Limits...
Predicting inpatient clinical order patterns with probabilistic topic models vs conventional order sets

View Article: PubMed Central - PubMed

ABSTRACT

Objective:: Build probabilistic topic model representations of hospital admissions processes and compare the ability of such models to predict clinical order patterns as compared to preconstructed order sets.

Materials and methods:: The authors evaluated the first 24 hours of structured electronic health record data for > 10 K inpatients. Drawing an analogy between structured items (e.g., clinical orders) to words in a text document, the authors performed latent Dirichlet allocation probabilistic topic modeling. These topic models use initial clinical information to predict clinical orders for a separate validation set of > 4 K patients. The authors evaluated these topic model-based predictions vs existing human-authored order sets by area under the receiver operating characteristic curve, precision, and recall for subsequent clinical orders.

Results:: Existing order sets predict clinical orders used within 24 hours with area under the receiver operating characteristic curve 0.81, precision 16%, and recall 35%. This can be improved to 0.90, 24%, and 47% (P < 10−20) by using probabilistic topic models to summarize clinical data into up to 32 topics. Many of these latent topics yield natural clinical interpretations (e.g., “critical care,” “pneumonia,” “neurologic evaluation”).

Discussion:: Existing order sets tend to provide nonspecific, process-oriented aid, with usability limitations impairing more precise, patient-focused support. Algorithmic summarization has the potential to breach this usability barrier by automatically inferring patient context, but with potential tradeoffs in interpretability.

Conclusion:: Probabilistic topic modeling provides an automated approach to detect thematic trends in patient care and generate decision support content. A potential use case finds related clinical orders for decision support.

No MeSH data available.


Related in: MedlinePlus

Example of 2 generated clinical topics plotted in a 2-dimensional space. Only clinical orders are plotted, based on their prominence in each of the topics. The top left reflects clinical orders most associated with TopicY, with little association with TopicX, suggestive of a workup for diarrhea and abdominal pain. The bottom right reflects clinical orders associated with TopicX, suggestive of a workup for an intentional (medication) overdose and involuntary psychiatric hospitalization. The top right reflects common clinical orders that are associated with both topics. For legibility, items whose score is < 0.2% for both topics are omitted and only a subsample of the bottom-left items are labeled. The diagonal arrow represents a hypothetical patient inferred to have P(TopicX/Patientk) = 80% and P(TopicY/Patientk) = 20%. The dashed lines reflect orthogonal P(Itemi/Patientk) isolines to visually illustrate how clinical order suggestions can be made from such a topic inference. In this case, orders farthest along the projected patient vector (e.g., serum acetaminophen) are predicted to be most relevant for the patient.
© Copyright Policy - cc-by-nc
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5391730&req=5

ocw136-F2: Example of 2 generated clinical topics plotted in a 2-dimensional space. Only clinical orders are plotted, based on their prominence in each of the topics. The top left reflects clinical orders most associated with TopicY, with little association with TopicX, suggestive of a workup for diarrhea and abdominal pain. The bottom right reflects clinical orders associated with TopicX, suggestive of a workup for an intentional (medication) overdose and involuntary psychiatric hospitalization. The top right reflects common clinical orders that are associated with both topics. For legibility, items whose score is < 0.2% for both topics are omitted and only a subsample of the bottom-left items are labeled. The diagonal arrow represents a hypothetical patient inferred to have P(TopicX/Patientk) = 80% and P(TopicY/Patientk) = 20%. The dashed lines reflect orthogonal P(Itemi/Patientk) isolines to visually illustrate how clinical order suggestions can be made from such a topic inference. In this case, orders farthest along the projected patient vector (e.g., serum acetaminophen) are predicted to be most relevant for the patient.

Mentions: Table 1 reports the names of the most commonly used human-authored inpatient order sets, while Table 2 reports summary usage statistics during the first 24 hours of hospitalization. Table 3 illustrates example clinical topics inferred from the structured clinical data. Figure 2 visualizes additional example topics and how patient-topic weights can be used to predict additional clinical orders. Figures 3 and 4 summarize clinical order prediction rates using clinical topic models vs human-authored order sets.Figure 2.


Predicting inpatient clinical order patterns with probabilistic topic models vs conventional order sets
Example of 2 generated clinical topics plotted in a 2-dimensional space. Only clinical orders are plotted, based on their prominence in each of the topics. The top left reflects clinical orders most associated with TopicY, with little association with TopicX, suggestive of a workup for diarrhea and abdominal pain. The bottom right reflects clinical orders associated with TopicX, suggestive of a workup for an intentional (medication) overdose and involuntary psychiatric hospitalization. The top right reflects common clinical orders that are associated with both topics. For legibility, items whose score is < 0.2% for both topics are omitted and only a subsample of the bottom-left items are labeled. The diagonal arrow represents a hypothetical patient inferred to have P(TopicX/Patientk) = 80% and P(TopicY/Patientk) = 20%. The dashed lines reflect orthogonal P(Itemi/Patientk) isolines to visually illustrate how clinical order suggestions can be made from such a topic inference. In this case, orders farthest along the projected patient vector (e.g., serum acetaminophen) are predicted to be most relevant for the patient.
© Copyright Policy - cc-by-nc
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5391730&req=5

ocw136-F2: Example of 2 generated clinical topics plotted in a 2-dimensional space. Only clinical orders are plotted, based on their prominence in each of the topics. The top left reflects clinical orders most associated with TopicY, with little association with TopicX, suggestive of a workup for diarrhea and abdominal pain. The bottom right reflects clinical orders associated with TopicX, suggestive of a workup for an intentional (medication) overdose and involuntary psychiatric hospitalization. The top right reflects common clinical orders that are associated with both topics. For legibility, items whose score is < 0.2% for both topics are omitted and only a subsample of the bottom-left items are labeled. The diagonal arrow represents a hypothetical patient inferred to have P(TopicX/Patientk) = 80% and P(TopicY/Patientk) = 20%. The dashed lines reflect orthogonal P(Itemi/Patientk) isolines to visually illustrate how clinical order suggestions can be made from such a topic inference. In this case, orders farthest along the projected patient vector (e.g., serum acetaminophen) are predicted to be most relevant for the patient.
Mentions: Table 1 reports the names of the most commonly used human-authored inpatient order sets, while Table 2 reports summary usage statistics during the first 24 hours of hospitalization. Table 3 illustrates example clinical topics inferred from the structured clinical data. Figure 2 visualizes additional example topics and how patient-topic weights can be used to predict additional clinical orders. Figures 3 and 4 summarize clinical order prediction rates using clinical topic models vs human-authored order sets.Figure 2.

View Article: PubMed Central - PubMed

ABSTRACT

Objective:: Build probabilistic topic model representations of hospital admissions processes and compare the ability of such models to predict clinical order patterns as compared to preconstructed order sets.

Materials and methods:: The authors evaluated the first 24 hours of structured electronic health record data for &gt;&thinsp;10&thinsp;K inpatients. Drawing an analogy between structured items (e.g., clinical orders) to words in a text document, the authors performed latent Dirichlet allocation probabilistic topic modeling. These topic models use initial clinical information to predict clinical orders for a separate validation set of &gt;&thinsp;4&thinsp;K patients. The authors evaluated these topic model-based predictions vs existing human-authored order sets by area under the receiver operating characteristic curve, precision, and recall for subsequent clinical orders.

Results:: Existing order sets predict clinical orders used within 24 hours with area under the receiver operating characteristic curve 0.81, precision 16%, and recall 35%. This can be improved to 0.90, 24%, and 47% (P&thinsp;&lt;&thinsp;10&minus;20) by using probabilistic topic models to summarize clinical data into up to 32 topics. Many of these latent topics yield natural clinical interpretations (e.g., &ldquo;critical care,&rdquo; &ldquo;pneumonia,&rdquo; &ldquo;neurologic evaluation&rdquo;).

Discussion:: Existing order sets tend to provide nonspecific, process-oriented aid, with usability limitations impairing more precise, patient-focused support. Algorithmic summarization has the potential to breach this usability barrier by automatically inferring patient context, but with potential tradeoffs in interpretability.

Conclusion:: Probabilistic topic modeling provides an automated approach to detect thematic trends in patient care and generate decision support content. A potential use case finds related clinical orders for decision support.

No MeSH data available.


Related in: MedlinePlus