Comparison of Integrated Outpatient Palliative Care With Standard Care in Patients With Parkinson Disease and Related Disorders

Key Points Question Is outpatient palliative care associated with improvements in patient or caregiver outcomes compared with current standards of care among persons with Parkinson disease and related disorders? Findings In this randomized clinical trial of 210 patients with Parkinson disease and related disorders and 175 caregivers, patients receiving palliative care had better quality of life at 6 months (primary outcome) as well as better symptom burden and rates of advance directive completion. No significant difference was found in caregiver burden at 6 months (coprimary outcome). Meaning Outpatient palliative care may improve certain patient and caregiver outcomes associated with Parkinson disease and related disorders.

With the aging of the US population this figure is expected to double by 2040. 2 While PD is traditionally defined by its motor symptoms (e.g. tremor, slowness), nonmotor symptoms including pain, depression, fatigue, and dementia are extremely common and contribute significantly to quality of life (QOL) and disability. 3 For example, over the course of the illness up to 80% of individuals with PD will develop dementia. 4 PD is currently ranked as the 14 th leading cause of death in the US and this may be an underestimate as PD contributes to other causes of death such as pneumonia and falls. 5,6 People living with PD are also five times more likely to be placed in a nursing home and die in hospitals significantly more often than their age-matched peers. 7,8 In addition to the direct impact of PD on patients, it also is associated with significant effects on their family caregivers including high rates of stress, increased incidence of depression and increased mortality. 9,10 Despite these advances in our understanding of the burden of PD on patients and caregivers, clinical care continues to be driven by a chronic illness model predominantly focused on managing motor symptoms through a patient-physician dyad. 11 Potential gaps in this model include addressing psychosocial and spiritual concerns; recognition and management of nonmotor symptoms; advance care planning; and caregiver support. 11,12 Palliative care focuses on the relief of suffering through the assessment and treatment of physical, psychosocial, and spiritual issues. 13 While traditionally associated with cancer, palliative approaches have been successfully applied to several chronic progressive illnesses including heart failure and pulmonary disease. 14,15 Several recent systematic review articles suggest palliative care approaches may be beneficial for PD patients and that PD patients and their family caregivers have unique palliative care needs compared with other populations (e.g. cancer). 16,17 In line with these articles, a few centers in the US and Canada have begun offering interdisciplinary outpatient palliative care for PD patients including the three sites heading this proposal.

A.2. Research Questions and Gaps:
As noted in recent reviews, 18 there are significant gaps in the evidence needed to support widespread adoption of this care model: 1) A need for randomized controlled comparative effectiveness trials of palliative care interventions in PD to determine whether such care does in fact improve patient-centered outcomes versus standard care.
2) Data to guide referral and selection of appropriate patients for this time and resource intensive intervention.
3) Data to standardize and optimize which services are provided and how they are delivered, optimally with direct input from PD patients and their caregivers.
This research has a significant potential to impact clinical care for persons living with PD and related disorders by providing the first empirical evidence of whether this model of care improves patient or caregiver outcomes. Positive results would influence care by: a) providing greater support for the assertion that PD and related disorders are conditions that merit PC; b) demonstrating that patient quality of life (QOL) and caregiver burden can be improved above current standards of care through a PC approach; c) providing data to optimize this care including characteristics of persons most likely to benefit from this care; and d) identifying a replicable model of care that could be implemented in other sites.
A.3. Specific Aims: Aim 1: Determine whether interdisciplinary outpatient palliative care improves patient QOL or caregiver burden in Parkinson's disease (PD) patients versus standard care with a neurologist. We hypothesized that PC would improve both patient QOL and caregiver burden by systematically addressing issues that contribute to patient and caregiver suffering.
Aim 2: Identify characteristics of PD patients most likely to benefit from a palliative care approach. We hypothesized that patients and caregivers with worse QOL would perceive greater benefits from this additional support.
Aim 3: Interview patients and caregivers to elicit their direct input on what palliative care services are most helpful, what additional services may be needed, and preferences for service delivery. We hypothesized that patients and caregivers will provide insights to optimize our model of care to better meet patient and caregiver needs.  19 This study showed a significant improvement in symptom burden as measured by the Edmonton Symptom Assessment Scale PD (ESAS-PD) between their baseline (mean + STD = 56 + 19) and first follow-up visit (40 + 17; paired ttest p < 0.0001). Significant improvements were seen in many individual symptoms including pain, anxiety, drowsiness, and dysphagia. Although there was no control group, these patients were largely referred by their treating neurologists because of limited success with standard care. Moreover, Dr. Miyasaki and Dr. Benzi Kluger (PI) published data showing that 80% of PD patients participating in outpatient palliative clinics died at home and less than 10% died in a hospital. 20 While there was no control group, historical data shows over 50% of PD patients typically die in hospitals and less than 10% die at home. 8 2) Patient and Caregiver-Centeredness of Intervention: Dr. Kluger (PI) obtained a Clinical

A.4. Preliminary Data and Feasibility
Effectiveness and Patient Safety Grant through the University of Colorado Hospital in July 2013 to optimize our outpatient palliative care clinic with the non-negotiable goal of achieving high quality, patient-centered care. We collected data through surveys and phone interviews with patients and caregivers and made ongoing adjustments in our clinic based on feedback received. Specifically, we revised clinic processes by: 1) Providing written information about palliative care and our clinic to patients prior to their first visit; 2) Clarifying team member roles to reduce redundancy and ensure coverage of key issues; 3) Providing written instructions to patients in real time from each team member; and 4) Following up on visit recommendations with phone calls within two weeks. We tracked patient and caregiver satisfaction through the course of this project and obtained mean scores of at least 9 (on a 10 point scale) on all items by the end of our 1-year improvement period including overall satisfaction, answering their questions, and likelihood of referring other patients to our program. Of note, palliative care as prescribed in the "intervention arm" of this trial is already available to and being utilized with PD patients at all sites. The "comparator arm" is meant to reflect the current standard of care in the community at large. Thus this trial represents a comparison of effectiveness of two currently-available standards of care.

A.5. Important Innovations of the Study and Intervention
Important innovations of the study and intervention include: a) Application of a PC approach in the outpatient setting to a non-cancer population.
b) Participant inclusion is based on perceived need rather than prognosis or disease-centered criteria of advanced illness. a neurologist. For Aim 1 mixed regression models will be used to assess our primary comparative effectiveness outcome: the difference in change on the QOL-AD and ZBI between the intervention and standard care groups at 6 months. We hypothesize that PC will improve both patient QOL and caregiver burden by systematically addressing issues that contribute to patient and caregiver suffering.
Aim 2: Identify characteristics of PD patients most likely to benefit from a palliative care approach. For Aim 2 our primary analysis will be built on regression models from Aim 1 to determine what baseline patient and caregiver characteristics are most strongly associated with caregiver and patient improvement in the intervention arm. We hypothesize that patients and caregivers with worse QOL will perceive greater benefits from this additional support. The sites chosen provide a broad demographic base, which is similar to the US as a whole in terms of racial and ethnic breakdown and includes persons from inner city and rural environments.
The study sites chosen also provide real-world settings relevant to future implementation efforts. These sites had already acted as models that had served to inspire other palliative clinics for PD in practice or in development. The University of Alberta site had been actively developed by an experienced clinician in this field (Dr. Miyasaki) and provided data on this implementation process potentially important to other new sites. Data from Aims 2 and 3 are specifically being collected with the goals of informing future implementation and dissemination efforts, including the potential for designing less intensive programs meeting essential needs for rural and community settings.

C.3. Participants:
Participants will be identified and recruited for this study through three main venues: 1) Referral from investigators' clinics or their colleagues at primary study sites; 2) Referral from community physicians who were notified of the study through existing personal contacts and advertisements; and 3) Self-referral by patients who learned of the study through community organizations, talks, advertisements, or websites (e.g. clinicaltrials.gov). To ensure inclusion of community and rural participants, each study site will partner with community neurologists and regional patient organizations which provide services to patients in rural locations. Study sites also provided services through inner city clinics serving indigent populations, which also tend to provide services to a larger proportion of minority populations.
We will select participants using the following inclusion and exclusion criteria: Inclusion Criteria: Potential participants must be fluent English speakers, over age 40, and meet UK Brain Bank criteria for a diagnosis of probable PD 30 or have another neurodegenerative cause of parkinsonism (progressive supranuclear palsy, corticobasal degeneration, Lewy Body Dementia of multiple system atrophy). Although PD and related disorders may be seen in individuals under age 40, it is more frequently related to specific genes and may display different clinical features than typical idiopathic PD. Patients must be at high risk for poor outcomes as identified by the Palliative Care Needs Assessment Tool (PC-NAT) 28 modified for PD, which screens patients on the basis of social factors (e.g. presence of caregiver), disease severity, and symptom burden. Caregivers, when present, will be identified by asking the patient: "Could you please tell us the one person who helps you the most with your PD outside of the clinic?" For patients with severe dementia, family caregivers can be self-identified and can be included in the study, even if the patient has communication limitations, in order to obtain data relevant to these highly vulnerable and underrepresented patients. Notably, our primary patient outcome measure has been used in PD patients with dementia and can be used with proxy caregiver reporting with good intraclass correlations. 31 Exclusion Criteria: Patients will be excluded if any of the following are present: 1) Immediate and urgent palliative care needs (these patients will not be randomized and will be offered appropriate services immediately); 2) Unable or unwilling to commit to study procedures including randomization, study visits, or addition of neurologist if in the standard care arm if not currently seeing a neurologist; 3) Presence of additional chronic medical illnesses that require palliative services (e.g. metastatic cancer); or 4) Already receiving palliative care and/or hospice. We have purposefully kept our inclusion/exclusion criteria broad to allow for greater generalizability of results and to ensure inclusion of potentially under-represented subgroups.
Potential participants will be first contacted by phone where they will be informed of the basic purpose of the study, their potential involvement, and inclusion criteria. Participants will undergo a scripted phone screening that includes queries on source of referral (e.g. from university or community physician, support group talk, website) and, if excluded at this stage, reasons for exclusion. Following phone screening, potential participants will proceed to an inperson visit for purposes of consent, further screening, and baseline data collection. Reasons for exclusion, including potential participants' reasons for declining to participate even if eligible, will be recorded as part of this visit.
Participants will be randomized in a 1:1 ratio to either the palliative or standard care group through a randomized block design stratified by site, presence of dementia (using standard criteria for PD-related dementia 32 ), and presence of a caregiver (patients can be enrolled without caregivers). Due to the nature of this study, participants and clinicians cannot be blinded to group assignment. We chose not to add to study expense with blinded assessment as our primary outcomes are self-administered patient-reported outcomes.

C.4. Interventions and Comparators or Controls:
Comparator: Our standard care arm will consist of care from the patient's primary care physician (PCP) and a neurologist. We consider this the current standard on the basis of recent evidence showing that PD care with a neurologist results in better outcomes (e.g. falls, mortality) than care with a PCP alone. 33 Patients not currently seeing a neurologist randomized to this arm will have appointments with a neurologist covered by their insurance or patient assistance programs arranged by the study team. These neurologists are board-certified and consist of either academic movement disorders specialists associated with the primary site (but not doing palliative care) or wellestablished community neurologists, many of whom also had completed movement disorders fellowships. Patients will be asked to see their neurologist at least every three months to match the frequency of visits in the intervention arm and saw their PCP at their typical frequency.
Intervention Arm: Our intervention will consist of Standard Care plus the addition of an outpatient interdisciplinary palliative care team. Palliative care visits will be performed in person at each site's academic university clinic (or telemedicine if needed for rural areas and/or limited transportation) every three months. Visits will last approximately 2 to 2.5 hours broken into one-on-one visits with the neurologist (60 minutes for new and 30 for returns) and 30minute visits with all other team members (RN, social worker, chaplain). Providers will communicate in-person or via notes during the clinic visit to increase interdisciplinary coordination and meet at the end of the day for discussion including follow-up plans. In addition, visits will be supplemented by phone calls one week after each visit to confirm understanding of team recommendations and at six weeks to check-in with patient and caregiver. Patients and caregivers can also contact the team as needed. Summaries of visits will be sent to the patient, caregiver, PCP, and neurologist and suggestions for care outside of palliative care issues will be left to the patients' Standard Care team.
The interdisciplinary team will consist of a neurologist with palliative care experience (consisting of direct experience, mentorship, and reading but without formal palliative medicine fellowship), a nurse, social worker, and a chaplain. Of note, the palliative care neurologist will play a consultative role for nonpalliative issues (e.g. PD medications) and will instead make recommendations to the patient's primary neurologist. A board-certified palliative care physician will be available at each site for coaching, phone, or in-person consultations at the discretion of the primary team and for periodic review of charts (after every 10 participants).
Visits will be standardized through the use of checklists (see Appendix 2) which will be used to guide all team member visits to ensure consistency across sites as summarized in Table 1. While these checklists will not be directly accessible in the medical record, they will be used to craft documentation related to study visits. All staff except the palliative care physician will be present for in-person visits and participate in summary meetings at the end of each clinic to ensure interdisciplinary care and completion of any needed referrals or medication changes. psychotherapy, physical therapy, etc.). In the intervention arm we additionally will assess fidelity with our standardized visit checklists and completion of study related phone calls to participants.

C.5. Study Outcomes:
Our co-primary outcomes were change in patient QOL, measured using the QOL-Alzheimer's disease (QOL-AD) scale and change in caregiver burden, measured using the Zarit Burden Inventory (ZBI), between the intervention and standard care groups at six months. The primary goal of palliative care in general, and our intervention specifically, is to improve patient QOL and to reduce patient suffering and caregiver distress. QOL refers to a persons' self-assessment of overall well-being including emotional, social, physical, and spiritual or existential aspects and is one of the cornerstones of the movement towards value-based medicine of which PCORI is a key proponent. 34,35 Patient QOL is clearly important to patients and is recognized by other stakeholders including clinicians, healthcare organizations, and insurance as the single most important patient-reported outcome, in some cases overriding even survival. 34 Caregiver burden refers to adverse effects of caregiving on the caregiver's overall wellbeing and functioning. 36 Caregiver burden is meaningful to caregivers not only as a measure of distress but also a marker for increased mortality and other health issues. 10 Caregiver burden also impacts quality of care for patients, likelihood of nursing home placement and, like QOL, is being increasingly recognized as an important value indicator for interventions among chronic progressive illnesses. 37 Patient QOL and caregiver burden are thus highly relevant and appropriate outcomes for this comparative effectiveness study and should provide relevant data to inform future research, stakeholder decisions, implementation, and dissemination efforts.
The QOL-AD was chosen for this study for several reasons including its brevity, validation in dementia including PD dementia, validated proxy reporting, sensitivity to change, and coverage of issues identified to be relevant to PD patients and caregivers in qualitative interviews, including with our patient advisory council. 24,[38][39][40][41] While the QOL-AD has been shown to be responsive to disease progression and the effect of interventions, the minimal clinically important difference (MCID) has not been defined for this instrument. 42 The ZBI is the most commonly used self-report measure of caregiver distress, including in PD. 9    intervention as a reflection of our underlying model of palliative care for a chronic illness whose progression is typically measured in years. 58 We chose the six month time point for our primary outcome as a pragmatic balance of several factors including time needed to build trust and rapport between a new clinical team and patients; our clinical experience and input from patients through quality improvement projects of when both patients and caregivers might experience benefit; a balance between accumulating benefits of the intervention and clinical progression; and a balance between accumulating benefits and study drop-outs. We chose quarterly follow-ups for our intervention as a reflection of current practice patterns in our clinical practices, and other groups who offer outpatient palliative care for PD and chose quarterly outcome assessment to track changes associated with these visits.

C.8. Data Collection and Sources:
For participants in the intervention group, we combined data collection visits with clinical care to minimize transportation and time burden. If this was not possible for participants' schedules or if they missed clinical intervention appointments, our coordinators reached out to participants to set up data collection. To minimize missing data and increase participation we offered multiple means of completing study outcomes including inperson visits, online data completion, mailed forms, telephone, and telemedicine visits. These could be combined (e.g. participant could do online data entry for most forms but do telemedicine for clinician administered portions) and, in recognition of the already high time burden for many patients and caregivers in medical-related tasks, we prioritized outcome measures to allow participants to do abbreviated reporting as they were able, including collecting only our co-primary outcomes. For participants in the standard care group, we followed the same procedure including scheduling data collection around clinic appointments if they were receiving neurology, primary, or other care at our academic facilities.
If participants elected to withdraw from the study we collected their reasons for withdrawal. If participants missed or were out of window for study procedures, attempts were made to contact participants to determine the reason(s) for missing study procedures. If participants could not be reached through multiple means (e.g. phone, email) over a reasonable timeframe and with multiple attempts, they were deemed simply lost to follow-up.

C.9.a. AIM 1:
The main objective of the study was to determine the comparative effectiveness of interdisciplinary outpatient palliative care versus standard care with a neurologist for PD regarding effects on patient QOL and caregiver burden. Participants were randomized in a 1:1 ratio to either the palliative or standard care group through a randomized block design stratified by site, presence of dementia (using standard criteria for PD-related dementia 32 ) and presence of a caregiver (patients did not require a caregiver for inclusion). This randomization strategy was chosen to minimize potential confounding effects in this relatively small sample based on time of study entry, site differences, and clinical characteristics expected to influence outcome of disease severity and dementia. The primary outcome variables were the QOL-AD scale and the Zarit Burden Interview (ZBI). 50,59 Secondary outcomes (see Table 2  Generalized mixed models account for repeated measures. We performed subgroup analyses for patients with dementia, severe disease, and depression and may perform exploratory subgroup analyses depending on primary results. Missing data approaches are described below. Since the study is relatively small for and lacks the power to investigate effect heterogeneities with precision, Aim 2 was largely exploratory and descriptive unless very large effect sizes were Initial coding was done independently by our qualitative researcher and her research assistant who then discussed codes, established inter-code reliability, and created an initial master code list. Drs. Kluger and/or Kutner were available for adjudication if coding differences persisted.

C.9.b. Aim 2 (Covers Heterogeneity of Treatment Effects -HTE):
The code list was revisited and revised with continued data collection and with input from the multidisciplinary team (Scientific Advisory Board). Text within and between codes was compared to develop themes. Tables were developed displaying counts of codes to search for patterns, similarities, and differences between caregivers and patients. Secondary analyses developed similar tables to determine whether there were potential differences in needs based on age, gender, race, disease stage, or cognitive status. Through this process we developed themes which were modified based on feedback from our multidisciplinary team of advisors and potentially through feedback at presentations prior to publication. Observer triangulation (using interdisciplinary team, stakeholders and patient advisors), participant triangulation (comparing caregiver and patient perspectives), and member checking (eliciting feedback on themes from subsamples of participants) were employed to increase validity.

C.9.d.1. Methods to prevent and monitor missing data (MD-1)
We took the following steps to prevent and monitor missing data: 1) Study participants were presented with full details of the study and their involvement at the beginning of the trial to avoid false expectations from participation in this study and screen out participants unlikely to be compliant with study visits.
2) Scales were collected in the same order at all visits with primary outcomes performed first to minimize data loss or inaccuracy due to fatigue.
3) Total duration of outcome assessments was 60-90 minutes per visit with allowance for breaks to minimize participant burden. to allow greater flexibility for participants regarding travel.
6) All study packets had a checklist on the front page to minimize missing forms.

7)
Coordinators reviewed all forms with patient/caregivers during study visits to minimize the chance of missing pages or items.

8) Schedules of study visits were provided to participants at their first visit and reminder calls
and emails were provided one week and one day prior to visits. 9) Database was set up with strict limits on values and we replicated study forms to minimize inaccurate entry and ensure consistency across sites.
10) Database was set up to flag missing data in real time.
11) Data was entered within one week of study visits to allow for timely follow-up of missing items if detected.
12) Data was verified by a research assistant independent from staff entering data to minimize missing or inaccurate entries.
The strategy was successful and the dropout rate in patients at 6months was approximately 10% (less than the rate used to determine sample size and power.

C.9.d.2. Statistical methods to handle missing data (MD-2)
Item non-response for the scales was dealt with according to the instructions of the validated scales or using the half rule. The half rule refers imputing missing items using the mean of the answered item when at least half of the items are answered. 57 Descriptive statistics have been compiled for missing scales and assessed for patterns. Logistic regression for missingness were used to demonstrate its dependency on variables. The initial analysis used the maximum likelihood for all available data for a response variable when data exists for all explanatory variables. In mixed models correlation on repeated outcomes provides some information on the times with missing outcome values. The data will be assumed missing at random. As a sensitivity analysis, joint models combined the longitudinal outcomes of patient and caregiver proxy reports on the PDQ-39 and used their correlations to provide additional information when a subset of the outcomes are missing. Finally, joint models for the longitudinal QOL scale outcomes and the time to drop out were fit to incorporate the drop out information. Even if the outcomes for one scale were not missing at random conditional on the observed data for the scale, measurements on other scales at the time, or time to drop out, could serve as proxies.
Models for different outcomes and time to drop out were joined through shared covariance structures and/or shared random effects.

C.9.d.3. Plans to use validated methods to deal with missing data that properly account for statistical uncertainty due to missingness (MD-3)
The half rule for scales is a standard method for imputing missing items for a scale when the items are not hierarchical. The scales we used had already been validated, and when available, we uses scale-specific missing data imputation rules. Mixed models by their nature utilize within subject correlation data, even if some participants are missing data for some time points.
Joint models connected through random effects is an established method for incorporating missingness including linkage of patient and caregiver proxy data as well as related scales within a participant. 57

C.9.d.4. Recording and reporting all reasons for dropout and missing data, and account for all patients in reports (MD-4)
See Section C.3 for inclusion and exclusion criteria. These criteria were chosen to maximize inclusion and generalizability while maintaining sound design and interpretability of results. We tracked study participants from the time of initial contact and record screening failures (including reasons for screen failure), as well as dropouts. Dropout data included obtainable specific reasons for dropping out, who decided patient would dropout, and whether dropout involved all or only some activities. This data has been reported using standard CONSORT diagram showing flow of participants from initial contact through study completion. 65 Missing data are recorded and methods to handle missing data, including dropping forms, items, or subjects and/or use of imputation methods are reported for analyses in manuscripts as are rates of missingness.

C.9.d.5. Examine sensitivity of inferences to missing data methods and assumptions, and incorporate into interpretation (MD-5)
The covariances and random effects connecting the joint models were tested for statistical significance. The joint model method is sensitive to the structure of the model. Sensitivity analyses compared the results for the different approaches. All methods assumed the data is only missing at random. Missingness conditioned on available data is not dependent on values of the missing variables. Another possible sensitivity analysis is to compare with results from a complete case model, but patients with complete data may differ from patients with incomplete data while the complete case model assumes missing completely at random and will have a smaller sample size. The missing data methodology anticipates two patterns of missingness: intermittent missingness for some scales, but not others, and permanent loss to follow up. The half rule assumes the data for the missing items is missing completely at random conditional on the non-missing items in the scale, but the scales (or subscales) have been validated for consistency. As assumptions for missing data methods are usually difficult, if not impossible to verify, we will plan to include a summary of methods used and potential influences on data in the discussion section of any manuscripts generated from this study. 3) Intervention Protocol: Interim phone calls were made optional one year through the study due to feedback from patients, caregivers, and the clinical team that they were often not clinically productive and were often perceived as being a greater burden than benefit to participants. 4) Analysis: Linear trajectories over time were not seemingly a good fit. Instead time was treated as a categorical variable, and unrestricted time trajectories were fit, except for imposing a common baseline between the treatment groups. The repeated measures covariance was modeled with unstructured covariance matrices, with different matrices for the different treatment groups. Satterthwaite degrees of freedom were used. The models allow for maximum flexibility based on the data, and compensate for missing data so long as a subject has some measurements included in the model and the data is not missing "not at random." The common baseline constraint was removed if there was evidence of a baseline treatment group differences. There was no evidence against the baseline assumption for most outcomes, except for QOL-AD (caregiver perspective) and UPDRS III.
The baseline covariates for the patient models were updated to include variables projected to influence QOL and caregiver burden: site, gender, age, disease duration, presence of caregiver, MOCA score (baseline), Hoehn and Yahr (baseline) (< 3 vs >= 3). Depression and UPDRS motor score were dropped from the covariates as we felt that depression could be an important effect modifier and that disease severity might be more accurately and simply captured through overall Hoehn and Yahr disease stage. In looking at patterns of missingness we found the following variables were significantly associated with missingness and were added as covariates to our adjusted model to reduce disruptions to the assumption of data missing at random: race (Caucasian vs Not Caucasian), Married (yes vs no), Education (less than college degree vs college degree or more).
The baseline covariates for caregivers included: site, caregiver gender, caregiver age, caregiver sharing the same residence as patient (yes vs no), duration of caregiving, patient MOCA score (baseline), patient Hoehn and Yahr (baseline) (< 3 vs >= 3), caregiver race (Caucasian vs Not Caucasian), caregiver married (yes vs no), caregiver education (less than college degree vs college degree or more).
The effect of the baseline covariates were allowed to vary with time point, but they were not effect modifiers for the differences between treatments groups unless specified. Race and marriage were included as covariates partly because of their association with discontinuation.
Education was a proxy for socio-economic status, but with less missingness than income, and its categories were more collapsible. Attempted remedies included both jointly modeling QOL-AD for patients and caregivers and using multiple imputation to fill in the missing data. From the joint model results, composite scores were constructed from weighted combinations of the patient and caregiver perspective.
One weighted patient:caregiver in a 2:1 ratio (as recommended in original publication but without statistical justification) 38 , and another weighted them equally.
Missing covariates at baseline caused considerable data loss in the longitudinal regression models, so missing baseline covariates were filled in with multiple imputation in the sensitivity analyses.
Partly for computational reasons, joint longitudinal/time to discontinuation models, with shared random effects, were impractical to apply to multi time point categorical time models with unstructured repeated measures covariance matrices. Instead multiple imputation was used to fill in missing variables. Baseline covariates aside, missingness was mostly monotonic, Results for imputed data sets were combined and tested using standard methods. Calculations were performed in SAS 9.4, using proc mi and proc mianalyze. For proc mianalyze the effective degrees of freedom input used the denominator degrees of freedom from the un-imputed model results Cross-overs patients, who for medical reasons were switched from controls to palliative care, were accounted for in a sensitivity analyses by assigning to them the same treatment effect corresponding to the amount of time since they started palliative treatment. For example, a decision might be made at the 3-month time point to switch a control patient to palliative care to relieve suffering. The first palliative care may then be administered at 6 months, time = 0 for palliative care treatment. At the 9-month visit the patient would be assigned the palliative care treatment effect for the 3-month visit, since palliative care has been in effect for the patient for three months.

D. Study Management across Sites
Laura Palmer will act as project coordinator across sites and will handle issues relating to budget, regulatory concerns, and conduct of the study in coordination with the PI, Dr. Kluger.
This will include annual site visits to review study related procedures and documentation. All data will be entered into a central electronic HIPAA compliant database (REDCAP) housed and managed at the University of Colorado and maintained by our UCD biostatisticians Drs. Stefan Sillau and Diane Fairclough. Dr. Sillau will routinely check database for accuracy and completeness and notify local PIs and coordinators of any issues. Hard copies of all study related documents will be housed at local sites and kept in locked filing cabinets in locked research offices and kept for a minimum of five years following completion of the study.
Ongoing and frequent coordination and communication between all three sites will occur with ad hoc, as well as regularly scheduled monthly calls and emails. This will allow for quick response to concerns if/as they arise during the study.

1.a. Human Subjects Involvement, Characteristics and Design
Due to the nature of our research questions regarding clinical, psychosocial, and spiritual needs in PD patients and their family caregivers we feel that involvement of human research subjects is necessary and justified. As we will discuss below, the risks of this research to human subjects are minimal and the potential for benefit for the involved human subjects, other people and family members living with PD, and society is significant.
We will recruit PD patients over age 40 diagnosed with probable PD using UK Brain Bank Criteria 30 or atypical parkinsonian disorders (APD) who are identified to have high potential for palliative care needs using the Palliative Care Needs Assessment Tool. 28 When present their family caregivers will also be included. We plan to recruit a total of 210 PD patients and project that at least 80% will have family caregivers who will also be included in this study.
We will select participants using the following inclusion and exclusion criteria described above.
PD patients with dementia (PDD) may be considered a vulnerable population but are necessary to answering our research questions. For all patients with cognitive impairment we will formally assess capacity to consent using the University of California Brief Assessment of Capacity to Consent (UBACC) modified for our specific study design where needed. 67 For patients who fail this screen or if investigators have additional concerns we will require that their guardian participate in the consenting process and sign an informed consent. We will also require that PDD subjects give consent, or at minimum assent if fully informed consent is not possible. Our protocol and consent forms will be approved by our local IRBs before approaching any potential subjects.
Caregivers will be consented and enrolled using a caregiver-specific consent form.
After enrollment, screening, and consent, patients will be randomized in a 1:1 ratio to either standard care or standard care supplemented by an outpatient interdisciplinary clinic. The interdisciplinary team will consist of a neurologist with palliative care experience, a nurse, social worker, and a chaplain. For the standard care arm, patients not currently seeing a neurologist randomized to this arm have appointments with a neurologist covered by their insurance or patient assistance programs arranged by the study team. Patients will be asked to see their neurologist at least every three months to match the frequency of visits in the intervention arm and may see their PCP at their typical frequency. In the intervention arm, palliative care visits will be performed in person (or telemedicine if needed for rural areas and/or limited transportation or health issues) every three months. In addition, visits will be supplemented by phone calls one week after each visit to confirm understanding of team recommendations and at six weeks to check-in with patient and caregiver. Patients and caregivers may also contact the team as needed. Summaries of visits will be sent to the patient, caregiver, PCP, and neurologist and suggestions for care outside of palliative care issues will be left to the patient's Standard Care team.

G.1.b. Sources of Materials
Data will be obtained from subjects non-invasively via interviews, questionnaires, cognitive testing, and neurological examination as described above.
This data will be accessible only to the research team and, if requested, by appropriate regulatory bodies such as the IRB and FDA. All documents generated by this study, including consent, will be stored in a locked filing cabinet in the locked office of the Principal Investigator (PI). Electronic data will be de-identified with the exception of a correlational spreadsheet including demographic information and study IDs for scheduling and safety contact purposes.
This, and all other data, will be stored on a password protected network drive and backed-up on the hard-drive of the PI, also password protected, and kept in the PI's locked office.
Identifying information will not be shared outside of the research team or regulatory bodies (e.g. IRB).

G.1.c. Potential Risks
There are minimal risks involved in the proposed research. There are no known significant risks from the intervention, interviews, or surveys proposed for this study. Experiences from collaborators on this study suggest that most patients and caregivers welcome the opportunity to talk about their experiences, even on these potentially delicate subjects. Our interviewers will be cognizant of signs of emotional distress in participants and will suggest breaks and remind them that participation is voluntary. Cognitive testing and surveys as proposed in this study may induce boredom or restlessness. Any new diagnoses detected during screening (e.g. depression or dementia) will be referred for appropriate treatment, including emergent treatment if indicated. There are no known significant risks to the proposed team-based intervention for addressing palliative care needs. Participants will be reminded of the voluntary nature of this research and informed that we can follow them outside of the intervention or they can completely discontinue the study if they find it excessively burdensome or otherwise troubling.

G.2.a. Recruitment and Informed Consent
Participants will be recruited from the academic medical centers (University of Colorado, University of California San Francisco, University of Alberta), associated clinics, community neurologists, and direct advertisements and contact with patients (e.g. clinicaltrials.gov, community seminars). Protocols will be approved by the relevant regulatory body at each institution before any contact is made with potential participants. We will obtain informed consent before any study related activities are performed. Potential subjects will be specifically told that their participation or lack of participation will not affect the clinical care they receive.
For PD Dementia subjects, special procedures will be in place to involve both guardians and patients in the informed consent process, and both guardians and subjects will be required to give consent in person or by phone if performing visit by telehealth (or assent for PDD subjects unable to give full informed consent). The consenting process will use standard IRB terminology and will include a full description of expected subject involvement, procedures, risks, rights, and benefits. Capacity will be formally assessed in any patients with cognitive impairments and guardian consent and patient assent will be required in these cases. 67

G.2.b. Protections Against Risks
All subjects will be reminded that they can take breaks or discontinue the study at any time if they find it upsetting. PD dementia subjects will be required to give consent (if they have capacity) or assent with caregiver consent to avoid potential manipulation of a vulnerable population. Regarding patient privacy, all research procedures will be performed in a confidential setting. All documents generated by this study, including consent, will be stored in a locked filing cabinet in the locked office of the PI. Electronic data will be de-identified with the exception of a correlational spreadsheet including demographic information and study IDs for scheduling and safety contact purposes. This, and all other data, will be stored on a password protected network drive and backed-up on the hard-drive of the PI, also password protected, and kept in the PI's locked office. Identifying information will not be shared outside of the research team or regulatory bodies (e.g. IRB).

G.3. Potential Benefits of the Proposed Research to Human Subjects and Others
All subjects participating will receive a free neurological examination and history from a movement disorder trained specialist and screening for dementia and depression. These results will be reviewed by a study neurologist and referred for appropriate care if abnormalities are identified. Participants randomized to the standard care arm will have care with a neurologist arranged if they do not already have this in place. Participants randomized to the palliative intervention will have access and care from an interdisciplinary team. This proposal has the potential to further our knowledge of care models in PD and has the potential to influence our ability to address many currently unmet needs for patients and caregivers. As discussed below, the knowledge gained by this proposal has a significant potential to advance our knowledge of palliative needs in PD which has large positive implications for PD patients, the scientific community, and society.

G.4. Importance of Knowledge to be Gained
Parkinson's disease (PD) affects 1-2% of adults over age 65 representing approximately 1.5 million Americans. Although PD is traditionally characterized by motor symptoms such as tremor and slowed movement more recent research shows that PD patients also commonly experience non-motor symptoms such as pain, fatigue, depression and dementia. PD interferes with patients' quality of life, leads to disability, decreases length of life, and causes significant distress for caregivers. Unfortunately, many of the needs most important to patients and their caregivers (e.g. depression, planning for the future) are poorly addressed under current models of care. The goal of this proposal is to take a patient-centered approach to answering these questions. Our specific objectives are to: 1) Determine whether an outpatient palliative care team improves patient quality of life and caregiver distress compared to standard care with a neurologist; 2) Determine what patient and caregiver characteristics best predict benefit from palliative care services; and 3) Interview patients and caregivers to optimize service delivery and selection. This project is important to patients and their family caregivers because it assesses a new approach to PD care that has the potential to improve quality of life, caregiver burden and other important outcomes (e.g. nursing home placement). By testing the effectiveness of this approach and determining who benefits most we hope to provide a new option to assist those PD patients at highest risk for poor outcomes. Information generated by this proposal will also be important for other stakeholders interested in PD including hospitals, insurers, and patient advocacy organizations.

G.5. Data and Safety Monitoring Plan
We will not assemble a Data and Safety Monitoring Board (DSMB) as this is a minimal risk study comparing the effectiveness of two currently available approaches to the care of PD and APD patients. The site PI will be notified of any adverse events occurring at all sites and will notify the study's central IRB (COMIRB) of all adverse events and any change in our appraisal of the risks and benefits of this study from our data or other published literature.
• Are the information needs of the patient and family congruous?

CARER AND FAMILY WELLBEING -"Do the carer or family…" Physical and psychosocial
• Experience physical strain, ill health, fatigue, disturbed sleep? Is there evidence of anxiety, depression or feelings of hopelessness?
• Have spiritual/existential issues that are of concern?
• Currently feel that caring has a net positive or negative affect for them personally and their relationship with the patient? Grief (pre and post death) • Experience intrusive images, severe emotion, denial of implications of loss to self and neglect of necessary adaptive activities at home or work?
•Know of the progressive nature of PD? Has future care planning been considered? (If patient not capacitous this may include best interest decisions) • Have access to support services (Such as PD Nurse Specialist, SPC, Local support groups, post bereavement support?) 49