Evaluating the full economic costs to workers of recovery from workplace injury and illness is crucial for the effective formulation and targeting of policy and programs for reducing workplace accidents and exposure to workplace hazards, for successfully bringing back injured workers into the workplace and the workforce, and for evaluating the adequacy and equity of workers' compensation. Administrative data, such as those found in workers' compensation claim databases, are frequently used to study the consequences and contexts of workplace injuries. Compared with administrative data, multi-purpose, nationally representative household survey data offer a far broader coverage of workplace injuries, and a broader range of information on their context and consequences. Longitudinal surveys are especially useful for investigating the roles of factors that influence a process such as recovery from a workplace injury or illness that occurs over a potentially extended time period. In particular, longitudinal survey data are more likely to accurately capture circumstances where return to work does not coincide with the end of disability benefits, or where an injury leads to multiple spells of time out of work. Any single longitudinal survey, however, has disadvantages that may include small samples of injured workers, limited frequency or duration of observation, a limited range of ages in the survey, and respondent attrition and error in their self-reports. The present study addresses these limitations by comparative evaluation of injuries and their consequences between surveys, and by developing and testing methods for pooling observations of injured workers across surveys. The study applies hazard modeling and multiple imputation methods to the combining of data from two nationally representative panel surveys, each including self-reports of work injury and resulting health condition and work limitation. Differentials in duration and sustainability of return to work are analyzed by occupation, industry, health condition, and the socio- economic characteristics of injured workers. The two panel studies' designs are overlapping but non-identical with respect to their population coverage, duration of observation, and mode of data capture (panel and retrospective observation). Using the two surveys together is expected to increase the power of the statistical inference about the differentials in return to work on key variables including type of injury of illness; and to strengthen our understanding of the return-to-work process by including the widest possible range of potentially important variables in the analysis of returning to work after a workplace injury. Differentials in duration and sustainability of return to work are analyzed by occupation, industry, health condition, and the socio-economic characteristics of injured workers. Hazard modeling and multiple imputation methods are applied to the combining of data from two nationally representative panel surveys, each including self-reports of work injury and resulting health condition and work limitation. Using the two surveys together is expected to increase the power of the statistical inference and to strengthen our understanding of the return-to- work process by including a broad range of variables and worker ages in the analysis. ? ? ?
| Rendall, Michael S; Ghosh-Dastidar, Bonnie; Weden, Margaret M et al. (2013) Multiple Imputation For Combined-Survey Estimation With Incomplete Regressors In One But Not Both Surveys. Sociol Methods Res 42: |
