Sage Journals: Discover world-class research

Abstract

French

Background:

An increased number of end-stage renal disease patients suffer psychosocial conditions and may experience delayed access to transplantation due to listing restrictions. However, it remains to be shown whether preexisting psychosocial conditions confer an independent risk factor of poor posttransplant outcomes.

Objective:

We addressed this gap in knowledge by conducting a retrospective cohort study to investigate an independent association between the risk of death after transplant and having a diagnosis of psychosocial conditions 1 year prior to starting dialysis.

Methods:

All cases of adult deceased-donor kidney transplantation performed in Ontario, Canada, between April 1, 2002, and March 31, 2013, were used. Propensity score matching was applied to adjust for potential endogenous bias of using predialysis psychosocial status to predict posttransplant mortality. Survival analysis techniques, including Kaplan-Meier curves and Cox proportional hazards modeling, were also used.

Results:

Our results indicate a 49.4% (hazard ratio [HR] = 1.494 [95% confidence interval (CI) = 1.168-1.913]) increased relative risk of posttransplant death to be associated with predialysis psychosocial conditions, when other factors are held constant. The effect is significant (P = .001) and is independent of other known predictors of death including advanced age.

Conclusions:

Findings from this study offered strong support for the development of psychosocial evaluation to screen candidates prior to transplant listing and early interventions for transplant candidates with psychosocial concerns.

Keywords

psychosocial conditions kidney transplantation transplant outcome

What was known before

Psychosocial conditions such as depression are common among end-stage renal disease patients and may adversely affect kidney transplant outcomes.

What this adds

Early diagnosis of psychosocial conditions at 1 year prior to dialysis initiation is an independent risk factor of all-cause death after deceased-donor kidney transplantation. Having predialysis psychosocial conditions is associated with a 49.4% increased relative risk of death after transplant.

Introduction

End-stage renal disease (ESRD) is associated with high symptom burdens that severely affect patients’ functional status and quality of life. Apart from physical conditions such as fatigue and dyspnea, psychosocial conditions (PCs) are highly prevalent among ESRD patients.¹ Depression is the most common PC that coexists with ESRD, affecting 20% to 57% of the population.^2-5 Other common PCs among ESRD patients include anxiety, personality disorders, psychotic disorders (eg, schizophrenia), and other mental health issues (eg, substance abuse).⁶

Transplantation is the most sought-after ESRD treatment that extends life expectancy^7,8 and improves life quality.^7,9 However, outcomes of transplant are sensitive to a range of factors, some of which are closely related to PCs. Correlates of depression among ESRD patients (eg, multiple comorbidities)^10,11 are usually prognostic of death after transplant.¹² Furthermore, having PCs may be associated with a higher likelihood of nonadherence after transplant,^13-18 which is a major risk factor of acute rejection,^19,20 graft failure,^19,21 and death.²² In Ontario, Canada, having a history of PCs (such as depression and schizophrenia) predisposed patients to a 17% decreased odds of completing transplant workup within 2 years of referral and a 12% decreased odds of transplantation.⁶ Hence, for patients with PCs who are transplanted, they are likely to have been exposed to an extended wait time on dialysis, which is a risk factor for poor survival after transplantation.^23-25

Studies seldom looked at transplant recipients with an early diagnosis of PCs as most of them focused on recipients who presented with PCs after transplantation.^26-28 As early PCs that were diagnosed before dialysis initiation may not translate into active PCs that persist after transplantation,²⁹ it remains unclear from the current literature whether early psychosocial symptoms constitute an independent risk factor of poor patient survival after transplant. A single-center study in Toronto, Canada,¹⁹ was unable to identify an effect of having a history of PCs on the rate of acute rejection, graft failure, or death-censored graft failure. However, as the literature has demonstrated, a history of PCs is likely to influence posttransplant outcomes via multiple pathways (eg, by extending pretransplant dialysis duration), a mechanism that introduces endogenous bias in the estimation of a true treatment effect.

In this study, we examined a 2-fold study objective, including (1) to test the hypothesis that kidney transplant recipients with a diagnosis of PCs at 1 year prior to initiating dialysis experienced a longer pretransplant dialysis time than their PC-free counterparts, and (2) to assess the relationship between the risk of death after kidney transplantation and a diagnosis of PCs at 1 year prior to initiating dialysis. We used a cohort (n = 4582) of deceased-donor kidney transplant recipients from Ontario, Canada, in the analysis as living-donor kidney transplants are often associated with different patient cohorts and allocation routines.³⁰ Propensity score matching method was applied to derive an unbiased relative risk of death due to predialysis PCs.

Materials and Methods

Study Design and Patient Cohort

We conducted a retrospective, population-based, cohort study of adult recipients of first-time, kidney-only transplantation from a deceased donor in Ontario, Canada, between April 1, 2002, and March 31, 2013 (n = 5922). This time period was chosen to align with Ontario’s adaptation of the International Statistical Classification of Disease and Related Health Problems, Tenth version, Canada (ie, ICD 10-CA) in 2002.³¹ Patients were followed up until death or till April 1, 2014. Patients who had received a preemptive transplant, living-donor transplant, multi-organ transplant, re-transplant, or had age <18 years at the time of transplantation were excluded.

Data Source

This study used a population-based data set derived by the Institute for Clinical Evaluative Sciences (ICES) that links various administrative databases to patient-level data from the Canadian Organ Replacement Register (CORR) by a validated unique patient identifier. Dialysis centers, transplant centers, and organ procurement organizations in Canada are obligated to report all cases of end-stage organ failure and the respective treatments to the CORR.³² Hence, the CORR contains complete follow-up data for ESRD patients from the first renal replacement therapy to death.³²

Tracking Comorbidity

The Johns Hopkins Adjusted Clinical Group (ACG) system,³³ a well-validated method of categorizing comorbidities and predicting mortality,³⁴ was used to classify patient’s comorbidity at 1 year prior to starting dialysis. This was computed using administrative records, including ICD 10-CA diagnosis codes from the Discharge Abstract Database (DAD), a database that includes acute care inpatient hospitalizations; physician billing codes from the Ontario Health Insurance Plan (OHIP); and records from the National Ambulatory Care Reporting System (NACRS). The ICD 10-CA codes were assigned to 1 of the 32 Aggregated Diagnosis Groups (ADGs) based on 5 clinical dimensions, including the duration of the condition, severity, diagnostic certainty, etiology, and involvement of special care.³³ ADGs were further collapsed into 12 Collapsed ADGs (CADGs) based on the likelihood that the condition would persist or recur, severity, and the type of health care services required.³³ Supplementary Table S1 summarizes each of the 12 CADGs and the ADGs as well as common ICD 10-CA codes that they comprise. In our study, each of the 12 CADGs was denoted using an indicator variable.

Outcome and Exposure Variables

All-cause death after transplant is the outcome of our study. Each patient’s date of death was obtained from Ontario’s Registered Person Database (RPDB), a population-based registry. The main exposure of interest is represented by a binary variable that indicates a diagnosis of PCs at 1 year before dialysis initiation, CADG 10.³³

Covariates

Additional patient-level covariates considered in our analysis included demographic characteristics (from the RPDB): sex (female or male), race (Caucasian, Asian, Black, Indian subcontinental, other races, or unknown race), and location of residence (membership of 1 of the 14 Local Health Integration Networks or LHINs)³⁵; transplant-related information (from CORR records): age at transplantation (age group of 21-30, 31-40, 41-50, 51-60, 61-70, or 71+), primary kidney diagnosis (glomerulonephritis, diabetes, renal vascular, congenital or hereditary, other diagnoses, or unknown diagnosis), peak class I panel reactive antibody or PRA (<20%, 20%-79%, ≥80%, or unknown), blood type (O, A, B, AB, or unknown), year of transplantation (2002-2013), transplant center (membership of 1 of the 6 transplant centers in Ontario)³⁶; comorbidity status: 11 CADGs. We calculated pretransplant dialysis duration as the interval in days between the dialysis initiation date and the transplantation date (both dates were obtained from the CORR).

Statistical Analysis

We denoted patients with CADG 10 = 1 to be in the PC group while those with CADG 10 = 0 to be in the PC-free group. Baseline characteristics of patients were summarized by group. A χ² test was used to assess the distribution of categorical variables in the 2 groups and a 2-sample t test was used for the mean of continuous variables.

Propensity Score Matching

An independent association between the risk of posttransplant death and the status of predialysis PCs cannot be estimated directly using a crude regression due to potential endogenous bias.³⁷ Propensity score methods aim to create a balanced distribution of observed covariates between the exposed and the unexposed subjects before estimating an unbiased causal effect. To do so, a propensity score or the probability of being exposed conditional on observed covariates is estimated.^38-41 Three primary techniques are then applied to calculate the effect, including matching, covariate-adjustment, and stratification (or subclassification).^40,41 Although there is no agreement on the most appropriate technique to use,⁴¹ matching is commonly applied when the number of unexposed subjects far exceeds the number of exposed subjects.⁴⁰ In our study, the ratio between patients in the PC and PC-free group is about 1:4, which is significant enough to necessitate a matching analysis.

We followed an analytical process demonstrated by Austin⁴² and Austin and Mamdani.⁴¹ First, propensity score was estimated by a logistic regression that modeled the probability of having PCs given measured confounders that were hypothesized to influence both the likelihood of having PCs and the risk of posttransplant death.⁴³ These confounders included patient sex, age, race, transplant year, primary kidney diagnosis, comorbidity, and pretransplant dialysis duration. We also included variables that were prognostically important correlates of posttransplant death but not of predialysis PCs because controlling for them would potentially reduce bias.^43-45 These covariates included patient’s location, transplant center, and peak PRA. We followed a structured, iterative algorithm³⁹ to determine a specification of the logistic model that produced balanced propensity scores in the PC/PC-free groups. Details of the construction of a propensity score and balance diagnostics were presented in the Supplementary Appendix. We applied greedy 1:1 matching without replacement that used a caliper distance that equaled 20% of the standard deviation of the logit of propensity score to form matched pairs.^46,47 Balance diagnostics were performed where McNemar tests (for categorical variables) or paired t tests (for continuous variables) were used to assess the distribution of characteristics in the 2 groups after matching.⁴⁷ Standardized differences were calculated to ensure no covariates exceeded the 10% threshold.⁴¹

Absolute and Relative Risk of Death

Kaplan-Meier curve was used to estimate patient survival in the matched sample. To avoid bias due to the matched nature of our sample, we used stratified log-rank test that was based on the 6 patient age groups rather than the conventional log-rank test to determine whether PC/PC-free patients had different survival functions.^42,48 The absolute increase in the risk of death due to PCs at different time points after transplant was computed.⁴² A univariate Cox proportional hazard model was applied in the matched sample to regress posttransplant survival on PC status. This estimated a hazard ratio (HR) of having PCs on death or the relative risk of death due to PCs.⁴² Robust estimation of the standard error of the HR was applied to reduce the bias from the matched sample.⁴⁹ A Cox proportional hazards analysis using the original, unmatched sample was conducted to compare the results derived from the matched analysis.

Sensitivity Analysis

Imputation of pretransplant dialysis duration (n = 288 missing) was based on patient’s blood type, a variable that has been demonstrated⁵⁰ to strongly predict pretransplant dialysis time. Specifically, the mean duration for patients with nonmissing value in the same blood type group was used. Analyses were conducted using the command psmatch2⁵¹ in Stata/MP 13.1.

Results

Cohort Description

Among 5922 deceased-donor kidney transplant recipients (Figure 1) in Ontario, Canada, 225 patients had received the transplant preemptively; 540, 63, and 7 patients had received a second, third, or fourth transplant, respectively; 55 and 159 patients had age between 1 and 10 and 11 and 20 years at transplantation, respectively; 2 and 288 patients had unknown sex or unknown pretransplant dialysis duration, respectively; and 1 patient was lost to follow-up. Exclusion of these patients resulted in 4582 patients in the final cohort. Mean follow-up time after transplant was 2588.6 days or 7.1 years (SD = 1273.1 days).

Figure 1.

Inclusion and exclusion of transplant recipients in the study cohort.

Baseline Characteristics

Table 1 summarizes baseline characteristics of patients based on PC status. At 1 year prior to initiating dialysis, 20.8% (n = 954) of patients had a diagnosis of PCs. About 16.1% of them (n = 154) died after transplantation within our observational period while 14.2% of PC-free patients (n = 517) died. This difference was not statistically significant (P = .14). Patients with PCs tended to be female (P = .06), had younger age at transplantation (P = .005), and had shorter pretransplant dialysis duration (P = .004). Notably, patients with PCs spent an averaged 1326.6 days (SD = 983.7 days) on pretransplant dialysis, which is 107 days less compared with PC-free patients (mean = 1434.1 days; SD = 1046.9 days) and 85 days less than the mean dialysis time of the entire cohort (mean = 1411.7 days; SD = 1034.9 days).

Table 1.

Comparing Baseline Characteristics of Transplant Recipients Based on Psychosocial Diagnosis (CADG 10) at 1 Year Prior to Initiating Dialysis, n = 4582.

Characteristics	PC group(CADG 10 = 1)	PC-free group(CADG 10 = 0)	Total	P value
No. of transplant recipients, n (%)	954 (20.8)	3628 (79.2)	4582	—
Cases of death after transplant, n (%)	154 (16.1)	517 (14.2)	671 (14.6)	.14
Follow-up after transplant, d	2524.9 ± 1266.9	2604.4 ± 1274.4	2588.6 ± 1273.1	.08
Pretransplant dialysis duration, d	1326.6 ± 983.7	1434.1 ± 1046.9	1411.7 ± 1034.9	.004
Female sex, n (%)	360 (37.7)	1250 (34.5)	1610 (35.2)	.06
Age at transplantation, year, n (%)
21-30	71 (7.4)	246 (6.8)	317 (6.9)
31-40	152 (15.9)	531 (14.6)	683 (14.9)	.005
41-50	245 (25.7)	834 (23)	1079 (23.5)
51-60	267 (28)	960 (26.5)	1227 (26.8)
61-70	184 (19.3)	839 (23.1)	1023 (22.3)
71 +	35 (3.7)	218 (6)	253 (5.5)
Blood type, n (%)
O	372 (39)	1542 (42.5)	1914 (41.8)
A	377 (39.5)	1331 (36.7)	1708 (37.3)	.12
B	136 (14.3)	469 (12.9)	605 (13.2)
AB	36 (3.8)	177 (4.9)	213 (4.6)
Unknown	33 (3.5)	109 (3)	142 (3.1)
Race, n (%)
Caucasian	672 (70.4)	2447 (67.4)	3119 (68.1)
Asian	53 (5.6)	255 (7)	308 (6.7)
Black	83 (8.7)	270 (7.4)	353 (7.7)	.11
Indian subcontinental	54 (5.7)	228 (6.3)	282 (6.2)
Other	56 (5.9)	275 (7.6)	331 (7.2)
Unknown	36 (3.8)	153 (4.2)	189 (4.1)
Transplantation year, n (%)
2002	44 (4.6)	251 (6.9)	295 (6.4)
2003	68 (7.1)	233 (6.4)	301 (6.6)
2004	62 (6.5)	223 (6.1)	285 (6.2)
2005	64 (6.7)	254 (7)	318 (6.9)
2006	84 (8.8)	272 (7.5)	356 (7.8)
2007	88 (9.2)	314 (8.7)	402 (8.8)	.53
2008	79 (8.8)	313 (8.6)	392 (8.6)
2009	94 (9.9)	374 (10.3)	468 (10.2)
2010	84 (8.8)	332 (9.1)	416 (9.1)
2011	97 (10.2)	337 (9.3)	434 (9.5)
2012	102 (10.7)	389 (10.7)	491 (10.7)
2013	88 (9.2)	336 (9.3)	424 (9.3)
Primary kidney diagnosis, n (%)
Glomerulonephritis	206 (21.6)	798 (22)	1004 (21.9)
Diabetes	210 (22)	743 (20.5)	953 (20.8)	.2
Renal vascular	57 (6)	268 (7.4)	325 (7.1)
Congenital/hereditary	107 (11.2)	484 (13.3)	591 (12.9)
Other	157 (16.5)	584 (16.1)	741 (16.2)
Unknown	217 (22.7)	751 (20.7)	968 (21.1)
Peak class I PRA, n (%)
<20%	637 (66.8)	2395 (66)	3032 (66.2)
20%-79%	124 (13)	515 (14.2)	639 (13.9)	.8
≥80%	44 (4.6)	160 (4.4)	204 (4.5)
Unknown	149 (15.6)	558 (15.4)	707 (15.4)
Other comorbid conditions, n (%)
CADG 1: Acute minor	728 (76.3)	2276 (62.7)	3004 (65.5)	<.001
CADG 2: Acute major	899 (94.2)	3236 (89.2)	4135 (90.2)	<.001
CADG 3: Likely to recur	606 (63.5)	1945 (53.6)	2551 (55.7)	<.001
CADG 4: Asthma	60 (6.3)	151 (4.2)	211 (4.6)	.005
CADG 5: Chronic medical, unstable	944 (99)	3563 (98.2)	4507 (98.3)	.1
CADG 6: Chronic medical, stable	647 (67.8)	2090 (57.6)	2737 (59.7)	<.001
CADG 7: Chronic specialty, stable	30 (3.1)	110 (3)	140 (3.1)	.86
CADG 8: Eye/dental	102 (10.7)	395 (10.9)	497 (10.8)	.86
CADG 9: Chronic specialty, unstable	106 (11.1)	404 (11.1)	510 (11.1)	.98
CADG 11: Preventive/administrative	455 (47.7)	1549 (42.7)	2004 (43.7)	.006
CADG 12: Pregnancy	13 (1.4)	11 (0.3)	24 (0.5)	<.001
Geographic location (censored), n (%)
LHIN A	128 (13.4)	459 (12.6)	587 (12.8)
LHIN B	99 (10.4)	394 (10.9)	493 (10.8)
LHIN C	43 (4.5)	187 (5.2)	230 (5)
LHIN D	27 (2.8)	125 (3.4)	152 (3.3)
LHIN E	36 (3.8)	132 (3.6)	168 (3.7)
LHIN F	8 (0.8)	37 (1)	45 (1)	.75
LHIN G	69 (7.2)	286 (7.9)	355 (7.7)
LHIN H	53 (5.6)	188 (5.2)	241 (5.3)
LHIN I	123 (12.9)	394 (10.9)	517 (11.3)
LHIN J	81 (8.5)	255 (7)	336 (7.3)
LHIN K	51 (5.3)	223 (6.1)	274 (6)
LHIN L	59 (6.25)	230 (6.3)	289 (6.3)
LHIN M	66 (6.9)	288 (7.9)	354 (7.7)
LHIN N	108 (11.3)	413 (11.4)	521 (11.4)
Unknown location	3 (0.3)	17 (0.5)	20 (0.4)
Transplant center (censored), n (%)
Center A	292 (30.6)	993 (27.4)	1285 (28)
Center B	17 (1.8)	59 (1.6)	76 (1.7)	.02
Center C	186 (19.5)	777 (21.4)	963 (21)
Center D	135 (14.2)	574 (15.8)	709 (15.5)
Center E	120 (12.6)	510 (14.1)	630 (13.7)
Center F	98 (10.3)	418 (11.5)	517 (11.3)
Unknown	106 (11.1)	297 (8.2)	403 (8.8)

Note. Categorical variables are summarized as counts and percentages in the respective PC or PC-free group, whereas continuous variables are reported as mean ± standard deviation. A χ² test is used to compare the distribution of categorical variables between the 2 groups of patients. The mean of continuous variables is assessed using a 2-sample t test. CADG = Collapsed Aggregated Disease Group; PC = psychosocial conditions; PRA = panel reactive antibody; LHIN = Local Health Integration Network.

Difference of other comorbid conditions was observed between patients with and without PCs. Compared with PC-free patients, patients with PCs were more likely to also have conditions that were acute minor (CADG 1, P < .001), acute major (CADG 2, P < .001), likely to recur (CADG 3, P < .001), asthma (CADG 4, P = .005), chronic medical/stable (CADG 6, P < .001), preventive/administrative (CADG 11, P = .006), and pregnancy (CADG 12, P < .001). The distribution of patients with PCs also differ among the 6 transplant centers in Ontario (P = .02).

Risk of Death Due to Predialysis PCs

Results of propensity score matching is detailed in the Supplementary Appendix. A 1:1 caliper matching yielded 945 pairs of PC/PC-free patients where 99.1% of PC patients (945 of 954) in the original sample were matched to a PC-free patient. About 0.9% (n = 9) of PC patients could not be matched due to their high propensity scores and were thus discarded from the analysis. Balance diagnostics in the matched sample were summarized in the Supplementary Appendix.

Figure 2 shows Kaplan-Meier curves that depicted patient survival after transplantation. Before matching, log-rank test suggested a trend toward a different survival (P = .09) between PC/PC-free patients. Graphically, the 2 curves crossed as time increased (top half of Figure 2). After matching, we observed different survival patterns for patients in the 2 groups as those with PCs had a lower survival probability at all time points after transplantation and the absolute difference increased with time (bottom half of Figure 2). Stratified log-rank test based on the age groups confirmed the difference in survival between the PC/PC-free patients (P = .003, .006, .007, .02, .02, and .04 for age groups of 21-30, 31-40, 41-50, 51-60, 61-70, and 71+ years, respectively).

Figure 2.

Kaplan-Meier curves of posttransplant patient survival before and after matching.

Table 2 presents the absolute risk of death due to predialyiss PCs at given time points after transplantation. While PC patients had a lower survival rate at all time points than PC-free patients, the absolute difference tended to grow with posttransplantation time, from 0.8% at 674 days (about 1.8 years) to 6.5% at 5385 days (about 14.7 years) after transplantation.

Table 2.

Kaplan-Meier Estimates of the Absolute Risk of Death Due to Predialysis PCs at Different Time Points After Transplantation Using the Matched Sample.

Days after transplantation	Survival probability (%), PC group	Survival probability (%), PC-free group	Absolute risk of death (%)
1	99.8	99.9	0.1
674	95.6	96.4	0.8
1347	91.2	94.3	3.2
2020	87.4	92.3	5.0
2693	83.2	88.7	5.5
3366	80.1	85.4	5.3
4039	78.2	83.7	4.8
4712	76.5	83.0	6.5
5385	76.6	83.0	6.5

Note. PC = psychosocial condition.

Table 3 presents the results of a univariate Cox proportional hazard model that assessed the relative risk of death due to PCs in the matched sample. Compared with PC-free patients, those with PCs were associated with a significant (P =.001) 49.4% (HR = 1.494 [95% robust confidence interval (CI) = 1.168-1.913]) increased relative risk of death after transplant. We also estimated the HR of PCs using the unmatched sample. All of the 5 Cox models that adjusted for different covariates suggested a significant (P = .005-.09) relative risk of death due to PCs, but the HRs were smaller than the one derived from the matched analysis, ranging from 1.164 (unadjusted) to 1.285 (comorbidity-adjusted) to 1.286 (demographic-adjusted) to 1.287 (adjusted for all covariates), and to 1.296 (adjusted for covariates with P < .05 in the log-rank test).

Table 3.

Relative Risk of Posttransplant Death Due to Predialysis Psychosocial Conditions Using Different Methods.

Methods	HR	95% CI, robust	P-value
Propensity score matching
Caliper matching, within 0.016 SD of the propensity score	1.494	1.168-1.913	.001
Unmatched Cox regression analysis
Univariate Cox regression (unadjusted)	1.164	0.972-1.395	.09
Crude Cox regression adjusted for all covariates	1.287	1.069-1.549	.008
Demographic-adjusted Cox regression^a	1.286	1.073-1.541	.006
Comorbidity-adjusted Cox regression^b	1.285	1.069-1.544	.008
Cox regression adjusted for significant covariates screened by a log-rank test^c	1.296	1.080-1.555	.005

Note. HR = hazard ratio; CI = confidence interval; CADG = Collapsed Aggregated Disease Group.

Adjusted for sex, age, race, and location.

In addition to all demographic variables, 11 binary variables of CADGs were adjusted.

This include sex, age, race, location, primary kidney diagnosis, transplant year, transplant center, CADG 6, and CADG 8.

Sensitivity Analysis

Of the 288 patients with unknown pretransplant dialysis duration, 118, 123, 35, and 12 patients had blood type O, A, B, and AB, respectively. For the other (n = 4582) patients, the mean pretransplant dialysis duration was 1515.4, 1278.9, 1629.1, and 1105.3 days for blood group O, A, B, and AB, respectively. After imputation, we re-run our original specification of the propensity score model and found no imbalance between the PC/PC-free group. Standard deviation of the logit of propensity score became 0.082 which suggested an optimal caliper distance of 0.016 to be used in the matching. This resulted in 947 pairs of PC/PC-free patients (99.3%) where 7 PC patients (0.7%) were excluded. Univariate Cox proportional hazard model estimated a 39.1% (HR = 1.391 [95% robust CI = 1.123-1.730]) significant (P = .03) relative increase of death due to PCs using the imputed matched sample. Hence, imputation of pretransplant dialysis duration using the blood group mean drove down the HR of PCs on death, which means our estimation is moderately sensitive to changes in patient’s length of dialysis before transplantation.

Discussion

To the best of our knowledge, this is the first study that used propensity score matching to estimate an unbiased effect of predialysis PCs on posttransplant death among deceased-donor kidney transplant recipients. We followed a rigorous analytical procedure and reported a 49.4% increased relative risk of death to be associated with a diagnosis of PCs at 1 year prior to dialysis initiation. This effect is independent of demographic characteristics, comorbidities, duration of pretransplant dialysis, transplant centers, and other clinical indicators of ESRD. In Ontario, Canada, the median wait time on dialysis before receiving a deceased-donor kidney transplant is 4.1 years (or 1507 days).⁵² Hence, our results suggested that being diagnosed with PCs at as early as 5 years prior to the expected date of deceased-donor kidney transplantation constituted an important marker of poor survival after transplant.

Contrary to our hypothesis, ESRD patients with predialysis PCs in our cohort experienced a shorter dialysis episode prior to transplantation than their PC-free counterparts. The difference was clinically meaningful due to an absolute value of standardized difference that exceeded 10% (15%). It is unknown to us why predialysis PCs appeared to give transplant candidates faster access to transplantation, but even with the protective effect of a shorter pretransplant dialysis duration, predialysis PCs still conferred a negative effect on posttransplant survival, which in turn made our conclusion more compelling.

We also conducted multiple survival analyses using the unmatched sample of patients, which returned lower estimates of HR of predialysis PCs on posttransplant death. Hence, we conclude that without properly adjusting for confounding, conventional methods are likely to underestimate the true causal effect of early PCs on patient survival after transplant.

Our conclusion was not in congruence with most studies that assessed the effect of early PCs on posttransplant death. A evaluation¹⁹ conducted at the Toronto General Hospital in Canada using 955 adult kidney transplant recipients reported no additional risk of acute rejection, graft loss, or death-censored graft failure among those with a history of mental health concerns. However, the study cohort was not exclusively nonpreemptive, deceased-donor transplant recipients. Rather, 56% and 12% of patients had received a living-donor transplant or a preemptive transplant. Hence, the worsened posttransplant outcome among recipients with pretransplant mental health issues may become less obvious when these patients also carried significant survival advantage related to receipt of a living-donor kidney and avoidance of dialysis.

A single-center study⁵³ that used the Stanford Integrated Psychosocial Assessment for Transplantation (SIPAT) to screen for problematic transplant candidates found no association between the rate of mortality at 1 year posttransplantation and a high SIPAT score (indicating severe PCs) before transplantation among 217 transplant recipients. They did, however, suggest higher rate of infections among patients with more severe pretransplant PCs, an event that is predictive of death.²³ Hence, we argue that although difference of mortality was not directly observed by the study possibly due to the short follow-up time and small sample of patients, their conclusions partially corroborated with ours.

Evans et al⁵⁴ analyzed 960 US veterans with mental health illnesses diagnosed at 1 year prior to a solid organ, including kidney (n = 396), or bone marrow transplantation. Three-year mortality was equal among patients with different pretransplant mental health status. However, they pointed out that one would expect a more divergent survival patterns after the fifth year where those with pretransplant mental health illnesses began to deteriorate much faster than their mentally healthy counterparts. In our analysis, we showed that the absolute risk of death due to PCs reached 5% after 2020 days (or 5.5 years) posttransplantation and continued to grow to 6.5% after 4712 days (12.9 years). Before 5.5 years, the absolute risk was less prominent (0.8% and 3.2% after 674 and 1347 days, respectively), which may in part explain why Evans et al were unable to conclude any effect. However, according to our results, the relative risk of posttransplant death due to PCs stayed constant throughout the trajectory of posttransplant follow-up at 49.4%.

There are possible explanations to our findings. Early PCs that were diagnosed before dialysis initiation may or may not translate into active PCs that persist after transplantation.²⁹ If they do, there is strong evidence from the literature that directly associates posttransplant PCs to elevated rates of nonadherence and death.^26-28 If these symptoms do not persist after transplant, possibly due to being cured or controlled by psychiatric treatments prior to transplant listing, we suspect that some permanent neurocognitive and physical damage may have been occurred that predispose these patients to an irreversibly disadvantaged state at the time of transplantation. For example, chronic and heavy cannabis use has been shown to lead to permanent loss of memory and brain function impairment that are irreversible by active detoxification and abstinence.⁵⁵ Hence, past cannabis abusers in receipt of a transplant may face a high risk of not taking their medications due to increased forgetfulness, which subsequently heightens their risk of graft loss and death.¹³

The high risk of posttransplant death due to predialysis PCs concluded in our analysis is not influenced by other predictors of death because of our use of propensity score matching. Hence, the study findings call for future policies and interventions that target the detection and correction of PCs prior to kidney transplantation. Our results suggest that early screening that occurred before dialysis initiation is likely to capture psychosocially impaired ESRD patients who may progress quickly to death after future receipt of a transplant. In Ontario, the determination of a patient’s psychosocial status largely relies on self-reported results during informal interviews, which are only conducted in preparation for transplant listing.¹⁹ Hence, stringent policies are needed to enforce a comprehensive psychosocial assessment of ESRD patients before dialysis initiation. Prescription of antidepressant medications for ESRD patients who are soon to be transplanted needs to be carefully monitored to reduce future risk of drug interactions. Furthermore, we showed in our analysis that transplant candidates who were PC-free actually experienced a longer pretransplant dialysis episode, which reflects a potential inadequacy in the current algorithm used for the management of waitlisted patients. Because patients with predialysis PCs constitute a group of high-risk transplant candidates, transplant programs should allow sufficient time for these patients to engage in active psychiatric therapies before transplantation. This will enable patients without the risk to obtain a timely transplant and gain the full benefit of transplantation.

Our study has several limitations. First, we did not have information on patients with predialysis PCs who did not receive a transplant or died on the wait list. Hence, the psychosocial effect concluded in our analysis is likely to be a conservative estimation of the true effect as those who eventually received a transplant probably have relatively mild PCs that passed pretransplant assessment.¹⁹ Second, we were unable to evaluate patient adherence after transplantation or if predialysis PCs persisted after transplantation. However, the objective of the study is to convey an independent association between early history of PCs and posttransplant death. Future studies with more comprehensive data need to elucidate the specific mechanism of how predialyiss PCs influence patient survival after transplant. Third, despite our acknowledgment on the heterogeneous nature of PCs among ESRD patients, we did not conduct any subgroup analysis to assess whether the risk of posttransplant mortality differed by types of PCs. This is due to our limited access to the person-level ICD 10-CA information in our data set which impeded us to make any further groupings of PCs. Similarly, although we concluded a statistically shorter duration of pretransplant dialysis experienced by patients with predialysis PCs compared with their PC-free counterparts, we were unable to provide any explanations to this observation using the study sample, largely due to our lack of information on patients’ specific PC characteristics. Hence, future investigators who have access to a more detailed tracking of transplant recipients may test the potential differences in pre- and posttransplant outcomes by groupings of early PCs.

Our study demonstrated some key strengths. First, we used a large provincial cohort of transplant recipients with complete follow-up data that enabled the tracking of patients from the year before dialysis initiation to death after transplantation. Second, patient’s predialysis psychosocial status was formally defined using ICD 10-CA diagnostic codes and the ACG system which allowed us to establish an explicit exposure variable in the analysis. Most importantly, we used propensity score matching method to effectively remove the confounding effect, a significant source of bias that we have demonstrated in our analysis to lower the independent effect estimation of predialysis PCs on posttransplant death.

Conclusions

The study finding indicates a strong and independent effect of early diagnosis of PCs on deceased-donor kidney recipient survival. Although the particular pathways by which predialysis PCs reduce posttransplant patient survival are unknown, future policies need to focus on early detection and correction of such conditions among ESRD patients, preferably before transplantation. Effective screening mechanisms for PCs among ESRD patients are required early in the trajectory of care to ensure ample treatment time to prepare patients for future transplantation.

Supplemental Material

PredialysisPCs_CJKHD_Rev2_Clean_supp_1 – Supplemental material for Impact of Predialysis Psychosocial Conditions on Kidney Transplant Recipient Survival: Evidence Using Propensity Score Matching

Supplemental material, PredialysisPCs_CJKHD_Rev2_Clean_supp_1 for Impact of Predialysis Psychosocial Conditions on Kidney Transplant Recipient Survival: Evidence Using Propensity Score Matching by Rui Fu and Peter C. Coyte in Canadian Journal of Kidney Health and Disease

Footnotes

Acknowledgements

The authors thank Dr. S. Joseph Kim for his guidance.

Ethics Approval and Consent to Participate

Institute for Clinical Evaluative Sciences (ICES) is a prescribed entity under section 45 of Ontario’s Personal Health Information Protection Act (PHIPA). Section 45 authorizes ICES to collect personal health information, without consent, for the purpose of analysis or compiling statistical information with respect to the management of, evaluation or monitoring of, the allocation of resources to or planning for all or part of the health system. Projects conducted under section 45, by definition, do no require review by a Research Ethics Board. This project was conducted under section 45, and approved by ICES’s Privacy and Compliance Office. ICES is a designated prescribed entity under section 45 of the PHIPA. Participant informed consent was not required for this study.

Consent for Publication

Not applicable.

Availability of Data and Materials

The data set from this study is held securely in coded form at the Institute for Clinical Evaluative Sciences (ICES). While data sharing agreements prohibit ICES from making the data set publicly available, access can be granted to those who meet prespecified criteria for confidential access, available at . The full data set creation plan is available from the authors upon request.

Author Contributions

R.F. and P.C.C. contributed to research idea, study design, and data acquisition; R.F. contributed to data analysis/interpretation and statistical analysis; and P.C.C. contributed to supervision or mentorship. Each author contributed important intellectual content during article drafting and accepted accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study is supported by the Institute for Clinical Evaluative Sciences (ICES), which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care. The opinions, results, and conclusions reported in this article are those of the authors and are independent from the funding sources. No endorsement by ICES or the Ontario Ministry of Health and Long-Term Care is intended or should be inferred.

ORCID iD

Rui Fu

Supplemental Material

Supplemental material for this article is available online.

References

Murtagh

FEM

Addington-Hall

Higginson

. The prevalence of symptoms in end-stage renal disease: a systematic review. Adv Chronic Kidney Dis. 2007;14(1):82-99. doi:10.1053/j.ackd.2006.10.001.

Palmer

Vecchio

Craig

et al . Prevalence of depression in chronic kidney disease: systematic review and meta-analysis of observational studies. Kidney Int. 2013;84(1):179-191. doi:10.1038/ki.2013.77.

Hedayati

Jiang

O’Connor

et al . The association between depression and chronic kidney disease and mortality among patients hospitalized with congestive heart failure. Am J Kidney Dis. 2004;44(2):207-215.

Watnick

Kirwin

Mahnensmith

Concato

The prevalence and treatment of depression among patients starting dialysis. Am J Kidney Dis. 2003;41(1):105-110. doi:10.1053/ajkd.2003.50029.

Walters

Hays

Spritzer

Fridman

Carter

WB.

Health-related quality of life, depressive symptoms, anemia, and malnutrition at hemodialysis initiation. Am J Kidney Dis. 2002;40(6):1185-1194. doi:10.1053/ajkd.2002.36879.

Mucsi

Bansal

Jeannette

et al . Mental health and behavioral barriers in access to kidney transplantation: a Canadian cohort study. Transplantation. 2017;101(6):1182-1190. doi:10.1097/TP.0000000000001362.

Wolfe

Ashby

Milford

et al . Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med. 1999;341(23):1725-1730. doi:10.1056/NEJM199912023412303.

Wolfe

McCullough

Leichtman

Predictability of survival models for waiting list and transplant patients: calculating LYFT. Am J Transplant. 2009;9(7):1523-1527. doi:10.1111/j.1600-6143.2009.02708.x.

Laupacis

Keown

Pus

et al . A study of the quality of life and cost-utility of renal transplantation. Kidney Int. 1996;50:235-242.

10.

Chilcot

Spencer

BWJ

Maple

Mamode

Depression and kidney transplantation. Transplantation. 2014;97(7):717-721. doi:10.1097/01.TP.0000438212.72960.ae.

11.

Veater

East

Exploring depression amongst kidney transplant recipients: a literature review. J Ren Care. 2016;42(3):172-184. doi:10.1111/jorc.12162.

12.

Gill

Tonelli

Johnson

Kiberd

Landsberg

Pereira

BJG

. The impact of waiting time and comorbid conditions on the survival benefit of kidney transplantation. Kidney Int. 2005;68(5):2345-2351. doi:10.1111/j.1523-1755.2005.00696.x.

13.

Coffman

KL.

The debate about marijuana usage in transplant candidates: recent medical evidence on marijuana health effects. Curr Opin Organ Transplant. 2008;13(2):189-195. doi:10.1097/MOT.0b013e3282f56139.

14.

Coffman

Crone

Rational guidelines for transplantation in patients with psychotic disorders. Curr Opin Organ Transplant. 2002;7(4):385-388. http://journals.lww.com/co-transplantation/Fulltext/2002/12000/Rational_guidelines_for_transplantation_in.15.aspx. Accessed December 27, 2018.

15.

Cukor

Newville

Jindal

Depression and immunosuppressive medication adherence in kidney transplant patients. Gen Hosp Psychiatry. 2008;30(4):386-387. doi:10.1016/j.genhosppsych.2007.12.003.

16.

Gorevski

Succop

Sachdeva

et al . Is there an association between immunosuppressant therapy medication adherence and depression, quality of life, and personality traits in the kidney and liver transplant population. Patient Prefer Adherence. 2013;7:301-307. doi:10.2147/PPA.S34945.

17.

DiMatteo

Lepper

Croghan

TW.

Depression is a risk factor for noncompliance with medical treatment: meta-analysis of the effects of anxiety and depression on patient adherence. Arch Intern Med. 2000;160(14):2101-2107. doi:10.1001/archinte.160.14.2101.

18.

Cukor

Rosenthal

Jindal

Brown

Kimmel

PL.

Depression is an important contributor to low medication adherence in hemodialyzed patients and transplant recipients. Kidney Int. 2009;75(11):1223-1229. doi:10.1038/ki.2009.51.

19.

Gumabay

Novak

Bansal

et al . Pre-transplant history of mental health concerns, non-adherence, and post-transplant outcomes in kidney transplant recipients. J Psychosom Res. 2018;105:115-124. doi:10.1016/j.jpsychores.2017.12.013.

20.

Denhaerynck

Dobbels

Cleemput

et al . Prevalence, consequences, and determinants of nonadherence in adult renal transplant patients: a literature review. Transpl Int. 2005;18(10):1121-1133. doi:10.1111/j.1432-2277.2005.00176.x.

21.

Butler

Roderick

Mullee

Mason

Peveler

RC.

Frequency and impact of nonadherence to immunosuppressants after renal transplantation: a systematic review. Transplantation. 2004;77(5):769-776.

22.

Levitsky

Goldberg

Smith

et al . Acute rejection increases risk of graft failure and death in recent liver transplant recipients. Clin Gastroenterol Hepatol. 2017;15(4):584-593.e2. doi:10.1016/j.cgh.2016.07.035.

23.

Cosio

Alamir

Yim

et al . Patient survival after renal transplantation: I. The impact of dialysis pre-transplant. Kidney Int. 1998;53(3):767-772. doi:10.1046/j.1523-1755.1998.00787.x.

24.

Meier-Kriesche

H-U

Kaplan

Waiting time on dialysis as the strongest modifiable risk factor for renal transplant outcomes: a paired donor kidney analysis. Transplantation. 2002;74(10):1377-1381. doi:10.1097/01.TP.0000034632.77029.91.

25.

Meier-Kriesche

H-U

Port

Ojo

et al . Effect of waiting time on renal transplant outcome. Kidney Int. 2000;58(3):1311-1317. doi:10.1046/j.1523-1755.2000.00287.x.

26.

Dobbels

Skeans

Snyder

Tuomari

Maclean

Kasiske

BL.

Depressive disorder in renal transplantation: an analysis of Medicare claims. Am J Kidney Dis. 2008;51(5):819-828. doi:10.1053/j.ajkd.2008.01.010.

27.

Zelle

Dorland

Rosmalen

JGM

et al . Impact of depression on long-term outcome after renal transplantation: a prospective cohort study. Transplantation. 2012;94(10):1033-1040. doi:10.1097/TP.0b013e31826bc3c8.

28.

Novak

Molnar

Szeifert

et al . Depressive symptoms and mortality in patients after kidney transplantation: a prospective prevalent cohort study. Psychosom Med. 2010;72(6):527-534. doi:10.1097/PSY.0b013e3181dbbb7d.

29.

Courtwright

Salomon

Lehmann

et al . The association between mood, anxiety and adjustment disorders and hospitalization following lung transplantation. Gen Hosp Psychiatry. 2016;41:1-5. doi:10.1016/j.genhosppsych.2016.04.002.

30.

Zaltzman

. The new kidney allocation in Ontario. St. Michael’s Hospital. http://www.stmichaelshospital.com/pdf/programs/renal-transplant/kidney-allocation-system-in-Ontario.pdf. Published 2014. Accessed June 14, 2019.

31.

Canadian Institute for Health Information. ICD-10-CA/CCI implementation schedule. Canadian Institute for Health Information. https://www.cihi.ca/en/icd-10-cacci-implementation-schedule. Published March 21, 2017. Accessed April 24, 2019.

32.

Canadian Institute for Health Information. Canadian Organ Replacement Register Metadata (CORR). Canadian Institute for Health Information. https://www.cihi.ca/en/canadian-organ-replacement-register-metadata-corr. Published December 18, 2018. Accessed December 26, 2018.

33.

The Johns Hopkins University. The Johns Hopkins ACG® system: excerpt from version 11.0 technical reference guide. https://www.healthpartners.com/ucm/groups/public/@hp/@public/documents/documents/cntrb_035024.pdf. Published 2015. Accessed June 14, 2019.

34.

Austin

van Walraven

Wodchis

Newman

Anderson

GM.

Using the Johns Hopkins Aggregated Diagnosis Groups (ADGs) to predict mortality in a general adult population cohort in Ontario, Canada. Med Care. 2011;49(10):932-939. doi:10.1097/MLR.0b013e318215d5e2.

35.

Local Health Integration Network. Date unknown. http://www.lhins.on.ca/. Accessed December 26, 2018.

36.

Ontario Renal Network. About transplantation. Ontario Renal Network. https://www.ontariorenalnetwork.ca/en/kidney-care-resources/living-with-chronic-kidney-disease/about-transplantation. Published April 25, 2018. Accessed December 17, 2018.

37.

Baiocchi

Cheng

Small

DS.

Tutorial in biostatistics: instrumental variable methods for causal inference. Stat Med. 2014;33(13):2297-2340. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4201653/. Accessed May 8, 2018.

38.

Rosenbaum

Rubin

DB.

The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41-55.

39.

Rosenbaum

Rubin

DB.

Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc. 1984;79(387):516-524. doi:10.2307/2288398.

40.

D’Agostino

RB.

Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17(19):2265-2281.

41.

Austin

Mamdani

MM.

A comparison of propensity score methods: a case-study estimating the effectiveness of post-AMI statin use. Stat Med. 2006;25(12):2084-2106. doi:10.1002/sim.2328.

42.

Austin

PC.

The use of propensity score methods with survival or time-to-event outcomes: reporting measures of effect similar to those used in randomized experiments. Stat Med. 2014;33(7):1242-1258. doi:10.1002/sim.5984.

43.

Garrido

Kelley

Paris

et al . Methods for constructing and assessing propensity scores. Health Serv Res. 2014;49(5):1701-1720. doi:10.1111/1475-6773.12182.

44.

Austin

PC.

A tutorial and case study in propensity score analysis: an application to estimating the effect of in-hospital smoking cessation counseling on mortality. Multivariate Behav Res. 2011;46(1):119-151. doi:10.1080/00273171.2011.540480.

45.

Brookhart

Schneeweiss

Rothman

Glynn

Avorn

Stürmer

Variable selection for propensity score models. Am J Epidemiol. 2006;163(12):1149-1156. doi:10.1093/aje/kwj149.

46.

Rosenbaum

Rubin

DB.

Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat. 1985;39(1):33-38. doi:10.2307/2683903.

47.

Austin

PC.

An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46(3):399-424. doi:10.1080/00273171.2011.568786.

48.

Klein

Moeschberger

Survival Analysis: Techniques for Censored and Truncated Data. New York, NY: Spring-Verlag; 1997.

49.

Lin

Wei

LJ.

The robust inference for the Cox proportional hazards model. J Am Stat Assoc. 1989;84(408):1074-1078. doi:10.1080/01621459.1989.10478874.

50.

Goldstein

Thomas

Zaroff

Nguyen

Menza

Khush

KK.

Assessment of heart transplant waitlist time and pre- and post-transplant failure: a mixed methods approach. Epidemiology. 2016;27(4):469-476. doi:10.1097/EDE.0000000000000472.

51.

Leuven

Sianesi

. PSMATCH2: Stata Module to Perform Full Mahalanobis and Propensity Score Matching, Common Support Graphing, and Covariate Imbalance Testing. Boston College Department of Economics. https://ideas.repec.org/c/boc/bocode/s432001.html. Published 2018. Accessed December 28, 2018.

52.

Canadian Institute for Health Information. Treatment of end-stage organ failure in Canada, Canadian Organ Replacement Register, 2008 to 2017: data tables, end-stage kidney disease and kidney transplant. https://www.cihi.ca/en/organ-replacement-in-canada-corr-annual-statistics-2018. Published 2018. Accessed June 14, 2019.

53.

Maldonado

Sher

Lolak

et al . The Stanford integrated psychosocial assessment for transplantation: a prospective study of medical and psychosocial outcomes. Psychosom Med. 2015;77(9):1018-1030. doi:10.1097/PSY.0000000000000241.

54.

Evans

Stock

Zeber

et al . Posttransplantation outcomes in veterans with serious mental illness. Transplantation. 2015;99(8):e57-e65. doi:10.1097/TP.0000000000000616.

55.

Crean

Tapert

Minassian

MacDonald

Crane

Mason

BJ.

Effects of chronic, heavy cannabis use on executive functions. J Addict Med. 2011;5(1):9-15. doi:10.1097/ADM.0b013e31820cdd57.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.10 MB