This study included multi-morbid patients from eight primary care practices in Noord-Brabant, the Netherlands. All patients with two or more registered chronic conditions (n = 413) were eligible to participate. Exclusion criteria were: too ill to participate or recently moved (and as a result no longer treated by the primary care practices under study). Based on information received from the GP, patient or their informal caregiver nineteen patients were not eligible to participate (death (n = 4), terminal illness (n = 2), incorrect address (n = 5), recent move (n = 2), inability to fill in the questionnaire due to poor cognitive functioning (n = 2), recent stroke (n = 1), or poor eyesight (n = 3)). Questionnaires were sent by mail to all remaining participants (n = 394). After a few weeks, reminders were sent to non-respondents. Another few weeks later, second reminders and duplicates of the questionnaire were sent to non-respondents. When no response was received after the second reminder, we called non-respondents for whom telephone numbers were available. In total, 216 patients filled in the questionnaire and consented to participate in the study. Thus, the response rate was 55% (216 out of 394 respondents). A sample size calculation revealed that 110 participants would be required in order to detect small to medium effects with 95% power and a type 1 error rate of 5% [17]. Having 216 respondents is therefore sufficient for valid results.
The medical ethics committee of the Erasmus Medical Centre, Rotterdam, the Netherlands, reviewed the research proposal (file number METC_2018_021) and decided that the rules laid down in the Dutch Medical Research Involving Human Subjects Act did not apply. Our research did not have a RCT design, participants were not subjected to procedures such as taking a blood sample, the research was not carried out with the intention of contributing to medical knowledge (e.g. etiology, pathogenesis, signs/symptoms, diagnosis) by systematically collecting and analyzing data. The main aim of the research was to investigate experiences of participants with care delivery, a process evaluation to improve quality of care delivery, which does not fall under the scope of Medical Research Involving Human Subjects Act (WMO) (see https://english.ccmo.nl/investigators/legal-framework-for-medical-scientific-research/your-research-is-it-subject-to-the-wmo-or-not). Written consent was obtained from all participants.
Measures
PCC for patients with multi-morbidity in the primary care setting
PCC for patients with multi-morbidity in the primary care setting was measured using the 36-item patient-centered primary care (PCPC) instrument, which assesses the eight dimensions of PCC [18]. The PCPC instrument builds on our earlier work, in which we investigated the eight dimensions of PCC in hospital and long-term care settings [19,20,21]. Responses of patients were measured on a 5-point scale ranging from 1 (totally disagree) to 5 (totally agree), with higher scores indicating greater PCC. Scores for each of the eight dimensions of PCC were derived by calculating the average score for all items in that particular dimension. The overall score of PCC, in turn, was derived by calculating the average score for the eight dimensions (mean of the eight subscales calculated in the previous step). In this study, the Cronbach’s alpha value for this instrument was 0.89, indicating good reliability.
Well-being
Well-being was measured with the 15-item version of the Social Production Function Instrument for the Level of Well-being (SPF-ILs) [22]. Levels of physical (comfort and stimulation) and social (status, behavioral confirmation, and affection) well-being were measured. Responses of patients were measured on a 4-point scale ranging from 1 to 4, with higher scores indicating greater well-being. Scores for physical and social well-being were derived by calculating the average score for all items in that particular subsection of items. In this study, the Cronbach’s alpha value for both physical and social well-being, measured with the SPF-ILs, was 0.83, indicating good reliability.
Co-creation of care
Co-creation of care was measured with the relational co-production instrument [23]. The instrument consists of seven items measuring four aspects of communication (timely, accurate, frequent, and problem-solving) and three aspects of the relationship (shared goals, shared knowledge, and mutual respect) between patients with multi-morbidity and the healthcare professionals treating them (GPs, nurse practitioners, and specialists). Responses of patients were measured on a 5-point Likert-scale ranging from 1 (never) to 5 (always), with higher scores indicating better co-creation of care. Scores for co-creation of care were derived by calculating the average score for all items in this instrument. In this study, the Cronbach’s alpha value for this instrument was 0.93, indicating excellent reliability.
Satisfaction with care
The adjusted version of the Satisfaction with Stroke Care questionnaire (SASC) was used to measure patients’ satisfaction with care [24]. Although the original 8-item SASC was used among stroke patients, this instrument contains generic questions about satisfaction with care and is not restricted to patients receiving stroke care. The SASC instrument is therefore often used in various patient populations in the hospital setting [25,26,27,28]. Given that the instrument was developed to assess satisfaction with care in the hospital setting, we did slightly adjust items for the primary care setting (e.g. ‘The doctors have done everything they can to make me well again’ was changed into ‘The staff has done everything they can to make me well again’). Furthermore, we removed irrelevant or overlapping items (e.g. ‘The hospitalization process went smoothly’ and ‘I have been treated with kindness and respect by the staff at the hospital’), which resulted in a final set of 6 items: ‘I have received all the information I want about the causes and nature of my illness(es)’, ‘The staff has done everything they can to make me well again’, ‘I am satisfied with the type of treatment they have given me (e. g. physiotherapy, occupational therapy)’, ‘I have had enough therapy (e.g. physiotherapy, occupational therapy)’, ‘I am happy about the effect treatments had on my disease progression’, and ‘I am satisfied with the treatment provided by the general practitioner who I visit’. Responses of patients were measured on a 4-point scale ranging from 1 (totally disagree) to 4 (totally agree), with higher scores indicating greater satisfaction with care. Satisfaction with care scores were derived by calculating the average score for all 6 items. In this study, the Cronbach’s alpha value for this instrument was 0.89, indicating good reliability.
Background characteristics
Patients were also asked to provide information on background characteristics, such as age, gender, education, and marital status. Dummy variables were created for marital status (1, living alone, widowed or divorced; 0, married/living with partner) and education (1, primary education or less; 0, preparatory school for vocational secondary education or higher).
Statistical analyses
SPSS software (version 23; IBM Corporation, Armonk, NY, USA) was used to analyze the data. Descriptive statistics were applied to all variables and involved the calculation of ns, means, minimums, maximums, standard deviations (SDs), and/or percentages. Pearson correlation analyses were performed to identify associations between PCC and background characteristics, co-creation of care, satisfaction with care, and physical and social well-being of patients with multi-morbidity. Regression analyses were performed to investigate multivariate relationships among these variables. Two-sided p values ≤0.05 were considered to be significant.
As data were missing for some PCC items due to occasional inapplicability, we additionally employed multiple imputation techniques (Markov chain Monte Carlo) and performed the regression analyses on pooled results based on the five imputed datasets (n = 216 each). Predictive mean matching was used as an imputation model to ensure that imputed values preserved the actual range of each variable.