An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Denise M. Hynes, PhD, MPH, RN, Jose Arruda, MD, Michael Berbaum, PhD, Ifeanyi Chukwudozie, MPH, Michael Fischer, MD, MSPH, Marian Fitzgibbon, PhD, Anna Porter, MD, and Linda Schiffer, MS, MPH.
Author Information and AffiliationsThe patient-centered medical home (PCMH) model has been proposed as a solution to fragmented care for patients requiring complex care. Patients with advanced kidney disease—with their high rates of morbidity, mortality, and health care use—have the potential to benefit from a PCMH model.
To design, implement, and evaluate the effect of an adaptation of a PCMH model for patients receiving chronic hemodialysis for end-stage renal disease. The design was informed by patients, caregivers, and clinical stakeholders.
We implemented a PCMH for kidney disease at 2 dialysis centers over 18 months. To the standard hemodialysis team (comprising a nephrologist, nurse, social worker, dietitian, and dialysis technician), we added a primary care physician, pharmacist, nurse coordinator, and community health worker. The primary outcome was patient-reported kidney disease quality of life (KDQOL). Secondary outcomes included patient-reported self-efficacy and disease knowledge, depression (Patient Health Questionnaire-9 [PHQ-9]), satisfaction with dialysis care (Consumer Assessment of Health Care Providers and System for in Center Hemodialysis [CAHPS-ICH] survey), and primary care assessment (PCAS), clinical laboratory measurements, health care use, and staff perceptions (team vitality score). We used descriptive and orthogonal regression analysis for repeated measures for the primary and secondary outcomes. Sensitivity analyses addressed missing values.
Of 248 eligible patients, 175 (71%) consented and participated; 97% were African American or Hispanic. At 12 and 18 months, the KDQOL mental component score improved from baseline (adjusted mean = 48.9; 95% CI, 47.14-50.73) by 2.5 (95% CI, 0.49-4.50; P = .01) and 2.8 (95% CI, 0.65-4.93; P = .01) points, respectively, adjusting for other factors. The KDQOL kidney disease effects improved from baseline (adjusted mean = 73.0; 95% CI, 69.38-76.53) to 6, 12, and 18 months by 4.3 (95% CI, 1.39-7.15; P = .004), 6.7 (95% CI, 3.63-9.74; P < .001), and 3.9 (95% CI, 0.51-7.29; P = .02), points, respectively, adjusting for other factors. KDQOL physical component summary and symptoms improved from baseline to 6 months only. PHQ-9 scores improved from baseline to 12 months; results for PCAS and CAHPS-ICH domains were mixed. Dialysis adequacy was stable. Inpatient and emergency department use at the primary medical center decreased.
This study is the first application of an adaptation of the PCMH for chronic hemodialysis patients. Patients in the study benefited from improved quality of life, particularly in the mental health and effects of kidney disease domains; improvements also occurred in some secondary outcomes. Benefits may need to be weighed against the relative costs of implementing an expanded care team. The findings should inform health care reorganization efforts aimed at improving outcomes for patients with complex chronic diseases.
As a before–after design and limited to 2 intervention sites with 1 group of academically based nephrologists, we cannot say with confidence that these effects were due to the intervention alone. With predominantly African American and Hispanic participants, results suggest that this model may be useful for a racially and ethnically diverse population. Information learned from this study can inform future study designs that include randomization at the site level with more than 1 clinical team and near–real time linkages with health care claims data.
About 400 000 people in the United States receive chronic hemodialysis to treat their end-stage renal disease (ESRD) each year.1 Even with hemodialysis treatment, ESRD patients experience substantial morbidity, mortality, hospitalizations, and health care costs.2-4 ESRD patients incur an average of 12 days of inpatient care per year and have annual mortality rates exceeding 150 deaths per 1000 patient years.1
In addition to prevention efforts, approaches are needed to reduce fragmented health care and the economic burden on ESRD patients and caregivers. Such patients have complex care needs associated with renal disease and common comorbidities (eg, diabetes) and receive fragmented care from multiple providers at multiple locations. ESRD patients undergo hemodialysis 3 days a week. The Centers for Medicare & Medicaid Services (Medicare) sets the requirements for the hemodialysis care team at dialysis facilities, which comprise a nephrologist, nurse, social worker, and dietitian. Scheduling and traveling to other appointments are difficult to manage, increase patients' burden, and reduce quality of life.
These challenges keep many ESRD patients from receiving care for other conditions outside of the dialysis setting, resulting in high rates of complications and emergency health care use.5
Over the past 15 years, the health care community has come to recognize the importance of care coordination for patients with chronic disease.6,7 Recent attempts to address the gaps in care coordination have focused on implementation of the patient-centered medical home (PCMH) model, as noted in previous literature reviews and studies.8 PCMH uses a team approach to provide comprehensive care for patients; has been implemented for patients with chronic complex illnesses such as diabetes; and has been found to reduce hospitalizations, emergency department visits, and health care costs.9-11 For example, among patients with chronic kidney disease not yet requiring dialysis, the use of a multidisciplinary care team, a key element of PCMH, reduced the rate of kidney function decline.12 However, to date, the implementation of a PCMH has not been tested among hemodialysis patients in the United States. Although the current US dialysis care team is multidisciplinary in its inclusion of a nephrologist, nurse, dietician, and social worker,13 it lacks integration with primary care.5 The current model also does not include other professionals, such as pharmacists, who have been recognized to improve care for other chronic illnesses.14,15 Moreover, the current care model does not include nonprofessional team members such as community health workers (CHWs) functioning as health promoters, who are individuals without a formal medical background but who receive specialized training as peer educators and liaisons between patients and health care professionals to assist in providing culturally sensitive care. In several studies of PCMH for chronic diseases other than ESRD, CHWs have been shown to improve clinical outcomes and reduce care costs.16-18
This project is the first systematic design, implementation, and evaluation of a PCMH model for kidney disease. For chronic hemodialysis patients we sought to establish a new model of care, as depicted in Figure 1, aimed at improving patient-centered outcomes. The proposed model supplements the current hemodialysis mandated team (ie, nephrologist, nurse, social worker, and dietician) by adding a general internist, pharmacist, nurse coordinator, and health promoter. We sought to address the following aims:
PCMH-KD Model of Care.
We anticipated that by implementing this new model, access to primary care and care coordination would increase. We hypothesized that these impacts would increase patient quality of life. We also anticipated that the intervention could improve patient knowledge and improve patient health by the prevention or early identification of emergent conditions and complications, leading to reduced emergency care and hospitalizations for conditions unrelated to kidney disease. This report provides an overview of the study and key results. Results from the study are also reported in Clinicaltrials.gov (NCT02270515).
We engaged (1) patients and their caregivers and (2) clinical stakeholders to participate in a stakeholder discussion group process. Our approach was informed by prior research that noted the importance of community-based participation to ensure a process for disseminating ideas and results back to the community of interest.19 Also, understanding the severity of symptoms and the progressive nature of ESRD,5 we were mindful to minimize burden on patients while seeking their input on the planning and the conduct of the study. Understanding the high proportion of African American and Spanish-speaking patients at our sites, our approach considered cultural and language support aspects in our staffing and procedures. With clinical stakeholders, we sought to gauge the interests and feasibility of an expanded care team within each dialysis center, so we focused on clinicians who were on staff at each study site. Each stakeholder group was involved during the preintervention design and intervention phase of the study.
The patient and caregiver stakeholder discussion group meetings comprised patients with ESRD receiving hemodialysis at 1 of 2 outpatient dialysis facilities and their caregiver/family member. Participants were invited based on the dialysis nurse coordinator recommendation of patients' reported health status and/or those who volunteered and were able to attend. We also invited 1 participant, who was identified from a list of patients recommended through discussions with the clinical team, to serve as a patient representative. The number of participants varied quarterly, ranging from 4 to 7 participants per session. Patient and caregiver stakeholder group sessions were a facilitated discussion, led by a trained facilitator before study commencement and during the study, about 3 to 4 times per year. The patient representative met with the study coordinator on an ad hoc basis and attended the stakeholder discussion groups. Patient and caregiver stakeholder discussion group participants shared experiences during the moderated discussions. Our patient representative additionally reviewed study protocol documents, discussed them with the project coordinator, and suggested format changes for the discussion group, to include feedback to dialysis center management. During the intervention and at the request of the participants, we modified the format of discussion groups to obtain feedback to report to dialysis center management.
During the preintervention phase we also sought input from CHWs. We invited experienced CHWs who had been involved in prior research projects on diabetes and cancer care at the University of Illinois to participate in a focus group format. Four CHWs agreed to participate and provided feedback on the PCMH-KD model and more specifically on the support that CHWs could provide to dialysis patient and their family members/caregivers. The roles and responsibilities of the health promoters in this project were adjusted, based on the feedback from this group.
In addition, we engaged clinical stakeholders experienced in dialysis care. During the preintervention design phase, the stakeholder discussion groups focused on refining the roles and responsibilities of the new clinical team members who would compose the PCMH-KD intervention team relative to the Centers for Medicare & Medicaid Services (CMS)-mandated hemodialysis care team. Participants included the current dialysis center nephrologists, study coinvestigators in nephrology and general medicine, dialysis nurse coordinators, and social workers. During the conduct of the study, the CHWs and the pharmacists were included in the discussion groups whose focus was on reinforcing and/or modifying clinician roles and responsibilities as well as clinical logistics to support the intervention. We discuss further details about the methods and the impact of stakeholder engagement efforts in Methods section below.
We considered a randomized controlled trial with randomization at the patient and site levels. From our preliminary work, we determined that patients in the dialysis center communicated frequently with each other, so a randomization model that selected some patients as treatment and others as control within the same center would not be practical. Randomization of sites was also impractical due to the need to hire the study providers and pay for them on the research budget. Moreover, since there had been in place an existing model of care mandated by Medicare at all dialysis centers, this model seemed more practical to use. The model was used as a comparison in a design that allowed us to collect information from the period of time when patients were exposed to the CMS-mandated model of care and then enroll in our study to be exposed to the new PCMH-KD care model in a controlled setting. Thus, using this quasi-experimental design—specifically, a before–after intervention study in which we collected baseline data from participants before the intervention—we sought to implement and evaluate the effect of the PCMH-KD model compared with the current Medicare-mandated model of care. Comparisons were within patients over time with repeated measures; baseline assessments for patient-reported primary and secondary outcomes before enrollment and 6-month retrospective data on health care use and laboratory data served as the comparator.
We sought to establish processes for patient and stakeholder input to develop and refine the care model, intervention procedures, and PCMH-KD training curricula. We used the start-up phase (year 1) for patient and clinical stakeholder engagement, project planning, clinician recruitment, and training of all participating clinicians and staff. The clinical intervention phase began in the second year.
We sought to address a primary research question (RQ) and several exploratory questions:
Throughout the report, we annotate methods and results that address these questions for easy reference.
The settings for the study were 2 urban dialysis centers affiliated with 1 academically based nephrology group. One site was a nonprofit, university-affiliated outpatient dialysis unit based at the University of Illinois Hospital and Health Sciences System (UIHS), and received a 3-star rating (out of 5) in Medicare Dialysis Compare for 2012-2015.20 The second was a for-profit, free-standing outpatient dialysis unit owned and operated by Fresenius Medical Care (FMC) at the Chicago Westside Dialysis Center, and received a 1-star rating in Medicare Dialysis Compare in 2012-2015. Eight nephrologists from UIHS composed the medical staff at both the UIHS and FMC units. Capacity at the 2 sites together for hemodialysis was 200 patients, with turnover of about 25% per year.
The patient population at the participating dialysis centers reflected the population in the centers' surrounding service area on the west side of Chicago and has a higher proportion of minority and low-income patients than the national average as well as a high burden of comorbid illness (Table 1). Patient eligibility criteria for the PCMH-KD clinical intervention required participants to be (1) fluent in English or Spanish language, (2) currently receiving maintenance hemodialysis at 1 of the 2 participating dialysis centers (UIHS or FMC), (3) aged 18 or older, and (4) able to provide informed consent for participation in the study. We excluded patients if they did not speak English or Spanish, were younger than 18 years, or were unable to give informed consent. Patients who left the participating dialysis center or who received a kidney transplant were no longer able to continue in the study. We recruited and enrolled patients over a 12-month period. Both sites had high percentages of African American and Hispanic patients compared with the national rates (See Table 1). The percentage of African Americans was higher at the UIHS site, whereas the percentage of Hispanic patients was higher at FMC.
Patient Population Characteristics at the Study Sites.
We held informational sessions about the study at each site. Recruitment was on a rolling basis (December 2014 through April 2016); the PCMH-KD intervention began in January 2015, with patients entering as they completed the baseline assessment, and continued for as long as the patient was eligible through August 2016. Patients who provided informed consent and enrolled in the study were offered the additional services of the PCMH-KD team. After patients reviewed information about the study, a research staff member approached them individually during their dialysis treatments to answer any additional questions and, if they were willing to participate, to obtain informed consent. Patients who did not wish to participate in the study were not eligible to schedule individual visits with the general internists or CHWs.
To assess quality of life, we used the Kidney Disease Quality of Life (KDQOL-36) questionnaire, a quality of life instrument that assesses both general mental and physical health quality of life domains as well as specific kidney disease and dialysis-related domains of quality of life.21 The 5 subscales of KDQOL-36 include mental component summary (MCS), physical component summary (PCS), burden of kidney disease (burden), effects of kidney disease (KDE), and symptoms and problems of kidney disease (symptoms). The MCS and PCS were derived from the Medical Outcomes Study Short Form 12, a generic health-related quality of life survey instrument that can be used in healthy individuals and across all disease states.21,22 While other versions of the KDQOL have been used over time, the KDQOL-36 has been widely used by Medicare as a performance measure for dialysis centers.3
In addition to the primary patient-reported outcome measure for quality of life, we also assessed additional patient-reported outcomes, clinical outcomes, health care use, and staff perceptions. Table 2 lists the measures included and the frequency of data collection. We obtained patient-reported information using structured interviews by a member of the research team. Demographic, medical history, and social characteristics were part of the initial intake. We used validated items and/or instruments, including a modified version of the Primary Care Assessment Survey (PCAS),23 the Consumer Assessment of Health Care Providers and System for in Center Hemodialysis (CAHPS-ICH) survey,24 the Self-efficacy for Managing Chronic Disease instrument,25 health literacy,26 the Chronic Hemodialysis Knowledge Survey (CHeKS),27 the Patient Health Questionnaire-9 (PHQ-9) for depression screening,28 the Morisky Medication Adherence instrument,29 the Renal Adherence Attitudes Questionnaire (RAAQ), and the Renal Adherence Behavior Questionnaire (RABQ).30
PCMH-KD Intervention Intake, Measures, and Schedule.
Follow-up interviews comprised the same assessments as the baseline interview except for health literacy, comorbidities, and dialysis history, and were conducted by trained interviewers at 6, 12, and 18 months. Each interview was about 60 to 90 minutes long and was conducted in either English or Spanish as per patient preference. Interviews took place in the dialysis center before, during, or after a patient's dialysis appointment and were recorded via live web-based data entry using Research Data Capture tools hosted at University of Illinois at Chicago on an Apple iPad 2 tablet.31 Patients were initially compensated for their participation at $10 per interview; this amount increased to $20 per interview in June 2016.
We assessed clinical measures for the period 6 months before patient enrollment through 6 months after the close of the study. Measures focused primarily on routine laboratory measurements related to blood glucose management, measured by blood levels of hemoglobin (Hgb) A1c; anemia management, measured by blood Hgb; nutrition status, measured by serum albumin levels; bone and mineral metabolism, measured by blood levels of calcium and phosphate; urea reduction ratio (URR); and kt/V (a dimensionless quantity used to assess dialysis adequacy). We also recorded average interdialytic weight gain, predialysis and postdialysis weight, and predialysis and postdialysis systolic and diastolic blood pressure from the dialysis center records for the first week of each calendar month. We used a combination of methods to collect these data—including a data extraction from each dialysis center's electronic records, provided securely in a series of batches to the research team. In addition, for some laboratory values we abstracted dialysis records for the dialysis-specific measures not found in electronic records (eg, predialysis and postdialysis blood pressures and weights).
We assessed health care use for the period 6 months before patient enrollment through 6 months after the close of the study. Measures focused on individual appointments with the members of the PCMH team and use of emergency department and inpatient admissions at the University of Illinois Hospital and Health System (UIHS). We collected data separately about patient visits with the general internists and CHWs. We abstracted emergency department visits and inpatient admissions data from documentation in the dialysis center and University of Illinois Health and Hospital Systems (UIHealth) electronic medical records, and from extracts from the UIHealth Center for Clinical and Translational Science Center Clinical Research Data Warehouse transmitted to the research team via secure data transfer.
All research procedures were reviewed and approved by the University of Illinois at Chicago Office of Protection of Research Subjects.
This phase included the planning period (pre-award) and the first year of the project. The pre-award period focused on obtaining input from dialysis patients and their caregivers about ways in which their dialysis care could be improved and about their opinions on the current project concept, including the proposed care model. We held 2 discussion groups in April and November 2012. The focus was on how patients felt their dialysis care could be modified to improve their well-being and that of their caregivers and family members, and on the proposed model of care and roles of new providers. Feedback addressed the aspects of their illness and the treatment that were difficult to manage, and the value and roles of potential new providers to include in the care model (Exhibits 1 and 2).
In October 2012, we convened a focus group comprising 4 experienced CHWs to gather feedback on the PCMH-KD model and, more specifically, on the support that CHWs could provide to dialysis patients and their family members/caregivers. We adjusted the roles and responsibilities of the health promoters in this project based on the feedback from this group.
In addition to patient and caregiver stakeholder input, we also sought input from clinical stakeholders. Notably the first year of the project focused on refining the roles and responsibilities of the new clinical team members who would compose the PCMH-KD intervention team relative to the CMS-mandated hemodialysis care team (Table 3). We developed the specification of these roles through an iterative process of discussions with the clinical stakeholders, including the current dialysis center nephrologists and study coinvestigators in nephrology and general medicine. We also considered input from the new PCMH-KD team members until consensus was reached. Discussions took place during the planning of the intervention, beginning in April 2012, in order to hire the appropriate clinical staff, and through December 2015. We held approximately 3 meetings during this preintervention period.
Roles of the Care Providers in the Usual Care and PCMH-KD Models.
The evolution of the general internist as a primary care physician (PCP) role was one example of how we incorporated the input from the new PCMH-KD members. Initially, we envisioned 1 half-time general internist and 1 half-time advanced practice nurse (master's level, certified clinical specialist or adult nurse practitioner) for the team. However, it was challenging to recruit an advanced practice nurse with sufficient expertise and interest to devote to the study. In addition, interviewees raised concern regarding general internal medicine coverage for vacation and leave. In response, we reformulated the team to comprise 2 general internists at half-time (0.5 full-time effort [FTE]) to co-lead the clinical team with the nephrologist. The general internists each spent the other 0.5 FTE in clinical responsibilities in the Division of General Internal Medicine. We modified the nurse coordinator role to be a generalist (BSN level) with more limited care coordination responsibilities.
The general internists functioned as each participant's PCP, providing management of comorbid medical conditions, preventive care, and coordination of subspecialty care, at the patient's request. A separate examination room was set up at each site in or near the dialysis center for the PCMH-KD PCP scheduled visits. The PCMH-KD PCPs completed patient intakes upon request; provided follow-up care; performed monthly patient rounds; attended monthly laboratory reviews with the dialysis center staff; documented care provided in the electronic medical record; and communicated with other clinical teams through flow sheets, email, and in-person meetings. For patients who indicated they had someone they considered a regular personal doctor (60% at the time of enrollment on the PCAS), the PCMH-KD PCPs remained available and participated in all activities except individual patient visits unless requested. In some cases, patients who indicated they had another established PCP also requested individual visits with our PCMH-KD PCP and communicated with other PCPs when requested by patients or on an as-needed basis as determined by the clinician. The PCMH-KD PCPs attended clinical research planning meetings.
One 0.25 FTE nurse coordinator coordinated care, monitored inpatient care, provided patient education, scheduled vascular access procedures with surgery and radiology, and monitored vascular access sites. Two 0.25 FTE pharmacists at each site monitored medication dosing, safety, compliance, and delivery. The nurse coordinator and pharmacists completed patient follow-up as needed, rounded on patients twice monthly, participated in laboratory reviews and clinical research planning meetings, and documented as needed in the electronic medical record and workflow sheets. Two CHWs, at 0.50 to 0.75 FTE, acted as liaisons between the patient, family, community, and care team. They helped overcome barriers of language, culture, and literacy, ensuring culturally competent care. One CHW was bilingual (English and Spanish) and the other was African American, and both had a bachelor's degree and prior experience as a CHW. The CHWs also provided logistical coordination necessary for patient adherence to care plans, such as transportation; completed initial intake of consented patients; checked in with patients monthly or as needed; attended clinical research planning meetings; and documented services provided in the electronic medical record and template reports.
The training for all PCMH-KD team members and dialysis center staff included introduction to the new intervention team and the aims of the research study. The Midwest Latino Health Research, Training, and Policy Center helped develop the training modules specifically for the clinical team. Dialysis center staff completed in-service training to familiarize themselves with the completely new clinical team of the intervention. The training covered topics related to the research project itself, such as timeline, key dates, new providers and their roles and responsibilities in the dialysis clinic, the new PCMH-KD model of care, and enhanced care coordination under this model. PCMH-KD clinical team members—in particular, the newly added providers—had undergone extensive training as part of the preparations for the project. Training was broken down into 6 separate sessions and covered topics specific to kidney disease and dialysis: (1) kidney disease and dialysis basics; (2) medication management for dialysis patients; (3) nutrition management; (4) social services for dialysis patients; (5) operation of the University of Illinois Health and Hospital Systems dialysis unit; and (6) operation of the FMC dialysis unit. Trainings were delivered by qualified clinicians who formed part of the dialysis care teams, and all trainings were recorded as webinars and made accessible online for future training needs.
During the second year, patient recruitment for the PCMH-KD intervention began on a rolling basis. Patients under care at each dialysis center were recruited and enrolled. A baseline assessment included patient-reported outcomes, clinical outcomes, and health care use. Once baseline assessment was completed for each patient, the intervention began for that patient (commencing first in January 2015). As patients were recruited and enrolled through April 2016, this process was repeated for each patient. The recruitment phase continued through April 2016, and the intervention ended for all patients in August 2016.
These groups were held at each center throughout the intervention and were a critical mechanism for obtaining patient and caregiver feedback and perceptions about the intervention. The 2 discussion groups held in the preintervention phase informed project planning and helped shape the roles and responsibilities of the PCMH-KD team members. These discussion groups continued during the intervention from March 2015 to August 2016. Patients enrolled in the study were invited to participate in discussion groups with a family member or a caregiver. We used a facilitator's guide to aid discussion, and we selected questions for each group discussion to reflect study changes, gather patients' and caregivers' perspectives and responses, and respond to emerging issues. After each discussion group, research staff generated notes to allow for rapid response to patients' and caregivers' expressed needs.
Twelve discussion groups were held and recorded and transcribed; 3 were in Spanish only.
The clinical stakeholder discussions initiated during the start-up phase transitioned to a clinical research planning meeting held about bimonthly during the Intervention phase. The clinical research planning meetings included the principal investigator (PI), research project coordinator, clinical coinvestigators, the new PCMH-KD team members, dialysis center managers, and social workers; meetings focused on reinforcing and/or modifying clinician roles and responsibilities as well as clinical logistics to support the intervention.
We used the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework as an organizing and monitoring tool.14,15,32 The 5 domains of RE-AIM are Reach, Effectiveness, Adoption, Implementation, and Maintenance. The domains apply to issues involving setting and participant representativeness, setting/site engagement with intervention, intervention adaptation during the study, program sustainability, and resource costs of an intervention. External validity was included later. The RE-AIM framework has been tested in diverse conditions and study settings.33
Table 4 depicts our key measures and concepts organized within the RE-AIM domains. Regarding Reach, we tracked the number of consented patients and the depth of their participation as per the number of PCMH-KD PCP and CHW visits. For Effectiveness, quality of life, as measured by the KDQOL-36, was the primary outcome measure, and secondary measures included health care use, including emergency department visits and inpatient admissions. For Adoption, we focused on feedback from both patients and staff. Quarterly discussion groups provided an opportunity for patients to give direct feedback regarding the study and provided context in the qualitative analysis. Clinic staff provided feedback at baseline and end of the intervention utilizing the Team Vitality Instrument, which, as noted previously, measures function of interdisciplinary teams34 and served as a quantitative measure of adoption. For Implementation, we monitored patient–clinical team member encounters for those providers who had individual visits with the patients. Specifically, we recorded the number of visits for those providers who had individual visits with the PCMH-KD PCPs and the CHWs. For the CHWs we tracked the themes of their consultations separately from the medical record; for the PCMH-KD PCPs we noted whether the visits were office (more intensive) or chairside (brief) visits. For the PCMH-KD nurse coordinator and pharmacist, they attended rounds only so individual visits were not relevant beyond discussions during rounds. Regarding the fidelity of the intervention, the specific quantitative measures we focused on pertained to the newest providers with individual patient visits and relevant to their specific roles; we measured whether the CHWs had an initial visit with each enrolled patient and whether the PCMH-KD PCPs had any visits with the enrolled patients. Regarding Maintenance, we ascertained the dialysis centers' likelihood of adopting the intervention, through a brief interview by the PI with each dialysis center's leadership.
RE-AIM Framework Measures.
In contrast to a typical clinical trial, this study involved no randomization of patients or clinics: The focus was on improving outcomes and measuring changes within each patient and within each clinic. At the outset of the study we were aware that 1 clinic handled a maximum of 138 patients, and the other 78, for a total of, at most, 216 patients available at a given occasion, and with turnover, conservatively estimated approximately 243 patients available over the course of the study. As the nature of a dialysis center is that new patients may enter and exit the center at different points over time, we used a periodization approach to take into account this patient turnover in our sample size calculations. Additionally, this approach accounted for the fact that patients' responses are highly correlated within themselves but that departing patients and the ones who arrive to take their places in a clinic are not correlated.
We employed the method of Rochon to calculate power and incorporate a cross-time correlation of outcomes rho = 0.70, which is appropriate for chronic disease.35-37 Using our primary outcome measure, KDQOL, we noted that typical ESRD patients score about 48.6 points on the KDQOL scale (SD, ≈11.3). With a goal to raise this figure by 10% to 53.5, which can be clinically meaningful, we calculated power based on a comparison of KDQOL score after 12 months of exposure to the intervention to baseline score (or score at entry for those who entered later). Using SASmacro GEESIZE (version 3.1), under a gaussian generalized estimating equation model for repeated measures, 77 patients total are required to test this hypothesis with power 0.80 at the α = 0.05 2-sided level of significance. Assuming a mild degree of clustering within clinics (intraclass rho = 0.01) and conservatively using the larger clinic size, the design effect is 1.37, yielding a required sample for 0.80 power of 1.37 × 77 = 106. Even if we then conservatively corrected for 10% patient loss, due to leaving the dialysis center, having a transplant, or dying during the study (106/0.90 = 118), ample statistical power remains assured.
Recordings of the patient and stakeholder discussion groups were transcribed, reviewed for specific quotes for context, and analyzed for content themes. We used content analysis, as described by Hsieh and Shannon (2005),38 to describe emerging themes. We used Atlas.TI (version 7.5) for the theme coding.39 One coder performed the theme coding; the PI reviewed the codes and themes for clarifications and refinement; the single coder performed revisions. Analyses were exploratory, and we used selected quotes to provide context for quantitative measures.
Descriptive analyses comprise simple means and standard deviations (continuous) and frequencies (categorical). For comparisons over time we report t tests for differences in means (continuous) and chi-square for differences in percentages (categorical) as appropriate for each measure. Analyses of scores or scales are indicated for each measure/instrument. We used SAS version 9.4 for all analyses.40
We examined change in primary (KDQOL) and secondary measures over time (baseline and 6, 12, and 18 months) and the factors that may have also influenced outcomes. We used a growth curve model to evaluate the changes over time in the outcome measures and combined it with a linear mixed models approach to account for other factors, referred to as an orthogonal polynomial adjusted modeling approach. A detailed description of these models follows.
We began with growth curve models that exhibit the responses as a function of an intercept and time alone. One way to do this is to employ 3 explanatory time variables: tt, tt2, and tt3; however, as these variables can be highly correlated, their corresponding coefficients are difficult to interpret. Therefore, we elected to employ 3 time variables that have very clear interpretations, because they are essentially uncorrelated as they derive from orthogonal polynomials. For the usual powers of time, the tt is represented as 0, 6, 12, and 18, and the tt2 is represented as 0, 36, 144, and 324. For the orthogonal polynomial, ttLLLLLLLLLLLL is represented as −3, −1, 1, and 3, and ttQQQQLLQQLLLLQQLLQQ is represented as 1, −1, −1, and 1. Each set of 4 numbers sums to zero. Moreover, the 2 sets are orthogonal (hence uncorrelated) because their cross-products sum to zero, per the definition of orthogonality: (−3 × 1) + (−1 × −1) + (1 × −1) + (3 × 1) = 0. The third time variable, ttCCQQCCLLQQ, is represented as −1, 3, −3, 1, and it is orthogonal with both ttQQQQLLQQLLLLQQLLQQ and ttCCQQCCLLQQ. The 3 time variables ttLLLLLLLLLLLL, ttQQQQLLQQLLLLQQLLQQ, and ttCCQQCCLLQQ are an orthogonal decomposition of a trend line, and the 3 components can be interpreted as separate additive elements that comprise the whole trend. When these 3 essentially uncorrelated representations of time form the growth curve model of observations on patients taken at 4 time points, their regression coefficients have clear interpretations and properties. For example, if 1 of the 3 variables is omitted, the coefficients and significance levels for the other 2 are not changed. This effect occurs due to the fact 3 time variables are a decomposition for the 4 time points. Moreover, by combining the 3 components, 1 completely recovers the original 4 time points. The 3 components have simple interpretations. Term ttLLLLLLLLLLLL captures a straight line; term ttQQQQLLQQLLLLQQLLQQ captures curvature (a “bowl” opening upward or a “dome” opening downward); and term ttCCQQCCLLQQ captures a bowl connected to a dome or vice versa—that is, 2 changes in direction, as in down-up-down or up-down-up. Quadratic terms often pick up on increasing returns or decreasing returns for each unit increase in the predictor. Cubic terms will reflect an antagonism between competing forces, which alternate in ascendancy over time.
For the primary outcome measure of quality of life, for each of a series of KDQOL components as dependent variables, we fitted the orthogonal polynomial growth curve models described above. Because the data are repeated measures (longitudinal), we included a single random effect on the intercept term that reflects the relative “height” of each patient's trend (4 connected observations) relative to the “heights” of other patients. For a patient whose overall level of response on a given KDQOL component is higher than that of other patients, the random effect raises the average response line to come closer to her level; for a patient whose overall level of response on a given KDQOL component is lower than that of other patients, the random effect lowers the average response line to come closer to his or her level. This is the standard random intercept model for longitudinal data.
We next introduced into the linear mixed model the time-constant covariates. For example, each patient is of a particular sex, and it might be that for a given dependent variable women have a higher average level of response. Thus, including sex as an explanatory variable will pick up on this fact by raising all fitted curves for women somewhat higher than the fitted curves for men. If sex is a statistically significant predictor, it means that this adjustment improved the overall fit of the model by bringing the fitted curves closer to the actual observed data. Each time-constant covariate works in the same fashion: It raises or lowers the fitted curve (combined average effect of ttLLLLLLLLLLLL, ttQQQQLLQQLLLLQQLLQQ, and ttCCQQCCLLQQ) to bring the curves closer to the observed data. Note that the average curve that is being adjusted higher or lower is the same for every patient. The idea is to use the time-constant covariates to bring that curve closer to the patients' observed trend lines—that is, to identify those factors that are significant predictive factors in relation to KDQOL. We also considered the covariance matrix for specific model selection.
We applied the same statistical orthogonal polynomial (adjusted) modeling approach to examine trends over time for secondary measures for self-efficacy (Self-efficacy for Managing Chronic Disease [SEMCD]), hemodialysis knowledge (CHeKS), PCAS, consumer assessment of providers and system in the dialysis centers (CAHPS-ICH), depression (PHQ-9), medication adherence (Morisky Medication), renal adherence attitudes (RAAQ), renal adherence behaviors (RABQ), and clinical measures (laboratory results, dialysis adequacy indicators, and vital signs).
To account for missingness of data in our analysis we applied the method of multiple imputation by chained equations.41 We performed multiple imputation with fully conditional specification using SAS Proc MI. We included in the imputation model all KDQOL scale scores and all covariates in the multivariable models except for the orthogonal time terms, along with visit as a class variable. After imputation, we ran the random-intercept mixed model for each of the 40 imputed data sets, and combined the results using SAS Proc MIAnalyze.
We screened 285 patients for eligibility at our 2 sites from December 2014 to April 2016. Of these patients, we determined 248 to be eligible to participate in the study (87%); 185 (75%) consented to participate. Ultimately, 175 (71% of those eligible) who consented to participate completed the baseline assessment and continued on in the study; 155 completed the 6-month assessment, 125 completed the 12-month assessment, and 103 completed the 18-month assessment. As noted in Figure 2, some patients enrolled later in the study and did not have the opportunity to have assessments after 6 or 12 months.
PCMH-KD Participant Recruitment and Enrollment.
Table 5 shows the demographic characteristics of patients at both sites. Patients ranged in age from 20 to 89 years (mean 54.4; SD, 15), with the majority male (55%). Participants were nearly all African American and Hispanic (97%). One-third of our subjects conducted interviews in Spanish. Most had at least a high school education (64%); 40% of patients were never married, while 35% were married or living with a partner. Regarding employment, a large majority were not employed (83%). Income levels were low, with 68% reporting incomes less than $40 000 per year.
Participant Characteristics by Site, Baseline.
Regarding dialysis history, length of time on dialysis averaged 4.4 (SD, 5.2) years, with long periods at their current dialysis center (mean 3.3 years; SD, 4.4), and three-quarters of patients at the same dialysis center for at least 6 months. Patients with prior transplants composed 19% of the study participants (Table 6).
Dialysis History by Site, Baseline.
Participants' health insurance was predominantly covered by Medicare or Medicaid (Table 7); only 11% reported some private insurance coverage that was not a Medicare supplement. Many patients were uncertain about their health insurance coverage.
Insurance Coverage by Site, Baseline.
Table 8 shows data for self-reported comorbidities. Patients mostly reported being diagnosed with 2 or more conditions (60%), with hypertension (82%) and diabetes (53%) having the highest occurrence. Most patients (60%) reported at least 2 or more comorbidities.
Self-reported Comorbidities by Site, Baseline.
KDQOL subscale scores, the primary outcome for the study, by sociodemographic and clinical factors and site, are reported in Table 9. The mean baseline (SD) KDQOL subscale scores were the following: PCS was 35.5 (±10.2); MCS was 49.2 (±10.7); burden was 46.5 (±27.1); symptoms was 76.5 (±15.9); and KDE was 72.3 (±20.6). Some highlights include that participants aged 55 years and older had a significantly lower PCS score at both UIHS and FMC sites compared with those patients younger than 55 years of age. Symptom scores were lower among those aged 55 years and older only at the UIHS site. None of the other KDQOL subscale scores were significantly different between age groups. Hispanic participants had significantly lower MCS and burden subscale scores at UIHS compared with African American and Caucasian/Asian participants, though all other subscale scores were similar between racial/ethnic groups. Participants who reported that their primary language was Spanish had significantly lower PCS scores at FMC and lower (poorer) burden scores at both sites compared with those whose primary language was English or those who used both Spanish and English equally; the remaining subscale scores were not significantly different. There were no significant differences in baseline KDQOL subscale scores related to sex, marital status, or having a PCP at study entry.
KDQOL by Characteristics and Site.
Table 10 shows changes (unadjusted) across all patients over time from baseline, 6 months, 12 months, and 18 months. Noteworthy is that the scores for the MCS and burden score trended upward over time, while other components varied in direction and magnitude over time.
KDQOL Mean Scale Scores at Baseline and at 6, 12, and 18 Months (Unadjusted).
Self-efficacy for Managing Chronic Disease, which has a scale of 1 (not confident) to 10 (totally confident), revealed that the study population had a mean self-efficacy score of 7.2 (SD, 2.0) for managing kidney disease (Table 11). Mean scores increased at 6 months, and although the absolute value declined slightly, the change from baseline was statistically significant for 12 and 18 months in unadjusted and adjusted models to test for trends (Table 11a). The adjusted model included covariates for the orthogonal linear (6-month), quadratic (12-month), and cubic (18-month) terms for visit, and baseline age, sex, race (African American [AA], all other), interview language, dialysis vintage (months), site, education (not high school [HS] grad, HS grad), marital status (married or living with partner, other), self-reported diabetes, current PCP for at least 6 months, and time-varying laboratory measures (URR, hemoglobin, and albumin).
Self-efficacy for Managing Chronic Disease Mean Scores at Baseline and at 6, 12, and 18 Months.
Trends in SEMCD Over Time: Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
Regarding health literacy, which is based on the 3-item screening questionnaire developed by Chew and colleagues26 and uses a scale from 3 (highest level of literacy) to 15 (inadequate health literacy), about 61% of the patients had adequate health literacy, while 39% had inadequate or marginal levels. Health literacy levels were statistically significantly lower in the FMC site compared with the UIHS site; measures were at baseline only (Table 12).
Health Literacy by Site, Baseline.
Patient's knowledge levels (Table 13), measured using CHeKS, ranged from 0.0 to 87.0, with 43% of subjects scoring less than 50%, indicating low general knowledge about kidney disease and hemodialysis at baseline. During the intervention period, the percentage of subjects scoring less than 50% decreased to 39%, although these improvements were not statistically significant (Table 13a).
Chronic Hemodialysis Knowledge Survey Percentage Correct at Baseline and at 6, 12, and 18 Months.
Trends in Hemodialysis Knowledge Over Time: Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
Of 174 enrolled patients who completed the PCAS, only 60% reported having a regular personal doctor—eg, primary care physician—at baseline. Table 14 shows unadjusted and adjusted scores for patients who had no PCP at baseline on the 5 scales of the Primary Care Assessment Survey. The integration of care assessed patients' perceptions of their PCP's role in coordinating care received by patients from specialists or when hospitalized. Overall, longitudinal continuity of care, which assessed the duration of patients' relationship with their PCP, had the lowest mean score (52.9 [SD, 37.2]) at baseline. The lack of primary care among many—and the variability in PCAS among subjects who indicated they had primary care doctors—suggested that there was an opportunity to improve care coordination for all patients. Notably, over the period of the intervention, the percentage of patients reporting they had someone they considered their personal doctor grew to 81%; however, the trend was statistically significant only at 6 months. Although longitudinal care seemed to remain somewhat flat over time, comprehensive knowledge, communication, interpersonal treatment, and integration of care ratings all improved over the period of the intervention in adjusted models (see Table 14a).
PCAS Mean Scale Scores at Baseline and at 6, 12, and 18 Months.
Trends in PCAS Scale Scores Over Time: Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
For the CAHPS-ICH survey (see Table 15) patients provided rating scores for the experience with their nephrologists (kidney doctors), dialysis staff, dialysis center, nephrologists' communication and caring, quality of dialysis center care and operations, and providing information to patients. Overall rating at baseline for kidney doctors averaged 2.3 (SD, 0.8); the highest rating was for quality of care (mean 3.4; SD, 0.6). Over the period of the intervention, ratings for the scales were mixed (see Table 15a). Ratings for the nephrologists improved from baseline at 12 months only; ratings for the dialysis staff improved at 18 months; rating of nephrologists' communication and the quality of dialysis center operations improved at 6 months only; and ratings for providing information to patients at 6 months and 18 months improved.
Consumer Assessment of Health Care Providers and System: In-Center Hemodialysis Survey Mean Scale Scores at Baseline and at 6, 12, and 18 Months.
Trends in Consumer Assessment of Health Care Providers and System: In-center Hemodialysis Survey Scale Scores Over Time; Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
Subject depression scores, as measured by the PHQ-9 (Table 16) assessment tool, revealed a baseline mean score of 5.6 (SD, 5.1), with 17% of the study population in the moderate to severe depression category. During the intervention, depression improved, as evidenced by lower PHQ-9 scores at 6 months, and by 12 months the trend was a statistically significant improvement (mean score, 4.1; SD 4.1), although this trend (see Table 16a) was not sustained at 18 months (mean score, 5.1; SD, 5.5). Over time, the percentage of patients who reported no depression increased from 50% to 63%.
PHQ-9 Depression Assessment Mean Scale Scores and Depression Severity Categories at Baseline and at 6, 12, and 18 Months.
Trends in PHQ-9 Depression Over Time: Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
Noteworthy is that only 26% of our study population was categorized as having a high medication adherence (Table 17), as measured by the Morisky Medication Adherence Questionnaire. This score indicates that patients had difficulty adhering to their medication/treatment at baseline. During the intervention period, the percentage who scored high on medication adherence increased to 33% at 18 months; medication adherence raw scores also increased from 5.7 (SD, 2.0) to 6.2 (SD, 1.9) from baseline to 18 months, respectively. The test for trend was statistically significant only for the improvement at 6 months (see Table 17a).
Morisky Medication Adherence Questionnaire Adherence Categories and Mean Score at Baseline and at 6, 12, and 18 Months.
Trends in Morisky Medication Adherence Over Time: Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
RAAQ scores (Table 18), measuring attitudes toward dietary and fluid restrictions and the impact of these restrictions to the patients, revealed that there were more positive attitudes toward well-being at baseline (mean 40.4; SD, 5.0) compared with the other scales (attitudes toward social restrictions and acceptance). Tests for trends over time were mixed (Table 18a).
Renal Adherence Attitudes Questionnaire Mean Scale Scores at Baseline and at 6, 12, and 18 Months.
Trends in Renal Adherence Attitudes Questionnaire Scale Scores Over Time: Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
For the RABQ (Table 19) baseline data showed greater adherence to fluid restrictions (mean, 39.6; SD, 6.6) compared with adherence to potassium/phosphate restrictions (mean, 20.3; SD, 3.3) and adherence to sodium intake (mean 8.2; SD, 1.9). The RABQ ratings for fluid restrictions showed gradual improvement at 6 and 12 months in tests for trends, whereas adherence to the potassium and sodium restrictions did not change significantly over the intervention period (see Table 19a).
Renal Adherence Behavior Questionnaire Mean Scale Scores at Baseline and at 6, 12, and 18 Months.
Trends in Renal Adherence Behavior Questionnaire Scale Scores Over Time: Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
Table 20 shows the monthly clinical data summarized at baseline and at 6, 12, and 18 months for vital signs, laboratory data for anemia management, bone and mineral metabolism, dialysis adequacy measures, and diabetes management. Values were relatively consistent over time. Tests for trends over time (see Table 20a) varied for specific measures, and although some changes reached statistical significance, no changes were clinically significant even in adjusted models that included covariates for baseline age, sex, race (AA, all other), interview language, dialysis vintage (months), site, education (not HS grad, HS grad), marital status (married or living with partner, other), self-reported diabetes at baseline, current PCP for at least 6 months at baseline, and baseline URR, hemoglobin (g/dL), and albumin (g/dL) with specific modifications in variables included as noted for the laboratory measures (see Table 20a).
Dialysis Clinical Care Outcomes at Baseline and at 6, 12, and 18 Months.
Trends in Dialysis Care Clinical Outcomes Over Time: Orthogonal Polynomial Regression Model Unadjusted and Adjusted Results.
The use of services provided by the PCMH-KD PCP and CHW relates to the reach and implementation of the clinical intervention. Of all the study patients, 53% (N = 93) made appointments with the PCMH-KD PCPs. The percentage of patients with at least 1 PCMH-KD PCP visit was higher at UIHS compared with the FMC site (Table 21). Overall, the number of PCMH-KD PCP appointments totaled 348 over the 18-month intervention period (January 2015 to August 2016); of those with PCMH-KD PCP appointments, the mean number of appointments was 2.0 over the study, with some patients having as many as 19 appointments (Table 22). Some visits were not completed due to patients canceling appointments or not showing up to due illness or other reasons. Individual visits were conducted in a separate clinical examination room (41%), at the chairside in the dialysis center (50%), or by phone (9%).
Health Care Use: Uptake of PCMH-KD Providers by Site, Reach.
Health Care Use: Uptake of the PCMH-KD Providers by Site, Implementation.
CHWs also conducted individual visits with 95% of enrolled study patients. Visit rates were similar at our 2 sites, demonstrating similar fidelity of the intervention insofar as CHWs were concerned. CHWs conducted 1508 visits with 166 patients; of the visits, intake visits and brief check-in visits accounted for 11% and 24%, respectively. The more intensive follow-up visits accounted for 66% of the visits, and mean visits per patient was 8.6 (SD, 4.1) and ranged from no visits to 16 visits. Patients at UIHS averaged more visits than those at the FMC site. Visits were conducted in either English or Spanish.
Of the follow-up visits conducted by CHWs, topics varied and are summarized in Table 23. Dialysis patients required assistance most often with scheduling appointments (38%).
Topics Discussed During Follow-up Community Health Worker Visits by Site: Implementation.
Inpatient admissions and emergency department (ED) use were tracked and validated at the primary medical center used by patients, UIHealth (Table 24). Inpatient stays for study participants at UIHealth deceased over time: 37% of patients had at least 1 inpatient stay at UIHealth before the intervention study compared with 13% during the last 6 months of the intervention. Length of stay also decreased, as indicated by the number of inpatient days before the intervention at 764 compared with 171 days during the last 6 months. ED visits also showed declines from the preintervention period compared with the last 6 months. The percentage of patients with an ED visit to UIHealth decreased from 28% to 12%. Among patients who had any ED visit to UIHealth, the average number of visits in the preintervention period was higher than that of the last 6 months of the intervention (2.1 versus 1.6, respectively). As these health care use data are limited to 1 institution, it is possible that patients may have utilized services at other facilities. It is also possible that the final reconciliation of billing data for July through August 2016 was incomplete owing to the data extraction timing (September 2016) needed for completion of the study. Additional information from Medicare and/or Medicaid claims may provide more complete view of health care use, although such data were not available for this study.
Health Care Use at UIHealth by Type and Time Period.
We assessed staff (ie, dialysis nurse coordinators, dialysis technicians, dialysis center nephrologists, PCMH-KD PCP, PCMH-KD CHW, and pharmacists) perceptions using the team vitality score34 (Tables 25 and 26). Overall scores were similar at sites, and item scores showed improvement from baseline to 18 months.
Staff Survey Team Vitality Summary Score by Site and Data Collection Period (Baseline and 18 Months).
Staff Survey Team Vitality Item Means by Site and Data Collection Period (Baseline and 12 Months).
Tables 27 and 28 show the KDQOL adjusted scores and adjusted change scores, 95% confidence intervals, and standard errors, respectively, for each time period estimated from the orthogonal regression models. For KDQOL PCS (baseline adjusted mean = 35.7; 95% CI, 33.89-37.55) the change from baseline to 6 months was 2.6 points (95% CI, 0.97-4.18; P = .002). Although the change from baseline PCS to 12 months (PCS change score at 12 months = 0.3; 95% CI, −1.43-1.95; P = .76) and to 18 months (PCS change score at 18 months = 1.0; 95% CI, −0.87-2.89; P = .29) was positive, the change was not statistically significant.
Adjusted Kidney Disease Quality of Life Mean Scale Scores at Baseline and at 6, 12, and 18 Months: Adjusted Random-intercept Models.
Adjusted Kidney Disease Quality of Life Scale Change Scores at 6, 12, and 18 Months: Adjusted Random-intercept Models.
KDQOL MCS improved from baseline (adjusted mean = 48.9; 95% CI, 47.14-50.73) to 6 months by 1.2 points (95% CI, −0.73-3.09; P = .23) to 12 months by 2.5 points (95% CI, 0.49-4.50; P = .01) and to 18 months by 2.79 points (95% CI, 0.65-4.93; P = .01). KDQOL burden (baseline adjusted mean = 45.8; 95% CI, 41.20-50.46) scores changed minimally from baseline to 6 months by 2.6 points (95% CI, −1.24-6.43; P = .18), to 12 months by 3.2 points (95% CI, −0.86-7.22; P = .12), and to 18-months by 3.8 points (95% CI, −0.40-8.03; P = .08). KDQOL symptoms score from baseline (adjusted mean = 77.1; 95% CI, 74.44-79.71) to 6 months improved by 2.6 points (95% CI, 0.35-4.78; P = .02), to 12 months improved by 2.2 points (95% CI, −0.14-4.54; P = .07), and to 18 months decreased by −0.6 points (95% CI, −3.20-1.90; P = .61). The KDQOL KDE showed improvement from baseline (adjusted mean = 73.0; 95% CI, 69.38-76.53) to 6, 12, and 18 months by 4.3 points (95% CI, 1.39-7.15; P = .004), 6.7 points (95% CI, 3.63-9.74; P < .001), and 3.9 points (95% CI, 0.51-7.29; P = .02), respectively. Except for the improvement in KDE from baseline to 12 months, the MCS and KDE improvements fell below our prespecified clinical level of significance of 10% (5-point change).
Table 29 shows the changes for the intervals from 0 to 6 months, 6 to 12 months, 12 to 18 months, and 0 to 18 months, for additional reference.
Estimated Kidney Disease Quality of Life Scale Change Scores for Each 6-Month Period and 0 to 18 Months: Adjusted Random-intercept Models.
The changes in the KDQOL domains over time, adjusted for the covariates described above, are shown in graph form in Figures 3 to 7.
Kidney Disease Quality of Life, Physical Health Composite, Mean at Baseline and at 6, 12, and 18 Months.
Kidney Disease Quality of Life, Mental Health Composite, Mean at Baseline and at 6, 12, and 18 Months.
Kidney Disease Quality of Life, Burden of Kidney Disease, Mean at Baseline and at 6, 12, and 18 Months.
Kidney Disease Quality of Life, Symptoms, Mean at Baseline and at 6, 12 and 18 Months.
Kidney Disease Qualify of Life, Disease Effects, Mean at Baseline and at 6, 12, and 18 Months.
Finally, for the 2 KDQOL subscales that improved over time, we show the factors contributing to the improvement. Results are shown in Table 30 for the adjusted random intercept models for KDQOL MCS and KDE change over time. Noteworthy is that in the MCS model, in addition to the significant effects for the KDQOL 12-month and 18-month visit parameters (ie, each representing the change score from baseline), only the hemoglobin and albumin levels were significant positive predictors (P = .02 and .06, respectively). Regarding subgroup analysis, it is noteworthy that in the KDE model, in addition to the significant effects for the 6-month, 12-month, and 18-month visit parameters, patients who were African American (P = .04), were a high school graduate (P = .06), and had a higher albumin (P = .02) were significant positive predictors.
Parameter Estimates From Adjusted Random-intercept Models of KDQOL MCS and KDE Change Over Time.
In sensitivity analyses, we explored our multivariable results using multiple imputation by chained equations for missing values or our primary outcome measure.41 Table 30 shows P values for the linear, quadratic, and cubic trend terms for all 5 KDQOL scales. The results for the models with and without imputation are similar. Tables 31 and 32 show all parameter estimates from the regression models with and without imputation for MCS and Kidney Disease Effects, respectively. The results for the models with and without imputation are similar.
Sensitivity Analysis-Parameter Estimates From Adjusted Random-intercept Models of Change Over Time in KDQOL Scale Scores.
PCMH-KD Final Research Report Acronym Table.
Results from the content analysis of the discussion groups suggested the following themes: advocacy, education, caregiver needs and activities, PCPs and PCP appointments, and health promoters/CHWs. Patients often reported that discussion groups were a time for them to discuss issues they were facing in the dialysis center. They encouraged one another to report difficulties with staff and brought concerns of care quality to the discussion group, with expectations that study team members would respond. This led to a meeting and a patient comment box at 1 of the dialysis sites. Education was also a primary concern of patients. Patients reported that much of the information they received needed to be repeated to help them remember it. Patients and caregivers expressed need for more education about dialysis, especially when first starting the process. One person specifically cited that caregivers needed to be part of the orientation process. Another person wondered why they were not allowed to sit with their family member during the dialysis process. Patients reported that they were too tired after treatments to communicate with family about what occurred at dialysis and that the dialysis center should help with educating them.
The PCMH-KD PCP and appointments were prominent in the discussions. Patients often cited that the PCMH-KD PCPs were easy to access and that the PCMH-KD PCP followed up with them through hospitalizations as well as helping them with forms. Additionally, Health promoters were frequently cited as part of the study feedback for the assistance they provided with appointment reminders. One patient cited gratefulness that the health promoter had helped schedule and cancel appointments. Patients reported that health promoters making appointments was beneficial and saved them time and effort from trying to navigate and make appointments when coming out of the hospital or on their off days from dialysis.
These qualitative results should be considered in the context in which they were conducted, as has been previously reported.42 It is noteworthy that we faced challenges in organizing and conducting these discussion groups. Participation of patients was difficult due to the nature of their health and its general toll on them and their family members. For example, confirming attendance for a planned discussion group and arranging transportation was time consuming for our study staff, and we were often uncertain about guaranteeing a quorum in advance of the session. We learned that patients' attendance could be uncertain due to variability in disease progression, symptom exacerbation, and complications. Therefore, we invited about 25% more patients than we expected to attend a given session. Having a large percentage of patients who were primarily Spanish speaking was another important consideration. We planned for bilingual staff and separate English and Spanish sessions to be held to accommodate as many patients as possible. Because there were a greater number of Spanish-speaking patients at FMC, we found that letting the patients know in advance that our bilingual staff would be in attendance during the discussion group was important to them in deciding whether to participate.
The PI conducted individual follow-up interviews with a dialysis center management representative from each site. These open-ended interviews focused on whether the organizations would consider pursuit of the care model used in this intervention study. We used probes to identify if members of the PCMH-KD team felt they might continue and to get their impressions of this model for general adoption. Both representatives felt that cost of the expanded team was a critical factor, and that a cost–benefit analysis would have been helpful. One representative felt that the medication reconciliation role of the pharmacist and coordination function of the nurse was possible for his division budget to support, but beyond that he would have to rely on the medical center to invest. Both representatives indicated that the cost of an expanded clinical team without compensation from insurance was an important consideration and could be important for other provider organizations, too.
We conducted a before–after intervention study of an adaptation of the PCMH for kidney disease focused on chronic hemodialysis patients at 2 dialysis centers in an urban area with a racially and ethnically diverse patient population. Patients completed baseline data collection before any intervention began, and we collected retrospective laboratory and health care use data; we did not include any other control group and thus limited our ability to separate changes due to the intervention and to time trends. However, our analyses controlled for patient sociodemographic and clinical characteristics and study site. This study was important as a means to explore a novel health system intervention aimed at improving patient-centered outcomes for a patient population with complex care needs. The PCMH-KD model introduced new providers in the dialysis setting and utilized new training materials developed specifically for this intervention. We modified aspects of the provider roles in the PCMH-KD in response to stakeholder feedback. Results from the study revealed that domains of the KDQOL related to mental health (MCS) and kidney disease effects (KDE) were significantly improved with the intervention over 18 months compared with baseline scores. To our knowledge, this is the first study to adapt the PCMH model for kidney disease. There are several noteworthy findings from our work.
We observed heterogeneity in the trend patterns of the 5 KDQOL components. All quality of life components increased early, some significantly. We found that 3 of the 5 domains were significantly improved from baseline to 6 months (PCS, KDS, KDE), and 2 domains were significantly improved from baseline to 18 months (MCS and KDE). For KDS, although it had a linear form, improving throughout the intervention, the improvements after 6 months never reached statistical significance. These patterns suggest that some KDQOL domains may be more sensitive to health system changes than others; another consideration is that the measures for these domains could be more unstable over time. The KDQOL MCS and PCS domains we used were based on the SF12, and some have reported that these domains may be subject to ceiling effects.43-45 Whether the SF36 version might be a better tool to address the potential ceiling effects should be evaluated in future studies.
The KDQOL component scores for our study population are consistent with other studies of ESRD among veterans for all domains.3,4 Previous studies of racial/ethnic differences in quality of life in ESRD patients have shown that African Americans report better quality of life than do other groups.46-48 As most of our patient participants were African American or Hispanic, whether these lower KDQOL component rates are related to race/ethnicity or other factors is worthy of additional investigation.
When we examined the factors that might be contributing to the changes in the KDQOL components, we found that, after adjusting for covariates, those who were African American, who had a less than high school education, and who had lower serum albumin scores (indicating poorer nutritional status) experienced significantly less improvement in their KDE component score. These 3 factors—race, education, and nutrition—point to a need that might be addressed together, such as through an education program that also provides information about diet as well as addressing cultural preferences for specific foods that might interfere with ESRD symptoms and dialysis effectiveness. Understanding the broader needs of specific patient populations should be carefully considered in designing interventions that aim to improve patients' well-being in a dialysis setting.
Patient assessment of primary care also improved during the intervention, with the percentage of patients reporting they had someone they considered their personal doctor growing from 60% to 81%. Although PCAS domain for longitudinal care seemed to remain somewhat flat over time, comprehensive knowledge, communication, interpersonal treatment, and integration of care ratings all improved over the period of the intervention. Simultaneously, the uptake of the PCMH-KD PCP averaged 3.7 visits per patient; 93 (53%) had at least 1 visit with a PCMH-KD PCP during the 18-month intervention period.
It is noteworthy that when provided with the opportunity for access to a PCMH-KD PCP at no economic cost, some patients without a regular PCP took advantage of it, although not all patients opted for this choice. This result is consistent with prior reports that showed the need for comprehensive and coordinated care.49,50 As expected, there was less uptake among those patients with an established PCP. The rationale behind this expectation included the patients' desire to maintain continuity of care and established relationships with their PCPs before the study. Yet some patients with an established PCP still chose to have individual visits with a PCMH-KD PCP. We postulate that prior experience with the benefits of a PCP, as well as the time-saving nature of the visits within the dialysis center, such as the dialysis chairside visits and colocation of the examination room at the dialysis center, and free cost were key to this decision. Anecdotal reports from patients in our patient stakeholder discussion groups support this theory; however, for patients who had an established PCP and who also used our PCMH-KD PCP, there may be concern that this second PCP may have worsened fragmentation of primary care. In this regard, it is noteworthy that the PCAS longitudinal continuity scale score was steady over time. Additionally, for those patients who received a referral from their personal doctor, they reported improved PCAS integration of care scale scores over time. Taken together, these PCAS results suggest that continuity of care was not compromised, and in cases when additional providers were consulted, that care integration improved. Further exploration of these choices in a larger population would be critical for expanded implementation of this novel care model.
The reason that patients with no established PCP had lower than expected uptake with a PCMH-KD PCP is unclear. While we provided information sessions for patients to understand the options and services with the implementation of the PCMH-KD, it is possible that a patient's lack of a prior PCP may be a choice made in the past that was not due to convenience factors or costs and that persisted during our study. This behavior is consistent with research on prior health care use influencing continued use patterns generally,51 although we are not aware of other studies of hemodialysis patients' primary care use. With the recent call for reform in the Medicare ESRD program to improve the patient centeredness of care,52 further research is needed to better understand patients' choices if care coordination is to be adequately addressed.
Our results can be viewed in the context of the RE-AIM framework domains. We found that patients and clinicians were engaged and participated in the model (Reach). Patients benefited from the intervention, as revealed in improvements in components of quality of life, depression, knowledge about kidney disease and dialysis, and satisfaction with primary care (Effectiveness). Patient and staff feedback about the intervention and how it was conducted was constructive, as reported in discussion groups and staff surveys (Adoption). The frequency and depth of patient–clinical team member encounters, the substance of the CHW encounters, and the fidelity of the intervention as conducted in the specific roles of the PCMH-KD team were satisfactory to the clinical stakeholders (Implementation). Finally, based on interviews with the leadership at each site, elements of the model related to medication management and enhanced care coordination may continue, although the cost of an expanded clinical team without compensation from insurance may offset any desire by provider organizations to do so (Maintenance). In summary, the RE-AIM framework provided a model to describe and monitor intervention implementation. Planned necessary adaptations to key aspects of our intervention and program content were important to ensure integrity of the implementation.
In conducting this study, we faced some challenges. First, we knew that in testing a health system intervention in the dialysis setting, it would not be practical to randomize patients due to the setting, as patients and families are in close communication about center operations. Using a before–after design with participants providing baseline data from before the intervention under the Medicare-mandated dialysis care model (ie, baseline care), we recognized this would limit the generalizability of our findings yet was necessary as a first step in determining if this intervention would have an effect.
Second, the budget for the funding announcement required that we make tradeoffs about staffing and the number of sites. At the time of study planning, the costs of setting up a PCMH-KD clinical team (2.5 FTE clinicians and health promoters) were prohibitive relative to the funds available for the contract to set up more than the 2 sites or more than 1 clinical team. Being limited to 1 team was a particular barrier with scheduling rounds, such that the part-time staff (general internists, nurse coordinator, pharmacists, and health promoters) had to split coverage for attending. It would have been ideal for the part time staff to attend all the rounds; however, they learned to rely on information from each other and notes for each patient in the dialysis records. It also would have been ideal to track the content of primary care provided by the PCMH-KD PCPs. Although this information is available in medical records, tracking this information in real time was beyond the scope of this study; however, it could be the focus of a future study. Forthcoming research should also carefully consider in its implementation plan the patient panel size, information about the content of primary care, and frequency of dialysis treatments and health care services that ESRD patients required.
Another, third, aspect that was especially challenging for our patients was the difficulty in scheduling appointments within and outside the medical center. Early on we revised the duties of the health promoter and nurse coordinator to assist patients with scheduling, and this reassignment limited the health promoter and nurse to meet individually with patients (health promoter) and the timeliness of tracking patients who missed dialysis (nurse coordinator) to determine if they were in an inpatient setting. These issues were regularly discussed and prioritized in our clinical stakeholder planning meetings, and yet a common theme in our discussions was that there were not enough hours in the day to balance all the support that patients were expecting and/or requiring. The health promoters also reported similar frustrations with “getting through the scheduling process.” With respect to study implementation, this workload meant that our team had to prioritize activities within their roles and to coordinate with other providers when needed. The regular clinical research planning meetings helped refocus and reassure staff to coordinate with other team members when needed.
Fourth, with a small intervention team, vacations and staff turnover had a major impact on the roles of other staff. When 1 of our general internists was on maternity leave for several weeks, the second general internist had to cover both schedules. When the nurse coordinator left, we were without a nurse coordinator for several months and so the health promoters covered much of the patient scheduling issues. Fortunately, team members had been cross-trained in many duties, so this knowledge likely limited the effects of missing staff. That said, it is also likely that there were delays on account of the higher workload for individual staff during these periods.
Aspects of our study also served as facilitators. Our training program during the first year supported clinical learning objectives for the new clinical team members and also helped foster team building and cross-training in care management of dialysis patients. This program was coordinated by an experienced team from the Midwest Center for The Midwest Latino Health Research, Training, and Policy Center, and included culturally sensitive training materials in a format that was readily available for retraining. We were able to use the training materials when new staff came into the study. These training materials may also be useful for other research or quality improvement projects embarking on reorganizing care delivery in other dialysis settings. We also developed an orientation program slide set for the regular dialysis staff to introduce team to the study and the roles of the new personnel and the research project in general.
Another advantage is that we recruited a clinical team that included highly experienced and skilled members; our first project coordinator was bilingual, and her input was critical in ensuring that our materials were accurate and culturally sensitive. Our health promoters had prior experience with diabetes patients and had benefited from extensive CHW training on at least 1 past study. Our nephrologists were very engaged throughout the project. During the study, we had 2 nurse coordinators (nonoverlapping) who had experience at the 2 study sites. This experience facilitated their orientation as well as their interactions with the existing dialysis center staff, which helped garner additional in-kind staff support when needed. For example, the nurse coordinator engaged the dialysis center clerk to occasionally help with scheduling patient appointments outside the dialysis center. The study clinicians also helped coordinate the timing of the dialysis patient interviews to limit interference with the dialysis treatments. The staff experience and the camaraderie they engendered with the existing dialysis center staff was very likely a factor in ensuring that patients had access to the study providers and the study interviewers when needed. In summary, having experienced clinical staff in a short-term study contributed to successful implementation; in a longer-term study, skills could be built up over time.
Support from the clinical and dialysis center leadership at each site was also an important facilitator. With bringing in new clinicians (physicians, pharmacist, and nurses), credentialing was required. Although the act was time consuming, with much coordination among multiple offices, both sites ultimately supported their clinical appointments and credentialing and their access to requisite information and systems to be a part of the clinical enterprise. Both sites supported the complex research approval and contract processes, authorizations for data exchange, and report reviews. In summary, to conduct a study that includes hiring new research clinicians, in addition to the hiring timeline, time must be planned for credentialing, orientation to the health system, and access to the information systems before individuals can begin clinical work. The cooperation and support from the study sites and their leadership were vital to these critical onboarding requirements.
Most importantly, our study participants and their families were engaged and supportive. Despite their frequent and severe symptoms, patients and their caregivers attended discussion groups and shared their perspective on their care, their quality of life, and the dialysis center operations. They helped us with coordinating best times to do interviews based on their treatment schedule. Their willingness to volunteer and to stay engaged in the research was invaluable to the implementation and evaluation of the intervention.
In terms of generalizability of the findings to other study populations and settings, the PCMH-KD study may provide the greatest insight for studies that include similar populations (ie, African American and Spanish-speaking dialysis patients) and settings (ie, urban academic and stand-alone outpatient dialysis centers). Although we controlled for many patient characteristics, including constant and time-varying factors and study site in our multivariable analyses, lack of a concurrent comparison group limits the ability to fully account for temporal trends.
In consideration of specific subpopulations, it is noteworthy that our study included predominantly African Americans and Spanish-speaking patients with ESRD (98%). In our multivariable analysis, we found that those who were African American, who had less than a high school education, and who had lower serum albumin scores (indicating poorer nutritional status) experienced significantly less improvement in their KDQOL KDE component score. The importance of these 3 factors—race, education, and nutrition—point to the need to address them together, such as through an education program that provides information about diet as well as addressing cultural preferences for specific foods that might interfere with ESRD symptoms and dialysis effectiveness.
We also conducted separate patient and caregiver stakeholder discussion groups for Spanish-speaking participants to ensure their participation. These discussions revealed a need for support for translation services in clinical care. Feedback was positive about our intervention having a CHW and a PCP who spoke Spanish.
Our study had limitations. First, we used a nonrandomized before–after design with participants' baseline measurements for patient-reported outcomes and 6-month measurements for health care use and clinical characteristics to represent their care under the Medicare-mandated dialysis care model (ie, usual care). Therefore, we did not have concurrent controls. Because this was a health system intervention, it was not practical to randomize patients to the intervention. Information learned about the planning and implementation process from our study can inform future study designs that might focus on randomization at the site level. Second, our implementation was limited to 2 sites with 1 group of academically based nephrologists. At the time of study planning, the costs of setting up a PCMH-KD clinical team (2.5 FTE clinicians and health promoters) were prohibitive relative to the funds available for the PCORI contract to set up more than the 2 sites; however, insights from the clinical stakeholders at our study sites provide a foundation for developing future multisite studies and with more than 1 clinical care team. Third, the information on health care use in our study was limited to 1 health system, due to patients' delays in reporting sufficient details about events, and the complexities of requesting medical billing records from multiple sites that were beyond the workflow and budget of this study. We believe the impacts are minimal because most patients reported using the health system for their care, albeit some patients may have also used nearby facilities. Although most of hemodialysis patients have their care covered by Medicare and/or Medicaid, claims data from Medicare cannot be obtained until 2018. This timeline was beyond the period of this project and is a limitation of all studies requiring such Medicare claims for research. Because patients in this study provided consent for all of their health insurance information and claims, a follow-up study is possible to ascertain other health care use when these Medicare data become available.
Our implementation and evaluation of an adaption of a PCMH model for kidney disease patients receiving hemodialysis offers important insights to inform future studies. Notably, future studies of dialysis patients should include quality of life outcomes that are informed by patients and that are sufficiently sensitive to changes over time. Study design with a concurrent comparison group would strengthen study outcomes.
In terms of specific subpopulations, recognizing that African Americans are at higher risk for ESRD and its sequalae,53,54 our findings offer insight for further study. That those who were African American, had less than a high school education, and had lower serum albumin scores (indicating poorer nutritional status) experienced significantly less improvement in their KDQOL KDE component score over time suggests a need for further research that addresses cultural issues, educational, and nutrition together. Also, the extent to which the availability of translation services versus having Spanish-speaking care providers makes a difference in patient-reported outcomes may be an area for further research. Our study participants were very receptive to the CHWs we employed. The individual role of the CHW in the care team and the extent to which the role enhances uptake of primary care services and coordination of care for dialysis patients also deserve further research.
Given the increased emphasis on both reducing unnecessary health care utilization and improving care from the patient perspective, there is an urgent need for novel health care interventions that address these issues. Time trends in care utilization in our study suggest decrease in ED care may be possible. Future research should seek to understand whether certain care can be handled in outpatient settings—rather than in ED or inpatient settings—to meet patients' urgent needs.
Finally, with the recent implementation of the CMS accountable care organization focused on ESRD, known as ESRD Seamless Care Organizations (ESCOs),55,56 our study may offer insight. Notably, several ESCO elements are similar to those of our PCMH-KD model, such as providing on-site colocation of different providers at dialysis facilities, assisting beneficiaries in scheduling nondialysis-related medical appointments, and assisting beneficiaries with transportation services.
Results from this study revealed that patient-reported outcomes for the KDQOL related to mental health and kidney disease effects were significantly improved with the intervention over 18 months compared with baseline scores; results for other KDQOL dimensions were mixed. As this study adopted a before–after design, we cannot say with confidence that these effects were due to the intervention alone. Regarding secondary outcomes, patient assessment of the primary care that they received showed improvement from baseline to 6 months in the PCAS domains of comprehensive knowledge, communication, interpersonal treatment, and integration of care ratings after accounting for other important factors. Even for patients who already had an established PCP at baseline, continuity of care was not compromised by the involvement of the PCMH-KD PCP. Observed decrease in ED use was limited to observations at 1 medical center and warrants further evaluation.
To our knowledge, this study is the first application of an adaptation of the PCMH for chronic hemodialysis patients and offers resources and lessons for future studies.11,56 While improvements in dialysis patients' quality of life and satisfaction with primary care are desperately needed, providers and payers should proceed cautiously with dialysis care reorganization, especially as health care organizations undergo sweeping changes in business partnerships and insurance reimbursement models, which may also affect dialysis care and patient outcomes. Future research using an experimental design with a concurrent control group is needed to better understand the potential for new dialysis care models to improve patient-centered outcomes.
Research reported in this report was [partially] funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (#IH-12-11-5420) Further information available at: https://www.pcori.org/research-results/2013/evaluating-patient-centered-medical-home-patients-receiving-dialysis-kidney
Hynes DM, Arruda J, Berbaum M, et al. (2019). Evaluating a Patient-Centered Medical Home for Patients Receiving Dialysis for Kidney Disease. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/5.2019.IH.12115420
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
