Study: For reliable step count estimates, use at least 3 days of fitness tracker data

When using wrist-worn devices or smartphones, at least 3–5 days’ worth of measurements is required to obtain a reliable average daily step count estimate, according to a recent Singapore study.

“Overall, when there is no restriction regarding the continuity of the measurement days, 3 days weekly and 4 days monthly are sufficient for the corresponding time windows,” the researchers said. “More measurement days are required for reliable estimates in participants with younger ages and those using smartphone-based step count tracking.”

Drawing from a 5-month physical activity programme in Singapore, the researchers divided data into weekly (n=22) and monthly (n=5) time windows. For each window, they determined the minimum number of measurements required in order to achieve an intraclass correlation coefficient (ICC) exceeding 80 percent for the estimation of mean daily step count.

The analysis of weekly and monthly time windows included 95,673 and 56,735 adults, respectively, who had complete daily step count available for analysis. Most participants (44 percent) used smartphones to record their step counts, as well as the wrist-worn tracking devices provided for free by Singapore’s Health Promotion Board (43.2 percent). [Sci Rep 2021;11;9633]

Looking at weekly time windows, achieving an ICC of at least 80 percent for mean daily step count would need 2–3 days’ worth of measurements, regardless of whether data were collected from consecutive days or not. Notably, however, slightly but significantly fewer data points were required when measurements were collected on simple random vs random consecutive days (mean, 2.5±0.5 vs 2.7±0.5; p=0.025).

In contrast, when considering monthly time windows, the researchers found that they needed more data to reliably estimate daily step count. To reach an ICC of ≥80 percent, step counts from 3–4 randomly sampled days were required; when obtained consecutively, step data from 4–5 days were needed to achieve reliable estimates.

“Our findings are robust over time under each study setting. While consecutive measurements require an extra day for monthly time windows, overall, a minimum of 3 days for a week and 5 days for a month allow reliable estimation of mean daily step counts using both random and consecutive measurement days,” the researchers said.

Consistent over both time window analyses, and for simple random and random consecutive sampling, the minimum number of days required to achieve a reliable daily step count estimate was higher for participants who were younger, obese, and used smartphones to measure their steps.

“Various factors could contribute to this observation, such as technical device characteristics, wearing location, and participant’s preference for tracking devices and physical activity behaviour,” the researchers said.

“Future studies on factors that influence the day-to-day variability of daily stepping measure, as well as their interaction effects, may foster the establishment of best practices in step count measurement and strengthen step count-based health research,” they added.