The Measures and Their Psychometric Properties

---

All the scales used in the analysis were developed from previously published measures. Adjustments were made to most of the scales modifying them for the specific computer-based medical information system studied. Any additional modifications to the scales are noted in the discussion of that particular scale. The scale measuring ease of system use was taken as previously published without modifications (Henry & Stone, 1999). The computer staff support measure was modified from a previously published scale of management support (Henry & Stone, 1994). The modification converted the management support scale into a measure focusing on the computer staff helping end users solve problems and make adjustments to the computer system. The item measuring the degree of system use was previously published as part of a computer experience scale (Henry & Stone, 1999). The item asked the respondent to report the percentage of time they spend using the system. The scale measuring past computer experience was a modified version of a published scale by the same name (Stone & Henry, 1998). The modification converted the scale from reporting the amount of prior computing experience to one encompassing the quality of these experiences. The items measuring computer self-efficacy and outcome expectancy were also developed from previously published scales (Henry & Stone, 1995). The computer self-efficacy measure was also modified so the scale was reversed in direction. In other words, individuals with high computer self-efficacy would have low scores on these questionnaire items. These items were reverse-coded before being used in the empirical analysis. Finally, the organizational commitment scale was developed from a previously published scale (Cook & Wall, 1980).

The next step in the empirical analysis was to evaluate the psychometric properties of these measures. The analysis was based on the results from a confirmatory factor analysis, using the structural equations approach Calis (i.e., Covariance Analysis of Linear Structural Equations) in PC SAS version 6.12. A detailed explanation of the confirmatory factor analysis is provided in the Appendix. However, the estimation results implied a good fit between the model and the data. Furthermore, the psychometric properties of the measures were evaluated using the standardized path coefficients from this confirmatory factor analysis. The individual questionnaire items, the estimated standardized path coefficients (i.e., factor loadings), and the calculated measures of these psychometric properties are shown in Table 2.

Table 2: The Measures, Items, Factor Loadings, Reliabilities, and Shared Variances

Questionnaire Item	Factor Loading	Composite Reliability	Percentage of Shared Vairance
Outcome Expectancy		0.94	65%
In general, the system makes it easier for me to perform my hospital duties.	0.89
Working with the system leads to a feeling of accomplishment.	0.90
Knowing how to use the system leads to higher quality work.	0.84
Working with the system results in my completing my work on time.	0.85
I believe I am more productive at work when using the system.	0.85
I believe my successful performance with the system increases my chance of promotion.	0.73
I expect the system to be easier to use as time goes by.	0.66
I think I will be able to use the system to produce high quality work.	0.69
Computer Self-Efficacy		0.76	51%
I really have very little sense how the system works.	0.66
If I were sitting before the system, I would not know how to use it.	0.74
I feel incompetent when I try to use the system.	0.74
Past Computer Experience		0.94	84%
My past performance with computers has been good.	0.91
My past experiences with computers have been good.	0.96
My past attitude toward computers has been positive.	0.87
Computer Staff Support		0.79	56%
When I need them, the system staff responds quickly to needed changes in the system	0.73
The system staff is always willing to help me solve problems.	0.84
The system staff is usually able to solve my problems.	0.66
Ease of System Use		0.83	61%
The system is easy to use.	0.76
I like to use the system.	0.83
The system is user friendly.	0.76
Organizational Commitment		0.88	78%
I would think about leaving this hospital if offered the same job with another hospital.	0.81
If another hospital offered me the same sort of job I have now and I was able to keep all the benefits I have now, I would accept the offer.	0.95

The three initial psychometric properties examined were item reliability, composite reliability, and shared or extracted variance. Item reliability examines how well an individual indicant "loads" on its construct. In other words, it examines the strength of the relationship between an indicant and its construct. Composite reliability measures the degree of internal consistency of all the indicants for the construct. That is, composite reliability is the degree to which the indicants, as a group, measure in common the construct. The percentage of variance extracted is also a measure of reliability. This percentage measures the amount of overall variance in the indicants accounted for by the construct. For additional information about these measures of psychometric properties see Hair, Anderson, Tatham & Black (1992).

The interpretation of the confirmatory factor analysis results indicates that the measures used in the study have desirable values of these three properties. Specifically, since the standardized path or factor loading for each indicant to its measure was at least as large as 0.66, item reliability was satisfied (Rainer & Harrison, 1993). Because all the composite reliability coefficients ranged from 0.76 to 0.94, composite reliability was satisfied (Nunnally, 1978). All the average percentages of shared variance were 51% or greater, demonstrating satisfactory levels of this trait (Rivard & Huff, 1988). Due to these desirable values, it can be concluded that convergent validity was satisfied for each measure (Rainer & Harrison, 1993; Igbaria & Greenhaus, 1992).

Discriminant validity was also examined using these results. This examination compared the squared correlation between each pair of measures to their average percentage of shared variances. Discriminant validity is satisfied if, for each measure pair, the average percentages of shared variance are greater than the corresponding squared correlation (Fornell & Larcker, 1981). The squared correlations ranged from 0.00 to 0.35 and are reported in Table 3. Since these squared correlations were less than all the average percentage of shared variance values, discriminant validity was satisfied (Fornell & Larcker, 1981). These results, coupled with convergent validity, imply that the measures satisfied construct validity (Rainer & Harrison, 1993). Thus, the developed measures had desirable psychometric properties.

Table 3: The Squared Correlations for Pairs of Measures

Measure Pair	Squared Correlation
Computer Self-Efficacy-Outcome Expectancy	0.27
Past Computer Use-Outcome Expectancy	0.35
Past Computer Use-Computer Self-Efficacy	0.19
Computer Staff Support-Outcome Expectancy	0.18
Computer Staff Support-Computer Self-Efficacy	0.04
Computer Staff Support-Past Computer Use	0.08
Ease of System Use-Outcome Expectancy	0.10
Ease of System Use-Computer Self-Efficacy	0.00
Ease of System Use-Past Computer Use	0.00
Ease of System Use-Computer Staff Support	0.04
Organizational Commitment-Outcome Expectancy	0.06
Organizational Commitment-Computer Self-Efficacy	0.08
Organizational Commitment-Past Computer Use	0.07
Organizational Commitment-Computer Staff Support	0.02
Organizational Commitment-Ease of System Use	0.00

---