POWER CONSIDERATIONS IN ESTIMATING SAMPLE SIZE FOR EVALUATION OF THE NIOSH LIFTING EQUATION

摘要

The efficient design of human studies requires careful consideration of issues of intended analysis and statistical power. A prospective epidemiological study is under way to test the ability of the NIOSH lifting equation to predict work-absence due to low back pain. Power calculations using a Proportional Hazards Model show that a sample of 650 would provide sufficient power for an overall evaluation of the equation. Sample sizes in the tens of thousands would be required to evaluate the correctness of the least influential multipliers in the equation. Due to a need to balance costs and benefits, a sample size of 1000 has been chosen for the study. This will permit the overall evaluation of the equation and of some of the constituent multipliers, but not of the smaller multipliers.