End-of-Life Care Among Patients With Kidney Failure on Maintenance Dialysis: A Retrospective Population-Based Study

Study Design and Data Sources

We conducted a retrospective cohort study using linked population-level health administrative data in Ontario, Canada (see Supplementary File 1 for a description of the datasets accessed). Administrative datasets are linked using unique encoded identifiers and analyzed at ICES (formerly known as the Institute for Clinical Evaluative Sciences).

Study Cohort

Our cohort included adult patients who died between January 1, 2017, and December 31, 2019, and who had any record of long-term maintenance dialysis initiation (including peritoneal dialysis) before their death date in the Canadian Organ Replacement Register (CORR) (see Supplementary File 1 for a description of data sources accessed). We excluded patients who were aged >105 years at death (in case of documentation errors in the patients’ birth or death dates) and those who were ineligible for the Ontario Health Insurance Plan (OHIP) at any point in the last year of life. Other exclusion criteria included patients who resided in a long-term care home in the last year of life, as these individuals receive different care patterns to patients residing in the community, and patients who did not have any weekly dialysis billing codes in the last year of life.

Physician Care Patterns

Our study exposure examines the mix of care provided by different physician specialties in the last year of life. Specifically, we focused on nephrology due to the nature of the population of interest, palliative care due to the end-of-life context, and primary care as primary care is responsible for most outpatient care provision in the last year of life for the general population. ⁹ We defined mutually exclusive categories of different outpatient physician patterns of care during the last year of life (Figure 2). First, patients who did not have an outpatient encounter with a primary care physician in the last year of life were categorized into the non-primary care pattern of care. Patients in this category may have any combination of encounters with all other specialties. Patients who did have an outpatient encounter with a primary care physician were categorized into the following categories: (1) primary care, nephrology, and palliative care triad and (2) primary care and nephrology care dyad. Patients were categorized into the care triad if they had at least one encounter with a primary care physician, nephrologist, and palliative care physician in the last year of life. Similarly, patients were categorized into the care dyad if they had at least one encounter with primary care and nephrology but did not have any palliative care encounters.

Physician specialty was obtained through OHIP billing records for all specialties except for palliative care as any physician can bill palliative care codes regardless of their specialty. We adapted a previously defined algorithm that is highly sensitive and has positive predictive value in identifying self-reported palliative care physicians. ¹⁵ The algorithm identified palliative care physicians as physicians with greater than 10% of their billings over 2 years as palliative care. We applied this definition to physicians who billed for palliative care services during the first calendar year of our study period (January 1, 2017, to December 31, 2017) and looked back 2-years before each of the unique palliative care billing codes to identify an average percentage of their billing patterns.

Outcomes

The outcomes of interest capture outpatient and inpatient care patterns in the last year of life. First, we measured the number of outpatient physician encounters by the outpatient physician care pattern categories in the last year of life, using individual outpatient billing codes in OHIP. We identified that this approach was inappropriate for nephrologist encounters who typically bill weekly maintenance dialysis billing codes and do not bill for each encounter. Hemodialysis patients will typically receive dialysis 3 times per week, wherein they interact with their dialysis care team, which can include the nephrologist and registered staff. As such, we supplemented the original definition by approximating outpatient encounters with nephrologists by the number of weekly maintenance dialysis OHIP billing codes (Supplementary File 2).

Next, we measured how end-of-life acute care utilization varies by patterns of physician care. Specifically, we examined the number of unique hospitalizations or emergency department visits in the last 30 days of life, the number of days spent in hospital or emergency department in the last 30 days of life, and the percentage of patients that died in acute care settings. We also reported patients’ acute care utilization in the last year of life, truncating the last 30 days of life. We used the Discharge Abstract Database and the National Ambulatory Care Reporting System to capture inpatient hospitalization and the Registered Persons Database to capture patient death locations: acute (i.e., emergency departments and inpatient hospital settings), sub-acute (i.e., complex continuing care and rehabilitation) or other (i.e., primarily in the community). ¹⁶

Patient Characteristics

Patient sociodemographic and clinical characteristics were reported. Neighborhood-level income and rurality status were identified using the patients’ postal code one year before their death date. Prevalence of chronic conditions was obtained using a 5-year look back period from the patients’ death date, using previously developed algorithms to assign the prevalence of 18 conditions.^17-25 Comorbidity status at 1 year before death was also captured using the Johns Hopkins Adjusted Clinical Group (ACG)® System Aggregated Diagnosis Groups, Version 10. ²⁶ All ICD-9/ICD-9-CM diagnosis codes assigned to patients are categorized into 32 functional groups, based on the diagnostic certainty, duration, severity, and etiology of the condition, and the likelihood of requiring specialist care. ²⁷ Dialysis location, type, and duration of time between dialysis initiation and death were captured using the CORR. Patients’ primary care physician roster status was captured using the Client Agency Program Enrolment table for patients formally attached to a primary care physician and the rostered primary care physician was the primary care physician with the highest proportion of encounters in the last year of life, for patients not formally attached. ²⁸

Analyses

Descriptive results were presented as proportions for categorical variables and as mean and standard deviation (SD), or medians and interquartile ranges (IQR) for skewed continuous variables. We reported study outcomes, stratified by the exposure variable, by producing means, SDs, medians, and IQRs. We also produced a time series plot reporting monthly percentages of outpatient encounters by physician specialties in the last year of life. We measured health utilization differences across patterns by conducting an Analysis of Variance (ANOVA) for numeric variables and Pearson’s chi-square tests for categorical variables. Statistical significance was set at P-value <.05 (two-tailed). All analyses were completed using SAS Enterprise Guide v. 7.15.

Ethics Approval

This study was approved by the Hamilton Integrated Research Ethics Board on March 28, 2022 (#14750-C). ICES is an independent, non-profit research institute whose legal status under Ontario’s health information privacy law (section 45 of Ontario’s Personal Health Information Protection Act) allows it to collect and analyze healthcare and demographic data, without consent, for health system evaluation and improvement.

Main Findings

In this study of patients who died with KF on maintenance dialysis in Ontario, Canada, patients received most of their outpatient care from nephrologists in the last year of life. This finding is unsurprising as patients with kidney failure who receive in-center hemodialysis receive dialysis 3 times a week and see a nephrologist once a week. ²⁹ Therefore, their outpatient care is anchored to their dialysis treatment, and therefore, they have high continuity with their nephrology team. ⁴ Among patients who experienced primary care encounters, primary care physicians were also consistently involved in the last year of life, albeit on a smaller scale compared to nephrologists. This phenomenon may be specific to patients with KF on maintenance dialysis as primary care plays a much more prominent role compared to specialists in end-of-life outpatient care among the general decedent population. ⁹ Patients with KF on maintenance dialysis have traditionally looked to their nephrology team for advance care planning, symptom management, and social and emotional support. ³⁰ This relationship is appropriate; however, primary care physicians may also be well-suited to provide these services due to their longitudinal relationships with their patients. ³¹ Given the potential benefits of primary care involvement in providing end-of-life care, it is notable that almost 15.5% of patients did not have any encounters with primary care during the last year of life; care was managed primarily by their dialysis team in these cases. Patients who wish to discontinue dialysis may experience a disconnect in care if primary care is not involved as they navigate non-dialysis-related healthcare services. Overall, most patients had continued contact with their primary care physician during their last year of life, in addition to other specialties.

Palliative care physician involvement increased as patients approached death, with large increases occurring in the last 3 months of life. Specialist palliative care is not being initiated early in the last year of life and is concentrated near death for patients dying with KF on maintenance dialysis. This finding aligns with other studies that identified that patients on dialysis access palliative care late in the end-of-life period.^1,2,32,33 Patients with primary care involvement experienced greater palliative care involvement from both provincial home care services and palliative care physicians, suggesting that patients without primary care involvement may experience barriers to palliative care access or that nephrologists are providing end-of-life care for their patients. Furthermore, almost 1 in 5 monthly encounters were attributed to other specialists, beyond primary care, nephrology, or palliative care. Therefore, patients dying with KF on maintenance dialysis are experiencing multiple comorbidities that require medical attention from other specialists, resulting in more outpatient encounters during the last year of life.

End-of-life healthcare utilization did not vary extensively for patients dying with KF on maintenance dialysis across patterns of care. Almost 80% of patients in each pattern of care experienced at least one hospitalization or emergency department visit in the last 30 days of life—which is higher than previously reported research. ¹ Patients dying with KF on maintenance dialysis may require more inpatient care provision at the end of life than other populations due to complications associated with reduced intervention effectiveness or other comorbidities (i.e., myocardial infarctions and infections). Patients who had palliative care physician involvement experienced lower rates of death in acute care settings, despite being more medically complex, as demonstrated by higher percentages of home care services enrolment, cancer prevalence, and spending more days in hospital or emergency departments in the last 30 days of life and having the greatest acute care utilization in the 11 months before the last 30 days of life. These findings complement existing literature that identified that patients on dialysis who received home palliative care services were less likely to die in acute or sub-acute care settings compared to those who did not receive home palliative care services. ¹ Therefore, involving palliative care physicians may facilitate patients being discharged from inpatient settings to pass away in their homes, which is often the preferred location of death for patients.^34-37 However, the timing and rationale of palliative care involvement remain uncertain, and future research ought to explore the characteristics associated with earlier palliative care referral.

Strengths and Limitations

This project has strengths, such as leveraging existing routinely collected health administrative data from Ontario, Canada. This data includes deidentified patient-level health information from approximately 13 million people with access to near-universal healthcare. As such, our findings may be relevant to other jurisdictions with public healthcare systems that also provide strong primary care services to their patients. This study also has limitations. First, we relied on different definitions for outpatient encounters with nephrology, as nephrologists do not typically bill for individual encounters for patients with KF on maintenance dialysis. As such, we may be overestimating the number of outpatient encounters that patients experience with their nephrologists by relying on individual encounter billing codes and weekly dialysis management billing codes. Some patients may receive primary care from nurse practitioners who do not have capturable OHIP physician billing codes. Therefore, patients in the non-primary care pattern may still benefit from having a primary care provider involved. Similarly, patients in the primary care and nephrology care dyad may have received palliative care services from their primary care physician or nephrologist. We also excluded patients who resided in long-term care homes during the last year of life as they may experience different outpatient care patterns compared to patients residing in the community. However, approximately 5% of chronic dialysis patients reside in long-term care homes in Ontario, and more long-term care homes have received funding to deliver peritoneal dialysis between 2012 and 2019, ³⁸ and therefore, the number of patients receiving dialysis in long-term care is expected to increase. Next, our study only includes data before the COVID-19 pandemic, and therefore, may not reflect contemporary outpatient patterns. Lastly, this was a descriptive study, and therefore, cannot infer causal relationships between patterns of outpatient physician care and end-of-life healthcare utilization as there is temporal ambiguity. As a result, a separate study will be conducted to measure the associations between patterns of physician care and end-of-life healthcare outcomes, adjusting for patients’ demographic and clinical characteristics.