Home versus In-Center Dialysis and Day of the Week Hospitalization: A Cohort Study

Key Points

• The rate of hospitalization differs across dialysis modalities; patients on home hemodialysis (HHD; including those on frequent HHD) have the lowest rate.

• The probability of a Monday/Tuesday admission is lowest for patients receiving frequent HHD and peritoneal dialysis compared with those receiving in-center hemodialysis.

• Having an awareness of this finding may help nephrologists tailor dialysis prescriptions, mainly for those who perform dialysis at home.

Visual Abstract

Abstract

Background

The dialysis treatment day after the 2-day interdialytic interval (Monday/Tuesday) is associated with a heightened risk of hospitalization for patients on in-center hemodialysis (ICHD). In this national cohort study, we sought to characterize hospitalizations by day of the week for patients receiving ICHD, home HD (HHD), and peritoneal dialysis (PD) and to identify whether there were differences in the probability of a Monday/Tuesday admission for each modality type.

Methods

Patients on maintenance dialysis in Canada were analyzed from 2005 to 2014 using the Canadian Organ Replacement Register. Patients on hemodialysis were categorized as those receiving ICHD, HHD, frequent ICHD, or frequent HHD (the latter two included short daily and nocturnal HD). Hospitalizations were attributed to the previous treatment if they occurred within 30 days of a treatment change. Differences in the proportion of patients experiencing a Monday/Tuesday admission with all other days of the week were compared using a generalized linear model with binomial distribution and reported using adjusted odds ratios (OR) with 95% CIs.

Results

Overall, 27,430 individuals experienced 111,748 hospitalization episodes. Rates per 1000 patient days were 3.76, 2.98, 2.71, 2.16, and 2.13 for each of frequent ICHD, ICHD, PD, HHD, and frequent HHD, respectively. Compared with those on ICHD, only patients receiving frequent HHD (OR, 0.89; 95% CI, 0.81 to 0.97) and PD (OR, 0.95; 95% CI, 0.93 to 0.97) had a lower odds of experiencing a Monday/Tuesday admission. The OR was lower when restricted to hospitalization episodes for cardiovascular reasons comparing frequent HHD with ICHD (OR, 0.68; 95% CI, 0.48 to 0.96).

Conclusion

In this nationally representative cohort, we identified that the probability of a Monday/Tuesday admission was lower for frequent HHD and PD compared with ICHD, most notably for hospitalizations due to cardiovascular causes. Gaining a better understanding of the reasons behind this observation may help to develop future strategies to reduce overall and cause-specific hospitalization for patients receiving dialysis.

Introduction

Patients receiving dialysis are at a high risk of hospitalization (1–14). Although patients receiving hemodialysis (HD) are generally more likely to be admitted to hospital compared with those receiving other modalities (2,4,7,10,11), the risk of hospitalization is not exclusive to in-center HD (ICHD). Patients receiving both peritoneal dialysis (PD) (8,9,14) and home HD (HHD) (12,13) are at risk for hospitalization, irrespective of home versus in-center location (4,5,9,12,13,15,16).

It is well recognized that the Monday/Tuesday after the long 2-day dialysis interval is a time of heightened risk for patients on ICHD. In a study using data from the United States Renal Data System, the risk of mortality or cardiac hospitalization was highest on the day after the long dialysis break (typically a Monday or Tuesday) (17). This has been shown in subsequent studies and, in fact, all-cause hospitalization occurs most frequently on the Monday/Tuesday relative to all other days of the week among patients receiving maintenance ICHD (18).

Although day of the week has an effect on mortality for ICHD, this risk is not observed in patients on HHD or PD. In an Australia/New Zealand registry study, the risk of cardiac death was more likely to occur on Monday for patients on ICHD receiving three or less treatments per week. In contrast, this daily variation was not observed for those on HHD, frequent HD (irrespective of location), or PD (19). To our knowledge, the presence or absence of daily variations in hospitalization have not been evaluated for patients receiving HHD versus ICHD. Understanding this association may be beneficial in helping nephrologists tailor dialysis prescriptions, particularly for those who have made a choice to perform their dialysis at home.

Therefore, in a national registry of patients receiving maintenance dialysis, the purpose of this study was to describe day of the week variations in hospitalization and to compare the risk of a Monday/Tuesday admission by dialysis modality. We hypothesized there would be a relatively higher risk of a Monday/Tuesday admission for patients on ICHD versus those receiving HHD and PD.

Materials and Methods

Design and Population

We conducted a retrospective cohort study of the Canadian Organ Replacement Register (CORR), a national administrative maintenance dialysis database, from January 1, 2005 to December 31, 2014. This study was approved by the Nova Scotia Health Research Ethics Board. Follow-up time was from the start of initiation of maintenance dialysis and censored at the first of either death or kidney transplantation. CORR is managed by the Canadian Institute for Health Information (CIHI) and has been previously validated (20). Characteristics included in the CORR are submitted at the time of dialysis initiation by each reporting facility and include baseline demographics, dialysis modality, comorbidity, cause of ESKD, baseline laboratory data, and outcome data for all patients initiating chronic dialysis in Canada (with the exception of Quebec). In addition to baseline data, each participating province also captures and submits data regarding any changes in modality that occur during follow-up, inclusive of the date of change.

Data Source: Hospitalizations

All hospitalization data were abstracted from the CIHI Discharge Abstract Database (DAD). The DAD captures administrative, clinical, and demographic information on hospital discharges, and most responsible diagnosis (using the Canadian International Classification of Diseases, 10th Revision [ICD-10-CA] codes) for all patients in Canada. For the purposes of this study, hospitalizations did not include same-day procedures, outpatient procedures, or isolated emergency department presentations without admission. All data linkages between the CORR and the DAD were performed at CIHI before data analysis, using a similar approach to a previous study (6). Patients residing in Manitoba and Quebec were excluded from the analysis; the former due to an inability to link hospitalization data, and the latter due to a lack of inclusion in CORR for the years covering the cohort. For all patients, we excluded any admissions that occurred within 7 days of the first dialysis treatment (i.e., at initiation of maintenance HD). We did this to avoid the possibility of misclassification (i.e., some early events may not have truly been readmissions) and to acknowledge that early admissions after dialysis initiation may be iatrogenic and not attributable to dialysis modality per se. However, in a prespecified secondary analysis, we did include early admissions, using the date of discharge as the index date for time at risk for those who initiated dialysis as an inpatient. Finally, admissions 30 days before transplantation were excluded because we assumed these may have represented a coding error in the date of admission for the purposes of transplantation (a censoring event).

Exposure of Interest

The primary exposure of interest was dialysis modality, and modality definitions were determined on the basis of standard CORR categories, which are provided to each reporting facility. Therefore, modality definitions relied on appropriate documentation by the nephrologist/health care provider completing the CORR form at each participating site. These categories were HD (3–6 hours, two to four times per week), frequent HD (inclusive of nocturnal HD, >6 hours, five to six nights per week; and short daily HD, 2–3 hours, 5–7 d/wk) and PD (inclusive of continuous ambulatory and automated PD). A separate variable included in the CORR captured home versus in-center location for each HD modality. Combining location with modality led to five possible exposure categories: ICHD, HHD, frequent ICHD, frequent HHD, and PD. In the primary analysis, all modality definitions included any changes in modality during follow-up, provided the change lasted >30 days. The number of patients with zero, one, two, three, and four or more modality switches for the overall population was reported, as was the median duration of treatment before switching (or end of follow-up for those that did not switch).

Outcome: Hospitalizations

For the primary analysis, we focused on all-cause hospitalizations. In three prespecified secondary analyses, we exclusively evaluated admissions for cardiovascular causes, dialysis-associated infections, and all infections using ICD-10 codes mapped to definitions previously established using ICD-9 codes (Supplemental Table 1) (12).

Statistical Analyses

Baseline characteristics of the overall study population were described using univariable statistics. Comparisons of baseline characteristics across modalities were not performed given the definition of modality included changes over time; however, baseline characteristics stratified by initial dialysis modality were also reported in a supplemental analysis. The total proportion and rate of hospitalizations (proportion/total follow-up time) was described for each modality overall and stratified by day of the week. Hospitalization rates were calculated as total count/cumulative time at risk, reported as count per 1000 patient days, and included those individuals who experienced no admissions. Hospitalization events that occurred within 30 days of a change were attributed to previous treatment (30 day lag). The number and proportion of hospitalizations occurring within the 30 days after transition to another modality was reported. In addition, in supplemental analyses, rates per 1000 patient days were reported in two scenarios: (1) without the 30-day lag, or (2) without both the 30-day lag or need for >30 days to consider a treatment change. Differences in the proportion of day of the week hospitalization were visually displayed using a radar plot for each modality category and analyzed for only those individuals who experienced one or more hospitalizations. The probability of a Monday/Tuesday admission versus all other days was compared across each modality using sequential adjusted generalized linear models with binomial distribution, logit link, and compound symmetry correlation structure for repeated events (21), and probability was reported using odds ratios (ORs) and 95% CIs with ICHD as the reference group. In prespecified sensitivity analyses, this was repeated (1) with inclusion of all admissions from the index date of dialysis initiation, (2) first year admissions only, and (3) without the 30 day lag in hospitalization. Variables included in all adjusted models were those captured at baseline in all patients using the CORR and included demographics (age, sex, race), body mass index, comorbid conditions (diabetes, coronary artery disease, congestive heart failure, cerebrovascular disease, peripheral vascular disease, prior malignancy), cause of ESKD, facility location, and select baseline blood work (serum albumin and hemoglobin). Missing data were handled with list-wise deletion in the primary analysis, but, in a final sensitivity analysis, we also addressed missing data with multiple imputation using chained equations (22). For all analyses, P<0.05 was the threshold for statistical significance. All analyses were conducted using SAS STAT 14.3 software, version 9.4 (SAS Institute, Cary, NC).

Results

Characteristics of the Cohort

The cohort consisted of 36,919 patients. Most patients either remained on a single treatment throughout or had only one modality switch (N=23,395; 63%). The proportion of patients with zero to four or more modality switches, and the duration of treatment before switch (or end of follow-up), is noted in Supplemental Table 2. Baseline characteristics are noted in Table 1. Median (interquartile range) age of patients was 67 (23) years, 61% were of male sex, and 70% were of White race. A total of 13,556 individuals (37%) had diabetic nephropathy as their cause of ESKD. Baseline characteristics stratified by initial modality are noted in Supplemental Table 3.

Table 1.

Characteristics of the study population

Variable	Value (N=36,919)a
Demographics
Age, yr, median (IQR)	67 (21)
Male sex, n (%)	22,492 (61)
Race, n (%)
White	25,957 (70)
Black	1407 (4)
Other	7755 (21)
Unknown	1800 (5)
Body mass index categories, kg/m2, n (%)
<18.5	1120 (3)
18.5 to <25	11,354 (34)
25 to <30	10,509 (31)
>30	10,546 (31)
Initial dialysis modality, n (%)
Hemodialysis	29,480 (80)
In-center	29,019 (79)
Frequent in-center	189 (1)
Home	136 (<1)
Frequent home	136 (<1)
Peritoneal dialysis	7439 (20)
Province of dialysis facility, n (%)
Ontario	20,992 (57)
British Columbia	6301 (17)
Alberta	4396 (12)
Saskatchewan	1656 (4)
Nova Scotia and Prince Edward Island	1544 (4)
New Brunswick	1059 (3)
Newfoundland	971 (3)
Cause of ESKD, n (%)
Diabetes	13,556 (37)
Other	7467 (20)
Ischemic/renovascular	6891 (19)
GN	4765 (13)
Unknown	2125 (6)
Congenital, hereditary, polycystic kidney	2115 (5)
Comorbidities, n (%)
Cardiac disease	10,928 (30)
Congestive heart failure	8376 (23)
Cerebrovascular disease	4808 (13)
Peripheral vascular disease	5835 (16)
Chronic obstructive lung disease	4084 (11)
Prior malignancy	4848 (13)
Diabetesb	4753 (13)
Laboratory investigations
Hemoglobin categories, g/L, n (%)
<95	13,815 (40)
95–115	15,143 (43)
>115	5927 (17)
Serum albumin, g/L, median (IQR)	33 (10)

IQR, interquartile range.

^aMissing values: body mass index, n=3390; hemoglobin, n=2034; albumin, n=3548.

^bComorbidity only, excluding diabetes as a cause of ESKD.

Hospitalization Data

Overall, 27,430 individuals (74%) experienced one or more hospitalizations during follow-up. Proportions/rates of admissions for each modality over follow-up time are noted in Table 2. Of the 111,748 admissions, 85,528 (77%) occurred in individuals receiving ICHD; 22,366 (20%) occurred in individuals receiving PD; and 3% occurred among individuals receiving either HHD, frequent HHD, or frequent ICHD. Overall, patients receiving frequent ICHD had the highest rate of hospitalizations (3.76 per 1000 patient days), followed by ICHD (2.98 per 1000 patient days), and PD (2.71 per 1000 patient days). Rates of hospitalization were lower for HHD (2.16 per 1000 patient days) and frequent HHD (2.13 per 1000 patient days). Overall, patients receiving PD, frequent HHD, and HHD had 347 (2%), 112 (7%), and 57 (7%) hospitalizations, respectively, that occurred within 30 days of a transition to another treatment. Patients receiving frequent ICHD had 94 (6%) of the hospitalizations that occurred within 30 days of transfer, of which 81 (86%) were after a transfer to ICHD. In contrast, only 0.1% of all hospitalizations for those receiving ICHD occurred within the 30 days after transfer to another treatment (Supplemental Table 4). As-treated rates were slightly lower in the sensitivity analysis after treatment restrictions were removed for all home dialysis modalities (Supplemental Table 5).

Table 2.

Daily variation in the number, proportion, and rate of admissions for each day of the week across dialysis modalities

Measure	Overall	Monday	Tuesday	Wednesday	Thursday	Friday	Saturday	Sunday
In-center hemodialysis (N=21,095)
Count	85,528	14,494	14,992	13,413	12,934	12,232	9103	8360
Proportion, %	—	17	18	16	15	14	11	10
Rate (admissions/1000 patient days)a	2.98	0.50	0.52	0.47	0.45	0.43	0.32	0.29
Frequent in-center hemodialysis (N=392)
Count	1458	260	241	217	220	202	172	146
Proportion, %	—	18	17	15	15	14	12	10
Rate (admissions/1000 patient days)a	3.76	0.67	0.62	0.56	0.57	0.52	0.44	0.38
Home hemodialysis (N=236)
Count	806	139	138	139	114	119	82	75
Proportion, %	—	17	17	17	14	15	10	10
Rate (admissions/1000 patient days)a	2.16	0.37	0.37	0.37	0.31	0.32	0.22	0.20
Frequent home hemodialysis (N=445)
Count	1590	226	267	284	261	229	162	161
Proportion, %	—	14	17	18	16	14	10	10
Rate (admissions/1000 patient days)a	2.13	0.30	0.36	0.38	0.35	0.31	0.22	0.22
Peritoneal dialysis (N=5262)
Count	22,366	3555	3663	3526	3464	3323	2368	2467
Proportion, %	—	16	16	16	16	15	11	11
Rate (admissions/1000 patient days)a	2.71	0.43	0.44	0.43	0.42	0.40	0.29	0.30

Hours per treatment and treatment frequencies for hemodialysis were as follows: in-center hemodialysis/home hemodialysis, 3–6 hours, two to four times per week; frequent in-center hemodialysis/frequent home hemodialysis, >6 hours, five to six nights per week, and 2–3 hours, 5–7 d/wk.

^aFollow-up time is inclusive of patients without an admission and accounted for all modality changes using the restrictions noted in the Materials and Methods.

Day of the Week Admission across Modalities

For all modality groups, both the proportion (Table 2) and rate (Table 2, Figure 1) of weekend day admissions (i.e., an admission on a Saturday or Sunday) was lower than the other days of the week. The highest rate of hospitalization for both patients on frequent ICHD and ICHD was on Monday (0.67 and 0.50 per 1000 patient days, respectively) and Tuesday (0.62 and 0.52 per 1000 patient days, respectively). In contrast, there was less variability in the rate of hospitalization for each nonweekend day of the week for HHD, frequent HHD, and PD. Variations in the proportion of hospitalizations for each day of the week among admitted patients are noted in Figure 2 (in-center modalities) and Figure 3 (home modalities). For ICHD, HHD, and frequent ICHD, the highest proportion of admissions occurred on Monday and Tuesday (35%, 34% and 34%, respectively, for each modality; Table 2). In contrast, patients receiving PD had a more consistent risk of admission for each nonweekend day of the week. The highest proportion of admissions occurred on Wednesday (18%), for patients receiving frequent HHD, but the proportion admitted for Monday and Tuesday was slightly lower than other HD modalities (31%).

Figure 1.

Rates of hospitalization for each day of the week, stratified by dialysis modality. For all modality groups, the rate of weekend day admissions (i.e., an admission on a Saturday or Sunday) was lower than the other days of the week. Hours per treatment and treatment frequencies for hemodialysis were as follows: in-center hemodialysis (ICHD)/home hemodialysis (HHD), 3–6 hours, two to four times per week; frequent ICHD/frequent HHD, >6 hours, five to six nights per week, and 2–3 hours, 5–7 d/wk.

Figure 2.

Radar plot of day of the week admission (proportion). (A) ICHD and (B) frequent ICHD. Patients receiving ICHD and frequent ICHD experienced the highest proportion of hospitalizations on Monday and Tuesday. Hours per treatment and treatment frequencies for ICHD, 3–6 hours, two to four times per week; frequent ICHD, >6 hours, five to six nights per week, and 2–3 hours, 5–7 d/wk. ICHD, in-center hemodialysis.

Figure 3.

Radar plots of day of the week admission (proportion). (A) frequent HHD, (B) HHD, and (C) peritoneal dialysis. Patients receiving HHD experienced the highest proportion of hospitalizations on Monday and Tuesday, but this finding was not observed for those receiving frequent HHD. Patients receiving peritoneal dialysis had a more consistent risk of admission for each non weekend day of the week. Hours per treatment and treatment frequencies were as follows: HHD, 3–6 hours, two to four times per week; frequent HHD, >6 hours, five to six nights per week, and 2–3 hours, 5–7 d/wk. HHD, home hemodialysis.

Primary Outcome

Patients receiving frequent HHD had a lower adjusted odds of a Monday/Tuesday admission (versus all other days) compared with those on ICHD (OR, 0.89; 95% CI, 0.81 to 0.97; P=0.007; Figure 4). Patients receiving PD also had a small, but statistically significant, lower odds of Monday/Tuesday admission (OR, 0.95; 95% CI, 0.93 to 0.97; P<0.001; Figure 4). There were no significant differences in the odds of a Monday/Tuesday admission for either HHD or frequent ICHD compared with ICHD. No other individual patient factors were associated with Monday/Tuesday admissions outside of Black race and albumin, and there were small differences for some, but not all, centers (Supplemental Table 6).

Figure 4.

Unadjusted and adjusted odds ratios (95% CIs) of a Monday/Tuesday admission comparing HHD, frequent ICHD, peritoneal dialysis, and frequent HHD (reference: ICHD). Patients receiving frequent HHD had a lower adjusted odds of a Monday/Tuesday admission (versus all other days) compared with those on ICHD. Patients receiving PD also had a small, but statistically significant, lower odds of Monday/Tuesday admission. Hours per treatment and treatment frequencies for hemodialysis were as follows: ICHD/HHD, 3–6 hours, two to four times per week; frequent ICHD/frequent HHD, >6 hours, five to six nights per week, and 2–3 hours, 5–7 d/wk. CIs, confidence intervals; ICHD, in-center hemodialysis; HHD, home hemodialysis.

Sensitivity and Secondary Analyses

Results were consistent with the results of the primary analysis when using multiple imputation (Table 3). Within the first year after dialysis initiation, the risk of a Monday/Tuesday admission was even lower for frequent HHD (OR, 0.79; 95% CI, 0.63 to 0.98) compared with ICHD, but no longer significantly different for those receiving PD (Table 3). In an analysis of admissions due to cardiovascular causes, the findings were even more pronounced (Table 4). The lowest risk of a Monday/Tuesday admission was observed for patients on frequent HHD compared with those on ICHD (OR, 0.68; 95% CI, 0.48 to 0.96). No significant differences were observed for hospitalizations associated with dialysis or nondialysis infections.

Table 3.

Risk of a Monday/Tuesday admission across dialysis modalities: Sensitivity analyses

Model	Odds Ratio (95% Confidence Interval)	P Value
Adjusted: Inclusive of all admissions from index date a
In-center HD	Reference	—
Home HD	1.05 (0.94 to 1.18)	0.40
Frequent in-center HD	0.97 (0.88 to 1.06)	0.49
Frequent home HD	0.87 (0.80 to 0.95)	0.003
Peritoneal dialysis	0.95 (0.92 to 0.97)	<0.001
Adjusted: Admissions within first year a
In-center HD	Reference	—
Home HD	0.97 (0.80 to 1.17)	0.74
Frequent in-center HD	0.89 (0.79 to 1.01)	0.07
Frequent home HD	0.79 (0.63 to 0.98)	0.04
Peritoneal dialysis	0.99 (0.96 to 1.03)	0.65
Adjusted: Without 30 day lag in events a
In-center HD	Reference	—
Home HD	1.00 (0.95 to 1.07)	0.87
Frequent in-center HD	0.98 (0.94 to 1.03)	0.41
Frequent home HD	0.94 (0.90 to 0.99)	0.02
Peritoneal dialysis	0.97 (0.96 to 0.98)	<0.001
Adjusted: Primary analysis using multiple imputation for missing data a , b
In-center HD	Reference	—
Home HD	1.00 (0.91 to 1.09)	>0.99
Frequent in-center HD	0.99 (0.93 to 1.07)	0.87
Frequent home HD	0.90 (0.84 to 0.97)	0.004
Peritoneal dialysis	0.95 (0.93 to 0.97)	<0.001

Hours per treatment and treatment frequencies for hemodialysis were as follows: in-center hemodialysis/home hemodialysis, 3–6 hours, two to four times per week; frequent in-center hemodialysis/frequent home hemodialysis, >6 hours, five to six nights per week, and 2–3 hours, 5–7 d/wk. HD, hemodialysis. —, no data.

^aAdjusted for baseline variables at start of dialysis including age, sex, race, cause of CKD, comorbidities (diabetes, coronary artery disease, congestive heart failure, cerebrovascular disease, peripheral vascular disease, malignancy, chronic obstructive lung disease), facility, hemoglobin, and albumin.

^bA total of 6459 individuals (17%) with any combination of missing data imputed.

Table 4.

Risk of a Monday/Tuesday admission for cause-specific admissions across dialysis modalities

Modela	Odds Ratio (95% Confidence Interval)	P Value
Cardiovascular (N=9681)
In-center HD	Reference	—
Home HD	0.96 (0.73 to 1.28)	0.79
Frequent in-center HD	1.06 (0.84 to 1.34)	0.62
Frequent home HD	0.68 (0.48 to 0.96)	0.03
Peritoneal dialysis	0.98 (0.91 to 1.06)	0.65
Dialysis infection (N=4452)
In-center HD	Reference	—
Home HD	1.08 (0.63 to 1.87)	0.79
Frequent in-center HD	0.72 (0.41 to 1.28)	0.26
Frequent home HD	1.17 (0.86 to 1.59)	0.33
Peritoneal dialysis	0.96 (0.87 to 1.07)	0.49
Nondialysis infection (N=13,403)
In-center HD	Reference
Home HD	1.08 (0.80 to 1.46)	0.62
Frequent in-center HD	1.04 (0.83 to 1.31)	0.74
Frequent home HD	0.86 (0.68 to 1.09)	0.21
Peritoneal dialysis	0.98 (0.91 to 1.05)	0.57

Hours per treatment and treatment frequencies for hemodialysis were as follows: in-center hemodialysis/home hemodialysis, 3–6 hours, two to four times per week; frequent in-center hemodialysis/frequent home hemodialysis, >6 hours, five to six nights per week, and 2–3 hours, 5–7 d/wk. HD, hemodialysis. —, no data.

^aAdjusted for baseline variables at start of dialysis including age, sex, race, cause of CKD, comorbidities (diabetes, coronary artery disease, congestive heart failure, cerebrovascular disease, peripheral vascular disease, malignancy, chronic obstructive lung disease), facility, hemoglobin, and albumin.

Discussion

In this national study, we identified differences in the pattern of hospitalization episodes across dialysis modalities. We found that patients receiving frequent HHD and PD had a lower relative risk of an early week (Monday/Tuesday) admission compared with ICHD. This lower risk was even more pronounced comparing frequent HHD with ICHD for admissions due to cardiovascular causes.

This study adds to previous literature identifying that, for patients receiving ICHD, Mondays and Tuesdays (often corresponding to the long dialysis break) are associated with a higher risk for poor outcomes (17,18,23). However, there are differences when evaluating dialysis modalities (19). In the study conducted by Krishnasamy et al. (19), there was no association between day of the week and cardiac mortality for HHD or PD. In contrast, they identified that, among those receiving ICHD, the risk of cardiac mortality was higher on Mondays compared with other days. Similarly, we found that those receiving frequent HHD or PD were at a relatively lower risk of a Monday/Tuesday admission that was even more pronounced for frequent HHD when exclusively analyzing admissions for cardiovascular causes. Mechanisms underpinning this lower risk may relate to an amelioration of putative risk factors for cardiac morbidity and mortality after the long dialysis interval, including higher interdialytic fluid gain (24), hyperkalemia (which is more common after the long dialysis interval) (19,23), and other electrolyte abnormalities (25,26). These risks would not be expected for patients receiving frequent HHD, where the long break was interrupted with a dialysis treatment.

Although it is established that frequent HHD may be associated with a higher risk of vascular access events (27), we did not identify a difference in the pattern of admission for dialysis-specific infection risk across modalities. We hypothesize this relates to the provision of home dialysis care, namely, that many factors that might otherwise lead to hospitalization if left unaddressed, are managed through outpatient units. For example, in a national survey of HHD facilities, 77% of facilities had 24-hour nursing on-call care (28), and provision of nursing support may allow successful continuation of dialysis treatments and avoidance of potential adverse events (29). Whether differences exist on the basis of the presence or absence of after-hour on-call nursing support is yet unknown.

Patients receiving PD had only a relatively small, reduced risk of Monday/Tuesday admission for all-cause hospitalization compared with those on ICHD, and there was no difference for cause-specific hospitalizations. Although the pattern of hospitalization was generally more consistent across weekdays, one might have expected a more pronounced relative reduction in hospitalization for Mondays/Tuesdays given the lack of a long dialysis interval. Importantly, both the proportion and rate of hospitalizations was the lowest for all modalities on the weekend days inclusive of PD. We hypothesize this may relate to patient avoidance or a lack of accessibility to PD service during the weekend. A prior study identified that peritonitis rates peak on Monday (30) and it was hypothesized that the lack of outpatient or on-call resources may have led to this finding. Similarly, in our study, rates of infection-related hospitalization were the same for PD versus ICHD. This would be expected to dampen any signal of a higher hospitalization risk on Mondays/Tuesdays for ICHD versus PD. Once again, a future study looking at the pattern of hospitalization (all cause and cause specific) on the basis of availability of weekend access to the home unit or on-call support would be valuable.

We identified that patients receiving ICHD and frequent ICHD had the highest baseline rates of hospitalization. Relative comparisons of hospitalizations for dialysis modalities are quite variable in their findings, likely due to the metric being used (hospital days versus hospitalizations) (15,31–33), timing of capture of the exposure definition (34), and variations in the practice of dialysis by country (10). Specific to frequent ICHD, in a prior comparison of survival across modalities, frequent or extended facility HD was not associated with a survival advantage (35). We hypothesize that some Canadian patients on dialysis who are receiving either short daily or nocturnal ICHD may be those where HD has been modified to address a baseline risk factor that, in and of itself, places that patient at a higher risk of hospitalization (indication bias). For example, a proportion of patients receive frequent ICHD to promote ultrafiltration or maximize clearance in response to volume overload or underdialysis. These patients are likely at a high baseline risk for frequent hospitalization due to their underlying comorbid conditions. Similarly, the criteria for entry into some in-center nocturnal HD programs are a lack of eligibility for HHD (which is a marker of a healthier patient) (36), and it has been shown, in a more recent Canadian cohort, that frailty (a marker of poor outcomes among patients on dialysis) is prevalent among patients receiving nocturnal ICHD (mean Clinical Frailty Scale score of four, 1 SD, corresponding to “vulnerable”) (37). The high baseline hospitalization rate among those receiving frequent ICHD may also be due to the lack of granularity around categories of treatment, leading to some patients being categorized as frequent ICHD who may not have been. Patients who require five treatments per week for fluid management would not have fit any category provided by CORR, but may have been arbitrarily included as either short daily ICHD or nocturnal HD. These patients are at a high risk of subsequent hospitalization. To provide a more accurate picture of hospitalizations related to a given treatment, we used a lag of 30 days, and we did not count modality switches unless they were ≥30 days. For the primary analysis, this likely led to fewer hospitalizations being attributed to ICHD or frequent ICHD because hospitalizations corresponding to technique failure would have related to the primary home dialysis modality. As expected, the rate of hospitalization for both HHD and frequent HHD was even lower when these restrictions were removed, appreciating that 7% of hospitalizations for both treatments occurred within 30 days of a transition to another modality.

We hypothesize that the lack of a fall on Monday/Tuesday admissions for those on frequent ICHD relates to treatment delivery. Particularly for those receiving dialysis five times per week, it is conceivable that the long dialysis break will remain (i.e., if an individual is scheduled to receive Monday through Friday or Tuesday through Saturday dialysis). Unfortunately, we were unable to determine the proportion of patients receiving five versus six treatments per week among those on frequent ICHD.

What are the implications of our findings? Identifying variability in the pattern of hospitalization for dialysis modalities is the first step toward developing strategies to reduce overall and cause-specific hospitalization. Specific to HHD for example, if one were at a high risk of cardiovascular morbidity on the basis of baseline characteristics, modifying the HHD schedule to avoid a 2-day break may ameliorate the established putative causes (i.e., fluid overload and hyperkalemia). This individualization of HHD therapy (38) need not imply a change in frequency or hours, but a redistribution of days of the week. Specific to PD, one may interpret these data in a different fashion. Acknowledging the observed lower risk of admission on weekends, but not early week admissions (which might have been expected), future studies should explore whether patients delay presentation to hospital on the weekend and what might underpin this decision. This may be especially important given the risk of delayed treatment for PD-related peritonitis on outcomes (39).

There are limitations to this study that should be considered. Although this was a large representative sample of patients receiving dialysis in Canada, we appreciate that, due to difficulties with obtaining or linking data, we did not capture patients from either Quebec or Manitoba. We did not have access to dialysis schedules to ascertain whether the Monday/Tuesday admissions were necessarily subsequent to the long break for a given patient. This also applies to those receiving frequent HHD schedules; a group in whom we did not have access to the exact frequency of dialysis. It is possible that the findings may have been more or less pronounced for modality-specific variations in the pattern of admission for different dialysis schedules. Furthermore, we did not have data on missed or shortened dialysis treatments, known risk factors for hospitalization (40). Although we defined modalities using registry-provided categories, we did not have detail on the hours or frequency of dialysis outside of CORR definitions, and this may have led to misclassification. For example, treatment frequencies up to four times per week were included in the CORR category for ICHD and HHD, but this could also have been categorized as a frequent schedule. Furthermore, it is possible that more granular treatment definitions would have demonstrated different results. For example, the data from the Australia and New Zealand Dialysis and Transplant Registry suggest that patients receiving quasi-intensive HHD (longer and more frequent, but less than five sessions per week) may have a lower risk of mortality (41), and more detail as to the effect of hours and days per week would have been valuable. In addition to issues around exposure definition, we acknowledge the possibility in this observational study that there may have been unmeasured or unknown confounders that, if included, would have changed the association between modality and day of the week hospitalization. Finally, as some variables were provider identified (in particular, race and sex), it is possible that misclassification occurred with this confounding variable and that using a self-identified definition might have had an effect on the results of our primary analysis.

In conclusion, we found differences in day of the week admission across dialysis modalities, such that patients receiving frequent HHD and PD were less likely to be admitted to hospital on a Monday/Tuesday compared with those receiving ICHD. Future studies to explore the implications and practical application of this finding will need to be conducted.

Disclosures

C.T. Chan reports having consultancy agreements with, and serving as a scientific advisor for, or member of, DaVita, Medtronic, and Quanta; and receiving research funding from Medtronic (through their external grant program). A.-C. Nadeau-Fredette reports having a current scholarship from Fonds de la recherche du Québec en Santé. J. Perl reports receiving research funding and salary support from Agency for Healthcare Research and Quality and Arbor Research Collaborative For Health; receiving honoraria from AstraZeneca, Baxter Healthcare USA/Canada, DaVita Healthcare Partners, DCI, Fresenius Medical Care, and US Renal Care; having consultancy agreements with AstraZeneca, Baxter Healthcare Canada, DaVita Healthcare Partners, Fresenius Medical Care, LiberDi, and Otsuka; and serving on a speakers bureau for Baxter Healthcare and Fresenius Medical Care. K.K. Tennankore reports receiving research funding from Astellas Canada and Otsuka Canada; serving on a speakers bureau for AstraZeneca, Baxter, and Bayer; having consultancy agreements with AstraZeneca, Baxter, Bayer, Janssen, and Otsuka; receiving honoraria from AstraZeneca, Bayer, and Otsuka; and serving as associate editor for the Canadian Journal of Kidney Health and Disease. E. Trinh reports having consultancy agreements with Baxter International. The remaining author has nothing to disclose.

Funding

None.

Author Contributions

All authors were responsible for investigation; C.T. Chan, A.-C. Nadeau-Fredette, J. Perl, K.K. Tennankore, and E. Trinh conceptualized the study; C.T. Chan, A.-C. Nadeau-Fredette, J. Perl, and E. Trinh reviewed and edited the manuscript; K. Matheson was responsible for formal analysis and validation; K.K. Tennankore and K. Matheson were responsible for visualization; K. Matheson and K.K. Tennankore wrote the original draft; K.K. Tennankore was responsible for project administration; K.K. Tennankore was responsible for data curation; and all authors were responsible for methodology.

Supplemental Material

This article contains supplemental material online at http://kidney360.asnjournals.org/lookup/suppl/doi:10.34067/KID.0003552021/-/DCSupplemental.

Supplemental Table 1

Admission diagnoses (using ICD-10 codes mapped to previously defined ICD-9 codes). Download Supplemental Table 1, PDF file, 253 KB ^{(252.9KB, pdf)}

Supplemental Table 2

Proportion of patients experiencing a modality switch and duration of treatment prior to switch (or until last follow-up). Download Supplemental Table 2, PDF file, 253 KB ^{(252.9KB, pdf)}

Supplemental Table 3

Characteristics of the study population stratified by initial modality. Download Supplemental Table 3, PDF file, 253 KB ^{(252.9KB, pdf)}

Supplemental Table 4

Number of hospitalizations that occurred within 30 days of a transition to another modality but attributed to the modality prior to transition. Download Supplemental Table 4, PDF file, 253 KB ^{(252.9KB, pdf)}

Supplemental Table 5

Rates of hospitalization for each day of the week stratified by modality: sensitivity analyses. Download Supplemental Table 5, PDF file, 253 KB ^{(252.9KB, pdf)}

Supplemental Table 6

Adjusted risk of a Monday/Tuesday admission across dialysis modalities (inclusive of all confounding variables). Download Supplemental Table 6, PDF file, 253 KB ^{(252.9KB, pdf)}