Causal Effects of Physical Activity and Sedentary Behavior on Healthcare Costs

Importance

While increased physical activity (PA) and decreased sedentary behavior (SB) are associated with favorable health outcomes, evidence regarding their causal effects on healthcare costs remains limited.

Objective

To assess the causal effects of PA and SB on healthcare costs.

Design

A two-sample Mendelian randomization (MR) study.

Setting

Separate, non-overlapping cohorts with genetic instruments for self-reported and device-based PA and SB, and healthcare costs.

Participants

The instruments used to assess self-reported PA were derived from a genome-wide meta-analysis of 606,820 individuals across 51 cohorts. Two large genome-wide association studies (GWASs) were used for self-reported SB (leisure screen time N=526,725; television watching N=408,815), while accelerometer-based GWASs (N=89,683–91,105) were used for device-based PA and SB. The instruments used to assess the outcome data were obtained from the FinnGen cohort (N=373,160).

Exposures

Genetically predicted PA and SB.

Main Outcomes and Measures

Validated genetic instruments for log-transformed annual healthcare costs derived from registers, including primary care, secondary care, and medication costs. Inverse variance weighting was used as the primary MR measure, while the sensitivity analyses included MR-Egger, weighted median, simple mode, weighted mode, F-score, Cochran’s Q, and leave-one-out analysis.

Results

Higher genetically predicted self-reported PA was associated with lower healthcare costs (causal estimate, β=-0.166; 95% CI, -0.270 to -0.062). In contrast, higher genetically predicted SB (leisure screen time or television watching) was associated with higher healthcare costs across self-reported datasets (β=0.097; 95% CI, 0.064 to 0.130; β=0.114; 95% CI, 0.063 to 0.165, respectively). No associations were observed for device-based PA (β=-0.014; 95% CI, -0.040 to 0.014) or SB (β=-0.009; 95% CI, -0.197 to 0.179).

Conclusions and Relevance

Findings based on genetically predicted PA and SB support a causal association between these behaviors and healthcare costs, suggesting that increasing population’s leisure-time PA and reducing SB may decrease healthcare expenditure. This highlights the importance of promoting PA for both population health and long-term sustainability of healthcare systems. However, causal evidence remains partly limited, particularly for device-based measures of these behaviors.

Key Points