Hospitalization Risk in the Integrated Home Dialysis Model

Visual Abstract

Abstract

Key Points

• Patients transitioning from peritoneal dialysis to home hemodialysis experience higher hospitalization rate during this transition period.

• After peritoneal dialysis-to-home hemodialysis transitions, hospitalization rates normalize and match those of patients starting home hemodialysis directly.

• Shorter transition periods are associated with an attenuated hospitalization risk, suggesting that timely peritoneal dialysis-to-home hemodialysis transitions may improve outcomes.

Background

Initiating dialysis with peritoneal dialysis (PD) followed by a transition to home hemodialysis (HHD) is known as integrated home dialysis. Hospitalization outcomes in this model are poorly known. We aimed to compare the hospitalization risk before, during, and after a PD-to-HHD transition with individuals receiving HHD as first home dialysis modality.

Methods

We analyzed the Canadian Organ Replacement Register and included individuals initiating home dialysis between 2005 and 2018. Patients transitioning from PD to HHD with <365 days of interim facility hemodialysis (“PD-HHD” group) were matched 1:1 to individuals receiving HHD as their first home dialysis modality (“HHD” group) using a propensity score. Patients were followed until December 2019, death, or a transfer to facility hemodialysis. In the PD-HHD group, the transition window was defined as the last 30 days of PD, the facility hemodialysis period and the first 30 days of HHD. Our primary outcome of all-cause hospitalizations (assessed before, during, and after transition) was analyzed using adjusted negative binomial regression.

Results

Three hundred and one patients underwent the PD-to-HHD transition and 711 initiated home dialysis in HHD. Hospitalization rates up to 2.17/patient-year (95% confidence interval [CI], 1.91 to 2.43) were observed during the PD-to-HHD transition, with infectious-related hospitalizations being most frequent (30%). Compared with HHD, the PD-HHD group had higher rates of hospitalization during (incidence rate ratio [IRR], 2.89 [95% CI, 2.01 to 4.16]) and before the transition (IRR, 1.38 [95% CI, 1.03 to 1.85]; predominantly in the year before) but not afterward (IRR, 1.04 [95% CI, 0.80 to 1.36]). Higher risk was not observed for cardiovascular-related hospitalizations and less pronounced in patients with short (<90 days in facility hemodialysis) transitions.

Conclusions

Most patients transitioning from PD to HHD experience a higher hospitalization risk before and during the transfer (relative to incident HHD). Yet, their hospitalization risk stabilizes after the transition. This higher risk of hospitalization is less apparent for patients with shorter interim facility hemodialysis.

Introduction

Offering home-based dialysis is a health care priority worldwide.¹ Home therapies (peritoneal dialysis [PD] and home hemodialysis [HHD]) result in greater patient satisfaction, treatment ownership and autonomy.² They are generally linked to improved quality of life and lower costs when compared with facility hemodialysis.^2–7 Yet the use of home dialysis remains limited, even in high-income countries.^8–10 Integrated home dialysis has emerged as a promising approach to improve home dialysis retention.¹¹ In this paradigm, patients initiate home dialysis with PD and then subsequently transition to HHD when PD is no longer feasible. Patients undergoing integrated home dialysis can benefit from both the initial advantages of PD (such as greater lifestyle, traveling,¹² preservation of vascular access,^13,14 preservation of residual kidney function,¹⁴ lower costs^3,4,15) and the technique longevity of HHD.¹⁶ This strategy prevents patients from transferring to longer-term facility hemodialysis when PD terminates, often occurring after only a few years of PD.¹⁷

Studies evaluating integrated home dialysis have first reported its feasibility in single-center settings.^18–21 Larger national studies have observed survival rates similar to patients initiating KRT directly with HHD and improved survival when compared with patients receiving PD or hemodialysis only.^22,23 Recently, we have shown using Canadian registry data that when controlling for dialysis vintage, patients receiving the PD-to-HHD strategy had slightly better survival than patients receiving HHD as a first home therapy.²⁴ However, hospitalization data in the integrated paradigm remain scare and have mostly focused on the period following the PD-to-HHD transition.²⁴

In this article, we therefore aimed to characterize the hospitalization rates of patients receiving integrated home dialysis care (i.e., transitioning from PD to HHD) before, during, and after the transition. We also aimed to compare these rates with patients receiving HHD as their first home KRT modality and to explore the impact of transition duration and cause of hospitalization on these rates. We hypothesized that hospitalizations would be higher during the PD-to-HHD transition and immediately following the transition, especially for patients with longer transition durations.

Methods

Population

This is an observational cohort study using data from the Canadian Organ Replacement Register (CORR). CORR is a validated register including all Canadian adults initiating KRT.²⁵ It is linked with a medico-administrative database (the Discharge Abstract Database from the Canadian Institute for Health Information [CIHI]) to provide data regarding hospitalizations. Similar to our previous work²⁴ using CORR, this study included individuals initiating KRT between January 1, 2005, and December 31, 2018. Patients from provinces of Quebec and Manitoba were excluded due to lack of CORR or hospitalization data during this time window. Exclusion criteria for this study were (1) patients with preemptive kidney transplant; (2) receiving both PD and HHD in the first 90 days of KRT; and (3) patients with <90 days of KRT. This study was authorized (20212531) by the research ethics committee of our institution. Owing to the registry nature of our data, individual-level informed consent was not required. Authors adhered to the Helsinki Declaration.

Exposure and Covariates

Study groups were defined based on to the dialysis modality received by each individual at day 90 of KRT (Figure 1).²⁴ The PD-to-HHD group included all individuals receiving PD at day 90 of KRT who subsequently transitioned to HHD. A maximum interval of 365 days of facility hemodialysis was allowed between periods of PD and HHD.²⁶ In comparison, the HHD group included all individuals receiving HHD at day 90 of KRT. The following variables were extracted from CORR at KRT initiation (Supplemental Table 1): demographics, clinical parameters, medication usage, and laboratory values.

Figure 1.

Study design. This figure displays the sequence of dialysis modalities in each study group and how the assessment periods (“before” transition, “during” transition, “after” transition) were created. Hatched intervals display the 30 days before and after the period spent in facility hemodialysis, which are included in the “transition” window. HD, hemodialysis; HHD, home hemodialysis; PD, peritoneal dialysis.

Outcomes

Hospitalizations were identified through linkage between the CORR and the Discharge Abstract Database by CIHI staff. All hospitalization episodes were assessed, and hospitalization causes were categorized (cardiovascular, infectious, CKD-related and dialysis-related, others) using the most responsible diagnosis (ICD-10-CA codes; Supplemental Table 2).²⁷ All patients were followed until December 31, 2019 (administrative censoring), and censored at death, kidney transplant, loss to follow-up (defined as 30 days without available data on KRT), or modality transfer (either transition to facility hemodialysis or a return in PD). Similarly to our previous work,²⁴ hospitalizations that occurred within 90 days of a modality change from HHD to another modality (return in PD or facility hemodialysis) were attributed to the exposure group to account for potential lag-time effects. Hospitalizations in the 7 days preceding kidney transplant were not considered. Three distinct follow-up periods were created according to the PD-HHD group (Figure 1). The transition window (during transition) began 30 days before the last day receiving PD and ended 30 days after the first HHD treatment. The period before transition began at the initiation of KRT and ended 31 days (−31) before the PD-to-HHD transition. The period after transition began at the 31st day of HHD after the PD-to-HHD transition and terminated at the end of follow-up (censoring or death). For patients with HHD, the corresponding dialysis vintages of the matched PD-to-HHD counterpart (days −30 and +30) were used to identify follow-up periods.

Statistical Analyses

Statistical analyses were conducted using R Software 4.2.3 (R Project for Statistical Computing). Baseline characteristics are expressed as means, medians, or counts and compared between study groups using standardized mean differences (SMDs). SMD values above 10% were considered clinically meaningful.²⁸ For other analyses, P values of 0.05 were considered statistically significant. Hospitalization incidence was examined descriptively in the entire PD-HHD group using incidence rates. Rates were stratified by follow-up period and hospitalization type.

Hospitalization rates were then compared between PD-HHD and HHD patients using a propensity score (PS) matching.²⁴ First, missing baseline covariate data (up to 16% of individuals) were addressed using multiple imputation by chained equations and predictive mean matching.^29,30 Distinct PSs estimating the probability of being attributed to the HHD or PD-HHD groups were then built in each imputed dataset using a logistic regression model including previously described covariates. Patients in both groups were matched with a 1:1 ratio employing a nearest-neighbor algorithm without replacement. Potential matches of a given PD-HHD patient were restricted to individuals in the HHD group with a dialysis vintage at least equivalent to the vintage at which the PD-HHD patient began its after transition period (i.e., day +30 after the first HHD treatment). This ensured that periods before and during the transition had the same duration for each match pair. Matches were restricted to individuals with a caliper width of 0.2 times the SD of logit(PS)³¹ and stratified for median age and geographical region. Matching adequacy was assessed using SMDs for each covariates and visual inspection of PS distributions in each group. Matching procedures were conducted separately in each imputed dataset and SMDs were combined using means, minimums, and maximums across imputed datasets.³² Since hospitalization counts were overdispersed (violating Poisson models assumptions), hospitalization incidence was compared between PD-HHD and HHD groups using negative binomial regression models and reported as incidence rate ratio (IRR) with 95% confidence intervals (CIs). Negative binomial models accounted for the repeated nature of hospitalization events. Overdispersion parameters ranged from 0.77 to 1.21 (before transition), 1.20 to 3.11 (during transition), and 0.83 to 1.15 (after transition). Models were run separately in each imputed dataset and follow-up period and combined using Rubin's rules.³³ For each patient pair, follow-up for the “before transition” period began at dialysis start (PD-HHD patients) or at the corresponding vintage (matched HHD individuals). Follow-up for the “during transition” period was strictly identical for each matched pair by design. In the post-transition period, follow-up was terminated independently for each pair according to censoring events or death. Clustering between pairs induced by the matching procedure was accounted with robust variances.³⁴ Similar models were run using hospitalized days as secondary outcome. Cause-specific analyses were also conducted by stratifying on hospitalization diagnosis codes. Finally, subgroup analyses were conducted by transition time (facility hemodialysis duration between PD and HHD; categorized as <90, 90–180, >180 days).

Results

Population Characteristics

From the 63,327 patients included in CORR, 13,726 were receiving PD at day 90 of KRT (Supplemental Figure 1). Of these, 301 (2%) patients transferred to HHD in <365 days after PD termination. These 301 patients represented 7% of the 4591 patients who transitioned from PD to either hemodialysis or HHD, compared with 995 receiving a kidney transplant (22%) and 2076 patients (45%) who died during follow-up. PD-to-HHD transitions accounted for 2% of patients initiating PD in Ontario, 3% in Western Canada, and 1% in Eastern Canada. PD+HHD patients transferred to HHD after a median of 1.8 years (interquartile range [IQR], 1.0–3.1) of KRT and received a median of 64 days of facility hemodialysis (IQR, 0–146 days) during their transition. The baseline characteristics of these patients are presented in Table 1 and stratified by facility hemodialysis duration in Supplemental Table 3.

Table 1.

Population characteristics

Characteristics	HHD (n=711)	PD-HHD (n=301)	SMD (%)
Demographics
Age, yr	52±14	52±14	4
Male sex	478 (67%)	203 (67%)	1
Non-White race	183 (26%)	113 (38%)	26
Region			65
Eastern Canada	23 (3%)	8 (3%)
Ontario	542 (76%)	142 (47%)
Western Canada	146 (21%)	151 (50%)
Comorbidities
Kidney disease			26
Diabetic	202 (28%)	104 (35%)
GN	173 (24%)	86 (29%)
Polycystic	109 (15%)	30 (10%)
Vascular-hypertensive	68 (10%)	34 (11%)
Other	159 (22%)	47 (16%)
Late nephrology referral	90 (13%)	38 (14%)	1
Active smoking	93 (13%)	31 (10%)	9
BMI, kg/m2	29±7	29±7	2
CAD	94 (13%)	37 (12%)	3
Prior stroke	45 (6%)	8 (3%)	18
PAD	58 (8%)	22 (7%)	3
Heart failure	69 (10%)	16 (5%)	17
Lung disease	27 (4%)	10 (3%)	3
Neoplasia	85 (12%)	14 (5%)	27
Diabetes	282 (40%)	133 (44%)	9
Antihypertensive use	583 (82%)	228 (76%)	15
Erythropoietin agonist use	316 (44%)	144 (48%)	7
Socio-economic quintile	3 (2–4)	3 (2–4)	9
Laboratory
Hemoglobin, g/dl	10.1±3.7	10.5±1.6	17
Bicarbonate, mEq/L	21±5	22±4	28
Albumin, g/dl	3.6 (3.1–3.9)	3.6 (3.2–4.0)	14
Corrected calcium, mg/dl	8.8±1.2	9.2±1.2	17
Phosphate, mg/dl	5.6 (4.7–6.8)	5.6 (4.7–6.5)	19
Parathyroid hormone, pg/ml	311 (179–538)	264 (151–443)	24

Characteristics are presented as count (percentage), mean±SD or median (interquartile range). A standardized mean difference value above 10% is considered significant. BMI, body mass index; CAD, coronary artery disease; HHD, home hemodialysis; PAD, peripheral artery disease; PD, peritoneal dialysis; SMD, standardized mean difference.

Laboratory values refer to the last values before dialysis initiation.

Western provinces include British Columbia, Alberta and Saskatchewan.

Eastern provinces include New Brunswick, Nova Scotia, Newfoundland and Labrador, and Prince Edward Island.

Socioeconomic quintiles refer to lowest (1) to highest (5).

Hospitalizations before, during, and after a PD-to-HHD Transition

Patients undergoing the PD-to-HHD transition experienced a total of 1277 hospitalizations (10,020 days) during a median of 4.4 years of follow-up. Two hundred fifty-nine (86%) patients experienced a hospitalization at any time during follow-up, with a median of 3 (IQR, 1–5) hospitalizations per patient and an overall rate of 0.88 hospitalizations per patient-year (95% CI, 0.83 to 0.92). Hospitalization rates before, during, and after the PD-to-HHD transition are presented in Figure 2. During the transition, 147 (49%) patients experienced a hospitalization (2.17/patient-year [95% CI, 1.91 to 2.43]); 30% of those were categorized as infectious (among which 59/79 [75%] subcategorized as dialysis-related infections), 25% as CKD-related, and 14% as cardiovascular (Figure 3B). Hospitalization rates were lower in the periods before (0.61/patient-year [95% CI, 0.55 to 0.67]) and after (0.89/patient-year [95% CI, 0.82 to 0.96]) the transition. Before the transition (Figure 3A), proportions of infection-related or cardiovascular-related hospitalizations were similar to those observed during the transition while CKD-related hospitalizations were lower (overall P value < 0.001). By contrast, the proportion of infection-related hospitalizations (22%) was smaller after the transition while cardiovascular-related hospitalizations (19%) were higher (Figure 3C). Findings regarding hospitalized days were similar (Supplemental Figure 2).

Figure 2.

Hospitalization rates in patients experiencing a PD-to-HHD transition. This figure displays the hospitalization rate (per patient-year) in the periods before, during and after the PD-to-HHD transition in the overall PD-HHD cohort. Plain circles indicate the point estimate, and bars indicate the 95% CI. One hundred and eighty days intervals were used for the periods before and after the transition. CI, confidence interval.

Figure 3.

Hospitalization main diagnoses. (A) Before, (B) during, and (C) after a PD-to-HHD transition.

Propensity-Score Match Analysis

Overall, 711 patients received HHD at day 90 of KRT, with a median time spent in HHD of 2.3 years (IQR, 1.2–4.2 years). Compared with the PD-HHD group (Table 1), patients receiving HHD were more often of non-White ethnicity, from Western Canada, had higher prevalence of diabetic kidney disease and GN, lower prevalence of comorbidities, and more favorable anemia and mineral parameters. Across imputed datasets, 221–230 PD+HHD patients (median 224) were matched to patients with HHD. The distributions of PS in matched groups were similar (Supplemental Figure 3). Matching improved baseline characteristics balance except for antihypertensive use (Supplemental Figure 4 and Supplemental Table 4). By design, follow-up periods were identical between both groups before (median, 1.5 years [IQR, 0.8–2.8]) and during (median, 0.3 years [0.2–0.5]) the transition. After the transition, the PD+HHD group was followed for a median 1.8 years (IQR, 1.0–3.1) and the HHD group for a median 1.1 years (0.4–2.3). Transplant incidences appeared comparable in both groups after matching with a median incidence across imputed datasets of 38% in the PD-to-HHD group and 36% in the HHD group.

After matching, hospitalization rates were significantly higher in the PD-to-HHD group during the transition, and in a smaller magnitude, before the transition (Table 2). This higher pretransition hospitalization rate was concentrated in the year immediately preceding it (Figure 4 and Supplemental Figure 5). Hospitalization rates were similar between HHD and PD-to-HHD patients after the transition. Similar findings were observed for hospitalized days (Table 2 and Supplemental Figures 6 and 7), albeit the between-group difference did not reach statistical significance before the transition. In cause-specific analyses (Table 3), all hospitalization types were higher during the transition for PD-to-HHD patients except for cardiovascular hospitalizations. By contrast, only hospitalization for “other” causes were higher before the transition (in this category, gastrointestinal hospitalizations were the most frequent [34%] among PD-to-HHD patients). Similar results were obtained when displayed year-by-year (Figure 5A).

Table 2.

Rates of Hospitalization and days in hospital before, during, and after the peritoneal dialysis-to-home hemodialysis transition in comparison with home hemodialysis

PD-HHD Transition Window	Rates (95% CI)		IRRs (95% CI)
	PD-to-HHD Group	HHD Group
Hospitalization rate
Before transition	0.61 (0.53 to 0.68)	0.50 (0.43 to 0.56)	1.38 (1.03 to 1.85)
During transition	2.14 (1.83 to 2.44)	0.72 (0.55 to 0.90)	2.89 (2.01 to 4.16)
After transition	0.87 (0.79 to 0.95)	0.80 (0.71 to 0.89)	1.04 (0.80 to 1.36)
Days in hospital
Before transition	3.68 (3.50 to 3.87)	3.06 (2.89 to 3.22)	1.47 (0.92 to 2.36)
During transition	20.63 (19.68 to 21.58)	4.19 (3.76 to 4.61)	4.41 (2.69 to 7.22)
After transition	6.75 (6.53 to 6.97)	7.38 (7.11 to 7.65)	0.78 (0.52 to 1.16)

Incidence rate ratios (95% confidence interval) are presented for the “peritoneal dialysis+home hemodialysis” with “home hemodialysis” group as reference and were combined across multiple imputations using Rubin rules. CI, confidence interval; HHD, home hemodialysis; IRR, incidence rate ratio; PD, peritoneal dialysis.

Figure 4.

Comparison of hospitalization rates for patients in PD-to-HHD and HHD groups. This figure displays the hospitalization rate (per patient-year) in the periods before, during, and after the PD-to-HHD transition for patients in the PD-to-HHD and HHD groups. Plain circles indicate the point estimate, and bars indicate the 95% CI. Three hundred and sixty-five days intervals were used for the periods before and after the transition.

Table 3.

Risk of hospitalizations in “peritoneal dialysis+home hemodialysis” compared with “home hemodialysis,” stratified by hospitalization type and time on interim facility hemodialysis

Subgroups	Before Transition	During Transition	After Transition
Hospitalization type
Cardiovascular	0.87 (0.52 to 1.46)	1.78 (0.82 to 3.85)	1.13 (0.73 to 1.76)
Infectious	1.32 (0.84 to 2.06)	4.59 (2.40 to 8.76)	1.01 (0.66 to 1.55)
CKD	1.40 (0.83 to 2.37)	5.34 (2.48 to 11.50)	0.99 (0.61 to 1.61)
Other	1.57 (1.01 to 2.45)	2.09 (1.20 to 3.64)	1.09 (0.75 to 1.59)
Time on facility hemodialysis, da
<90	0.92 (0.62 to 1.37)	1.86 (0.89 to 3.90)	0.96 (0.65 to 1.43)
90–180	1.36 (0.83 to 2.24)	3.67 (1.93 to 6.95)	0.98 (0.63 to 1.53)
>180	2.23 (1.23 to 4.03)	3.23 (1.84 to 5.67)	1.20 (0.72 to 2.03)

Incidence rate ratios (95% confidence interval) for hospitalizations are presented for the “peritoneal dialysis+home hemodialysis” with “home hemodialysis” group as reference and were combined across multiple imputations using Rubin rules.

^aTime on facility hemodialysis corresponded to the time spent in facility hemodialysis for patients experiencing peritoneal dialysis to home hemodialysis. Patients in the home hemodialysis group were categorized according to the time on facility hemodialysis of their matched peritoneal dialysis-to-home hemodialysis counterpart.

Figure 5.

IRRs comparing hospitalizations in PD-HHD and HHD groups. Comparison by (A) hospitalization type or (B) transition duration. This figure displays IRRs comparing hospitalization rates between the PD-to-HHD and HHD groups. Plain circles indicate the point estimate, and bars indicate the 95% CI. Three hundred and sixty-five days intervals were used for the periods before and after the transition. IRR, incidence rate ratio.

After stratification for transition duration (Table 3), hospitalization rates were higher before the transition solely among patients with the longest transitions (>180 days). During the transition, both patients with medium (90–180 days) and long (>180 days) transition had higher hospitalization rates, while those with short transition (<90 days) did not have statistically higher hospitalization rates. Neither subgroup of PD-to-HHD transition duration displayed higher hospitalization after its transition when compared with patients on HHD. In year-by-year analyses (Figure 5B), patients with the longest transition duration were the only subgroup with higher hospitalization rates in the year before transitioning.

Discussion

In this study, we characterized hospitalization for patients transitioning from PD to HHD in a large Canadian medico-administrative registry. We observed higher hospitalizations during the transition period, mostly from infectious or CKD-related diagnoses, when compared with patients initiating KRT in HHD. Statistically, higher hospitalizations were not observed in patients with interim facility hemodialysis duration shorter than 90 days. The excess hospitalization risk seen in patients with longer transition periods did not translate to higher hospitalization after the transition.

Our findings build on previous studies regarding the outcomes of integrated home dialysis. Among these, an analysis of United States Renal Data System reported a 24% lower mortality in patients experiencing PD-to-HHD compared with individuals receiving facility hemodialysis.²³ A registry analysis from Australia and New Zealand observed that PD-to-HHD patients had survival comparable with patients initiating KRT directly on HHD.²² Recently, we used CORR to compare the technique survival of PD-to-HHD with HHD “only” patients.²⁴ In contrast to Australia and New Zealand, we observed that patients in the PD-to-HHD group had lower death and hemodialysis-transfer rates than patients receiving only HHD when accounting for dialysis vintage. Although we briefly explored hospitalization rates in that previous study, this present work differs from our previous one in several ways, including a more inclusive definition of PD-to-HHD transitions and the analysis of hospitalizations during the entire transfer spectrum (before, during, and after the transition). Among the other studies that assessed hospitalizations in the integrated model, one single-center report (conducted by some authors of this study using a distinct cohort) evaluated the time to first hospitalization in patients completing HHD training with or without previous PD exposure.¹⁸ They observed a higher hospitalization risk in patients with previous PD exposition. These results contrast with our findings in the post-transition period but may be explained by several methodological and populational discrepancies, including direct adjustment rather than PS matching, baseline imbalances between groups, and no assessment of the entire hospitalization burden (including before and during the transition).

By contrast, hospitalizations related to PD-to-hemodialysis transitions have been more thoroughly evaluated. A gradual higher rate of hospitalization episodes in the year preceding the transition was observed in an American study, culminating in a sharply higher rate of hospitalization the month before.³⁵ Hospitalizations due to PD catheter complications were the most frequent cause of admission before the transition, followed by “other morbidity” and cardiovascular diseases. Of note, cause-specific categorization differs slightly between our two studies (most notably related to mechanical and infectious PD complications). In line with that prior data, we observed that infectious diagnoses were the most frequent ones (30%) among the hospitalizations occurring during the transition window, with peritonitis as the leading subdiagnosis in this category (75% of infectious diagnoses). Although these combined findings underscore the known role of peritonitis as a leading cause of PD terminations,^17,36–38 a majority of the overall hospitalization burden (208 out of 267 episodes) during the transition period in the PD-HHD group was in fact not caused by dialysis-related infections. Some of these noninfectious hospitalizations might reflect medical complications that directly caused PD discontinuation, but we hypothesize that others are the proxy of a heightened comorbidity profile that concomitantly resulted in a transition to hemodialysis. Unfortunately, several of these diagnoses were coded with a CKD-related code (e.g., CKD or diabetic nephropathy), which precluded us from clearly discerning the hospitalization main cause. The relatively small number of PD-to-HHD patients also prevented the conduct of more specific analyses of coding patterns across dialysis centers. Interestingly, cardiovascular hospitalizations were not statistically increased in PD-to-HHD patients compared with HHD, reassuring findings in light of previous reports of higher cardiovascular events or hospitalizations in individuals receiving PD.^39–41

Among our novel observations, we showed that patients with shorter periods of facility hemodialysis (<90 days) between PD and HHD did not have statistically significant higher rates of hospitalization before or during the transition, in contrast to patients with longer hemodialysis periods. This group of patients with “short transitions” was younger, less often male or from non-White; we hypothesize that the lower hospitalization risk in these patients reflects an overall more favorable health status. Importantly, these findings can also reflect planned PD-to-HHD transitions, in which clinicians anticipate the need to terminate PD before a clinical event (often a hospitalization) precipitates the patient into facility hemodialysis. By analogy, creating an arteriovenous access before a modality transition (proxy of its planned nature) was associated with better survival in patients undergoing a PD-to-hemodialysis transfer.⁴² In addition, shorter hemodialysis periods might highlight closer monitoring and follow-up after transferring to hemodialysis, such as nurse navigators or transitional care units.^11,43 Because our IRR estimate in the subgroup with “short transitions” had broad CIs in the transition period, our findings cannot be definitively interpreted as the absence of higher hospitalizations for PD-HHD patients, but rather as an attenuated risk compared with patients with longer transition times. Evaluating registry data also prevented us from distinguishing which PD-to-HHD transfers resulted from institutional integrated home dialysis programs rather than opportunistic transfers. Nevertheless, based on our observations of heterogenous transfer durations (short durations can reflect better anticipation deriving from formal integrated initiatives) and varying regional proportions of PD-to-HHD transitions, we hypothesize that a non-negligible proportion of transfers was clinically driven rather than part of formal programs.

Our study has several strengths. We analyzed hospitalization incidence, a patient-centered outcome which is scarcely reported in medico-administrative studies. We evaluated a large register of patients initiating KRT and used an inclusive definition of PD-to-HHD transitions, which better reflect the heterogeneity in transition practices. Furthermore, we report hospitalization risk both before, during, and after a PD-to-HHD transition, which had not been done in previous studies. We also compared this risk with a control group of patients initiating KRT directly in HHD using an elaborate methodological approach including PS matching. Finally, we stratified our findings according to transition duration and hospitalization causes which further refines our understanding of patient outcomes during a PD-to-HHD transition.

This study also has some limitations.

1. Despite evaluating a large national registry over a long period, our PD-to-HHD sample remained small, reflecting the low incidence of this care model.

2. The medico-administrative nature of our study prevented us from analyzing some useful patient-level data such as the transfer rationale, nonhospitalized peritonitis events, or vascular access. Proxies for patient-level information might be subject to classification bias, most notably regarding transfer duration (dialysis centers could encode HHD training time differently) and hospitalization diagnoses. In addition, our analyses cannot easily distinguish the factors that caused a PD-to-HHD transition from the events that result from the same causes.

3. Comorbidities were only assessed at dialysis initiation, which might not fully reflect patient status after some years of KRT vintage.

4. Our PS matching cannot fully exclude residual confounding due to data unavailability (e.g., frailty, health seeking behaviors, prior hospitalization patterns). For measured confounders, perfect balance between groups was not obtained by matching, which was expected based on our sample size and the use of geographical-related and vintage-related restrictions while matching. Since some PD-HHD individuals were unmatched in our PS process, we reported the crude hospitalization rates in the entire PD-to-HHD population.

5. We opted for a comprehensive definition of the PD-to-HHD transition window (up to 365 days in facility hemodialysis) which might not reflect what all practitioners consider as the transition period. Despite these comprehensive inclusion criteria, the PD-to-HHD population remains a clinically selected subgroup and consequently, our results should not be generalized to the entire PD population. Nonetheless, in addition to patients with PD eligible for the PD-to-HHD model, our results can help patients and clinicians select a home dialysis modality at KRT initiation (the integrated approach with PD or HHD directly). Our choice of HHD as comparator hence reflects the clinical decisions of patients initiating home dialysis or terminating PD; however, our comparisons cannot be extrapolated to facility hemodialysis.

6. While follow-up was identical by design between study groups before and during the transition, patients in the PD-to-HHD were followed-up for approximatively 6 months longer after the transition in comparison with the HHD group. Such findings were presumably related to better “technique survival” in PD-to-HHD patients²⁴ and lower mortality, as estimated life expectancy was likely incorporated in the clinical decision of transitioning to HHD. Interestingly, transplant rates were similar in both groups. Nevertheless, we evaluated the robustness of our main findings in year-stratified analyses, which yielded similar results for the post-transition period.

In conclusion, in Canadian individuals initiating KRT, we investigated the hospitalization profile associated with PD-to-HHD transitions compared with patients initiating KRT directly on HHD. We observed that hospitalizations were higher during the PD-to-HHD transition period and in the year preceding it. The extent of hospitalization risk before and during the transition varied by diagnosis and transition duration, with attenuated rates of higher for cardiovascular hospitalizations and in patients with shorter transitions. Importantly, hospitalizations were not higher in the years following a PD-to-HHD transfer regardless of their transition duration or hospitalization type. These findings expand our understanding of the integrated home dialysis paradigm. They highlight the favorable long-term outcomes of this model since expected higher rates of hospitalizations before and during the transition are not translated to the years following it. These results underscore the importance of planned PD-to-HHD transitions among eligible patients to potentially minimize hospitalizations in the transfer window. Further work is hence needed to promote and better identify patients eligible for a PD-to-HHD transition, streamline the transition process and minimize transfer-related burden.

Supplementary Material

cjasn-21-1177-s001.pdf ^{(1.4MB, pdf)}

cjasn-21-1177-s002.pdf ^{(497.4KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C713.

Author Contributions

Conceptualization: Christopher T. Chan, Louis-Charles Desbiens, Rémi Goupil, Annie-Claire Nadeau-Fredette, Jeffrey Perl, Karthik K. Tennankore, Emilie Trinh.

Data curation: Annie-Claire Nadeau-Fredette.

Formal analysis: Louis-Charles Desbiens.

Funding acquisition: Annie-Claire Nadeau-Fredette.

Investigation: Annie-Claire Nadeau-Fredette.

Methodology: Louis-Charles Desbiens, Annie-Claire Nadeau-Fredette.

Supervision: Annie-Claire Nadeau-Fredette.

Validation: Annie-Claire Nadeau-Fredette.

Writing – original draft: Louis-Charles Desbiens.

Writing – review & editing: Christopher T. Chan, Louis-Charles Desbiens, Rémi Goupil, Annie-Claire Nadeau-Fredette, Jeffrey Perl, Karthik K. Tennankore, Emilie Trinh.

Funding

A.-C. Nadeau-Fredette: Fonds de Recherche du Québec - Santé.

Declarative Statements

This study includes clinical experimentation and received Institutional Review Board or Ethics Committee approval. The need to obtain informed patient consent was waived. This study complies with the Declaration of Helsinki.

Data Availability Statements

Data belong to a third party, and authors are not authorized to share the data. Third Party: The CORR and CIHI. Reason for Restriction: Data request for the CORR and the Discharge Abstract Database can be made to CIHI using the following link. https://www.cihi.ca/en/access-data-and-reports/data-holdings/make-a-data-request.

Supplemental Material

This article contains supplemental material online, published as provided by the authors, at http://links.lww.com/CJN/C714.

Supplemental Table 1. Study variables and definitions.

Supplemental Table 2. Cause-specific International Classification of Diseases version 10 diagnosis codes.

Supplemental Table 3. PD population characteristics stratified according to facility hemodialysis time during the PD-to-HHD transition.

Supplemental Table 4. Baseline characteristics after multiple imputation and PS matching.

Supplemental Figure 1. Study flow chart.

Supplemental Figure 2. Hospital days rates in patients experiencing a PD-to-HHD transition.

Supplemental Figure 3. PS distributions before and after matching.

Supplemental Figure 4. SMD for baseline population characteristics before and after PS matching.

Supplemental Figure 5. IRRs comparing hospitalizations in PD-HHD and HHD groups.

Supplemental Figure 6. Comparison of hospitalized days rates for patients in PD-to-HHD and HHD groups.

Supplemental Figure 7. IRRs comparing hospitalized days rate in PD-HHD and HHD groups.