documentation gaps. With appropriate algorithm definitions, NLP can serve as an adjunct tool to improving documentation and in turn improve hospital reimbursement.