The Accuracy of a SmartBottle Device for Objective Assessment of Changes in Bottle Weight During Infant Feeding

Current infant feeding assessment methods are often labor-intensive for caregivers and subject to bias.

This study assessed the accuracy of the SmartBottle device, which attaches to an infant feeding bottle and automatically records the amount consumed during bottle-feeding.

This was a proof-of-concept study wherein changes in bottle weight during infant feeding interactions were simulated and compared with criterion weights. Research assistants conducted 4 trials to obtain SmartBottle-derived test weights and corresponding scale-derived criterion weights for bottles of various sizes, shapes, and formula volumes. Each trial focused on a different bottle: (1) Dr. Brown's 120 mL, (2) Dr. Brown's 250 mL, (3) Avent (Philips) 125 mL, and (4) Avent 260 mL. Twenty-five corresponding test and criterion weights were collected for each bottle, resulting in 100 measurements total.

Three research assistants collected data within a controlled laboratory setting.

SmartBottle-derived test weights and scale-derived criterion weights were the main outcome measures.

Two 1-sided t tests were used to compare the equivalence of SmartBottle-derived test weights to scale-derived criterion weights at ±5%, ±7.5%, and ±10% of scale-derived criterion weights. Bland-Altman plots were examined to assess whether there were systematic errors in SmartBottle-derived test weights.

The mean difference between SmartBottle-derived test weights and scale-derived criterion weights was -0.01 (90% CI -0.02 to 0.00) and was equivalent within ±5% bounds (P < .0001). Bland-Altman plots illustrated that there was no evidence of systematic bias related to bottle weight, size, or shape.

SmartBottle-derived test weights were equivalent to scale-derived criterion weights within ±5% equivalence bounds with no evidence of bias. Further testing within real-world infant feeding contexts is warranted.