# A retrospective study of a weight estimation model for hospitalized bedridden patients based on anthropometric parameters

## TL;DR

This study created a reliable model to estimate the weight of bedridden patients using easily measured body measurements.

## Contribution

A novel weight estimation model for bedridden patients using accessible anthropometric parameters was developed and validated.

## Key findings

- The male model had an adjusted-R2 of 0.901 and RMSE of 3.81 kg.
- The female model achieved an adjusted-R2 of 0.829 and RMSE of 3.81 kg.
- Bland-Altman analysis showed strong agreement between estimated and actual weights.

## Abstract

The aim of this study was to construct a weight estimation model for bedridden patients using anthropometric parameters that are readily obtainable during routine clinical care.

A retrospective study was conducted involving 494 bedridden inpatients from the Department of Endocrinology of a tertiary general hospital (February 2023–February 2024). Weight was measured via a calibrated wheelchair scale. Anthropometric parameters (age, height, wrist, lower limb, waist, and hip circumferences) were measured in the supine position by trained researchers using standardized tools and specific anatomical landmarks. The estimation models were developed using stepwise regression.

The final models demonstrated excellent performance. The male model achieved an adjusted-R2 of 0.901 and Root Mean Square Error (RMSE) of 3.81 kg; the female model achieved an adjusted-R2 of 0.829 and RMSE of 3.81 kg. Bland-Altman analysis confirmed strong agreement between the actual and estimated weigh values, with a mean difference close to 0, no significant proportional bias, and most differences residing within the 95% limits of agreement.

The developed models provide a reliable, cost-effective method for weight estimation in bedridden patients, using parameters that can be integrated into routine clinical assessments, offering a practical alternative to specialized equipment in resource-limited settings.