HipScreen mobile app for the measurement of hip migration percentage in children with cerebral palsy: Accuracy, reliability, and discriminatory ability

Abstract

Aim: To assess the accuracy, reliability, and discriminatory ability of a mobile app for measurement of migration percentage in hip surveillance radiographs of children with cerebral palsy (CP).

Method: The free mobile app HipScreen (www.hipscreen.org) was utilized by a diverse group of users to measure the migration percentage of 40 hips at two time points after completing an online tutorial and competency test. The mean absolute error (MAE) was calculated against the reference standard obtained on a radiology workstation. Statistical analyses included linear regression, intraclass correlation coefficient (ICC), and area under receiver-operating characteristic curve (AUC).

Results: Thirty-seven users completed the study, with 30 having a healthcare professional background, but only 15 with x-ray interpretation expertise. The overall MAE of migration percentage measurement using the HipScreen app was 5.72% (95% confidence interval [CI]: 5.38-6.06), with good reliability between time points (ICC = 0.83). With a migration percentage less than 30% considered as a positive case, the HipScreen app had a sensitivity of 94% (95% CI: 87-97), specificity of 66% (95% CI: 61-77), and an AUC of 0.92 (95% CI: 0.88-0.96).

Interpretation: Users from a broad range of backgrounds can utilize the HipScreen app to measure hip surveillance radiographs with clinically acceptable accuracy, reliability, and discriminatory ability.

What this paper adds: The HipScreen app can accurately and reliably measure migration percentage. As a screening tool, HipScreen has excellent sensitivity and discriminatory ability. A broad range of HipScreen users achieve clinically acceptable performance.

Figure 1.

For quantification of migration percentage (MP), three lines for each hip are drawn perpendicular to the pelvis horizontal “H” line: L, a line touching the lateral aspect of the femoral head ossific nucleus; M, a line touching the medial aspect of the femoral head ossific nucleus; and P, a line touching the lateral edge of the acetabulum. The MP of each hip is distance A divided by distance B, multiplied by 100%.

Figure 2.

Receiver-operating characteristic curve (ROC) and area under the ROC curve (AUC) of the HipScreen app based on a positive case defined as migration percentage (MP) > 30% as measured on the reference standard radiology workstation

Figure 3.

With positioning of the HipScreen MP Ruler overlay on the x-ray image such that the white and black lines touch the lateral and medial aspects of the femoral head ossific nucleus, the femoral head is divided into 10% increments by vertical lines. (A) The position of the lateral edge of the acetabulum (yellow arrow) is half-way between the first and second vertical lines from the edge of the screen, so the MP is 15%. (B) The lateral edge of the acetabulum (yellow arrow) touches the fifth vertical line, so the MP is 50%. For rapid screening, if the lateral edge of the acetabulum is inside the red box on either side of the HipScreen MP Ruler, the MP is >30%, triggering a referral.