Rethinking Missing Data Handling in Accelerometer-Based Physical Activity Research: Barriers, Current Methods, and Future Directions.

Abstract

Accelerometers provide objective measures of physical activity, yet missing data from nonwear, noncompliance, and technical failure are pervasive, and the field's predominant response-excluding days below wear-time thresholds-yields heterogeneous, biased estimates. The purpose of this review was to advance missing-data handling in accelerometer-based physical activity research by identifying barriers, synthesizing the methodological literature, and recommending future directions. In this conceptual review, we synthesized 26 methodological papers across six method families to characterize current practice. Three interacting barriers appear to shape it: ambiguity about the unit and mechanism of missingness, a disconnect between methodological development and applied uptake, and the absence of consensus reporting standards. Situated within Rubin's missing-data theory, the synthesis yields two organizing conceptions: a unit-and-mechanism alignment conception and a purpose-driven, decision-oriented framework linking analytic choice to a study's purpose and design. We recommend a data-driven framework; consensus reporting standards; and applied demonstration to improve validity, transparency, and comparability.

Abstract as supplied by the indexing source. © the authors / publisher.