Validity and Reliability/Rigor

A total of 3,532 samples were used to assess the measurement and structural models. Initially, attention was focused on ensuring the reliability and validity of the reflective constructs (nursing participation in hospital affairs, nursing foundations for quality of care, nursing manager ability, leadership, and support of nurses, staffing and resource adequacy, and nurse-physician relations). This was extended to include the four reflective dimensions of caring behavior (CR): assurance of human presence, knowledge and skill, respectful deference to the other, and positive connectedness.

Next, the evaluation of reliability and convergent validity were carried out. In order to verify reliability, the threshold value of composite reliability (CR) and Cronbach's alpha (CA) should be higher than 0.7, while the minimum cutoff value for outer loading is 0.5. Also, the average variance extracted (AVE) should be higher than 0.5 (45) to confirm convergent validity. The CR and CA of all constructs in this study were above 0.70. In addition, all item loadings were above the value of 0.5, which, assuming that the CR and AVE met the required thresholds, was acceptable (45). Thirteen indicators were deleted due to their low loadings. Table 1 provides an overview of these results for all reflective constructs in stage 1, demonstrating that reliability and convergent validity had been established.

Results: assessment of reflective measurement and composite models.

CR, composite reliability; AVE, average variance extracted; VIF, variance inflation factor; ASSU, assurance of human presence; NR, nurse–physician relationship; CON, positive connectedness; KAS, knowledge and skill; FQ, nursing foundations for quality of care; NM, nursing manager ability, leadership, and support of nurses; HA, nursing participation in hospital affairs; RES, respectful deference to the other; SR, staffing and resource adequacy.

Following this, discriminant validity was examined. For this, the Fornell-Larcker criterion and heterotrait-monotrait (HTMT) approaches were employed (47). Extant research suggests that acceptable HTMT values can be lower than either 0.85 or 0.9 (48); this study adopted the 0.9 HTMT value. Table 2 shows that discriminant validity was acceptable. Further, as per the Fornell and Larcker (49) criterion, the results demonstrated that the square root of the AVE for each construct was greater than its correlation with all other constructs, again demonstrating discriminant validity (Table 3).

Discriminant validity: HTMT.

ASSU, assurance of human presence; NR, nurse–physician relationship; CON, positive connectedness; KAS, knowledge and skill; FQ, nursing foundations for quality of care; NM, nursing manager ability, leadership, and support of nurses; HA, nursing participation in hospital affairs; RES, respectful deference to the other; SR, staffing and resource adequacy.

Discriminant validity: Fornell–Larcker.

ASSU, assurance of human presence; NR, nurse–physician relationship; CON, positive connectedness; KAS, knowledge and skill; FQ, nursing foundations for quality of care; NM, nursing manager ability, leadership, and support of nurses; HA, nursing participation in hospital affairs; RES, respectful deference to the other; SR, staffing and resource adequacy. The bold numbers in the diagonal are the square root of AVE of each construct, and other numbers are correlations between constructs.

Next, the measurement model of caring behavior as a second-order composite construct was assessed. To assess the measurement model of a composite construct, three criteria should be checked: multicollinearity, via variance inflation factors (VIFs), should be <5; the outer weights of associated items of the composite construct should be significant; and nomological validity should be established (50). Table 1 demonstrates that all VIF values were acceptable as they were <5 (38). Additionally, the significance of all outer weights was established via the confidence interval bias corrected approach (0.95). Further, to assess the composite construct, its nomological validity was examined (50). Following the inclusion of the composite construct, the fit indices should not be worse than prior to including them in the model (50). The SRMR for the saturated model before and after including the composite construct was 0.05, below the recommended threshold (0.08) (51), indicating an acceptable model fit and acceptable nomological validity for the composite second-order CB construct.