Association of Primary Care Involvement with Death or Hospitalizations for Patients Starting Dialysis

Abstract

Visual Abstract

Background and objectives 

It is uncertain whether primary care physician continuity of care associates with a lower risk of death and hospitalization among patients transitioning to maintenance dialysis.

Design, setting, participants, & measurements 

Using provincial-linked administrative databases in Ontario, Canada, we conducted a population-based study of incident patients who initiated maintenance dialysis between 2005 and 2014 and survived for at least 90 days. We defined high primary care physician continuity as both a high usual provider of care index (where >75% of primary care physician visits occurred with the same primary care physician) in the 2 years before dialysis (an established measure of primary care physician continuity) and at least one visit with the same primary care physician in the 90 days after dialysis initiation. We used propensity scores to match a group of patients with high and low continuity so that indicators of baseline health were similar. The primary outcome was all-cause mortality, and secondary outcomes included all-cause and disease-specific hospitalizations during the 2 years after maintenance dialysis initiation.

Results 

We identified 19,099 eligible patients. There were 6612 patients with high primary care physician continuity, of whom 6391 (97%) were matched to 6391 patients with low primary care physician continuity. High primary care physician continuity was not associated with a lower risk of mortality (14.5 deaths per 100 person-years versus 15.2 deaths per 100 person-years; hazard ratio, 0.96; 95% confidence interval, 0.89 to 1.02). There was no difference in the rate of all-cause hospitalizations (hazard ratio, 0.96; 95% confidence interval, 0.92 to 1.01), and high primary care physician continuity was not associated with a lower risk of any disease-specific hospitalization, except for those related to diabetes (hazard ratio, 0.88; 95% confidence interval, 0.80 to 0.97).

Conclusions 

High primary care physician continuity before and during the transition to maintenance dialysis was not associated with a lower risk of mortality or all-cause hospitalization.

Introduction

Recipients of maintenance dialysis are a high-risk population; on average, they are hospitalized nearly twice a year, and the survival of incident patients is only 40% at 5 years (¹). These risks are accentuated during the first 90 days after commencing dialysis (²), with most adverse events attributable to complications unrelated to their ESKD, which may fall outside the traditional skill set of nephrologists (³,⁴). For example, the risk of hypoglycemia and psychosocial stresses both increase after the transition to dialysis (⁵,⁶), yet many nephrologists do not routinely manage diabetes-related medications or treat mental health concerns (³). One potential solution to address these care gaps is ensuring the maintenance of strong ties with primary care providers, similar to how more frequent nephrologist visits has been associated with reduced 30-day rehospitalization and improved vascular access care (^7–9).

A recent systematic review on the role of primary care for patients receiving maintenance dialysis identified 14 studies that primarily used qualitative and survey-based designs (³). Although many patients reported having a primary care physician (PCP), most patients still relied on their nephrologists for primary care services. Major gaps in the literature include limited study size and scope, particularly with respect to addressing whether greater PCP involvement improves patient-relevant outcomes. This information is necessary to inform quality improvement efforts and outcome measures for the delivery of high-quality primary care to patients on dialysis in a manner that respects both the patient experience and health care costs.

In this population-based study from Ontario, Canada, we sought to investigate the relationship between PCP continuity of care and outcomes. We defined high PCP continuity as both a high usual provider of care (UPC) index (where >75% of PCP visits occurred with the same PCP) in the 2 years before dialysis (an established measure of PCP continuity) and at least one visit with the same PCP in the 90 days after dialysis initiation (^10–12). Our primary objective was to determine whether PCP continuity of care associates with a lower risk of all-cause mortality and all-cause hospitalizations during the 2 years following dialysis initiation.

Materials and Methods

Study Design

We conducted a population-based retrospective cohort study with province-wide administrative databases linked using unique encoded identifiers and analyzed at ICES (¹³). In Ontario, Canada (population of over 14 million residents), the presence of a single-payer, universal health care system that encompasses physician services, ambulatory care, and in-hospital care enabled complete capture of physician visits and the study outcomes of interest. Emigration from the province is the only reason for lost follow-up, and the rate is low (<0.5% per year) (¹⁴).

ICES is an independent, nonprofit research institute with legal status under Ontario’s health information privacy law that allows it to collect and analyze health care and demographic data without consent for health system evaluation and improvement. The use of data in this project was authorized under section 45 of Ontario’s Personal Health Information Protection Act, which does not require review by a research ethics board. The reporting of this study follows the Reporting of Studies Conducted Using Observational Routinely Collected Health Data guidelines for observational studies (Supplemental Table 1) (¹⁵).

Databases

We used the following provincial-linked databases to define key study criteria, exposures, outcomes, and covariates (Supplemental Table 2). The Canadian Organ Replacement Register (CORR) contains validated information on dialysis modality, including kidney transplantation (¹⁶,¹⁷). The Ontario Health Insurance Plan database contains records of all physician claims for outpatient and inpatient services (e.g., dialysis and clinic visits), including service date, procedure performed, and diagnosis. The ICES Physician Database comprises information on physician specialty (ascertained with annual telephone surveys combined with billing information) and physician characteristics. We defined PCPs as physicians whose main specialty was “general practitioner/family physician” because internal medicine specialists usually do not provide primary care in Canada (¹⁸). We obtained information on hospitalizations using the Canadian Institute for Health Information Discharge Abstract Database, which contains the date of admission and discharge, hospital procedures, and up to 25 diagnoses in the International Classification of Diseases, Ninth Revision (prior to 2002) or Tenth Revision (2002 onward). We ascertained vital status using the Registered Persons Database.

Study Population

We identified all adult patients aged ≥18 years old with ESKD in the CORR whose first ever dialysis (hemodialysis or peritoneal dialysis) session occurred between January 1, 2005 and December 31, 2014; this date was ascertained with validated physician billing codes (¹⁹,²⁰). To avoid immortal time bias (where patients who visited their PCP after initiating dialysis were by definition alive), we established the index date as being day 91 after the initial dialysis session to ensure that all cohort members were alive and could in theory visit their PCP.

We excluded patients who received a kidney transplant in the 5 years prior to the index date and patients living in a long-term care facility (because the delivery of primary care differs in these settings). We also excluded patients hospitalized for ≥30 days between their first dialysis session and the index date because prolonged hospitalizations would interfere with PCP follow-up (²¹). To ascertain the UPC index (¹⁰,¹²), an established measure of PCP continuity, we excluded patients with fewer than three PCP visits in the 2 years prior to their first dialysis session.

Exposure

We defined high PCP continuity as requiring both (1) high UPC index, where >75% of PCP visits occurred with the same PCP in the 2 years before the first maintenance dialysis session, and (2) at least one visit with the same PCP in the 90 days after the first maintenance dialysis session. We defined low PCP continuity as failure to satisfy either criteria.

We used the UPC index because it is the most established tool to ascertain continuity of care as well as because of its association with high-quality physician-patient interactions (¹⁰,¹²). It is calculated as the proportion of PCP visits to the most frequently visited PCP over a 2-year period, excluding patients with fewer than three primary care visits (¹¹). Therefore, patients must have three or more PCP visits over a 2-year period to calculate the UPC index. Because the UPC can only be ascertained in the past (i.e., prior to the first maintenance dialysis session), we also required patients to visit the same PCP at least once after their first maintenance dialysis session to strengthen our definition of high PCP continuity. To assess whether our results were dependent on this definition of PCP involvement, we conducted sensitivity analyses using alternate definitions as described below.

Outcomes

The primary outcome was all-cause mortality during the 2 years after dialysis initiation censored for kidney transplantation. Secondary outcomes included all-cause hospitalizations and disease-specific hospitalization for heart failure, myocardial infarction, stroke/transient ischemic attack, sepsis, diabetes, chronic obstructive pulmonary disease, chronic pain, delirium, or fracture. We selected these reasons for hospitalization because they either commonly occur for patients on dialysis (i.e., heart failure and sepsis) or may potentially be affected by PCP continuity of care (i.e., diabetes and chronic pain). We also included hospitalizations for previously defined kidney-specific ambulatory care–sensitive conditions (²²,²³). Follow-up commenced on day 91 after the first maintenance dialysis session and continued for up to 2 years thereafter.

Propensity Score Development

We developed a multivariable logistic regression model to estimate propensity scores for high PCP continuity (²⁴). Clinical significance guided the choice of covariates in the model and included age, sex, rural residence (community with population ≤10,000), income (estimated using mean neighborhood income by quintile), ethnicity, primary cause of ESKD, comorbidities in the preceding 5 years, the number of hospitalizations and outpatient physician visits in the prior 12 months, and PCP characteristics (age, sex, number of years since graduation, hospital affiliation, rural practice, and volume of patients on dialysis). We used a structured, iterative approach to refine this model and achieve covariate balance within the matched pairs prior to initiating any analysis of the outcomes (²⁵). We measured covariate balance by the standardized difference (Sdiff), in which an absolute Sdiff >10% represents meaningful imbalance (²⁶). We aimed to match each patient with high PCP continuity to a patient with low PCP continuity 1:1 by means of a greedy matching algorithm without replacement using the logit of the propensity score (±0.2 SD). We forced an exact match on dialysis modality (in-center hemodialysis, peritoneal dialysis, or home hemodialysis).

Statistical Analyses

We summarized baseline characteristics using descriptive statistics. We expressed continuous variables as the mean (SD) or median (25th, 75th percentile) and categorical variables as frequency (percentage). We calculated the mortality rate per 100 person-years for patients with high and low PCP continuity and derived hazard ratios (HRs) and 95% confidence intervals (95% CIs) from Cox proportional hazards models with a robust sandwich variance to account for correlation within the matched pairs (²⁷). We assessed the proportional hazards assumption with the Kolmogorov-type supremum test and generated survival curves using the Kaplan–Meier estimator. We used a similar approach for disease-specific hospitalizations, deriving HRs from Cox proportional hazards models with a robust sandwich variance and censoring for death. For all-cause hospitalizations, we used the Andersen–Gill model to allow for multiple hospitalizations per patient, censoring for death (²⁸). We stratified by year of cohort entry to explore any potential effect modification by era.

Because the importance of PCP continuity may differ by dialysis modality (e.g., in-center hemodialysis, peritoneal dialysis, or home hemodialysis), we performed a prespecified subgroup analysis stratified by the initial maintenance dialysis modality. We also performed multiple sensitivity analyses to test alternative definitions of PCP involvement. We compared patients with any PCP visit in the 90 days after their first maintenance dialysis session with patients who did not visit with any PCP during the same time period. We also repeated this analysis including patients with fewer than three PCP visits in the 2 years prior to their first maintenance dialysis session. To determine whether the results differed with a more sustained definition of PCP continuity, we recalculated the UPC using 1 year of PCP contact before dialysis initiation and 1 year of PCP contact after dialysis initiation, with outcome ascertainment starting 1 year after dialysis initiation. We repeated the same propensity score–matching process as the primary analysis for all sensitivity analyses. We considered a two-sided P value <0.05 as statistically significant. We performed all analyses using SAS version 9.4 (SAS Institute, Cary, NC).

Results

Patient Characteristics

Between January 1, 2005 and December 31, 2014, a total of 25,158 patients with ESKD started maintenance dialysis in Ontario, Canada (Figure 1). Of the 19,099 patients included in the study cohort, 10,084 (53%) had a UPC index >75%. Of the 10,084 patients with a UPC index >75%, 6612 (66%) also saw their UPC physician within 90 days following their first maintenance dialysis session. Thus, we identified 6612 of 19,099 (35%) patients with high PCP continuity and 12,487 (65%) patients with low PCP continuity.

Figure 1.: Study flow diagram. PCP, primary care physician.

In the unmatched cohort of 19,099 patients (Supplemental Table 3), patients with high PCP continuity were significantly older (68 [13] versus 64 [15]; Sdiff = 0.27) and had lower Charlson comorbidity scores (3.6 [1.8] versus 3.8 [1.9]; Sdiff = 0.11). Patients with high PCP continuity also had fewer hospitalization days over the prior year (12.9 [18.3] versus 19.0 [26.0]; Sdiff = 0.27) and fewer hospitalization days within the 90 days that followed their first maintenance dialysis session (9.7 [16.7] versus 15.1 [28.9]; Sdiff = 0.23).

We matched 6391 (97%) patients with high PCP continuity 1:1 to similar patients with low PCP continuity (Table 1). In each group, 4901 (76.7%) patients started on in-center hemodialysis, 1442 (22.6%) patients started on peritoneal dialysis, and 48 (0.7%) patients started on home hemodialysis. The total person-years of follow-up were 10,852 years in the high-PCP continuity group and 10,760 years in the low PCP continuity group.

Table 1. - Characteristics of patients initiating maintenance dialysis in Ontario, Canada, propensity matched on the likelihood of high primary care physician continuity

Baseline Characteristics	Low Continuity, n=6391	High Continuity, n=6391	Standardized Difference
Demographics
Age (yr), mean (SD)	68 (14)	67 (13)	0.01
Women, %	2454 (38)	2445 (38)	0.00
Ethnicity, %
White	4653 (73)	4625 (72)	0.01
Black	285 (5)	289 (5)	0.00
Asian	453 (7)	459 (7)	0.00
Indian subcontinent	353 (5)	362 (6)	0.01
Other/unknown	647 (10)	656 (10)	0.01
Income quintile, %
1 (lowest income)	1479 (23)	1521 (24)	0.02
2	1461 (23)	1465 (23)	0.00
3	1310 (20)	1274 (20)	0.01
4	1146 (18)	1142 (18)	0.00
5 (highest income)	995 (16)	989 (15)	0.00
Rural residence, %	754 (12)	780 (12)	0.01
Rostered to a primary care physician, %	5456 (85)	5419 (85)	0.02
Primary cause of ESKD, %
GN	743 (11)	743 (12)	0.00
Cystic kidney disease	254 (4)	246 (4)	0.01
Diabetes	2529 (40)	2508 (39)	0.01
Vascular disease	1252 (20)	1246 (20)	0.00
Other	782 (12)	800 (12)	0.01
Unknown	831 (13)	848 (13)	0.01
Initial dialysis modality, %
In-center hemodialysis	4901 (76.7)	4901 (76.7)	0.00
Home hemodialysis	48 (0.8)	48 (0.8)	0.00
Peritoneal dialysis	1442 (22.6)	1442 (22.6)	0.00
Year of cohort entry, %
2005	386 (6)	474 (7)	0.06
2006	582 (9)	589 (9)	0.00
2007	616 (10)	602 (9)	0.01
2008	652 (10)	581 (9)	0.04
2009	599 (9)	587 (9)	0.01
2010	671 (11)	665 (10)	0.00
2011	607 (10)	605 (10)	0.00
2012	652 (10)	666 (10)	0.01
2013	718 (11)	713 (11)	0.00
2014	730 (11)	738 (12)	0.00
2015	178 (3)	171 (3)	0.01
Comorbidities, %
Diabetes	3921 (61)	3882 (61)	0.01
Heart failure	2607 (41)	2559 (40)	0.02
Myocardial infarction	864 (14)	856 (13)	0.00
Stroke/transient ischemic attack	290 (5)	290 (5)	0.00
Chronic obstructive pulmonary disease	625 (10)	610 (10)	0.01
Hypertension	6035 (94)	6019 (94)	0.01
Cancer	2462 (39)	2449 (38)	0.00
Chronic liver disease	614 (10)	616 (10)	0.00
Peripheral vascular disease	555 (9)	540 (8)	0.01
Depression	204 (3)	199 (3)	0.01
Chronic pain	428 (7)	428 (7)	0.00
Dementia	315 (5)	311 (5)	0.00
Charlson comorbidity index, mean (SD)	3.7 (1.9)	3.6 (1.8)	0.07
Health care utilization in the previous 1 yr, mean (SD)
No. of visits to all primary care physicians	15.4 (17.7)	15.4 (12.6)	0.00
No. of visits to most frequent primary care physician	7.2 (8.9)	12.8 (10.4)	0.58
No. of internal medicine visits	5.3 (8.3)	5.2 (7.8)	0.01
No. of nephrology visits	23.6 (11.3)	23.4 (10.9)	0.02
No. of endocrinology visits	0.9 (2.7)	0.9 (2.7)	0.01
No. of cardiology visits	5.2 (8.9)	5.2 (8.3)	0.00
No. of geriatric medicine visits	0.4 (2.14)	0.3 (1.9)	0.01
No. of psychiatry visits	0.3 (2.2)	0.3 (2.2)	0.00
No. of hospitalization days over prior 1 yr	13.5 (18.7)	13.2 (18.5)	0.02
No. of hospitalization days between first dialysis session and index date	10.2 (20.4)	9.9 (16.9)	0.01
Home care, %	2723 (43)	2672 (42)	0.02
Physician characteristics
Age (yr), mean (SD)	55 (11)	55 (10)	0.01
Women, %	1251 (20)	1242 (19)	0.01
No. of years since graduation, mean (SD)	29 (11)	29 (10)	0.01
International medical graduate, %	1231 (19)	1217 (19)	0.01
Hospital affiliation, %	2119 (33)	2131 (33)	0.00
Rural practice, %	656 (10)	658 (10)	0.00
Volume of patients on dialysis in the previous 1 yr, %
0	1182 (19)	1160 (18)	0.01
1	1408 (22)	1410 (22)	0.00
2	1223 (19)	1223 (19)	0.00
3	870 (14)	897 (14)	0.01
4	551 (9)	548 (9)	0.00
≥5	1157 (18)	1153 (18)	0.00

During the 2-year follow-up period, the high-PCP continuity group continued to have more continuity of care with their PCP with an average of 19.8 (21.0) total PCP visits, of which 12.3 (13.5) were to the UPC physician. Corresponding visits in the low-PCP continuity group included 17.8 (24.2) total PCP visits, of which 5.8 (10.1) were to the UPC physician.

Outcomes

High PCP continuity was not associated with a reduction in all-cause mortality (14.5 deaths per 100 person-years versus 15.2 deaths per 100 person-years; HR, 0.96; 95% CI, 0.89 to 1.02) (Figure 2, Table 2). There was also no difference in the rate of all-cause hospitalizations (HR, 0.96; 95% CI, 0.92 to 1.01). The most common reasons for hospitalization were complications of diabetes (n=1542, 12%; 7.5 events per 100 person-years), myocardial infarction (n=737, 6%; 3.5 events per 100 person-years), and heart failure (n=645, 5%; 3.1 events per 100 person-years) (Figure 3). Compared with patients with low PCP continuity, high PCP continuity was not associated with any disease-specific hospitalizations except for those related to diabetes (7.0 events per 100 person-years versus 8.0 events per 100 person-years; HR, 0.88; 95% CI, 0.80 to 0.97; P = 0.01). The results did not differ by the initial dialysis modality (interaction P = 0.54 for all-cause mortality and P = 0.13 for rate of hospitalizations) (Supplemental Figure 1), year of cohort entry (2005–2010 versus 2011–2015; interaction P = 0.46 for all-cause mortality and P = 0.87 for rate of hospitalizations), or hospitalization for ambulatory care–sensitive conditions (Supplemental Table 4).

Figure 2.: Kaplan–Meier survival curves for all-cause mortality, indicating no association between patients with high and low primary care physician (PCP) continuity.

Table 2. - Association of all-cause mortality and hospitalizations with primary care physician continuity in 6391 propensity-matched patients

Outcome and Exposure	No. (%) of Events	Event Rate (per 100 person-yr)	Hazard Ratio (95% Confidence Interval)
All-cause mortality
High continuity	1573 (25)	14.5	0.96 (0.89 to 1.02)
Low continuity	1631 (26)	15.2	Referent
Rate of hospitalizations
High continuity	10,675	154.6	0.96 (0.92 to 1.01)
Low continuity	10,997	162.1	Referent

Figure 3.: Association of time to first disease-specific hospitalization with primary care physician (PCP) continuity; compared with patients with low PCP continuity, high PCP continuity was not associated with any disease-specific hospitalizations except for those related to diabetes. Each patient could contribute to multiple disease-specific hospitalizations as long as it was their first hospitalization for the disease in question. Error bars indicate 95% confidence intervals (95% CIs). COPD, chronic obstructive pulmonary disease; MI, myocardial infarction; TIA, transient ischemic attack.

Sensitivity Analyses

When we changed the primary exposure to the receipt of any PCP visit versus no PCP visit during the 90 days after the first maintenance dialysis session (Supplemental Table 5), the results resembled the primary analysis (HR for mortality, 1.03; 95% CI, 0.94 to 1.13 and HR of rate of hospitalizations, 1.03; 95% CI, 0.97 to 1.09). These results did not change after including patients with fewer than three PCP visits in the 2 years prior to their first maintenance dialysis session (Supplemental Table 6). A definition of PCP continuity of care that incorporated PCP visits for 1 year after dialysis initiation found similar results as the primary analysis for mortality (HR, 0.97; 95% CI, 0.91 to 1.05), with a significant reduction in the rate of hospitalizations (HR, 0.89; 95% CI, 0.85 to 0.93) (Supplemental Table 7).

Discussion

In patients newly started on maintenance dialysis, PCP continuity of care before and during the transition to dialysis was not associated with a reduction in mortality or hospitalizations at 2 years. These results remained consistent across initial dialysis modalities (i.e., hemodialysis versus peritoneal dialysis) and different definitions of PCP involvement, except for a small reduction in hospitalizations when the definition of PCP continuity was changed to incorporate PCP visits for 1 year after dialysis initiation.

These results provide further insights into the controversy surrounding the role of PCPs in the care of maintenance dialysis recipients. Previous work has highlighted the marginalization of PCP involvement after patients commence dialysis (²⁹), which has been attributed to logistical (dialysis schedules and travel challenges) and clinical (complexity of patients) factors (³). In a survey of 233 nephrologists randomly selected from the American Society of Nephrology directory (²⁹), 90% provided primary care to patients on dialysis, which constituted 38% of their time devoted to such patients. Only 20% of nephrologists reported that their patients also had access to another PCP. A Canadian study in 196 nephrologists, 167 family physicians, and 147 patients on dialysis similarly demonstrated that most nephrologists spend one third of their time addressing issues that would conventionally be considered in the realm of primary care (³⁰). Most patients in this study visited a PCP and preferred going to their PCPs for general medical care, although 95% of PCPs surveyed provided care for fewer than or equal to two patients on dialysis and 40% lacked confidence in their knowledge and training to care for patients on dialysis. However, it remains unclear whether patients who consistently visit their PCP experience better outcomes at the same level as patients who are frequently seen by their nephrologist (⁸,²¹).

Despite the primary results showing that high PCP continuity likely has limited effect on mortality and hospitalizations for patients newly started on dialysis, we cannot rule out relative reductions below 15% as illustrated by the 95% CIs of the primary and sensitivity analyses. We also cannot rule out benefits for certain disease-specific hospitalizations and subgroups. We observed a reduction in hospitalizations for diabetes, which may relate to both an existing care gap and evidence-based guidelines for this population. For example, Winkelmayer et al. (³¹) previously demonstrated low rates of diabetic eye examinations and hemoglobin A1c testing after dialysis initiation, and it is possible that high PCP continuity in our study increased these preventive services to reduce diabetes-related hospitalizations. We also observed a reduction in hospitalizations when the definition of PCP continuity was changed to incorporate PCP visits for 1 year after dialysis initiation. This finding suggests that high PCP continuity may have some benefit in patients able to survive the first year of dialysis, but it requires confirmation in studies of prevalent patients on dialysis.

The limited benefit of primary care for patients newly started on dialysis may have several explanations: the adequacy of primary care provided by nephrologists may minimize the effect of primary care by PCPs; PCPs may require more awareness into the needs of patients on maintenance dialysis; or primary care itself may have limited abilities to reduce mortality and hospitalizations in this comorbid population. Our findings could help inform strategies to guide primary care for patients on dialysis, focusing on simple, low-cost tests and interventions given the minimal benefit from primary care observed. For patients with a PCP who want to maintain this relationship following the initiation of maintenance dialysis, knowledge translation activities are needed to promote evidence-based primary care practices that are beneficial, such as immunizations (³²), mental health (³³), and smoking cessation, as well as avoidance of practices that are unnecessary and costly, such as annual health examinations (³⁴) and cancer screening (³⁵,³⁶). Maintaining these PCP relationships may also have beneficial effects on access to health care resources (e.g., psychotherapy, physiotherapy, and home care) and treatment adherence, which may improve patient-centered outcomes beyond mortality and hospitalizations, such as symptom burden and psychosocial stress (³,³⁷).

For patients without a PCP or who are unable or unwilling to see their PCP after dialysis initiation, options include nephrologist-led primary care, patient management (³⁸), and the provision of comprehensive primary care by trained primary care providers in the dialysis unit (³⁹). The latter approach could be performed by physician assistants, nurse practitioners, or PCPs, with the additional benefit of maintaining continuity and expertise in primary care while also reducing patient visits and travel time to health care providers. An important message from these results given the limited observed benefit on mortality and hospitalizations is that decisions on who provides primary care for patients on dialysis should be patient driven rather than specialty driven.

This study has several strengths. We identified over 12,000 matched patients newly started on dialysis over the past 10 years across Ontario, Canada, which highlights the generalizability of this work. This includes patients on home dialysis, for whom little prior data exist. Our linked health care databases allowed for the ascertainment of PCP involvement, with little missing data and virtually complete patient follow-up. We had access to extensive data on sociodemographic characteristics, preexisting comorbidities, resource utilization, and dialysis modality, which allowed us to adjust for a broad array of potential confounders. We also focused only on incident patients to further reduce confounding. Lastly, we captured objectively measured patient-centered outcomes to complement the existing qualitative and survey literature.

Our study also has limitations. First, the results are from a single Canadian province with a single-payer, universal health care system, and they may not be generalizable to other populations, including prevalent patients on dialysis and health systems with less overall physician involvement where nephrologists do not already provide some degree of primary care. Second, there is no standardized definition for PCP continuity of care. The high-PCP continuity group continued to have more UPC visits during the 2-year follow-up, suggesting sustained PCP involvement, but the rate of hospitalizations was lower when we used a definition of continuity that incorporated PCP visits for 1 year after dialysis initiation. This sensitivity analysis should be interpreted cautiously given that it is a secondary outcome. We also could not determine nurse practitioner or physician extender involvement in our definition of primary care, but solo nurse practitioner/physician extender models are uncommon in Ontario (⁴⁰,⁴¹). Third, it is possible that PCP involvement may improve symptom burden, mental health, and quality of life; these data were not available in ICES but warrant further study. Fourth, we did not determine if differences existed in PCP ordering practices or access to health care. Now that we have demonstrated little effect on mortality and hospitalizations, future work will focus on describing PCP resource use and its associated costs. Fifth, some residual confounding is likely despite our best efforts because administrative data limit the ability to capture the severity of comorbid conditions.

Our population-based study found that high PCP continuity before and during the transition to maintenance dialysis was not associated with a reduction in mortality or hospitalizations at 2 years. These results did not substantially differ on the basis of dialysis modality. These findings suggest that calls for more PCP involvement in maintenance dialysis recipients may not necessarily confer better outcomes. In light of these results, additional work is needed to quantify the effect of PCP involvement on symptom burden and quality of life as well as simplify primary care for patients and providers in a manner that respects patient preferences, patient time, and health care resources.

Disclosures

Dr. Clemens reports receiving a Diabetes Canada Award sponsored by AstraZeneca and honoraria from Sutherland Global Services for an accredited Continuing Medical Education talk and reports attending conferences sponsored by Merck outside of the submitted work. Dr. Bota, Dr. Garg, Dr. Harel, Dr. McArthur, Dr. Naylor, Dr. Silver, Dr. Sood, and Dr. Wald have nothing to disclose.

Funding

This study was supported by the ICES Western site. ICES is funded by an Ontario Ministry of Health and Long-Term Care annual grant. Core funding for ICES Western is provided by the Academic Medical Organization of Southwestern Ontario, the Schulich School of Medicine and Dentistry, Western University, and the Lawson Health Research Institute. Dr. Silver is supported by the cofunded Kidney Foundation of Canada, Canadian Society of Nephrology, and Canadian Institutes of Health ResearchKidney Research Scientist Core Education and National Training Program New Investigator Award. Dr. Garg is supported by the Dr. Adam Linton Chair in Kidney Health Analytics and a Canadian Institutes of Health ResearchClinician Investigator Award.