Predictors of Health Deterioration Among Older New Zealanders Undergoing Dialysis: A Three-Year Accelerated Longitudinal Cohort Study.

BACKGROUND: Patient involvement in dialysis decision-making is crucial, yet little is known about patient-reported outcomes over time on dialysis.
OBJECTIVE: To examine health-related outcomes over 24 and 36 months in an older cohort of dialysis patients.
DESIGN: The "Dialysis outcomes in those aged ≥65 years study" is a prospective longitudinal cohort study of New Zealanders with kidney failure.
SETTING: Three New Zealand nephrology units. PATIENTS: Kidney failure (dialysis and predialysis) patients aged 65 or above. We have previously described outcomes after 12 months of dialysis therapy relative to baseline. MEASUREMENTS: Patient-reported social and health factors using the SF-36, EQ-5D, and Kidney Symptom Score questionnaires.
METHODS: This article describes and compares characteristics of 120 older kidney failure patients according to whether they report "Same/better" or "Worse" health 24 and 36 months later, and identifies predictors of "worse health." Modified Poisson regression modeling estimated relative risks (RR) of worse health.
RESULTS: Of 120 patients at 12 months, 47.5% had worse health or had died by 24 months. Of those surviving at 24 months (n = 80), 40% had "Worse health" or had died at 36 months. Variables independently associated with reduced risk of "Worse health" (24 months) were as follows: Māori ethnicity (RR = 0.44; 95% CI = 0.26-0.75), Pacific ethnicity (RR = 0.39; 95% CI = 0.33-0.46); greater social satisfaction (RR = 0.57; 95% CI = 0.46-0.7). Variables associated with an increased risk of "Worse health" were as follows: problems with usual activities (RR = 1.32; 95% CI = 1.04-1.37); pain or discomfort (RR = 1.48; 95% CI = 1.34, 1.63). At 36 months, lack of sense of community (RR = 1.41; 95% CI = 1.18-1.69), 2 or more comorbidities (RR = 1.21; 95% CI = 1.13-1.29), and problems with poor health (RR = 1.47; 95% CI = 1.41-1.54) were associated with "Worse health." LIMITATIONS: Participant numbers restricted the number of variables able to be included in the multivariable model, and hence there may have been insufficient power to detect certain associations.
CONCLUSIONS: In this study, the majority of older dialyzing patients report "Same/better health" at 24 and 36 months. Māori and Pacific people report better outcomes on dialysis. Social and/or clinical interventions aimed at improving social satisfaction, sense of community, and help with usual activities may impact favorably on the experiences for older dialysis patients. TRIAL REGISTRATION: Australian and New Zealand clinical trials registry: ACTRN12611000024943.

Introduction

New Zealand has an increasing number of people who are aged above 65 years and dialyzing. Older age is independently associated with a 5-fold increase in death within 90 days of dialysis therapy initiation. It has been reported that initiation of dialysis is linked to a decline in functional status among elderly patients with end-stage kidney disease and, therefore, can negatively affect their quality of life. Consideration of these factors in shared decision-making with patients at the time of initiation of invasive treatment like dialysis is essential. To inform such shared decision making, data about older patient populations’ experiences, and patient-reported outcomes of dialysis are required.

The DOS65+ (Dialysis Outcomes in Those Aged ≥65 Years Study) is a study of older New Zealanders (NZers) with chronic kidney disease stage 5 (CKD5).[4 -6] Previously, we have investigated global self-reported health status at 12 months compared to time of recruitment (baseline) to DOS65+ for 150 patients who were dialyzing at time of recruitment. A significant proportion of participants with poorer health at 12 months, reported less satisfaction with social relationships, little sense of community, and/or had mobility problems at baseline. Multivariable analysis found participants reporting Māori or Pacific ethnicities were less likely to report poorer health at 12 months compared to NZ European participants. Patients with higher number of comorbidities, poor health and/or problems with EQ-5D-3L anxiety or depression at baseline reported worse health outcomes. With regards to modality of dialysis, patients on peritoneal dialysis were at reduced risk of reporting worse health outcomes. This article reports global self-reported health outcomes of DOS65+ participants on dialysis therapy to medium and longer-term follow-up points (ie, 24 months and 36 months) relative to 12 months earlier. Specifically, we aim to investigate which, if any, demographic, health and functioning, social, and clinical characteristics are associated with “poorer health” at 24 and 36 months compared to the 12 months before. Prospective long term data related to factors which are important determinants of quality of life outcomes, as well as survival for older individuals, will help to inform patient focused discussions around dialysis.

Methods

The DOS65+ protocol and baseline data have previously been described.[5,6] DOS65+ is an “accelerated longitudinal design” prospective cohort study Participants were eligible if they were aged ≥65 years, had kidney failure, were on an active conservative pathway, or had commenced kidney replacement therapy, or preparing to commence kidney replacement therapy having received kidney replacement education. Three New Zealand District Health Board (DHB) nephrology units (Counties Manukau, Hawkes Bay, and Southern) were involved in the study. Counties Manukau DHB has a tertiary nephrology unit serving a large urban population with high proportions of Māori, Pacific and lower socio-economic status patients. Hawkes Bay DHB is a provincial rural center with a relatively high proportion of Māori. Southern DHB has a tertiary nephrology center with a more geographically dispersed population, with an exclusive home dialysis policy.[4 -6] New Zealand has a tax-funded public health care system for all citizens. As such, New Zealand is well suited to outcomes research for patients with kidney failure because there are neither direct health care costs related to dialysis incurred by patients, nor financial incentives for health care professionals affecting treatment choices or service provision.[7,8]

Eligible patients, identified from treatment center databases, were contacted by telephone to arrange an interview. All interviews were completed either by telephone or face-to-face by trained DOS65+ research interviewers, which were independent of the nephrology team providing patient care.[4 -6] Interviews were conducted at baseline (ie, recruitment), with follow-up interviews occurring 12, 24, and 36 months later. At the time of each follow-up, participants had to be clinically stable with no recent inter-current illness requiring hospitalization within four weeks. Follow-up interviews were not done if the treating nephrologist determined an interview to be inappropriate (eg, because the patient had a terminal diagnosis or a serious cognitive impairment making an interview impossible).[4 -6]

DOS65+ received ethical approval from the New Zealand Multi-Regional Ethics Committee (MEC/10/084), and was registered with the Australian and New Zealand clinical trials registry: ACTRN12611000024943. Study enrolment ran from 2010 to 2014.

Variables Collected at Each Interview

Information collected included demographics, socio-economic, health, and functioning, dialysis modality (either hemodialysis or peritoneal dialysis), and end-stage kidney disease symptoms. Demographic information included age, sex, ethnicity,[9,10] and living arrangements. The socio-economic information included household income where participants were asked to respond as “more than enough, enough or not enough money to meet their everyday needs, such as accommodation, food, clothing, or other necessities?” Participants were asked about their satisfaction with social relationships, family involvement, and their sense of community and responses were categorized as reported in Table 1, consistent with earlier analyses.[4,13-15]

Table 1.

12-Month Characteristics of Patients Stratified by Health Status at 24 Months.

12 months characteristics	Health state at 24 months relative to 12 months
	Same or better (N = 67)		Worse or died (N = 53)		P-value
	n	%	n	%
Demographics
Age a
65-69	33	70.2	14	29.8
70-74	20	52.6	18	47.4
75-79	10	43.5	13	56.5	.05
80+	4	33.3	8	66.7
Sex a
Male	39	52.0	36	48.0
Female	28	62.2	17	37.8	.28
Body mass index a
18.5-24.9	17	51.5	16	48.5
25.0-29.9	17	53.1	15	46.9
30+	33	60.0	22	40.0	.69
Ethnicity (prioritized)
Maori	21	77.8	6	22.2
Pacific	17	70.8	7	29.2
European	22	40.7	32	59.3	.01
Other	7	46.7	8	53.3
Social factors
Social relationships
Completely/mostly satisfied	61	55.5	49	44.5
Others	4	80.0	1	20.0	.39
Sense of community
Some/little	32	51.6	30	48.4
Strong	33	62.3	20	37.7	.36
Living arrangement
With others	58	57.4	43	42.6
Alone	7	50.0	7	50.0	.78
Family involvement
Small/very small	5	55.6	4	44.4
Large/very large	60	56.1	47	43.9	1.00
Adequacy of household income
Not enough/just enough	34	59.6	23	40.4
Enough/more than enough	31	54.4	26	45.6	.70
General health measures
Comorbidities
CVD	40	51.9	37	48.1	0.44
Cerebrovascular diseases	3	33.3	6	66.7	0.30
Peripheral vascular diseases	12	46.2	14	53.8	0.38
Diabetes	32	53.3	28	46.7	0.85
Lung diseases	13	65.0	10	43.5	1.00
Musculoskeletal	23	56.1	18	43.9	0.84
Cancer (except skin cancer)	12	48.0	13	52.0	0.50
Other comorbidities	36	51.4	34	48.6	0.45
Comorbidity count
0-2	32	61.5	20	38.5
3+	32	49.2	33	50.8	0.18
Overall general health
Fair/poor	10	32.3	21	67.7
Excellent/very good/good	55	64.7	30	35.3	<0.01
EQ5D mobility
No problems	32	65.3	17	34.7
Some problems	32	48.5	30	48.4	0.10
Extreme problems	1		3	75.0
EQ5D selfcare
No problems	50	61.0	32	39.0
Some problems	13	39.4	12	48.0	0.13
Extreme problems	2		6	75.0
EQ5D usual activities
No problems	30	66.7	15	33.3
Some problems	29	41.4	27	48.2	0.80
Extreme problems	6		8	57.1
EQ5D pain/discomfort
No problems	42	62.7	25	37.3
Some problems	21	43.8	24	53.3	0.11
Extreme problems	2		1	33.3
EQ5D anxiety/depression
No problems	52	59.1	36	40.9
Some problems	12	44.4	14	53.8	0.32
Extreme problems	1		0	0.0
Intellectual ability
No problems	49	62.8	29	37.2
Some problems	13	35.1	21	61.8	0.05
Extreme problems	3		0	0.0
Chronic kidney disease (CKD)-related factors
Type of dialysis
Hemodialysis	47	65.3	25	34.7
PD	16	39.0	25	61.0	0.01
Not dialyzing now	4	57.1	3	42.9
Vintage
0 to <2 years	21	61.8	13	38.2
2 to <6 years	38	53.5	33	46.5
6+ years	8	53.3	7	46.7	0.72
Dialysis dose at baseline(minutes dialyze per week)
<120	2	66.7	1	33.3
120-239	6	50.0	6	50.0
240-359	43	61.4	27	38.6
360+	14	45.2	17	54.8	0.45
Cause for CKD
Glomerulonephritis	16	69.6	7	30.4
Hypertensive vascular	6	28.6	15	71.4
Polycystic	5	62.5	3	37.5
Diabetics	27	60.0	18	40.0
Other	13	56.5	10	43.5	0.07
Kidney Disease Quality Of Life score (KDQOL)
70+	49	59.8	33	40.2
<70	16	47.1	18	52.9	0.22

Note. P-value is from chi-square test or Fisher’s exact as appropriate.

Age, sex, body mass index, and ethnicity were as at baseline.

Clinical information was collected by the treating DHB nephrologists, with participants’ consent, from the clinical records at recruitment to DOS65+ and throughout follow-up. This included the etiology of kidney failure and comorbidity which was categorized according to the Dialysis Outcomes and Practice Patterns study (DOPPS) and the Australian and New Zealand Dialysis and Transplant Registry (ANZDATA).[1,16] Patients were recorded as having, or not having, the following conditions: cardiovascular disease, cerebrovascular disease, peripheral vascular disease, lung disease, musculoskeletal disease, diabetes, cancers (other than skin cancer).[1,16]

For general health status, we used the EQ-5D-3L. EQ-5D-3L asks about problems with mobility, self-care, usual activities, pain or discomfort, anxiety or depression; for multivariable analysis people were categorized as having “No problems” or “Any problems.”

We assessed kidney symptom burden using the KDQOL questionnaire. For multivariable modeling, this score was dichotomized as greater bother (≥31) and less or no bother (<31) as done previously.

The key outcome of interest for this article was determined by asking participants: “Compared to 1 year ago, how would you rate your health in general now?” People who responded that their health is “About the same as 1 year ago,” “Somewhat better now than 1 year ago,” or “Much better than 1 year ago” were classified as having the “Same or better health”; those responding that their health is “Somewhat worse than 1 year ago” or “Much worse now than 1 year ago” were classified as having “Worse health.” Participants who faced the previous interview but had died before the next follow-up were also included in the “Worse health” group.

Analysis

All eligible participants experiencing “same/better” or “worse/died” health outcomes at 24 or 36 months relative to their health state 12 months previously were examined according to each potential explanatory variable using the Chi-square test or Fisher’s Exact test as appropriate. To identify possible predictors of “Worse health” at 24 and 36 months, we used univariate and multivariable modified Poisson regression models with robust standard errors to estimate relative risks (RRs) and 95% confidence intervals (CIs). To overcome the insufficiency of the sample size to implement a variable selection from the large list of prospective predictor variables in multivariable model building, and as a practical approach to prevent model overfitting, for 24 and 36 months separately, we used a 2-step process to develop multivariable models. At the first step, we developed separate sub-models for each of the following 4 groups of characteristics: demographics (age, sex, and ethnicity), socioeconomic (social satisfaction, sense of community, living arrangements, household income, and family involvement), health and functioning (comorbidities, overall general health, and 5 EQ-5D-3L domains), and kidney-related variables (dialysis modality, dialysis vintage, dialysis dose, cause for kidney failure, and KDQOL score). This step used backward stepwise procedures to identify “important” variables using P-value thresholds of P = .20 to enter variables to the sub-model and P = .25 to remove variables from the sub-model. This step was a screening of variables that are unlikely to be related to the outcome. These very large P-value thresholds were used to retain variables that may be “important” when adjusting for variables used in other sub-models even though not “important” when adjusted for variables used in the current sub-model. At the second step, the final model was developed using variables identified in any of the 4 sub-models. Backward stepwise procedures were used again but with P-value thresholds of .10 and .12 used respectively to enter and remove variables. As done in previous studies,[4,5] comorbidities, dialysis modality, and dialysis vintage were not allowed to be dropped from any of the models they were used in (irrespective of P-values), and possible clustering by treating nephrology center was accounted for in the modeling.[4,5] Analyses were undertaken using Stata 15.1 software.

Results

Study Participants: The DOS65+ study recruited 225 participants. Of these, 56 participants were not on dialysis at baseline. Of the remaining 169 participants, 26 had died in the first 12 months (recruitment to 12-month interview) and 23 had missing outcome data. We have previously described the outcomes at 12 months of dialysis therapy relative to baseline. The flow chart of participants is described in Figure 1. Data from 120 participants was available for analysis at 24 months.

Figure 1.

Flow chart of DOS65+ (Dialysis Outcomes in Those Aged ≥ 65 Years).

Note. Study participants recruited for analysis at 24 months and 36 months.

24-Month Analyses

Table 1 presents participants’ characteristics at 12 months according to health outcome at 24 months follow-up. Among demographic characteristics, a higher proportion of those in the “Worse health/died” group were above the age of 70 years. A higher proportion of NZ Europeans (n = 32; 59%) were in the “Worse health/died” group compared to Māori (n = 6; 22%) and Pacific peoples (n = 7; 29%). Participants with cerebrovascular disease at 12 months reported worse health outcomes compared to those without cerebrovascular disease. The EQ-5D-3L found a higher proportion of patients who reported problems with mobility, usual activities at 12 months were in the “Worse health/died ‘ at 24 months. Those with problems with intellectual activity were also in ‘Worse health.”

Multivariate analysis of characteristics at 12 months associated with “Worse health/died” at 24 months: Of the 120 participants, 6 had missing information for all 12 month variables retained in the multivariable model. Table 2 presents relative risk (RRs) of worse health at 24 months based on data from 114 participants who had non-missing information for all 12 month variables retained in the multivariable model. Participants older than 75 had worse health outcomes (RR = 1.32; 95% CI = 1.25-1.39) compared to their younger counterparts. Pacific (RR = 0.39; 95% CI = 0.33-0.46) and Māori (RR = 0.44; 95% CI = 0.26-0.75) participants had lower risk of “Worse health/died” at 24 months compared to New Zealand Europeans and other ethnicities. Among health and health status characteristics at 12 months, problems with EQ-5D usual activities (RR = 1.32; 95% CI = 1.04-1.67) and EQ-5D pain/discomfort (RR = 1.48; 95% CI = 1.34-1.63) independently placed participants in the “Worse health/died” group. EQ-5D mobility was not retained in the final model; nor was intellectual functioning.

Table 2.

Showing Multivariate Analysis of Outcomes at 24 Months Relative to Characteristics at 12 Months: Final Models. Outcome is the Health Status Becoming “Worse” or Death Occurring Within Past 12 Months.

Characteristics (measured 12 months prior to measuring the outcome)	24 months (N = 114)
	RR	95% CI for RR		P-value
Demographics
Age**				<.01
65-74	Ref
75+	1.32	1.25	1.39
Ethnicity(prioritized)**				<.01
Maori	0.44	0.26	0.75
Pacific	0.39	0.33	0.46
European/Other	Ref
Social factors
Social relationships
Completely/mostly satisfied	Ref
Others	0.57	0.46	0.70	<.01
General health factors
Comorbidity types
0-2	Ref
3+	1.34	0.75	2.39	.33
EQ5D usual activities
No problems	Ref
Some problems	1.32	1.04	1.67	.03
EQ5D pain/discomfort
No problems	Ref
Some problems	1.48	1.34	1.63	<.01
Chronic kidney disease factors
Type of dialysis
Hemodialysis	Ref			<.01
PD	1.21	1.08	1.35
Not dialyzing now	0.72	0.36	1.45
Vintage (years)#	1.01	0.99	1.03	.38
Kidney Disease Quality Of Life score (KDQOL)
70+ (less bother)	Ref
<70 (more bother)	0.87	0.79	0.96	.01

Note. RR = relative risks.

refers to age or ethnicity as recorded at entry into the study.

Among dialysis characteristics, more PD patients were in the “Worse health/died” group at 24 months (RR = 1.21; 95% CI = 1.08-1.35) relative to hemodialysis (HD) patients. Participants reporting greater bother with symptoms as per KDQOL symptom score at 12 months were at reduced risk for worse health outcomes at 24 months (RR = 0.87; 95% CI = 0.79-0.96).

36-Month Analyses

We had 120 participants for analysis at 24 months potentially available for analyses of health related outcomes at 36 months. Of these, 34 participants died between 12 and 24 months. Of the remaining 86, 6 had missing outcome data at 36 months. Thus, 80 participants were included for analysis at 36 months.

Table 3 presents participants’ characteristics at 24 months according to health outcomes at 36 months follow-up period. A higher proportion of Māori were in the “Same/Better health” group compared to New Zealand Europeans and Pacific ethnicities. NZ Europeans tended to experience “Worse health/died” outcome as compared to both Māori and Pacific. Patients with problems with intellectual ability at 24 months were in the “Worse health” group at 36 months.

Table 3.

24-Month Characteristics of Patients Stratified by Health State at 36 Months.

24-month characteristics	Health state at 36 months relative to 24 months
	Same or better (N = 48)		Worse or died (N = 32)		P-value
	n	%	n	%
Demographics
Age a
65-69	22	62.9	13	37.1
70-74	13	56.5	10	43.5
75-79	10	62.5	6	37.5	.90
80+	3	50.0	3	50.0
Sex a
Male	23	47.9	25	52.1
Female	25	78.1	7	21.9	.01
Body mass index a
18.5-24.9	14	63.6	8	36.4
25.0-29.9	12	63.2	7	36.8
30+	22	56.4	17	43.6	.87
Ethnicity (prioritized)
Maori	14	73.7	5	26.3
Pacific	10	62.5	6	37.5
European	21	61.8	13	38.2
Other	3	27.3	8	72.7	.10
Social factors
Social relationships
Completely/mostly satisfied	42	59.2	29	40.8
Others	2	66.7	1	33.3	1.00
Sense of community
Some/little	21	51.2	20	48.8
Strong	23	65.7	12	34.3	.25
Living arrangement
With others	41	58.6	29	41.4
Along	3	50.0	3	50.0	.69
Family involvement
Small/very small	3	42.9	4	57.1
Large/very large	41	59.4	28	40.6	.45
Adequacy of household income
Not enough/just enough	24	64.9	13	35.1
Enough/more than enough	22	55.0	18	45.0	.49
General health measures
Comorbidities
CVD	22	51.2	21	48.8	.31
Cerebrovascular diseases	2	50.0	2	50.0	1.00
Peripheral vascular diseases	8	53.3	7	46.7	.77
Diabetes	15	48.4	16	51.6	.23
Lung diseases	12	70.6	5	29.4	.26
Musculoskeletal	16	55.2	13	44.8	1.00
Cancer (except skin cancer)	10	62.5	6	37.5	.77
Other comorbidities	19	39.6	20	51.3	.14
Comorbidity count
0-2	17	60.7	11	39.3
3+	21	53.8	18	46.2	.57
Overall general health
Fair/poor	10	47.6	11	52.4
Excellent/very good/good	34	61.8	21	38.2	.31
EQ5D mobility
No problems	19	65.5	10	34.5
Some problems	24	51.1	20	45.5	.34
Extreme problems	1		2	66.7
EQ5D selfcare
No problems	30	56.6	23	43.4
Some problems	13	56.5	5	27.8	.80
Extreme problems	1		4	80.0
EQ5D usual activities
No problems	20	66.7	10	33.3
Some problems	20	41.7	15	42.9	.24
Extreme problems	4		7	63.6
EQ5D pain/discomfort
No problems	19	39.6	14	42.4
Some problems	21	48.8	17	44.7	1.00
Extreme problems	4		1	20.0
EQ5D anxiety/depression
No problems	32	56.1	25	43.9
Some problems	10	52.6	6	37.5	.79
Extreme problems	2		1	33.3
Intellectual ability
No problems	25	55.6	20	44.4
Some problems	19	61.3	11	36.7	.64
Extreme problems	0		1	100.0
Type of dialysis
Hemodialysis	29	56.9	22	43.1
PD	14	58.3	10	41.7	1.00
Not dialyzing now	5	100.0	0	0.0
Vintage
0 to <2 years	—	—	—	—
2 to <6 years	42	60.0	28	40.0
6+ years	6	60.0	4	40.0	1.00
Dialysis dose at baseline(minutes dialyze per week)
<120	2	50.0	2	50.0
120 - 239	5	62.5	3	37.5
240 - 359	30	65.2	16	34.8
360+	6	37.5	10	62.5	0.28
Cause for chronic kidney disease
Glomerulonephritis	13	72.2	5	27.8
Hypertensive vascular	6	54.5	5	45.5
Polycystic	5	71.4	2	28.6
Diabetics	14	51.9	13	48.1	0.66
Other	10	58.8	7	41.2
Kidney Disease Quality Of Life score (KDQOL)
70+	26	61.9	16	38.1
<70	18	52.9	16	47.1	0.49

Note. P-value is from chi-square test or Fisher’s exact test as appropriate.

Age, sex, BMI, and ethnicity were as at baseline.

Multivariate analysis of characteristics at 24 months associated with worse health at 36 months follow-up period: Of the 80 participants included for analysis at 36 months, 17 had missing data on variables retained in the final model. Hence, data from 63 participants was analyzed at 36 months for multivariable analysis. Table 4 presents the relative risk of “Worse health/died” outcomes at 36 months follow-up period based on data from 63 participants who had non-missing data on all variables retained in the final model. Age and ethnicity were not in the mini model stage. Among social characteristics participants reporting no satisfaction (RR = 1.20; 95% CI = 0.99-1.46), without a strong sense of community (RR = 1.41; 95% CI = 1.18-1.69) and without large family involvement in management of chronic kidney disease (RR = 1.68; 95% CI = 0.90-3.12) experienced “Worse health/died.” Among health characteristics, patients with three or more comorbidities were in the “Worse health/died” group (RR = 1.21; 95% CI = 1.13-1.29). Those who reported overall fair or poor health at 24 months reported “Worse health/died” outcomes at 36 months (RR = 1.47; 95% CI = 1.41-1.54).

Table 4.

Showing Multivariable Analysis at 36 Months Relative to Characteristics at 24 Months: Final Models.

Characteristics (measured 12 months prior to measuring the outcome)	36 months (N = 63)
	RR	95% CI for RR		P-value
Demographics
Sex**
Male	Ref
Female	0.42	0.30	0.59	<.01
Social factors
Social relationships
Completely/mostly satisfied	Ref
Others	1.20	0.99	1.46	.06
Sense of community
Some/little	1.41	1.18	1.69	<.01
Strong	Ref
Family involvement
Small/very small	1.68	0.90	3.12	.10
Large/very large	Ref
General health factors
Comorbidity types
0-2	Ref
3+	1.21	1.13	1.29	<.01
Overall general health
Fair/poor	1.47	1.41	1.54	<.01
Excellent/very good/good	Ref
EQ5D selfcare
No problems	Ref
Some problems	0.74	0.66	0.83	<.01
Chronic kidney disease factors
Type of dialysis
Hemodialysis	Ref
PD	0.82	0.29	2.38	.72
Not dialyzing now	Drop due to everyone had “same/better” outcome
Vintage (years)#	1.06	1.03	1.08	<.01
Dialysis dose (minutes per week)#	1.0005	1.0003	1.0007	<.01

Note. Outcome is the health status becoming “worse” or death occurring within past 12 months. RR = relative risks.

# refers to dialysis vintage and dialysis dose at baseline entry into the study.

Among dialysis characteristics, those on PD were at reduced risk of “Worse health/died” outcomes compared to HD patients (RR = 0.82, 95% CI = 0.29-2.38). Longer dialysis vintage was associated with reporting worse health outcome (RR = 1.06; 95% CI = 1.03-1.08) Cause of chronic kidney disease and KDQOL symptom score were not retained in the final model.

Discussion

According to ANZDATA registry figures, 54% of NZ dialysis patients are over the age of 65 years. The incidence of renal replacement therapy in patients above the age of 65 years is on the rise with the highest incidence of dialysis in the 65- to 74-year age group. The median survival for patients above the age of 65 years on dialysis is 2 to 3.7 years, with a 5-year survival of 38% in patients 65 to 75 years and 20% in patients above 75 years of age. This study provides insights into the older patients’ self-reported health after 24 and 36 months of dialysis therapy. A large proportion of patients in our study reported the same or better health outcome at both time points.

In our study, Māori and Pacific participants experienced better health at both 24 months and 36 months compared to New Zealand Europeans (although it was only statistically significant at 24 months, probably due to the smaller numbers at 36 months). Of interest, there was a strong association between same or better health with a sense of community and family support. Among the Indigenous (ie, Māori) and Pacific participants, the cultural importance of family and elders is very strong and is an important component of general well-being. Future studies with larger numbers of Māori and Pacific participants are required to further understand these relationships.

Kimmel et al studied the impact of social factors like living arrangements and household income on mortality in dialysis patients. The authors concluded that residence in highly segregated areas and lower median household income are associated with higher mortality in dialysis patients. Increased perception of social support is associated with improved survival in hemodialysis patients. In our study, over the 36 months of follow-up, patients who were not satisfied with their social relationships reported worse health outcomes. A similar finding was noted for patients who reported a strong sense of community. These participants tend to report less worse outcomes at 12 months, and 24 months. These results reflect the importance of social connectedness in older dialyzing individuals and its impact on self-reported health outcomes.

Previous studies show that the benefit of dialysis over conservative management is lost in patients more than 75 years of age with high comorbidities and poor performance status.[25 -27] In our study, participants with 3 or more comorbidities reported worse health outcomes, however, it was only statistically significant at 36 months. Of those who reported fair or poor general health at 24 months, 74% were at risk of reporting poor health outcomes at 36 months. Thus, the assessment of patient’s overall health and burden of comorbidities is important in decision-making regarding the benefit of dialysis over the longer term in this population.

In New Zealand since 2013, approximately 40% to 43% of dialysis patients aged above 65 years are on peritoneal dialysis. Survival on peritoneal dialysis in this age group in NZ is 92% at 1 year, 67% at 3 years and 42% at 5 years. When we look at hemodialysis patients above 65 years of age, survival is 90% at 1 year, 65% at 3 years, and 46% at 5 years. These data suggest that there is no survival difference in patients on hemodialysis and peritoneal dialysis. The North Thames Dialysis Study, conducted in patients more than 70 years of age looking at quality of life, indicated that survival, hospitalization, and quality of life were similar between patients on hemodialysis and peritoneal dialysis.[28-30] The BOLDE study (Broadening Options for Long term Dialysis in the Elderly) found the older patients on peritoneal dialysis have significantly less perception of intrusion of disease in their lives. In our study, patients on peritoneal dialysis tended to experience less “Worse health” at 12 months, and 36 months as compared to hemodialysis patients. Our results are in keeping with these previously reported studies where patients on peritoneal dialysis reported better health outcomes than patients on haemodialysis.[28-30] Longitudinal analysis specifically to examine the changes in health states over time by dialysis modality has been planned.

There is an increase in the relative risk of death with increasing dialysis vintage both in peritoneal dialysis and hemodialysis patients.[32,33] In this study, longer dialysis vintage was associated with a higher risk of reporting worse health outcomes at both 24 months and 36 months analysis. The higher mortality in patients with longer dialysis vintage may be due to increased incidence of major cardiovascular events and infections in this group.[34,35]

One of the strengths of this study is its prospective design and focus on the patient’s perspectives and patient-reported outcomes. Engaging patients in decision making about initiating or continuing dialysis therapy is essential. Our study highlights the importance of factors like sense of community, satisfaction with social relationships, and functional problems such as issues with mobility, self-care, and intellectual ability and the impact of these factors on patient-reported health outcomes. Clinicians and health care providers should focus on interventions at improving social satisfaction, sense of community, mobility, which may improve experiences in older patients on dialysis. It also highlights the importance of patient-centered data collection.

There are a number of limitations from this study. Although DOS65+ study has a larger cohort (n = 225), participants who had died in the preceding 12-month period were removed from subsequent analysis, reducing the patient population included in the analysis at 24 months to 120 and 36 months to 80 in total. Smaller patient population limits the number of variables that were able to be included in the multivariable model, which likely means we are underpowered to detect associations, which do, in truth, exist. Another significant limitation is the basement or floor effect.[36,37] The question asked the participants, that was used to derive our response variable, was whether they felt their health is the “same or better” or “worse” compared to 1 year ago. For example, if the participant reported feeling worse at 24 months and continued to feel worse at 36 months, they would report being feeling the same. Although some people can actually improve from “worse” at 24 months to “same/better” at 36 months, such reporting could be at least partly due to floor effect (ie; of the people at “worse health” state at 24 months, if some continue to remain at that state for next 12 months, then they will fall into the “same/better” group at 36 months even though they have not actually improved). This floor effect was observed to be larger among those who reported EQ5D self-care problems than those who not reported such problems at 24 months (50% vs 36%). This basement effect can produce on-the-face counter-intuitive results (Table 4).

In the multivariable analysis model, there is also a possibility of having inflated type 1 and/or type 2 errors due to 2-step model building with 2 backward selection procedures. To mitigate this, we purposely selected large P-value thresholds at all levels of model building; however, we do not claim our mitigation measures completely eliminated the possible inflation of type1 and type2 errors. Also, in the multivariable analyses, the list of variables retained at the 2-time points is not the same, and a possible reason for this is that patients that could be used in these models were not the same even though data were collected longitudinally over the years from the same cohort. As a result, we cannot interpret the results as trends over time but have to interpret the models separately at each time point. Longitudinal studies with more focused research questions are planned so that trends over time can be analyzed in more detail. However, multiple imputation or any other way of handling missing data to retain everyone in a longitudinal analysis is likely to amplify the imperfections of imputation procedures given the large proportion of (>30%) missing data. Also, because some people were not included in the 36 months multivariable model due to item missingness, we assessed possible bias due to this by comparing those included and not included in the final model. This found, the vintage is larger (P = .01) among those included in the model than others, while no difference was observed in any of the other variables. Therefore we may have overestimated the influence of vintage.

Conclusion

In summary, patient-reported outcome measures are critical in informed shared decision making regarding the impact and on-going maintenance of dialysis, as previously highlighted by the SONG (Standardized Outcomes In Nephrology) initiative. Our study highlights the importance of consideration of factors like burden of comorbidities, social characteristics like satisfaction with relationships, and sense of community amongst other factors in predicting health outcomes.