Effect of the ASCENT Intervention to Increase Knowledge of Kidney Allocation Policy Changes Among Dialysis Providers

Abstract

Introduction

The Allocation System Changes for Equity in Kidney Transplantation (ASCENT) trial was a cluster-randomized pragmatic, effectiveness-implementation study designed to test whether a multicomponent educational intervention targeting leadership, clinic staff, and patients in dialysis facilities improved knowledge and awareness of the 2014 Kidney Allocation System (KAS) change.

Methods

Participants included 690 dialysis facility medical directors, nephrologists, social workers, and other staff within 655 US dialysis facilities, with 51% (n = 334) in the intervention group and 49% (n = 321) in the control group. Intervention activities included a webinar targeting medical directors and facility staff, an approximately 10-minute educational video targeting dialysis staff, an approximately 10-minute educational video targeting patients, and a facility-specific audit and feedback report of transplant performance. The control group received a standard United Network for Organ Sharing brochure. Provider knowledge was a secondary outcome of the ASCENT trial and the primary outcome of this study; knowledge was assessed as a cumulative score on a 5-point Likert scale (higher score = greater knowledge). Intention-to-treat analysis was used.

Results

At baseline, nonintervention providers had a higher mean knowledge score (mean ± SD, 2.45 ± 1.43) than intervention providers (mean ± SD, 2.31 ± 1.46). After 3 months, the average knowledge score was slightly higher in the intervention (mean ± SD, 3.14 ± 1.28) versus nonintervention providers (mean ± SD, 3.07 ± 1.24), and the estimated mean difference in knowledge scores between the groups at follow-up minus the mean difference at baseline was 0.25 (95% confidence interval [CI], 0.11–0.48; P = 0.039). The effect size (0.41) was low to moderate.

Conclusion

Dialysis facility provider education could help extend the impact of a national policy change in organ allocation.

Graphical abstract

See Commentary on Page 1385

In 2014, the United Network for Organ Sharing implemented a major change to how deceased donor kidneys are allocated to patients with an overhaul of the KAS policy. The policy change was intended in part to increase patient access to kidney transplantation and in part to reduce racial disparities in transplantation.¹ In the previous policy, time on the waitlist begun when the patient was deemed transplant-eligible and was added to the waitlist. In the new KAS, patient waiting time, an important determinant of receiving an organ, now includes time spent on dialysis. In 2016, hemodialysis was the initial renal replacement therapy for 87.3% of patients with incident end-stage renal disease (ESRD).² Therefore, dialysis facilities remain an important point of access to renal transplantation. However, the extent to which dialysis providers are educated about the impact of KAS on patient waitlisting and access to transplantation is largely unknown.

It is essential that clinical providers who are responsible for referring patients for kidney transplantation are aware of the change in the policy because patients who have been on dialysis for multiple years would likely go to the top of the waiting list and receive a transplant more quickly than in the past. This is particularly important for minority patients, such as African American patients with ESRD, who typically have longer times on dialysis and substantial racial disparities in transplant access.3, 4, 5 Lags in access to transplantation implies barriers to its associated benefits, including (i) better long-term survival,^6,7 (ii) superior quality of life,6, 7, 8, 9 (iii) a lower risk of cardiovascular events,⁶ and (iv) substantial health care cost savings to the health care system,^10,11 among other benefits. Despite changes in the KAS designed to increase transplant access and reduce racial disparities, Kim et al.¹² found that 81% of dialysis providers in facilities with low waitlisting were unaware of existing racial disparities in waitlisting.

The Centers for Medicare and Medicaid Services, as a part of conditions for coverage for patients with ESRD, requires that education on kidney transplantation be provided to the approximately 500,000 patients in the >7000 US dialysis facilities.13, 14, 15 However, evidence suggests that dialysis providers, including nephrologists, social workers, clinic managers, and/or nurses, have limited knowledge of transplantation. For example, research has shown that 52% of staff had insufficient knowledge to answer most patient questions about transplantation.¹⁶ Inadequate care provider education on transplantation,^12,16 provider beliefs,^17,18 and provider’s differential provision of patient education on transplantation¹⁹ may preclude patient education and contribute to disparities in transplantation.

Previous research suggests that system-level interventions targeting dialysis facilities can improve care at the regional and national levels.20, 21, 22, 23 Evidence also suggests that audit and feedback reports combined with multicomponent and multilevel educational interventions in dialysis facilities can reduce racial disparities in transplant access.24, 25, 26 In this study, we assess the effectiveness of a multicomponent educational intervention on changes in knowledge of the impact new KAS on transplantation among dialysis facility providers in US dialysis facilities with low waitlisting. The outcome of the present analysis, provider knowledge, is a secondary outcome in the ASCENT trial. Increasing provider knowledge on KAS and transplantation is a requisite step toward increasing the quality of patient education on transplantation, increasing access to transplantation, and reducing inequities in kidney transplantation. We hypothesized that providers in dialysis facilities allocated to the multicomponent educational intervention group would have a greater knowledge of transplantation and KAS compared with those in dialysis facilities receiving standard educational brochures. Our results may inform the reach of the KAS policy among providers as related to renal transplantation knowledge and access.

Methods

Study Design

The ASCENT study²⁷ was a cluster-randomized,^28,29 pragmatic, effectiveness-implementation study designed to test whether a multicomponent educational intervention targeting medical directors and facility leadership, clinical staff, and patients with ESRD in US dialysis facilities with low waitlisting increased patient waitlisting and reduced waitlisting disparities at 1 year. Data collection for these longer-term endpoints at the patient level is ongoing. The ASCENT study was also designed to evaluate an earlier secondary endpoint—whether the multicomponent educational intervention increased dialysis facility provider knowledge, the focus of this work. Participant enrollment, allocation, and follow-up are presented in Figure 1.²⁹ Following e-mailed consent for participation, the baseline survey was administered to dialysis facility providers from October to December 2016, with a follow-up survey administered about 3 months after each facility completed the baseline survey. Educational materials were made available within a few days of baseline survey completion. The study was approved by the Emory University Institutional Review Board (IRB#81580; trial registration: ClinicalTrials.gov identifier NCT02879812).

Figure 1.

Flow diagram illustrating the number of dialysis facilities that (i) met the inclusion/exclusion criteria during enrollment, (ii) were allocated to the study groups, (iii) participated during follow-up, and (iv) were included as part of the analytic sample. ESRD, end-stage renal disease.

Eligibility, Settings, and Participants

Among all US dialysis facilities in 2014 (n = 7822), we excluded those with <11 patients (n = 1324) and <4 African American patients (n = 1519). Facilities that ranked in the lowest national tertile for proportion of patients waitlisted (<15.2%) were defined as low waitlisting and considered eligible (n = 1529; Figure 2).³⁰ Eight hundred twelve facilities did not respond to the request to participate via e-mail and 62 facilities had erroneous contact information, leaving 655 dialysis facilities with low waitlisting (42.8%) consenting to participate in the study before randomization. A total of 334 (51%) and 321 (49%) dialysis facilities were randomly allocated to the intervention and waitlist control groups, respectively (Figure 1).

Figure 2.

Dialysis facility locations were overlaid on cartographic boundaries provided by the US Census Bureau (public domain) and implemented using the USAboundaries R package.³⁰

Recruitment

ESRD Network 6 leadership led the recruitment process by sending the baseline and follow-up surveys to eligible dialysis facility medical directors or nurse managers.²⁷ Dialysis facility providers were aware at the time of the survey that their identifiable data were blinded to network leaders. ASCENT research staff sent e-mails and made calls to invite facilities that did not respond to the initial or follow-up e-mails to participate. We tested whether the distribution of participating dialysis facilities differed from nonparticipating facilities based on cluster level characteristics: profit status (profit vs. nonprofit), census region (Northeast, South, Midwest, or West), and percent of families considered to be below the national poverty level (<20%, 10%–20%, or >20%).

Randomization and Intervention

A random number generator was used to assigned facilities to the group (cluster/dialysis facility). The assignment sequence remained concealed until assignment was complete. To minimize spillover, we assigned facilities with the same medical director or nurse manager to the same group. There were no other deviations from the random assignment protocol.

Facilities in the intervention group received a webinar targeting medical directors and other dialysis facility staff, an approximately 10-minute educational video targeting dialysis staff, and a facility-specific audit and feedback report of their transplant performance as hard copies and by e-mail within 2 weeks after the baseline surveys were completed.²⁷ The interventions were developed by a multidisciplinary advisory board described in detail elsewhere.²⁷ All intervention group educational materials were made available online for unlimited access (www.ascenttotransplant.org). Given the nature of the intervention, providers were not blinded. The power of our study was based on the original ASCENT trial primary outcome.²⁷ The control group received a standard United Network for Organ Sharing brochure.

The outcome variable, measured using 5 items on a survey, was designed a priori to be a secondary outcome of the trial: staff knowledge of KAS and transplantation. This survey was developed by a multidisciplinary stakeholder group as previously described²⁷ and was used to quantify knowledge of KAS and transplantation using a cumulative knowledge score (0 = least knowledge, 5 = highest level of knowledge). The survey items can be found in Section S1 of the Supplementary Table S1.

Analytical Methods

As part of the pragmatic trial design, we applied a provider-level, intention-to-treat analysis³¹ and used linear mixed-effects models to estimate the treatment effect of the intervention with 95% CIs using R software’s robustlmm,³² lme4,³³ and lmerTest³⁴ packages. To address potential baseline differences in outcome, we estimated the treatment effect as the mean difference in knowledge between the intervention and control group at follow-up minus the mean difference in knowledge between the study groups at baseline. We adjusted for the following individual-level covariates: dialysis care provider, sex, race, care role, and length of time in role. We examined the proportion of model variability attributable to facilities, geographic areas (ZIP codes), and ESRD networks each via a random effect. Dialysis facilities were nested within ZIP codes, which were nested in ESRD networks. We obtained restricted maximum likelihood model estimates.³³ P < 0.05 was considered statistically significant. The estimated effect size was defined as the treatment effect divided by the dispersion between dialysis facilities.

In an exploratory stratified analysis, we investigated the potential differential gain in intervention effects using facility characteristics that may influence health outcomes and contribute to waitlisting disparities: facility size,³⁵ profit status (profit vs. nonprofit),¹⁹ number of social workers,³⁶ geographic region (Northeast, South, Midwest, or West),³⁷ level of ZIP code poverty,³⁸ and percent of patients in the facilities with identified comorbidities.³⁹ The facility-level characteristics not included in the study data were obtained by linking to United States Renal Data System facility-level data, US census data, and Dialysis Facility Report data.

Missing data patterns were investigated using R software’s VIM package.⁴⁰ For primary analyses, multiple imputation was implemented using the MICE package in R.⁴¹ We imputed missing values using the classification and regression trees method in MICE, a nonparametric approach that can be applied to data with complex distributions. We imputed both missing covariate and response variables, creating 10 datasets over 50 iterations. and subsequently applied our analytic model to the pooled dataset.⁴² In a sensitivity analysis, we performed a complete case analysis to investigate possible bias in model estimates.

Results

The number of respondents per facility ranged between 1 and 5. Of the 655 facilities, 94% and 82% had 1 respondent at baseline and follow-up, respectively. We included data from facilities with ≥1 participant. Baseline facility-level characteristics were similarly distributed across study groups by profit status, census region, percentage of families below the national poverty level, numbers of patient and staff, and distributions of comorbid conditions (Table 1). Based on cluster-level characteristics, the facilities that were eligible for study inclusion but did not respond to the invitation to participate (n = 812) were similarly distributed to those that participated in the study (n = 655) by profit status and by percent of families considered to be below the national poverty level, and differently distributed by census region (χ² = 32.6, P < 0.001).

Table 1.

Baseline characteristics among US dialysis facilities included in the ASCENT study by study group, October 2016.

Baseline characteristics	Baseline participation, N = 655	Control group, n = 321 (49%)	Intervention group, n = 334 (51%)
Profit status, n (%)
For-profit	64 (9.8)	29 (9.0)	35 (10.5)
Nonprofit	591 (90.2)	292 (91.0)	299 (89.5)
Census region, n (%)
Northeast	27 (4.1)	13 (4.0)	14 (4.2)
South	424 (64.7)	223 (69.5)	231 (69.2)
Midwest	120 (18.3)	66 (20.6)	54 (16.2)
West	54 (8.2)	19 (5.9)	35 (10.5)
Percentage of families considered to be below the national poverty level (based on ZIP code US Census data), n (%)
<5	52 (7.9)	22 (6.9)	30 (9.0)
5–10	137 (20.9)	73 (22.7)	64 (19.2)
10–15	174 (26.6)	85 (26.5)	89 (26.6)
15–20	127 (19.4)	60 (18.7)	67 (20.1)
>20	154 (23.5)	80 (24.9)	74 (22.2)
Missing	11 (1.7)	1 (0.3)	10 (3.0)
Patients, n (%)
≤50	187 (28.5)	92 (28.7)	95 (28.4)
51–70	157 (24.0)	76 (23.7)	81 (24.3)
71–95	147 (22.4)	73 (22.7)	74 (22.2)
<95	148 (22.6)	73 (22.7)	75 (22.5)
Missing	16 (2.4)	7 (2.2)	9 (2.7)
Staff, n (%)
≤10	170 (26.0)	87 (27.1)	83 (24.9)
11–15	215 (32.8)	97 (30.2)	118 (35.3)
16–20	124 (18.9)	66 (20.6)	58 (17.4)
>20	123 (18.8)	61 (19.0)	62 (18.6)
Missing	23 (3.5)	10 (3.1)	13 (3.9)
Percent of patients with comorbidities, quartiles
Diabetes, n (%)
<50	149 (22.7)	71 (22.1)	78 (23.4)
50–55.34	166 (25.3)	83 (25.9)	83 (24.9)
55.34–61.32	160 (24.4)	74 (23.1)	86 (25.7)
>61.32	159 (24.3)	85 (26.5)	74 (22.2)
Missing	21 (3.2)	8 (2.5)	13 (3.9)
Hypertension, n (%)
<84.71	158 (24.1)	72 (22.4)	86 (25.7)
84.71–89.19	164 (25.0)	90 (28.0)	74 (22.2)
89.19–92.86	156 (23.8)	70 (21.8)	86 (25.7)
>92.86	156 (23.8)	81 (25.2)	75 (22.5)
Missing	21 (3.2)	8 (2.5)	13 (3.9)
Congestive heart failure, n (%)
<22.50	155 (23.7)	71 (22.1)	84 (25.1)
22.50–29.17	161 (24.6)	85 (26.5)	76 (22.8)
29.17–36.84	157 (24.0)	80 (24.9)	77 (23.1)
>36.84	160 (24.4)	77 (24.0)	83 (24.9)
Missing	22 (3.4)	8 (2.5)	14 (4.2)
Chronic obstructive pulmonary disease, n (%)
<5.33	148 (22.6)	71 (22.1)	77 (23.1)
5.33–9.00	149 (22.7)	78 (24.3)	71 (21.3)
9.00–13.31	150 (22.9)	75 (23.4)	75 (22.5)
>13.31	155 (23.7)	75 (23.4)	80 (24.0)
Missing	53 (8.1)	22 (6.9)	31 (9.3)
Cancer, n (%)
<3.7740	152 (23.2)	67 (20.9)	85 (25.4)
3.77–6.25	135 (20.6)	74 (23.1	61 (18.3)
6.25–9.09	153 (23.4)	74 (23.1)	79 (23.7)
>9.09	143 (21.8)	69 (21.5)	74 (22.2)
Missing	72 (11.0)	37 (11.5)	35 (10.5)
Insurance type, mean (SD)
Medicare	43.7 (35.0)	45.4 (37.3)	42.2 (32.7)
Medicaid	30.9 (26.3)	30.8 (26.4)	31.0 (26.2)
Employer-based	22.0 (19.9)	22.5 (20.7)	21.5 (19.2)
Other insurance	6.5 (11.0)	6.3 (8.3)	6.7 (13.2)
Uninsured	10.4 (11.6)	10.7 (11.7)	10.1 (11.5)

ASCENT, Allocation System Changes for Equity in Kidney Transplantation.

Among the 690 providers who completed the baseline survey (Table 2), 17.2% were medical directors, 49.1% were nurse managers, 17.4% were facility administrators, and 13.2% were social workers. Most participants (73.6%) were female. Approximately 62.1% of participants in the intervention group were white, and 11.8% were African American, compared with 58.2% and 13.5% in the control group. Almost half of participants (48.3%) had been in their professional role for 1–5 years. All provider baseline characteristics were similarly distributed across study groups by sex, race, provider role, and length of time in the role (Table 2).

Table 2.

Characteristics of providers who completed the baseline survey from 655 dialysis facilities in all 18 ESRD Networks by study group, October 2016

Baseline characteristics, n (%)	Baseline participation, N = 690	Control group, n = 342 (49.6%)	Intervention group, n = 348 (50.4%)
Sex
Male	147 (21.3)	74 (21.6)	73 (21.0)
Female	508 (73.6)	245 (71.6)	263 (75.6)
Missing	35 (5.1)	23 (6.7)	12 (3.4)
Race
White	415 (60.1)	199 (58.2)	216 (62.1)
African American	87 (12.6)	46 (13.5)	41 (11.8)
Hispanic	42 (6.1)	19 (5.6)	23 (6.6)
Asian	58 (8.4)	27 (7.9)	31 (8.9)
Other	20 (2.9)	12 (3.5)	8 (2.3)
Missing	68 (9.9)	39 (11.4)	29 (8.3)
Professional role
Medical director	119 (17.2)	59 (17.3)	60 (17.2)
Nurse manager	339 (49.1)	175 (51.2)	164 (47.1)
Facility administrator	120 (17.4)	59 (17.3)	61 (17.5)
Social worker	91 (13.2)	40 (11.7)	51 (14.7)
Other staff	21 (3.0)	9 (2.6)	12 (3.4)
Length of time in role, yr
<1	108 (15.7)	51 (14.9)	57 (16.4)
1–5	333 (48.3)	170 (49.7)	163 (46.8)
5–10	120 (17.4)	52 (15.2)	68 (19.5)
≥10	129 (18.7)	69 (20.2)	60 (17.2)

ESRD, end-stage renal disease.

Approximately 55% of the sample had no missing information; 28% had missing provider role variable, and 3% of the sample had missing race information. Approximately ≤3% of the sample had ≥3 variables with missing values. The missing data were assumed to be missing at random.

Among 655 facilities, a total of 23 (7%) facilities (n = 24 staff respondents) in the intervention and 19 (6%) facilities (n = 20 staff respondents) in the control groups were considered lost to follow-up. In the follow-up survey (n = 613 facilities), providers from 73% and 65% of the dialysis facilities in the intervention and control group, respectively, reported to have received the allocated intervention. The remaining providers may have not received the intervention or may have received the intervention but remained unaware or nonadherent. However, using intention-to-treat analysis served to maintain the balance in our original allocation and yield conservative treatment effects.³¹

At baseline, the control group had a higher crude average knowledge score (mean ± SD, 2.45 ± 1.43) than the intervention group (mean ± SD, 2.31 ± 1.46). Investigating baseline differences using both adjusted and unadjusted linear mixed effects models, we found no statistically significant difference in baseline cumulative knowledge scores by study group. At follow-up, the intervention group had a higher crude average knowledge score (mean ± SD, 3.14 ± 1.28) than the control group (mean ± SD, 3.07 ± 1.24). The primary analysis model estimates obtained from using multiple imputation pooling 10 datasets over 50 iterations showed that the mean change of the knowledge score over time in the intervention group compared with the model-estimated mean change of the knowledge score over time in the control group was 0.25 (95% CI, 0.11–0.48; P = 0.039; Table 3), with a low to moderate effect size of 0.41.

Table 3.

Mean difference in staff knowledge gap between the dialysis facility intervention (50.4%) and control group (49.6%) at follow-up minus mean difference of knowledge gap at baseline using multiple imputation pooling 10 datasets over 50 iterations for providers from 655 dialysis in all 18 ESRD Networks in the lowest national tertile (<15.2%), October 2016 to May 2017

Description	Estimate	95% CI	P Value
Estimated treatment effecta	0.25	0.11–0.48	0.039
Intervention vs. control groupb	−0.16	−0.36 to 0.03	0.104
Follow-up vs. baselineb	0.63	0.47–0.80	<0.001
Covariates: sex, role, length of time in role, race
Male vs. female	−0.07	−0.33 to 0.18	0.562
Provider role vs. medical director
Nurse manager	−1.00	−1.30 to −0.69	<0.001
Facility administrator	−0.92	−1.23 to −0.61	<0.001
Social worker	−0.48	−0.82 to −0.13	0.008
Other	−0.98	−1.40 to −0.57	<0.001
Length of time in role vs. shortest length of time in role (<1 yr)
1 to <5	0.22	0.04–0.40	0.016
5 to <10	0.12	−0.10 to 0.34	0.282
≥10	0.35	0.12–0.58	0.003
Race vs. white
African American	−0.43	−0.64 to −0.22	<0.001
Hispanic	−0.56	−0.85 to −0.27	<0.001
Asian	−0.33	−0.61 to −0.05	0.023
Other	−0.04	−0.40 to 0.32	0.819

CI, confidence interval; ESRD, end-stage renal disease.

^aThe treatment effect is defined as the mean difference in knowledge between the intervention and the control group at follow-up minus the mean difference in knowledge between the intervention and control group at baseline.

^bThe lower-order terms are a result of the treatment effect, which is modeled as an interaction of time and study group.

In the sensitivity analysis using complete case model estimates with robust standard errors, we found the increase in knowledge score from baseline to follow-up among intervention providers minus the increase of the knowledge score from baseline to follow-up among control providers to be slightly higher at 0.33 (95% CI, 0.04–0.62; P = 0.025; Supplementary Table S2) than primary analysis model estimates (Table 3). The direction of the model parameter estimates in the imputed data was the same as that in the complete case analysis. Medical directors had higher knowledge overall at any point compared with providers in other roles regardless of study group. Of the total model variability, about 20% was attributable to the dialysis facilities. We found no variability attributable to geographic regions or ESRD networks.

In an exploratory stratified analysis, we examined the facility characteristics associated with a greater impact of the intervention, measured as a comparison of the difference in the average cumulative provider knowledge score between the intervention and control groups at follow-up and at baseline. A higher intervention impact was associated with facilities in which <50% (lowest quartile) of patients had diabetes (mean difference in knowledge, 0.89 [95% CI, 0.33–1.42]; P = 0.001), with facilities that had >20 (largest quartile) staff members (mean difference in knowledge, 0.63 [95% CI, 0.70–1.11], P = 0.017) and with nonprofit facilities (mean difference in knowledge, 1.28 [95% CI, 0.64–1.89]; P < 0.001).

Discussion

This study provides evidence of the effectiveness of a pragmatic, multicomponent educational intervention in increasing knowledge of KAS and transplantation among nephrologists, social workers, nurses, and clinic managers working in dialysis facilities with historically low waitlisting performance. While the estimated effect on knowledge change was low to moderate, when applied to a large number of facilities across the country, even small effects can have impact on improving knowledge and awareness about changes in kidney allocation policies on a population level.

Under the oversight of the Centers for Medicare and Medicaid Services, ESRD Networks develop and evaluate adherence to standards that govern quality of care in facilities and assignment of patients to other renal replacement modalities, including transplantation.⁴³ Evidence indicates that multicomponent provider education interventions can improve the quality of care of patients with ESRD in dialysis facilities.^44,45 These gains may be understood through the theory of reasoned action, which postulates that attitude is a mediating mechanism between knowledge and action.46, 47, 48, 49 Previous studies have shown that the engagement of institutional leaders in quality improvement initiatives increases the level of engagement and enhances the attitude and compliance of physicians and staff in those initiatives.^50,51

ESRD networks, dialysis facilities, UNOS, transplant center outreach staff, and/or Centers for Medicare and Medicaid Services could use this freely available multicomponent educational intervention to improve knowledge of KAS and transplantation among dialysis providers as a first step to improve patient education and transplantation access. Furthermore, such interventions may extend the reach of the new KAS policy. The results of this study are particularly relevant because comprehensive patient education on transplantation is predicated on the level of provider knowledge of KAS and transplantation. Our study examined 655 dialysis facilities in which an average of approximately 60 patients with ESRD were treated. If the ASCENT intervention were applied to all facilities, providers in these facilities could translate this knowledge to the approximately 40,000 patients with ESRD within these facilities. While additional resources for successful implementation may be required, we demonstrated that combining pragmatic techniques such as leveraging leadership engagement,^16,50,51 encouraging staff to watch videos during “lunch and learn” sessions as part of quality improvement initiatives, and offering continuing medical education credit as an incentive for participation can support successful implementation of such complex educational interventions compared to a single intervention component.⁵² Future research on the cost-effectiveness of such a multicomponent intervention may be worthwhile to further inform implementation strategies.

Our study targeted facilities with the lowest waitlisting in the nation, which are the facilities that most need to improve patient interest and knowledge of transplantation and to increase provider knowledge on their role at every step of the care process leading up to transplantation. Research indicates that patients who receive comprehensive education on transplantation are more likely to complete transplantation evaluation,⁵³ a requisite step for waitlisting. Furthermore, existing racial disparities in waitlisting³⁸ may in part be attributable to the lower likelihood of racial minorities to receive adequate patient education.⁵⁴

There has been a longstanding recognition of the need for provider education interventions to improve quality of care and to reduce disparities in care processes and outcomes since the Institute of Medicine’s Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care.⁵⁵ Extant research suggests that provider educational interventions can improve care processes and health outcomes of patients.⁵⁶ However, passive interventions such as printed material and traditional continuous medical education alone tend to be less effective in increasing knowledge or in changing the behavior of physicians compared with interactive interventions.⁵⁷ Our intervention encouraged interaction through questions and answers during the webinar and used visual illustrations of content during the video presentations and webinar session, and may explain why we observed an effect of the intervention.

Our study results also suggest that medical directors, regardless of group or time (baseline or follow-up) had higher cumulative knowledge scores than providers in other roles. Combined evidence suggests that facility leaders can potentially serve as champions in the sustained implementation of quality improvement programs.^50,51 Our exploratory stratified analysis also suggests that the intervention was most effective in dialysis facilities characterized by lower diabetes burden, more staff members, and nonprofit facilities. Future studies are needed to investigate institutional strategies such as incentives that promote consistent provider engagement in quality initiatives.

There are several potential limitations of this study. We analyzed the secondary outcome for the ASCENT trial because the primary outcome data are not yet available, and our original power calculation was for the primary outcome.²⁷ Nonetheless, our study shows a low to moderate effect size of 0.41 with improvement in knowledge among both the control and intervention groups, indicating that any provider education is an improvement of the status quo. It is possible that the effect size may have been smaller with a larger size of respondents. Nevertheless, extant meta-analysis suggests that educational and behavioral interventions with low effect sizes can be meaningful.^58,59 Future patient-level data for the primary outcome of the ASCENT trial (waitlisting) will be available to determine if provider change in knowledge was associated with a population-level impact on access to transplantation and in reduced disparities on the patient level.

The outcome as measured in this study is not continuous. However, resulting departures from normality do not much impact fixed effects and variance component estimators obtained using linear mixed effects models, and they can be corrected by using robust standard errors.⁶⁰ We showed in our sensitivity analysis that model estimates remain comparable before and after robust standard errors are applied to correct for departures from normality. Our results may only apply to US dialysis facilities with low waitlisting. The generalizability of the study was improved given that facilities were well distributed across the country and by numerous facility attributes related to facility level random allocation. Also, the proportion of missing data was large. However, the use of multiple imputation has been shown to reduce bias in missing at random data even when the proportion of missing data is large.⁶¹ Furthermore, selection bias resulting from differential nonresponse between the intervention and nonintervention groups is mitigated by multiple imputation.^62,63

The study also relied on self-reported survey data from providers. We provided anonymity of responses to network leaders to mitigate any social desirability bias. Over 25% of respondents did not recall receiving the intervention; results of a process evaluation study to examine delivery, common implementation approaches, and reactions to inform understanding of the effectiveness of intervention are ongoing. It is possible that the impact of an intervention such as ours may be attenuated by high provider turnover. However, evidence-based long-term quality improvement interventions have been shown to increase health care professional morale and reduce turnover.⁶⁴

In this effectiveness-implementation trial, we found that the ASCENT intervention increased dialysis provider knowledge of the KAS policy change and transplantation. The need for long-term education interventions designed to increase the knowledge of KAS and transplantation among dialysis providers remains. This study is poised to inform such sustainable and pragmatic intervention designs and implementation. Our study demonstrates that a pragmatic education intervention may be useful in extending the reach of a national policy change through provider education. The educational intervention we tested in this study is brief, freely available, and accessible. Centers for Medicare and Medicaid Services, United Network for Organ Sharing, transplant center outreach staff, ESRD Networks, and/or individual nephrology clinics and dialysis facilities may find these interventions valuable to implement in practice.

Disclosure

All the authors declared no competing interests.