We applied a mixed-method approach using a sequential research design , which combines the strengths of quantitative and qualitative approaches. Quantitative research allows us to assess factors impacting caregiver burden in a large and potentially generalizable sample of participants but falls short by trying to fit real-life experiences of people into pre-defined categories. Qualitative research, by contrast, allows people to describe their lived experiences in their own words, but it is limited to a relatively small number of participants whose characteristics do not necessarily represent the source population, which in turn affects the generalizability of the research findings. A mixed methods approach allows us to complement research findings that might be omitted by using only quantitative or qualitative research.
For the purpose of this study, we used the baseline data of people with PD and informal caregivers from a longitudinal research project called PRIME-NL . The PRIME-NL aims to evaluate a new integrated and proactive care management model that is implemented in a certain region within The Netherlands and is compared to usual care . For this purpose, questionnaire data among people with PD, informal caregivers and health care providers is collected over multiple years.
An initial analysis of these data revealed that caregiver burden scores were generally very low in our study population, relative to previously published data and clinical experiences of health care professionals in The Netherlands. In order to find an explanation, we performed a literature research on PubMed to assess whether certain population characteristics could explain the difference. This was not the case. We compared the preliminary results in our study population with previous studies in informal caregivers of people with PD that had also used the Zarit Burden Interview (ZBI) to measure caregiver burden. This search for research-based literature was performed by one author [ADG] in September 2021 using the Medical Subject Heading (MeSH) “Parkinson’s Disease” and relevant keywords including “Burden”, “Caregiver” and “Zarit”. The used PubMed search strategy is presented in Additional file 3: Supplemental file 1. Abstracts and titles were screened to identify relevant studies. Search limits were applied to include only studies reporting on caregiver burden measured by ZBI in a PD population. After the abstract review, full-text articles were assessed for relevance and summarized (Table 1). Therefore, we added personal interviews with informal caregivers to assess whether those would confirm, expand or contradict with our quantitative findings. By using a qualitative approach, participants were able to describe their own lived experiences in their own words instead of trying to fit them into predefined constructs. Through this, we were able to retrieve additional information about the impact of PD and their perceived caregiver burden. The final step consisted of integrating the findings from our quantitative and qualitative studies. Figure 1 displays the applied mixed methods structure for this paper.
Study sample and setting
In the present study, we report results from baseline data of the PRIME-NL study, a prospective observational study among people with PD, informal caregivers and health care professionals in The Netherlands . The study investigates the effect of the novel PRIME Parkinson care model on (I) the health of people with PD, (II) the experience of people with PD and their informal caregivers, (III) the experience of health care professionals involved in the care for people with PD and (IV) the costs of care. A detailed description of the study design has previously been published .
In short, the baseline assessment of the study took place between February and December 2020. To be eligible for this study, informal caregivers were required to self-identify themselves as the primary informal caregiver of a person with PD or atypical parkinsonism. “Informal caregiver” was thereby defined as a relative, friend or neighbour who feels involved in the care of a person with PD, for instance, someone who attends medical appointments together with the person with PD, or who is aware of the situation and feels involved. Although a person with PD can have several informal caregivers, the PRIME-NL study strived for the inclusion of one caregiver per person with PD. As the term “caregiver” can have different meanings for different people and can be linked with positive as well as negative associations, potential study participants did not need to define themselves as a caregiver. Actually, some people do not identify as caregivers although they do manage and provide informal care to a person with PD. The inclusion criteria for people with PD were (a) clinical diagnosis of PD or atypical parkinsonism and (b) visited the neurology outpatient clinic of a community hospital at least once during the previous year. Data were gathered through self-administered digital or paper questionnaires. In total, 988 patients and 564 caregivers of people with PD completed the questionnaires. In the PRIME-NL study, there were also cases in which the caregiver participated, but not the person with PD.
We did not perform a sample size calculation for the quantitative analysis in advance, because we assumed that the sample size of the cohort (PRIME-NL) in which the current analyses were embedded would be sufficient to detect meaningful associations. To limit the possibility of a type I error, we performed exploratory factor analysis on the caregiver burden metric in this study (ZBI), which reduced the outcome data from 22 items to 3 factors. This meant that, in our analyses of determinants of caregiver burden, we tested considerably fewer hypotheses than we would have if we had performed no factor analyses. Given the correlation between these factors, and because we had a specific hypothesis based on previous literature for each potential determinant of caregiver we selected, we did not adjust the alpha threshold for multiple-hypothesis testing.
Due to the design of the electronic questionnaire environment, participants were obliged to answer all items of the questionnaires in order to proceed and submit the questionnaire. For the paper-based version of the questionnaire, the PRIME-NL assessment team checked whether all questionnaires were completed, and if in case answers were missing, participants were contacted via telephone to retrieve missing data. Therefore, in the current manuscript, no missing data is reported.
In this current study, we focused on the dyads of persons with PD and their informal caregivers (N = 504). We included both people with idiopathic PD as well as atypical parkinsonism. Although people with atypical parkinsonism may experience additional symptoms and signs that are not typically present in PD, such as orthostatic hypotension, early postural instability and faster disease progression, there are still large similarities between the diseases. Next to those similarities in signs and symptoms, the frequently delayed diagnosis of atypical parkinsonism has supported our decision to include caregivers of people with atypical parkinsonism as well in this study. The study was conducted in The Netherlands, where the government encourages informal care to ensure that people are able to live independently (e.g. able to perform basic day-to-day activities such as getting in and out of bed, eating and drinking, getting (un-)dressed and personal hygiene) and participate in social society (e.g. participate in social activities and able to meet other people) take as long as possible . In case people are not able to live independent and participate in society, the assistance of family, friends and neighbours is required which is then defined as providing informal care. Municipalities can help by offering home care services (i.e. receiving additional resources as rental lifting devices or getting assistance for meal services and doing the groceries) or supporting informal caregivers.
The PRIME-NL study was reviewed and approved by the Ethical Board of the Radboud University Medical Center (file number: 2019–5618). All participants provided digital or written informed consent before they took part in the study.
In the baseline caregivers’ questionnaire of the PRIME-NL study, participants were asked if they agreed to be separately contacted for possible participation in additional research on caregiver burden. A total of 289 out of 504 informal caregivers indicated that they were willing to participate in an additional interview study. To prevent over-recruitment and unnecessary extra burden on informal caregivers, and to guarantee a diverse sample, only 48 informal caregivers were pre-selected based on varying ZBI scores and distribution among age and gender. Those 48 informal caregivers received a letter containing information about the study in which they were invited to participate in an interview regarding the impact of PD on their life. We purposefully sampled participants for the interviews based on variation in the distribution of the following characteristics in the total study population: age group, gender, years since diagnosis of PD and caregiver burden scores. In total, seventeen interviews were conducted with informal caregivers. All participants provided written informed consent before the interviews were conducted.
Data collection and measurement
Caregiver burden was measured by the Zarit Burden Interview (ZBI). ZBI is a validated and most widely used measurement for caregiver burden and has been validated in informal caregivers of people with PD [11, 16]. It covers various domains of caregiver burden, including personal life, social life, interpersonal relationships, health and finances. The scale consists of 22 items which are rated on a 5-point Likert scale ranging from 0 (never) to 4 (nearly always). The ZBI is scored by summing the responses of the individual items (range 0–88). A higher score indicates a higher perceived level of caregiver burden. The severity of the burden was classified into four levels of the total ZBI scores: “little or no burden” (< 21), “mild” (21–40), “moderate” (41–60) and “severe” (> 60) .
Assessments among informal caregivers
Brief Coping Orientation to Problems Experienced Inventory (COPE). The brief COPE is an abbreviated version of the original 60-item COPE measure which is an inventory of individuals’ coping behaviours . It is a self-reported 28-item scale regarding 14 coping style subscales, with two items assessing each coping strategy. For each statement, participants indicate the extent to which they have been using the different coping strategies on a 4-point Likert-scale ranging from 0 (I usually do not do this) to 3 (I usually do this a lot). Total scores range from 0 to 3 for each coping strategy. The higher the score for a particular coping strategy, the more preferred it is by the participants. The 14 coping styles can further be summarized into three overarching coping styles: (a) problem-focused, (b) emotion-focused and (c) avoidant coping .
Caregiver activation measurement [CG-PAM]. The CG-PAM was used to assess caregivers’ knowledge, skills and confidence for the management of health conditions . It is a validated 13-item instrument with scores ranging from 0 to 100 and a higher score indicating a greater level of activation, which is a positive attribute. Each item is rated on a 4-point Likert scale ranging from 1 (disagree strongly) to 4 (strongly agree).
Multidimensional scale of perceived social support (MSPSS). The MSPSS consist of 12 items which can be scored on a 7-point Likert scale ranging from 1 (very strongly disagree) to 7 (very strongly agree) . It has three subscales, each assessed by four items, to assess perceptions of perceived social support from three sources: family, friends and significant others. Overall scores range from 12 to 84 with higher scores indicating greater perceived social support.
Assessments among people with PD
Beck depression inventory (BDI). The BDI is a 21-item self-reported questionnaire used to measure the severity of depressive symptoms . Each item is composed of four statements, depicting a particular symptom. Participants can score on each item on a 4-point Likert scale ranging from 0 (no symptoms) to 3 (very intense symptoms). The total score indicated minimal, mild, moderate or severe depression.
State-Trait Anxiety Inventory (STAI). The STAI is a tool to evaluate anxiety  measuring two dimensions: (1) state anxiety, the current emotional state of anxiety, and (2) trait anxiety, the type of anxiety characteristics for the individual’s personality. It is a self-reported questionnaire composed of 40 items and is based on a 4-point Likert scale ranging from 0 (almost never) to 3 (almost always). The total score of each dimension ranges from 20 to 80, and higher scores indicate greater anxiety.
Acceptance of illness scale (AIS). The AIS was used to assess to what extent people with PD were able to accept the illness without experiencing negative reactions . The scale consists of 8 items with a 5-point Likert scale ranging from 1 (I strongly agree) to 5 (I strongly disagree). Total score ranges from 8 to 40 with lower scores indicating poorer adaption to the illness.
Motor symptoms. To assess the impact of motor symptoms on daily living, part II of the MDS-Unified Parkinson’s Disease Rating Scale (MDS-UDPRS) was used . The scale consists of 13 self-reported items with a 5-point Likert-scale ranging from 0 (normal) to 4 (severe). A higher score therefore indicates greater difficulties with coping with activities of daily living.
Self-assessment Parkinson’s Disease Disability Score (SPDDS). The SPDDS is an instrument to assess disabilities in activities of daily living in Parkinson’s disease . The scale consists of 24 self-reported items with a 5-point Likert scale ranging from 1 (able to do alone without difficulties) to 5 (unable to do at all). Total score ranges from 24 to 120 with higher scores indicating more severe impairment.
Socio-demographic and other variables
The following sociodemographic data of informal caregivers were assessed and included in this study: gender, age, education, marital status, working status, relationship to people with PD, living situation, years of caregiving and caregiving involvement. In addition, the following sociodemographic data of people with PD or atypical parkinsonism were examined: sex, age, education, marital status, working status, living situation, type of diagnosis and time since diagnosis.
An open-ended interview guide was developed after quantitative analyses of the data set were performed (Additional file 3: Supplemental file 2). The interview guide was divided into three parts, covering different stages of the PD journey: time of diagnosis, current situation and future expectations. The interview guide covered topics raised by the quantitative research findings in greater depth but also allowed flexibility in the interviews so that new themes could emerge. Interviews were conducted by a research assistant [CJS] trained in qualitative research techniques. Although we initially planned to conduct the interviews at the informal caregivers’ house, COVID-19 regulations in The Netherlands at the time of this study made that impossible. Therefore, all interviews were conducted by telephone. Interviews lasted between 30 and 60 min.
The distribution of characteristics of informal caregivers and people with PD was expressed with mean and standard deviation values for continuous variables and percentages for categorical variables. Log transformation was used to handle the right-skewed distribution of ZBI values.
To determine the factors of the ZBI in this study population, we performed an exploratory factor analysis (EFA). Promax rotation was used as the correlation between the factors and was assumed. Kaiser–Meyer–Olkin (KMO) and Bartlett’s test of sphericity were measures used to indicate sampling adequacy. We selected factors with an eigenvalue greater than 1 and by inspecting the scree plot . Items that loaded highly on two factors or had a loading ≤ 0.35 were excluded. The fit of the models was evaluated using the comparative fit index (CFI), and a Cronbach’s alpha value ≥ 0.70 was considered acceptable. Raw scores of ordinal variables were used.
Separate multivariable linear regression models were used to identify predictors of caregiver burden based on knowledge from available literature from which a causal diagram was drawn. In model 1, the following background characteristics of informal caregivers were included in the model: age, gender, marital status, educational level, working status, caregiver involvement, caregiver activation, perceived social support and coping strategies. In model 2, the following background characteristics of people with PD were included: age, gender, educational level, working status, coping strategies, disease duration, depression, anxiety, motor symptoms and activities of daily living. Finally, in model 3, we modelled both characteristics of caregivers and people with PD. We used variance inflation factor (VIF) to detect multicollinearity between the predictors of the model. We found high VIF values (> 5) for gender and age variables of people with PD and informal caregivers indicating a severe correlation. After removing gender and age of people with PD as variables, VIF values were between close to 1 and 3 indicating a moderate correlation. Furthermore, linear regression models were repeated for the factors resulting from the EFA. The level of significance was set at a value of p < 0.05. Statistical analyses were carried out using R studio, version 3.6.2.
Contrary to our expectations from clinical observations and existing literature on this topic, a low perceived caregiver burden was reported among included informal caregivers. We therefore wanted to find an explanation by adding in-person interviews. Those interviews allowed us to examine whether the quantitative findings could be confirmed or additional factors impacting caregiver burden identified. All interviews were audio-taped, transcribed verbatim, and entered in Atlas.ti version 9.1.6 for data analysis. For ethical considerations, all identifying information of the participant was removed from the transcripts and pseudonyms were used. Based on the grounded theory methodology, the transcripts were first open-coded to allow for new theoretical possibilities and followed by axial coding to group the codes into a boarder category from which main themes were identified. We followed the six steps of thematic analysis of qualitative research as proposed by Braun and Clark consisting of (1) gaining familiar with the data by reading and re-reading the transcripts, (2) generating initial codes, (3) grouping codes into potential themes, (4) reviewing themes by checking if themes match with the extracted codes, (5) naming the themes and (6) producing the report by finalizing analysis of the selected data extracts . In Additional file 3: Supplemental file 3 the coding tree is displayed. Descriptive statistics were used to summarize the demographic and disease-related variables.
Integrating qualitative and quantitative findings
Visual joint displays were used to integrate quantitative and qualitative findings after initial separate analyses . The data are visually brought together to allow comparison for confirmation, discordance or expansion of findings from each strand.
Validity and reliability
Construct validity was previously examined for the quantitative PRIME-NL study by discussing and testing the questionnaires with an expert panel consisting of patient researchers, informal caregivers, researchers experienced in qualitative health care research and health care professionals. In the PRIME-NL study, we carefully selected instruments that were specific for people with PD and informal caregivers of people with PD. To ensure the trustworthiness of qualitative findings, Guba’s and Lincoln’s criteria were adopted: credibility, transferability, dependability and confirmability . For the qualitative part, the interview guide was developed by researchers [ADG and WMK] and discussed with a former informal caregiver to check for construct validity, i.e. credibility. To enhance transferability, interview participants were recruited from different geographical locations within The Netherlands. To ensure dependability, the coding of the interviews was independently performed by two researchers experienced in qualitative research [ADG and CJS]. Disagreements were resolved during a consensus meeting with other co-authors reached [WMK, SKLD], who were not involved in the analysis process until an agreement was reached. Credibility was ensured by discussing the final codes, themes and interpretation of the content with the research team until an agreement was reached. Confirmability was researched by an expert panel meeting where the findings were presented and discussed with the current and former informal caregivers and researchers with experience and expertise in the area of caregivers of people with PD.