A study conducted by a group of Chinese researchers has shown that incorporating latent treatment pattern mining during clinical pathway execution can help clinicians better account for comorbidities likely to occur.
Although clinical pathways have become an important facet of health care, the occurrence of comorbidities presents a unique challenge. Understanding the effects of these comorbidities can be an important part of understanding what the best next step is for treatment, but this is a topic that has not been explored thoroughly.
In a study published in the Journal of Biomedical Informatics, researchers tested the use of a complex statistical model for extracting underlying treatment patterns, unveiling the latent associations between diagnostic labels and treatments, and calculating the contributions of comorbidities to treatment patterns.
Researchers developed the model by looking at clinical pathway execution data, which contain information about the specific therapies and treatments selected for particular patients in a hospital setting. To test the model, the researchers evaluated data from 12,120 patients who were treated for chest pain at the cardiology department of Chinese PLA General Hospital. Researchers identified an initial diagnosis and any set of comorbidities (hypertension, diabetes, etc). Further statistical analysis was then conducted to compare the outcomes of various treatments and determine any associations between these outcomes and the presence of comorbidities.
The model was used to successfully identify the top 8 comorbidities and top 50 treatment activities associated with the sample population. With this data, researchers were then able to use other models to determine common treatment patterns for patients with these comorbidities.
The authors, led by Zhengxing Huang, PhD (Zhejiang University, Zhejiang, and Xinjiang Medical University, Xinjiang, China), explained that their statistical model could be used to determine the contribution of comorbidities on the adoption of different treatments within a clinical pathway. In turn, this data could be used to better adapt clinical pathways to unique patient conditions so that the recommended treatment options are better suited to each patient, particularly those with comorbidities.
The authors concluded: “The discovered patterns can help clinicians better understand their specialty and learn [from] previous experiences from real health care data. In particular, clinicians can utilize the discovered treatment patterns to redesign composite clinical pathways with specific comorbidities [taken] into consideration.”
Future studies will be done to explore expansions of the model’s application as well as to investigate additional information sources that could enhance the model’s utility.
------
