We conducted a cross-sectional survey across six sites as part of a mixed methods multiple-case study in Québec, Canada. Québec is the second largest province in Canada in terms of population. NP roles in primary care were implemented in 2007 [16] in Québec. NPs in primary care practice in offices, home care, and long-term care [16]. We purposefully selected sites to represent practices with NPs in Québec across a range of characteristics [17], including rural and urban settings, type of practice, and team size.
Data collection and recruitment
Data were collected as part of a larger study examining how teams with NPs optimize roles of members of the healthcare team, including patients and families (2016 to 2019). The survey was administered between March 2018 and April 2019, following research ethics board approval. The minimum sample size of 400 was required to obtain significant results from a logistic regression with 6 significant independent variables (IVs) in the final model [29].
Each site included between one to three NPs, and each NP worked with one healthcare team in the organization. Respondents in four of the six sites identified their location as rural. Patients needed to be cared for by a healthcare team with an NP. Family included persons significant to the patient who were involved in their care. Participants were randomly selected from the NPs’ caseload using a computer-generated random sequence. We recruited 200 patients and families where the NP’s caseload included at least 200 patients. The three NPs working in home care and long-term care had a caseload of less than 200 patients, and all the patients in their caseload were recruited. We used Dillman’s method with four mailing cycles [30]. Participants could complete the questionnaire on paper or on-line. A self-addressed stamped return envelope was included. A total of 982 questionnaires were mailed out, 75 were returned undelivered, and 485 questionnaires were completed (response rate: 53%). Questionnaires completed on-line represented 28% (n = 134) of responses. A small voucher incentive ($5-gift card at a national coffee house) was included with the first mailing as a thank you gesture.
Instrument
The Patient-PTE Questionnaire is a 43-item self-report instrument that measures how teams function. It is available in English and in French and takes approximately 10 min to complete [4]. Responses range from 1 (strongly disagree) to 6 (strongly agree) on a scale with no neutral point. Higher scores indicate improved PTE and team functioning. One question (Q. 14) specifically examines beliefs about team effectiveness (BE). One question (Q. 8) describes the NP role and asks respondents if their team includes an NP. One open-ended question (Q. 43) gathers additional comments. Psychometric testing (n = 355) of the instrument included construct validity using known groups with additional testing in teams with physiotherapists in extended roles [4]. Differences were identified by clinical specialty, patient education, length of follow-up, and reason of the health visit. Reliability was assessed using Cronbach α with values ranging from 0.76 to 0.94. Rho coefficients (rs) for processes included in the Team Processes subscale ranged from 0.55 to 0.79 (p < 0.001) indicating that these processes were highly correlated with BE. Responsiveness was assessed and differences were noted in low and high functioning teams (p < 0.001). Individual results can be aggregated to the team, organization, or system level. The questionnaire measures teamwork by including key dimensions of team functioning consistent with our conceptual framework based on the structures-processes-outcomes model and by situating patients and families as members of the healthcare team. Using the current sample, reliability was assessed for the 15-item Team Processes (PREM) and Outcomes of Care (PROM) subscales using Cronbach α. Values ranged from 0.771 (PROM) to 0.877 (PREM).
Operational description of variables
All the variables needed for the analysis were included in the questionnaire and consistent with our conceptual framework. Given that age is not specific to team care in primary care, we included age as a descriptive variable rather than a variable in the predictive models.
Dependant variable
The Outcomes of Care subscale represents a PROM of team functioning. Six items measure timely care, promptly dealing with potential or actual complications, medication, access to information, answers to questions about care, and knowledge and skill of the healthcare team.
Independent variables
Individual variables included patient and family characteristics with type of respondent (patient, family), age (years), sex (male/female), perceived health status (low: poor, fair; high: good, very good, excellent), marital status (living with a partner: yes/no), education (completed high school: yes/no), employment status (not employed/employed), financial status (perceived income adequate/ inadequate), length of follow-up (less than 24 months/ greater than 24 months), routine health visit (yes/no), country of birth (Canada: yes/Other: no).
Team characteristics included healthcare team size with small (less than 5 team members), medium (5 to 10 team members), large (more than 10 team members). Team size was dichotomized as small/ medium-large.
Organizational characteristics were measured using two items. Healthcare setting included 15 possible response options in the questionnaire. Survey responses were grouped into primary care and home care/long-term care. Location including urban, rural, remote and rurban, was dichotomized as urban/non urban.
Role clarity was measured using two items (i.e., how well roles are defined in teams, how work is divided among team members).
Mediator variable
Perception of Team Effectiveness includes 15 items: decision-making (information is shared, ideas are valued), communication (plan of care is communicated, health record is up-to-date, flow of information), care coordination (next steps in care plan, care adjusted to change in patient’s condition, care well organized), cohesion (working together), problem-solving (differences of opinion are respected), patient and family focus (patient/family has a role in the team, contribution is valued, working with family), trust (trust in team), and belief about team effectiveness (team is effective).
Analysis
Logistic regression and mediation analyses were used to examine the influence of the IVs (i.e., individual, team, and organizational characteristics) and role clarity on outcomes of care mediated by team processes. We then examined the influence of team processes on outcomes of care, the dependent variable (DV) [31, 32]. Analyses were completed with IBM Statistical Package for Social Sciences version 27 (2020) [33] and STATA version 13 (2013) [34]. The 5% threshold was used to determine significance. Negatively worded items were reverse-coded prior to analysis. As described above, dichotomous variables were created for all the independent variables in the model. Numeric responses from 1 to 4 were recoded as low, and responses 5 and 6 were recoded as high. Descriptive statistics (number and proportion) were generated. Bi-variate analyses were conducted to examine significant relationships between the independent variables, team processes, and outcomes of care. Multicolinearity between the individual characteristics was examined [31]. Perceived health status and education were highly correlated (p < 0.001 with other individual characteristics and were not included in the models. Unadjusted odds ratios (UORs) were estimated. Finally, we performed three multivariate binary logistic regression and mediation analyses to identify in primary healthcare teams with NPs the (i) IVs (i.e., individual, team and organizational characteristics, and role clarity) that influence team processes (Model 1); (ii) IVs that influence outcomes of care (Model 2); (iii) influence of the IVs and team processes on outcomes of care (Model 3). Coefficients for the indirect effects were calculated using unstandardized variables to identify the potential mediator effect of team processes [32]. The indirect effect (mediation) of team processes was considered when the effect of X on Y was weaker in Model 3 than in Model 2. This was considered complete indirect effect if X no longer had an effect on Y when the effect of the mediator was controlled (Model 3). Sensitivity analyses were performed to examine the indirect effect using Sobel test, Aroian test, and Goodman test [32, 35, 36]. As proposed by Rijnhart et al. (2019) [32], we used the following equations:
$$\mathrm{M}={\mathrm{i}}_2+\mathrm{aX}$$
(1)
$$\mathrm{Y}={\mathrm{i}}_1+\mathrm{cX}$$
(2)
$$\mathrm{Y}={\mathrm{i}}_3+\mathrm{c}'\mathrm{X}+\mathrm{bM}$$
(3)
Where: X represents the characteristics and role clarity (IV); Y represents the outcome (PROM); M represents the mediator (PREM); \(i\) represent the intercepts; aX represents the slope of each independent variable; bM represents the slope of the mediator; cX represents the slope of outcome; and c’X represents the slope of outcome (when the IV and PREM are also a predictor of the PROM); ab represents indirect effect.
The 95% confidence interval (95% CI) of the indirect effect ab was determined by the set of values between the 2.5th percentile and the 97.5th percentile of the distribution of ab obtained by the hierarchical Bayesian method (simulation). Using the computational program developed by Falk and Biesanz (2016) [37] the confidence interval was calculated.
The variables that were included in the final model were kept because of their identified conceptual significance. We did not impute data. All the patients and families with at least one missing datum were removed during the multivariate analysis. The Stepwise procedure was used for the selection of variables retained in the models (using Likelihood Ratio) [38]. The Hosmer–Lemeshow statistic allowed us to determine the quality of the models [31].
Patient and public involvement
Our research team's expert patient (NF) was involved in all phases of the study: development and validation of the Patient-PTE questionnaire, study conception and design, acquisition and analysis of study data, and dissemination.