A Comparison of Hospitalization Outcomes Between Peritoneal Dialysis and Home Hemodialysis Patients by Sex and Race

Abstract

Introduction

As interest in home dialysis as an initial dialysis modality grows, it remains unclear how the different home dialysis modalities may impact hospitalization outcomes, or how this relationship may change depending on patient sex and race.

Methods

We compared all-cause hospitalization rates and days in hospital between incident peritoneal dialysis (PD, n = 14,643) and home hemodialysis (HHD patients, n = 875) between January 2005 and December 2018 (last follow-up was in July 2020) using a nationally representative cohort of incident dialysis patients.

Results

The overall hospitalization rate was 0.82 hospitalization events per patient-year. Compared with those initiated on PD, HHD patients had a lower hospitalization rate (incident rate ratio [IRR] = 0.78, 95% confidence interval [CI] 0.71–0.85), and spent fewer days in hospital (IRR = 0.68, 95% CI: 0.59–0.78). This was more pronounced in more contemporary cohorts and for males. The protective effect of HHD was stronger for Black patients. When hospitalizations were analyzed by cause, the protective effect of HHD was stronger for infection-related admissions, with Black patients seeing the largest benefit.

Conclusion

The type of home modality at dialysis initiation is associated with differences in hospitalization outcomes, an association that is stronger in selected racial groups and sexes. While exploratory in nature, our work highlights the importance of further study on the differential impact of PD and HHD on hospitalization outcomes so that patients incident to dialysis may make an informed decision.

Graphical abstract

Home dialysis use, particularly HHD is on the rise globally,¹ including in Canada, where a “home first” approach is often used.² Considering that the choice of modality is typically left to the patient when medical contraindications are absent,³ this increase may be attributable to patient preference because of greater treatment flexibility, as well as a growing body of evidence suggesting improved clinical outcomes and reduced healthcare costs as compared with in-center hemodialysis.^1,2,4,5

Despite this, there remains a paucity of studies comparing patients treated with HHD with those receiving PD.6, 7, 8, 9 When HHD and PD have been compared, HHD patients have been found to have a lower mortality risk,^7,9, 10, 11, 12 with a study of incident dialysis patients in Canada reporting a 36% lower mortality than those on PD. However, this difference was attenuated in contemporary eras.¹³

Patients receiving dialysis experience poor health outcomes, including high rates of morbidity, mortality, high symptom burden, and frequent and prolonged hospitalizations.14, 15, 16, 17 In Canada, patients treated with dialysis are hospitalized 10 times more frequently than the general population.^6,18 To date, only a few observational studies comparing hospitalization outcomes between PD and HHD have been reported, all of which found that HHD was associated with lower hospitalization rates and/or fewer days spent in hospital.^7,8,19,20 Many possible explanations exist to explain this; many HHD patients use an arteriovenous fistula for access, that has been associated with fewer infectious complications.²¹ In addition, cardiovascular complication rates are lower in those on HHD than in those on PD,²² likely contributing to this differential risk of hospitalization.

Although the results of these studies are valuable, there continue to be important knowledge gaps, both for patients and providers. One such gap is how outcomes between different sex and racial groups differ for patients receiving home dialysis.²³ Further efforts to expand the uptake of home dialysis modalities have led to an increase in patient complexity, with higher risk populations now being considered for home dialysis in contemporary eras. However, the impact of era on hospitalization risk (comparing HHD with PD) has not been extensively studied.^13,24 Acknowledging these gaps, this study aimed to do the following: (i) investigate the differences in all-cause hospitalization outcomes between incident HHD and PD patients across Canada and (ii) assess if the risk of hospitalization was modified by sex, race, or era of dialysis initiation.

Methods

Data Source and Study Population

We conducted a retrospective cohort study of all adult patients with kidney failure who initiated home dialysis within 90 days of kidney replacement therapy (KRT) start, between January 1, 2005 and December 31, 2018, in Canada. We obtained the cohort data from the Canadian Organ Replacement Register (CORR) and the Discharge Abstract Database; all participating centers have a formal pathway for assessing and educating patients regarding dialysis modalities. The CORR retrospectively captures demographic and clinical outcome data on patients initiating KRT, including dialysis and transplantation for all individuals living with end-stage organ failure in Canada,²⁵ and has been previously validated for clinical research.²⁶ The Discharge Abstract Database collects administrative, clinical, and demographic information on hospital discharges and the most responsible diagnosis (using the International Classification of Diseases Version 10 CA codes) for all patients in Canada.²⁷ This study was approved by the Nova Scotia Health Research Ethics Board. Patients residing in Quebec and Manitoba were excluded because of the lack of inclusion in the CORR database for the years covering the cohort (Quebec) and the inability to link hospitalization data (Manitoba). Patients who had not initiated a home dialysis modality within the first 90 days after starting KRT, had previously had a kidney transplant,²⁸ patients who belonged to the age group 15 to 19 years or less at dialysis initiation, and those who had sex coded as “other” were excluded.

Exposure Assessment

The exposure variable of interest was incident home dialysis modality (i.e., PD or HHD). HHD was defined as any hemodialysis regimen completed at home, and PD included both continuous ambulatory PD and automated PD. We elected to combine all HHD prescription types because it was not anticipated that this would impact hospitalization outcomes based on previous study,²⁹ and because of the possibility of inaccuracy with the classification of each subtype of HHD (leading to misrepresentation of results). Patients who transitioned to a home dialysis modality within 90 days of their dialysis start date were also included.

Outcomes

The outcomes of interest were all-cause hospitalization rate, cumulative days in hospital, and time to the first hospitalization. Hospitalization rates and days in hospital included all hospitalizations that occurred in patients while receiving home dialysis during the follow-up period. Patients were followed-up with from initiation of home dialysis until the earliest occurrence of a censoring event, including transplant, death, withdrawal from dialysis, transition to a nonhome-based dialysis modality for greater than 30 days, loss to follow-up or end of the study period (July 1, 2020).

Covariable Assessment

Covariates included in the adjusted models were assessed at KRT initiation and included age group, sex, race, cause of end-stage kidney disease (ESKD), relevant comorbidities, body mass index, era of dialysis initiation, late referral to a nephrologist, and income quintile. For privacy reasons, age is captured in the CORR as a continuous variable coded as the median value of 5-year age groups. Both sex and race were clinician-identified on the initial registration form. Racial groups were defined as White and non-White (Black, Asian, Indian Sub-Continent, Pacific Islander, Middle Eastern/Arabian, Latin American, and other/multiracial) when used as a covariate using the standard categories provided by the CORR form. The primary cause of ESKD was based on the diagnosis coded on the CORR registration form. A detailed breakdown of race categories as defined by the CORR can be found in Supplementary Table S1. Comorbidities included the presence or absence of coronary artery disease, congestive heart failure, cerebrovascular disease, peripheral vascular disease, chronic obstructive lung disease, prior malignancy, and hypertension. Late nephrology referral was defined as a referral to a nephrologist < 90 days before KRT initiation. Other covariables considered in sensitivity analyses included access type at dialysis initiation (central venous catheter, arteriovenous fistula or graft) and estimated glomerular filtration rate calculated using the 2021 Chronic Kidney Disease Epidemiology Collaboration formula.³⁰

Statistical Analysis

Descriptive statistics were reported stratified according to home modality. Hospitalization rates (hospitalization events/total time at risk) and days in hospital (cumulative days in hospital/total time at risk) were reported as a count per patient-year and stratified according to home modality. Patients who experienced no hospitalization events only contributed to the total time at risk. Patients who switched dialysis modalities were censored only if the switch lasted at least 30 days or resulted in death. In addition, a sensitivity analysis using a 60-day definition for treatment failure was conducted with comparable results (data not shown). Comparison of the hospitalization rates and days in hospital between patients treated with PD or HHD was done using multivariable negative binomial regression models and reported using adjusted IRRs with 95% CIs. Time to first hospitalization was modelled using a multivariable adjusted Cox proportional hazards model with 95% CIs. Missing data were minimal (approximately < 5%) and were handled through multiple imputation by chained equations.

Prespecified interactions were tested between home dialysis modality and the following covariables: sex (male and female), race (White, Black, and Other), and era of dialysis initiation based on the previous literature.¹³ All interactions were stratified regardless of statistical significance. The eras were selected by dividing the 13 years of follow-up into the following categories: era 1 (2005–2009), era 2 (2010–2014), and era 3 (2015–2018).

We repeated all analyses with the exposure defined as those who initiated home dialysis within 180 days and 365 days of KRT, respectively. In addition, the primary outcome was repeated with estimated glomerular filtration rate added to the models with the a priori decision to report both models if adding estimated glomerular filtration rate resulted in a ≥ 10% change in effect size. We assessed if the type of baseline HHD access the patient used at dialysis initiation influenced hospitalization outcomes. Lastly, the primary outcome was repeated as a cause-specific analysis, looking specifically at cardiovascular and infection-related hospitalizations. Analyses were stratified according to HHD access type to compare those receiving HHD via catheter to PD and with those receiving HHD via a surgically created arteriovenous access to PD. All statistical analyses were conducted using Stata/SE software version 18.0 (StataCorp. 2021. Stata Statistical Software: Release 17.0. College Station, TX: StataCorp LLC).

Results

In total, 63,327 patients were registered as having started KRT within the study period. Of these patients, 47,560 were excluded for not initiating a home dialysis modality within 90 days of dialysis start. An additional 247 patients were excluded for age < 19 years, and 2 were excluded for not having a sex listed. This left a total of 15,518 patients, with 14,643 initiating PD with another 875 initiating HHD (Figure 1).

Figure 1.

Derivation of study cohort for primary analysis. KRT, kidney replacement therapy.

Patients treated with HHD were younger, more likely to be male, White, belong to a high-income quintile, and had fewer comorbidities compared with patients treated with PD. The most common cause of ESKD was diabetes for both PD and HHD (Table 1).

Table 1.

Baseline characteristics of the study population stratified by home modality

Characteristica	Home Dialysis		P	Missing Data
	PD (n = 14,643)	HHD (n = 875)
Age, median yrs (IQR)	62 (52–72)	52 (47–62)	< 0.001	0
Sex, male	8990 (61.4)	587 (67.1)	0.001	0
White race	9384 (66.3)	637 (75.2)	< 0.001	523 (3.4)
BMI (kg/m2)			< 0.001	936 (6.0)
≤18.5	344 (2.5)	18 (2.3)
18.5–24.9	4858 (35.2)	225 (28.6)
25–29.9	4592 (33.3)	229 (29.1)
≥30	4002 (29.0)	314 (40.0)
Cause of kidney failure			< 0.001	0
Diabetes	5756 (39.3)	254 (29.0)
Glomerulonephritis/ autoimmune diseases	2509 (17.1)	197 (22.5)
Renovascular diseases	2480 (16.9)	90 (10.3)
Congenital/hereditary/ polycystic kidney disease	1192 (8.1)	145 (16.6)
Other/unknown	2706 (18.5)	189 (21.6)
Comorbidities
Coronary artery disease	3214 (23.0)	124 (14.6)	<0.001	714 (4.6)
Congestive heart failure	1800 (13.1)	87 (10.4)	0.023	955 (6.2)
Cerebrovascular disease	1405 (10.2)	56 (6.6)	0.001	861 (5.6)
Peripheral vascular disease	1627 (11.8)	71 (8.4)	0.003	914 (5.9)
Chronic obstructive lung disease	969 (7.1)	40 (4.8)	0.012	959 (6.2)
Prior malignancy	1377 (10.1)	115 (14.0)	< 0.001	1108 (7.1)
Hypertension	11,818 (84.9)	729 (85.8)	0.511	754 (4.9)
Era			< 0.001	0
2005–2009	4514 (30.8)	166 (19.0)
2010–2014	4902 (33.5)	396 (45.3)
2015–2018	5227 (35.7)	313 (35.8)
Late nephrology referral (<3 mo)	1476 (10.5)	105 (12.6)	0.049	594 (3.8)
Income quintile			< 0.001	366 (2.4)
5 (high)	2177 (15.2)	187 (22.0)
4	2448 (17.1)	185 (21.8)
3	2901 (20.3)	150 (17.7)
2	3281 (22.9)	179 (21.1)
1 (low)	3495 (24.4)	149 (17.5)
eGFR (ml/min), median (IQR)b	8.2 (6.3–10.6)	7.5 (5.8–10.1)	< 0.001	489 (3.2)
Dialysis access			< 0.001	140 (0.9)
Central venous line		410 (48.2)
Arteriovenous fistula or graft		441 (51.8)

BMI, body mass index; eGFR, estimated glomerular filtration rate; HHD, home hemodialysis; IQR, interquartile range; PD, peritoneal dialysis.

^aResults are presented as count (percentage) unless specified.

^beGFR at kidney replacement therapy initiation.

Hospitalizations and Cumulative Hospital Days

Overall, 10,112 patients (65.2%) experienced ≥ 1 hospitalizations during the follow-up period (47,036 total person-years at risk). Patients receiving HHD spent a median of 5 cumulative days in the hospital (interquartile range: 0–14) and patients receiving PD spent a median of 10 cumulative days in the hospital (interquartile range: 1–29). Crude hospitalization outcomes for patients who experienced ≥ 1 admissions during follow-up can be found in Supplementary Table S2.

Patients' incident to HHD experienced a lower crude hospitalization rate of 0.64 per patient-year compared with a rate of 0.84 per patient-year for PD (IRR = 0.78, 95% CI: 0.71–0.85), and spent 5.1 days per patient-year compared with 9.1 days per patient-year for PD (IRR = 0.68, 95% CI 0.59–0.78) (Table 2).

Table 2.

Rate of admission and days in hospital comparing PD to HHD in the entire cohort and across eras

	Entire cohort (n = 15,518)		Era 2005–2009 (n = 4680)		Era 2010–2014 (n = 5298)		Era 2015-2018 (n = 5540)
	Crude ratea	IRR (95% CI)	Crude ratea	IRR (95% CI)	Crude ratea	IRR (95% CI)	Crude ratea	IRR (95% CI)
Admissionb,c
PD	0.84	1.00 (ref)	1.03	1.00 (ref)	0.91	1.00 (ref)	0.59	1.00 (ref)
HHD	0.64	0.78 (0.71–0.85)	1.15	0.89 (0.76–1.06)	0.61	0.66 (0.59–0.76)	0.40	0.84 (0.71–1.00)
Days in hospitalb,d
PD	9.1	1.00 (ref)	12.1	1.00 (ref)	9.4	1.00 (ref)	6.2	1.00 (ref)
HHD	5.1	0.68 (0.59–0.78)	11.0	0.71 (0.55–0.91)	4.4	0.55 (0.45–0.66)	2.86	0.82 (0.62–1.07)

CI, confidence interval; HHD, home hemodialysis; IRR, incident rate ratio; PD, peritoneal dialysis.

^aExpressed as the rate per patient-year.

^bAdjusted for sex, age, race, body mass index, kidney failure cause, coronary artery disease, congestive heart failure, cerebrovascular disease, peripheral vascular disease, chronic obstructive pulmonary disease, prior malignancy, hypertension, era of dialysis initiation, late nephrology referral, and income quintile.

^cP-value for era interaction: 2010–2014 = 0.064; 2015–2018 = 0.905.

^dP-value for era interaction: 2010–2014 = 0.503; 2015–2018 = 0.265.

The median time to first hospitalization was 535 days (1.47 years) for patients using HHD compared with 414 days (1.13 years) for patients using PD. Compared with those using PD, patients using HHD had a hazard ratio of 0.85 (95% CI: 0.77–0.92) for experiencing a hospitalization event (Table 3).

Table 3.

Time to first hospitalization comparing PD to HHD and accounting for the competing risks death and kidney transplant

	PD	HHD
	HR (95% CI)
Entire cohorta (n = 15,041)	1.00 (ref)	0.85 (0.77–0.92)
Erab
2005–2009 (n = 4539)	1.00	0.90 (0.75–1.08)
2010–2014 (n = 5128)	1.00	0.73 (0.64–0.83)
2015–2018 (n = 5374)	1.00	1.01 (0.86–1.19)
Sexc
Female (n = 5736)	1.00	0.94 (0.80–1.09)
Male (n = 9305)	1.00	0.80 (0.72–0.89)
Racial groupd
Non-White (n = 4463)	1.00	0.79 (0.66–0.96)
White (n = 8960)	1.00	0.85 (0.76–0.95)

HHD, home hemodialysis; HR, hazard ratio; PD, peritoneal dialysis.

^aAdjusted for sex, age, race, body mass index, kidney failure cause, coronary artery disease, congestive heart failure, cerebrovascular disease, peripheral vascular disease, chronic obstructive pulmonary disease, prior malignancy, hypertension, era of dialysis initiation, late nephrology referral, and income quintile.

^bAdjusted for all variables above excluding era of dialysis initiation.

^cAdjusted for all variables above excluding sex.

^dAdjusted for all variables above excluding racial group. Done as complete case analysis.

Era Effect and Hospitalizations

When patients were separated according to era of dialysis start, the relationship between dialysis modality and number of admissions was only significant in more contemporary eras. With regard to length of stay, this was only statistically different between 2005–2010 and 2010–2014 (Table 2).

Subgroup Analysis

There was no statistically significant interaction between home dialysis modality and sex for any of the models. However, the association between HHD and shorter length of stay was significant for both women (IRR: 0.76, 95% CI: 0.60–0.97), and men (IRR: 0.65, 95% CI: 0.55–0.76) (Figure 2). There was a statistically significant interaction between home dialysis modality and Black race when modeling for days in hospital (interaction P < 0.001). Although all racial groups showed a statistically significant reduction in total admissions and days in hospital when initiated on HHD, the effect was most pronounced for Black patients, in particular when modeling for days in hospital (Figure 2).

Figure 2.

Adjusted incident rate ratios of all-cause hospitalization and days in hospital for patients using HHD compared with patients using PD for the entire cohort, stratified by sex and race. Adjusted P values for sex were 0.1 for all-cause hospitalizations, 0.774 for days in hospital. Adjusted P values for Black race were 0.056 for all-cause hospitalizations, < 0.001 for days in hospital. HHD, home hemodialysis; PD, peritoneal dialysis.

Sensitivity Analyses

Analyses were repeated restricting the exposure to those who initiated home dialysis within 180 days of KRT initiation (n = 15,097) and 365 days of KRT initiation (n = 16,396), respectively. These exposure definitions resulted in consistent findings to the primary analysis exposure definition (home dialysis use by 90 days of KRT initiation) (Supplementary Table S4). In the additional sensitivity analysis comparing hospitalizations by initial HHD access type, patients receiving HHD who initiated dialysis using a central venous catheter had higher hospitalization rates and longer hospital stays than those who initiated dialysis with a fistula or graft (Supplementary Table S3). Adding estimated glomerular filtration rate to our adjusted models did not result in any significant change in effect size (data not shown).

The primary analysis was repeated for cause-specific hospitalizations, considering admissions related to cardiovascular and infectious causes using International Classification of Diseases Version 10 codes (Supplementary Table S5, with crude rates of hospitalizations and days in hospital in Supplementary Table S6). When only cardiovascular-related admissions were considered, there was no significant interaction between total number of hospitalizations and sex or race. However, a significant interaction was found between total days in hospital and sex (adjusted P value: 0.034). In addition, the association between HHD and shorter length of stay was only significant for women (IRR: 0.40, 95% CI: 0.23–0.69), and not men (Figure 3).

Figure 3.

Incident rate ratios of cause-specific hospitalization and days in hospital for patients using HHD compared with patients using PD for the entire cohort and stratified by sex and race. Panel (a): cardiovascular hospitalizations, Panel (b): infection-related hospitalizations. Adjusted P values for sex were 0.58 for cardiovascular hospitalizations, 0.034 for days in hospital. Adjusted P values for Black race were 0.617 for cardiovascular hospitalizations, 0.068 for days in hospital. Adjusted P values for sex were 0.514 for infection-related hospitalizations, 0.027 for days in hospital. Adjusted P values for Black race were 0.095 for infection-related hospitalizations, 0.001 for days in hospital. HHD, home hemodialysis; PD, peritoneal dialysis.

When only infection-related admissions were considered, there was no significant interaction between total number of hospitalizations and sex or race. However, significant interactions were found between total days in hospital for infection-related admissions and sex (adjusted P value: 0.027) and Black race (adjusted P value: 0.001). In addition, the association between incident HHD patients and length of stay was significant for men (IRR: 0.57, 95% CI: 0.42–0.77), but not women. The association was also stronger for Black patients (IRR: 0.16, 95% CI: 0.06–0.44) than for White patients (IRR: 0.62, 95% CI: 0.46–0.84) (Figure 3).

Discussion

In this national cohort study, incident home dialysis initiation was associated with decreased hospitalization rates, fewer days in hospital, and a longer time to first hospitalization compared with patients initiating PD. These findings were most evident in our subgroup analyses, with the association being stronger for Black patients. Our results were consistent across eras of time and in several sensitivity analyses.

Our overall hospitalization rate for this home dialysis cohort fits well with previous reports in similar populations. A US study from 2008 comparing hospitalization rates for daily hemodialysis (0.68 hospitalizations per patient-year) and PD (0.76 per patient-year) were comparable with ours (0.64 for HHD and 0.84 for PD).¹⁹ A more recent study using United States Renal Data System data found slightly higher rates (0.93 admission per patient-year for daily hemodialysis, 1.35 for PD).⁸ Although rates were overall comparable, a possible explanation for the lower rates of hospitalization between the US and Canada for home dialysis patients may relate to clinical experience, given the relatively higher proportion of Canadian patients with ESKD receiving dialysis at home.³¹ In addition, CORR data is gathered from a single-payer model (Canada), whereas United States Renal Data System data reflects only traditional Medicare; therefore, patients with private insurance, who may be at lower risk for hospitalization, might be excluded.

Interestingly, we found a significant decline in hospitalization rates for both HHD and PD patients over eras. Although our hospitalization rates were overall lower than those reported for similar patients in the US, a similar trend has been reported in PD patients in the US, with United States Renal Data System data showing a decrease in hospitalization rates for PD patients from 1.73 hospitalizations per patient-year in 2010 to 1.43 hospitalizations per patient-year in 2020.³² Over a similar period (era 2 to era 3), we found that crude hospitalization rates for PD patients decreased from 0.91 hospitalizations per patient-year in era 2 to 0.59 hospitalizations per patient-year in era 3, representing a similar net change in hospitalizations across countries. Further study is needed to better understand the potential contributors to improved hospitalization outcomes in this population.

Our findings of reduced hospitalization events in patients receiving HHD adds to the growing body of literature that HHD is associated with improved patient outcomes compared with other modalities.^7,8,19,20 Multiple rationales for this association have been proposed. Some HHD modalities may involve a more intensive dialysis prescription, leading to optimization of middle molecule clearance, fluid balance,³³ phosphate control,³⁴ and cardiovascular health (particularly regression of left ventricular hypertrophy).³⁵ In contrast, it is likely that our findings are influenced by differences in patient characteristics between those on PD and HHD. Previous research has shown that there are systematic differences between patients who initiate PD and those who initiate HHD (both in characteristics and process of initiation of chronic dialysis) that may in turn contribute to hospitalization rates.^9,10,36,37 Although we adjusted for many of these factors, the likelihood of residual confounding is high.

Although the interaction of sex did not meet statistical significance, the point estimate for hospitalization was lower for males than for females. Interestingly, there was no difference in the crude hospitalization rate between sexes using PD, and very little difference for days in hospital; however, in those using HHD, males had both a significantly lower crude hospitalization rate and fewer days in hospital. Socioeconomic factors such as education level, access to care, social supports, and appropriate housing conditions are potential reasons for this difference.³⁸ We were unable to assess for these factors because of the nature of our data; however, a recent study comparing sex differences in mortality for patients using dialysis found that more males were married (72% of males vs. 47% of females) and employed (53% of males vs. 39% of females) whereas more females were widowed (26% of females vs. 7% of males) and had less than high school education (26% of females vs. 18% of males), which may indirectly influence hospitalization outcomes.^8,20,37,39

In this study, Black patients on HHD experienced significantly fewer hospitalizations and days in hospital than those on PD. Previous work has shown that dialysis patients belonging to racial minorities have fewer hospitalizations and improved survival.^33,40, 41, 42 One study of patients receiving in-center hemodialysis found that Black patients had an adjusted relative risk of 0.95 (95% CI: 0.94–0.96) for hospital admissions compared with white patients. However, they did not compare admission rates or time in hospital by dialysis modality, as we did in our study, and were limited by a short follow-up time (median of 1 year follow-up).

There are several potential explanations for the more pronounced impact of home dialysis modality on hospitalization outcomes for Black patients. First, individuals belonging to a racial minority group are less likely to receive a kidney transplant than White patients.⁴² As a result, it is possible that the racial minority group on dialysis are “healthier” patients, because the healthiest White patients have been transplanted. Alternatively, it is known that chronic kidney disease is more prevalent in racial minority groups and progresses to ESKD faster³⁷; therefore, another important consideration is that Black patients who had a higher illness burden at the time of dialysis initiation may have died before they received appropriate care and initiated dialysis, resulting in a survival bias. Regardless, we still observed a strong protective effect of HHD over PD in this population, suggesting that the choice of modality has a larger impact in some patient groups. Previous work showing an increased rate of peritonitis in Black patients may help explain this finding,⁴³ because a proportionally higher rate of admissions related to PD peritonitis in Black patients would amplify the protective impact of HHD in this population. This is supported by the results of our cause-specific analysis, where Black patients had a significantly reduced risk of infection-related hospitalizations, as well as time spent in hospital if they were on HHD.

We had hypothesized that an increase in hospitalization events in more modern eras would occur because older patients with a higher comorbidity burden and frailty are being placed on home modalities, specifically HHD. Contrary to our hypothesis, hospitalization outcomes did not consistently increase across eras. For patients using PD, the number of crude cumulative days in hospital decreased over time, something not replicated in patients using HHD. This could be explained by a differential improvement in the quality of care for patients receiving PD, because of more rapid and robust development of evidence-based clinical practice guidelines for this more frequently used home modality.⁴⁴

Other studies have looked at the change in outcomes for home dialysis patients across eras. A recent study comparing survival between patients using either PD or HHD across eras found that mortality differences attenuated over time.⁹ Another study looking at HHD patients found no change in survival across eras.⁴⁵ However, the most modern era in this study extended only to 2012; it is possible that they did not capture changing patient characteristics occurring in more contemporary time periods. Another study exploring differences in hospitalization rates between patients on home and in-center hemodialysis in the US found no statistical difference in hospitalization rates across eras (2006–2007 and 2008–2009); however, the follow-up time was short and would likely not have shown an era effect over this limited time period.²⁴ To our knowledge, our study is the most contemporary examination of the difference in hospitalization outcomes between incident HHD and PD patients.

The benefit seen for patients incident to HHD may be the result of improvements to HHD technology that have acted to mitigate the risk of hospitalization. Today, many patients using home dialysis have access to 24-hour on-call nursing support to assist with any issues and help solve problems at home.⁴⁶ HHD equipment continues to become more user-friendly and both patients and physicians may be better at troubleshooting and managing small problems at home because of extensive HHD training before initiation.⁴⁷ Virtual care treatment and management options may have also contributed to fewer hospital admissions in the more modern eras because telemedicine and remote patient monitoring systems have become more common.48, 49, 50, 51, 52

Despite this, there are some notable limitations. First, using an administrative database means our study was observational in nature, and our analyses were limited to the data collected. The CORR does not collect information on psychosocial characteristics, and there was limited information regarding socioeconomic status. It is possible that patients who have a caregiver to assist them with home dialysis have different hospitalization outcomes than those who lack caregiver support; however, we were unable to account for this. The variables, race and sex were provider-identified, which leaves the potential for misclassification bias. Dialysis modality was only assessed at baseline, thus increasing the potential for misclassification errors; however, any modality switches lasting longer than 30 days were treated as censoring events. Nevertheless, there is likely variability in how modality change dates are captured within CORR, potentially introducing bias. In addition, it is understood that this type of study has unavoidable potential for survivor bias⁵³ because patients must have survived up to the point of home dialysis initiation to be included in the study. We attempted to reduce the impact of this bias by choosing a short time frame to include those who switched to home dialysis (i.e., 90 days) and the results of our sensitivity analyses were consistent with our primary results, suggesting that the extent of this bias was likely minimal. In addition, it is beyond the scope of our study to imply a cause and effect on the relationship between home dialysis and hospitalizations. There is likely residual confounding contributing to patients treated with HHD having lower hospitalization outcomes because this population tends to be healthier and therefore is less likely to experience hospitalization events.⁵⁴ Finally, though the exploration of the impact of race and sex on the differential risk of cause-specific hospitalization by home dialysis modality is informative, the conclusions must be tempered because of the overall small number of HHD patients per cohort.

In conclusion, this national Canadian cohort study of incident home dialysis patients found that incident HHD initiation was associated with fewer hospital admissions, fewer days in hospital, and a longer time to first hospitalization when compared with patients using PD. These findings were most pronounced in male patients and in Black patients, primarily because of a significant reduction in hospitalizations and days in hospital from infection-related causes.

Disclosure

ACNF hods a scholarship from Fond de recherche du Québec – Santé and received honorary for CME from Baxter. KT has conducted advisory board and CME work with Otsuka, AstraZeneca, Bayer, Baxter and Vifor Pharmaceuticals, but none of this work is relevant to the current submission. ET has received honoraria for CME from Davita. All the other authors declared no competing interests.

Data Availability Statement

The linked data from Canadian Organ Replacement Register (and the Discharge Abstract Database supporting the findings of this study are openly available through CIHI at https://www.cihi.ca/en/access-data-and-reports/.

Supplementary Material

Supplementary File (PDF)

Table S1. Detailed breakdown of race groups as coded in the CORR stratified by home modality. Table S2. Crude outcomes for those who experienced one or more hospitalization event during the total follow-up period. Table S3. Rate of admission, and days in hospital between those who initiated home hemodialysis with a graft or fistula compared with those who initiated home hemodialysis with a CV line. Table S4. Rate of admission and days in hospital comparing peritoneal dialysis with home hemodialysis using 180-day and 365-day exposure definitions. Table S5. International Classification of Diseases Version 10 codes used for cause-specific hospitalization sensitivity analysis. Table S6. Rate of cardiovascular and infection-related admission and days in hospital comparing peritoneal dialysis to home hemodialysis.