医疗保健中的知识共享和创新工作行为英文文献和中文翻译(5)
The four items measure an inpidual’s moti-vation to effectively and frequently shareknowledge with co-workers. Opportunity wasmeasured with four items adapted from Bocket al. (2005), and captures an inpidual’scontrol over knowledge sharing as a functionof organizational climate and workload. Thefour-item scale for ability was adapted fromArmitage and Conner (1999), and measuresthe extent to which inpiduals are able toshare knowledge. Knowledge sharing behaviourwas measured with four items adapted fromDaft (2001) and Wasko and Faraj (2005) toassess the extent to which inpiduals engagein knowledge sharing activities in differentwork-related situations. Knowledge reciproca-tion was measured with two items adaptedfrom Wasko and Faraj (2005). This scale cap-tures the extent to which an inpidualreceives new data, information and ideas inreciprocation for his/her knowledge sharingeffort. Finally, the scale for Innovative workbehaviour was adapted from Janssen (2000).The resulting four-item measure captures howfrequently inpiduals engage inmicro innova-tions to improve their practice.We also included control variables, namely:age, professional experience, professionalexperience within the specific PCO (all meas-ured as the natural logarithm of the number ofyears), gender and professional role (physi-cian, psychologist, physiotherapist, nurse,other; all measured as dummies).AnalysesWe validated our research hypotheses usingpartial least squares (PLS). This structuralequation modelling (SEM) technique has beenwidely employed in past research because ofits advantages for latent variable modelling(see Chin, 1998 and Reinartz, Haenlein &Henseler, 2009, who exhaustively addressedthis issue). We adopted PLS because it pro-vides higher statistical power than covariance-based SEM when dealing with samples ofsmall or moderate size (Reinartz, Haenlein &Henseler, 2009). Also, it has minimal condi-tions on residual distributions and does notrequire assumptions onmultivariate normalityof the data.All the analyses were run using thesoftware Smart-PLS 2.0 (beta) (Ringle, Wende& Will, 2005).ResultsMeasurement ModelTo assess the psychometric properties ofmodel constructs, we performed validity andreliability tests using PLS confirmatory factoranalysis, as suggested by previous research inthe field (Chin, 1998; Gefen, Straub &Boudreau, 2000; Vinzi et al., 2010). Withrespect to convergent validity, all the indica-tors were found to load well on their hypoth-esized factor (factor loadings were above the0.70 threshold) except for item KS4. However,as KS4 did not cross-load relevantly, wedecided to retain it in the subsequent analysis.Also, the loadings of the indicators on theirrespective constructs were higher than thecross-loadings on other constructs. Based onthese results, convergent validity was sup-ported (see Table 2). Discriminant validity was assessed by showing that the square root of theaverage variance extracted (AVE) was greaterthan all of the inter-construct correlations. Theresults in Table 3 suggest that our measure-ment model has sufficient discriminant valid-ity. The composite reliability coefficients andthe AVE were calculated to assess the reliabil-ity of model constructs. The results in Table 2show that the composite reliability of all con-structs was above the recommended 0.80threshold. Also, the average variancesextracted by our measures were all above the0.50 acceptability level. All these resultssuggest that the measurement model exhibitsgood psychometric properties.
These analysesalso led us to exclude two items (OPP2 and ABIL1) that did not satisfy the proposed cri-teria (see the Appendix).As all the data were self-reported, and thesame inpidual reported both the dependentand independent variables, we addressed theconcern of common method bias as recom-mended in the literature. To this end, wereverse-coded one item to reduce respond-ents’ bias (Podsakoff et al., 2003). In addition,we used the Harman’s one-factor test, whichindicates that common method variance ispresent when one factor accounts for a major-ity of the covariance in the variables. Applyingthe test to our data showed that six factorsarise with eigenvalues higher than one, andeach factor explains roughly equal variance.Therefore, our data do not exhibit substantialcommon method bias.Hypotheses TestFigure 2 shows the standardized PLS pathcoefficients. The control variables used in thisstudy did not show significant relations, andare therefore not reported in the figure. Weinitially included in the model a direct linkbetween motivation to share and IWB which,as expected, was not significant (β= 0.003,p > 0.05), and was thus dropped from themodel. To assess the statistical significance ofthe path coefficients, which are standardizedbetas, a bootstrap analysis (with 300 replica-tions) was performed.The model explained 47 per cent of the vari-ance in knowledge sharing behaviour, 38 percent of the variance in innovative behaviour,and 10 per cent of the variance in our constructcapturing the level of reciprocation.The results support Hypothesis 1, whichposits a direct effect of knowledge sharing oninnovative behaviour (β= 0.303, p < 0.01), butthey do not support Hypothesis 2, whichposits a mediating effect of reciprocation onthe link between knowledge sharing and inno-vative behaviour. The results show that knowl-edge sharing positively affects reciprocation(β= 0.313, p < 0.001); however, reciprocationdoes not affect innovative behaviour (β= 0.023,p > 0.05).These findings fully support our thirdhypothesis, since all three MOA variablespositively affect knowledge sharing (β= 0.512,p < 0.001 for motivation; β= 0.187, p < 0.01 foropportunity; β= 0.234, p < 0.001 for ability).Finally, the results also fully supportHypotheses 4 and 5. Both the relationshipbetween ability and innovative behaviour(β= 0.258, p < 0.001), and that between oppor-tunity and innovative behaviour (β= 0.247,p < 0.001), are positive and statisticallysignificant.DiscussionThe involvement of employees in generating,promoting and applying new ideas is crucialfor organizational development, especially inprofessionalized service contexts where top-down control and management are unfeasible(Anand, Gardner & Morris, 2007; McNulty &Ferlie, 2004).Most of the research on this ques-tion to date has pursued amanagerial perspec-tive, investigating how organizations can elicitemployees’ abilities and opportunities to inno-vate through job design, knowledge manage-ment systems and HRM practices (Abstein &Spieth, 2014;