CohortDevelopmental medicine and child neurology2023

Gross motor function prediction using natural language processing in cerebral palsy.

Kelly Greve, Yizhao Ni, Amy F Bailes and 6 others

PMID 35665923

WHAT IT FOUND

Natural language processing of clinical notes accurately predicted whether children with cerebral palsy were ambulatory or non-ambulatory, distinguishing these broad groups well.

It was less reliable for separating specific GMFCS levels, particularly within the ambulatory range.

Key findings

01The model achieved high accuracy in distinguishing ambulatory from non-ambulatory status, with area under the curve values of 0.88 to 0.99 for non-ambulatory detection.

02Performance dropped when trying to differentiate specific GMFCS levels, such as distinguishing level III from levels I-II, where sensitivity was low.

03Distinguishing between GMFCS levels IV and V was more successful than within the ambulatory groups, likely due to clearer documentation of equipment and mobility differences.

STILL TO COME

How it was doneWhat they found

Read the rest of this summary

You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.

Already have one?

What it does not show

The study was conducted at a single center, so the results may not apply to other hospitals or health systems. The population was primarily White (75%), limiting generalizability to more diverse groups. GMFCS level III patients made up only 17% of the data, which may have reduced the model's ability to accurately identify this specific group. The models relied on existing documentation quality; if clinicians did not record specific details like equipment or transfer needs, the model could not use that information.

Declared interests

The authors declared no conflicts of interest. The study was supported by grants from the National Institutes of Health.

The easy way to misread this

Do not assume this tool can accurately assign a specific GMFCS level to a patient. While it reliably identifies broad ambulatory status, it struggles to distinguish between adjacent levels like I and II or to consistently identify level III patients.

Read it on PubMed →


The study

Participants
2483 patients
Certainty of evidence
Low

Browse

    Cite

    Kelly Greve, Yizhao Ni, Amy F Bailes, et al. Gross motor function prediction using natural language processing in cerebral palsy. Developmental medicine and child neurology. 2023.

    Read the original