Construct validation of a tool for measuring job satisfaction for nurses

The purpose of this study was to investigate the construct validity of the Index of Work Satisfaction developed by Slavitt and others (1978) to measure job satisfaction for nurses. Based upon need satisfaction theories, the scale contained seven subscales which were believed to explain job satisfaction. A global satisfaction level item was added in an attempt to obtain a separate measure of the dependent variable. A review of the literature revealed that a valid tool for measuring nurses' job satisfaction does not presently exist. The instrument was modified, pilot-tested for reliability and after a second phase of modification, was administered to a volunteer sample of 177 staff nurses representing several hospital and community work settings. Multiple regression and discriminant function analyses were performed on the data, and the results of these analyses were interpreted in terms of the construct validity of the job satisfaction scale. Results indicated that the scale is highly reliable, and that three of the subscales explained approximately 30 percent of the variance in the scores on the global satisfaction measure of job satisfaction. High intercorrelation of the subscales with each other and with the total scores hindered the interpretation of the variance in the total scores explained by each of the significant variables. The results of the analyses suggest that the high reliability of this version of the tool makes it a psychometrically useful measurement of job satisfaction for nurses, to the extent that job satisfaction is comprised of the seven components contained in the scale. Regarding its construct validity, there is conclusive evidence that the linear additive model of job satisfaction on which the instrument is based does not allow a complete view of the construct. Whether the three significant predictors of the total score, Professional Status, Administration, and Interaction, are part of one broader construct, or whether they interact in some unique way, could not be determined because of the multicollinearity problem. A major difficulty throughout the study was the lack of a reliable alternate measure of the criterion. The global satisfaction item responses did not correlate highly with the total test scores, and this self-report, Likert-type item would be subject to the same response bias as the questionnaire itself. Apart from the demonstration that the Professional Status, Administration, and Interaction components appear to contribute to the measurement of the construct, the study failed to gather evidence in support of the construct validity of the modified Slavitt scale. It can be concluded, therefore, that the scale 'reliably measures some aspects of job satisfaction for nurses, but one cannot state with confidence that it actually measures the complex attitude which comprises the construct "job satisfaction." Recommendations have been made for appropriate use of the tool and for further construct validation studies.Applied Science, Faculty ofNursing, School ofGraduat