Cost-Efficient Designs for Assessing Work-Related Biomechanical Exposures

Abstract: Work-related disorders due to biomechanical exposures have been subject to extensive research. Studies addressing these exposures have, however, paid limited attention to an efficient use of resources in exposure assessment. The present thesis investigates cost-efficient procedures for assessment of work-related biomechanical exposures, i.e. procedures aiming at a proper balance between statistical and economic performance.Paper I is a systematic review of tools used in literature providing cost-efficient data collection designs. Two main approaches were identified in nine publications, i.e. comparing cost efficiency among alternative data collection designs, and optimizing resource allocation between different stages of data collection, e.g. subjects and samples within subjects. The studies presented, in general, simplified analyses, in particular with respect to economics.Paper II compared the cost-efficiency of four video-based techniques for assessing upper arm postures. The comparison was based both on a comprehensive model of cost and error and additionally on two simplified models. Labour costs were a dominant factor in the cost efficiency comparison. Measurement bias and costs other than labour cost influenced the rank and economic evaluation of the assessment techniques.Paper III compared the cost efficiency of different combinations of direct and indirect methods for exposure assessments. Although a combination of methods could significantly reduce the total cost of obtaining a desired level of precision, the total cost was, in the investigated scenario, lowest when only direct measurements were performed. However, when the total number of measurements was fixed, a combination was the most cost efficient choice.In Paper IV, demand functions were derived for a four-stage measurement strategy with the focus of either minimizing the cost for a required precision, or maximizing the precision for a predetermined budget. The paper presents algorithms for identifying optimal values of measurement inputs at all four stages, adjusted to integers, as necessary for practical application.In summary, the thesis shows that it is important to address all sources of costs and errors associated with alternative measurement designs in any particular study, and that an optimal determination of samples at different stages can be identified in several cases not previously addressed in the literature.