Variables

I utilized two archival measures of performance derived from hospital incident report data: the number of medication errors and patient falls occurring in a unit in a given month. These data were collected for the month after all surveys were returned. That is, if all surveys were returned by January 31, the corresponding dependent variable would be for the following month (February). I also conducted analyses at three and six months (not reported) after the surveys were returned, and the results reported below are substantively similar. As a result of missing or otherwise incomplete data, the final sample used in the regression analyses was reduced to seventy-five units.

As recommended by prior research in strategic human resource management (for example, MacDuffie 1995), I measure HPHR practices as an additive index of six clusters of HR practices including selective staffing, extensive training, developmental performance appraisal, performance-based rewards, employee empowerment, and job security (alpha = 0.76). The survey items comprising each of these clusters were adapted from previous research (Snell and Dean 1992). To capture respectful interaction and mindful organizing, I developed two new measures. I used four survey items to measure respectful interaction (for example, "We demonstrate a great deal of mutual respect for each other," alpha = 0.84) and nine survey items to measure mindful organizing (for example, "We regularly discuss alternatives as to how to go about our normal work activities," alpha = 0.88). The measures of respectful interaction and mindful organizing have been shown to demonstrate adequate psychometric properties and individual-level survey items acceptably aggregated to the unit level (Vogus 2004; Vogus and Sutcliffe 2006). Due to space constraints, I refer interested readers to these papers for a fuller discussion.

To ensure that the effects of HPHR practices and mindful organizing were robust, I controlled for RN behavior as well as RN and unit characteristics. First, I controlled for three unit characteristics in my regression analyses—unit size (natural logarithm of the number of beds), task interdependence (a fiveitem measure adapted from Pearce and Gregerson [1991], alpha = 0.81), and the complexity of the work processes on the unit (three survey items adapted from Preuss [2003], alpha = 0.70). Second, I controlled for organizational citizenship behavior (six items adapted from Podsakoff et al. 2000, alpha = 0.89) in the models of medication errors and patient falls to ensure that my measure of mindful organizing was not merely capturing the effects of RN effort and willingness to go above and beyond the call of duty. To ensure that HPHR practices and mindful organizing were influencing future performance and not merely the artifact of prior performance, I controlled for prior performance (the number of errors or falls from the prior month). In additional models not reported I also controlled for unit type (dummy variables for if the unit was an intensive care unit, surgical unit, or medical unit), the average patient load per nurse (as reported by the nurses), the use of clinical pathways (dummy variable), and the presence of a no-fault error reporting system (dummy variable); the results were substantively identical.

Results

Table 1 shows the correlations between all the variables used in this study. Most notable is the fact that respectful interaction and mindful organizing are very highly correlated (r = 0.80). While these two variables should be closely related given their theorized relationship, it could pose a problem to the extent that it creates multicollinearity. Multicollinearity is not a concern here because these variables do not appear simultaneously in the same models, and even if they were to appear in the same model additional analyses (not reported) revealed that variance inflation factors, which quantify how much the variance of the estimated regression coefficient is inflated by the existence of multicollinearity, were well within acceptable levels (Chatterjee and Price 1991).

I utilized negative binomial for analysis of the count variables (medication errors and patient falls; Long [1997]), regression equations, and ordinary least squares for the models of respectful interaction and mindful organizing to test my hypotheses. Given that my unit of analysis (the nursing unit) is nested within a hospital and, as such, shares hospital-level properties that may affect the variables of interest, units within the same hospital cannot be considered independent. To correct for this, I use robust standard errors with clustering by hospital in all models. Table 2 reports the results of these analyses. In Model 1 I find that HPHR practices are positively associated with respectful interaction (B = 0.10, p < .05) supporting hypothesis 1. Model 2 reveals respectful interaction is strongly associated with mindful organizing (B = 0.57, p < .001) supporting hypothesis 2, and in Models 3 and 4

I find that mindful organizing is associated with lower incidence of medication errors (B = -.48, p < .05; a 16 percent decrease in errors for a standard deviation increase in mindful organizing) and patient falls (B = -0.95, p < .001; a 29 percent decrease in falls for a standard deviation increase in mindful organizing), supporting hypothesis 3. Interestingly, HPHR practices have no direct effect on the processes of mindful organizing (p > 0.56) or the incidence of medication errors (p > 0.65) but have a strong negative association with the occurrence of patient falls. Lastly, and surprisingly, organizational citizenship behavior was positively associated with medication errors and falls (both p < .02). This suggests that in dynamic and interdependent work like acute-care nursing, typically performance-enhancing behaviors can actually have negative unintended consequences.

Conclusion

In this paper I developed a behavioral approach to understanding how HPHR practices generate social capital and performance. My analyses revealed that implementing HPHR practices creates the opportunities for and signals the importance of respectful interaction. In turn, these respectful interactions create the relational infrastructure upon which the unit-level capabilities for rapid error detection and correction (that is, mindful organizing) can be built, which results in substantially lower levels of medication errors and patient falls. These findings are consistent with my hypotheses as well as an emerging body of research demonstrating the importance of relationships and relational processes on performance (see Gittell 2000) and provide evidence that the behaviors occurring within networks are also an essential mechanism for understanding the multifaceted effects of HPHR practices and organizational performance. Although my results suggest that respectful interaction and mindful organizing underlie social capital, future research needs to directly examine the relationship among network structure (for example, density), respectful interaction, mindful organizing, and performance. Such research could help further understand the complex ways in which HPHR practices affect performance as well as how network structure influences individual and collective action.

References