The coordination of care is a significant challenge in the fragmented US health care system [1]. For US patients receiving dialysis, who are hospitalized for an average of 9.5 days per year (associated with annual Medicare costs of approximately US $12 billion) [2], transitions between the outpatient dialysis facility and the hospital present unique challenges. In addition to the usual elements of successful care transitions [3], dialysis-specific issues must be coordinated with providers during hospitalization and at discharge, such as identification and mitigation of difficulty with dialysis adherence; maintenance of patients’ dialysis schedules during hospitalization; and communication of changes in dry weight, vascular access status, or medications (particularly those that should be administered intravenously during outpatient dialysis treatments, such as antibiotics) to dialysis providers [4]. Although the frequency of dialysis treatment presents an opportunity to improve timely coordination for patients receiving dialysis [4], patients are likely to return to their facilities without having seen other outpatient providers who could help coordinate postdischarge care, and patients themselves may be unable to provide the reason for hospitalization or updated medical information after discharge. These care transition challenges contribute to poor outcomes, such as 30-day readmissions, which occur after approximately one-third of hospitalizations among patients receiving dialysis [2] and often before the patient presents to the dialysis facility [5]. Thus, early follow-up care after discharge is a necessary component of successful care transitions for these patients [6]: we and others have shown in previous work that time-sensitive clinical factors such as documented changes in dry weight [7], timely medication reconciliation [8], and increased physician encounters after discharge [9] are all associated with lower hospital readmission risk among patient receiving dialysis.

Electronic health records (EHRs) represent a possible means of instant information exchange between dialysis and hospital providers to address these issues; however, US outpatient dialysis facilities are usually managed independently and do not share EHRs with the hospitals to which their patients are admitted. Regional health care information exchanges are meant to circumvent these issues and close the patient information gap across health care settings [10]; however, the evidence for better outcomes is mixed [11], given that communication may not be timely, there may be no context for the information shared, and the information may not include what the provider is seeking (eg, information in medical notes rather than in the EHR). Thus, although the Centers for Medicare and Medicaid Services have explicitly prioritized the reduction of hospital readmissions in patients receiving dialysis [12-14]—and, implicitly, the improvement of care coordination between US hospitals and dialysis facilities—the lack of tools to ensure timely exchange of critical information during the hospital–dialysis facility transition remains a primary barrier to improving hospital outcomes in this population [15].

To address this barrier, we conducted a pragmatic pilot study in which we developed DialysisConnect, a secure web-based provider communication platform, and implemented it at our 4 Emory Dialysis outpatient facilities and Emory University Hospital Midtown (EUHM). Emory Dialysis and EUHM share an academic affiliation but do not share health care management or, importantly, an EHR. We aimed to (1) gather information on the critical components of the system from stakeholders and potential users, including hospital providers (hospitalists, advanced practice providers [APPs; nurse practitioners or physician assistants], and social workers or care coordinators), and dialysis providers (nephrologists, APPs, nurses, and social workers); (2) develop and introduce the system into the clinical setting with the help of site champions to determine which potential users engaged with the system and how; and (3) examine provider perceptions of the system. Together, this information on the feasibility, acceptability, and potential sustainability of this system, as piloted, is important for future research addressing communication gaps and care coordination between settings for patients receiving dialysis. Here, we describe the development and pilot implementation of DialysisConnect.

The initial wireframes for DialysisConnect, which were created by the technical team (led by RM) and showed each intended task (admission, communication, and discharge), were based on an existing platform for the exchange of transplant referral information for patients receiving dialysis [16]. The proposed workflow process for DialysisConnect was developed by the research team (including experts in epidemiology and health services research, hospital medicine, nephrology, gerontology, and engineering) [17]. Initial feedback on this DialysisConnect prototype (video simulation using static wireframes) was collected from 4 focus groups of stakeholders and potential users of DialysisConnect (excluding the site champions), as previously described [17]. The purpose of the focus groups (led by AEV) was to gather feedback on the desired elements and features of the system and increase future buy-in by the inclusion of potential users of DialysisConnect in its creation. Data regarding feedback on the proposed system were collated from the transcripts for the technical team (Multimedia Appendix 1). Incorporation of this initial feedback into the test system was prioritized by its importance to the focus group participants, the degree of automation that was possible, and the ease of building features into the system over a short period. Research team members thoroughly evaluated this test system and presented it to the clinical teams at the hospital and dialysis facilities in February 2020; disruptions to clinical care early in the COVID-19 pandemic led to an 8-month delay in the rollout, during which additional planned future upgrades to the system and user testing (n=9 potential users, including the site champions [JPL and TM at Emory Dialysis and CMO and KJ at EUHM]) were performed. Each participant was asked to perform a standardized set of tasks on the test system, with guidance from the research team (LCP and CH). Observations and feedback (Multimedia Appendix 2) were provided to the technical team for a final round of fixes and updates to the system before rollout.

Just before rollout, the team conducted remote group training sessions using the pilot version of DialysisConnect (Figure 1; Multimedia Appendix 3), and group emails were sent to all potential DialysisConnect users (as identified by champions) to announce the live rollout, along with detailed user guides (Multimedia Appendices 4 and 5) and contact information for questions. Throughout the pilot study, the list of potential users was updated as needed, and all significant upgrades (Textbox 1) were announced via email, including reminders about the availability of on-demand training sessions and revised user guides (Multimedia Appendices 4 and 5). The user guides included an idealized scenario of use for both hospital and dialysis providers, in which (1) a hospital provider identified an Emory Dialysis patient at admission; (2) the hospital provider reviewed any relevant, automatically uploaded information about the patient (eg, nephrologist, clinic, recent laboratories, and medications) and checked the reasons for admission; (3) a dialysis provider received an automated message about admission generated by the hospital provider’s entry and logged in to review the patient and status; (4) hospital and dialysis providers communicated via instant messages as needed or desired about the patient’s status throughout the hospitalization; (5) a hospital provider clicked through the discharge elements at the patient’s discharge; and (6) dialysis providers reviewed the discharge elements and, as needed, confirmed receipt (eg, for antibiotic orders). Given that this was a pilot in which we hoped to learn how staff engaged with the system, no specific expectations or benchmarks for the use of the system were communicated by the research team, although site champions were free to encourage use through whatever means they preferred. DialysisConnect remains available to all users; however, there has been no active encouragement of the use of DialysisConnect since May 2021.

Provider participants in the initial focus groups and midpilot and end-of-pilot surveys provided informed consent and were incentivized with meals (focus groups) and nominal (US $10) gift cards (surveys). In this pragmatic study, potential users in the pilot were not incentivized to use the system. The need for consent was waived for the EHR and system data, which were only reported in aggregated form. The entire study protocol was approved by the Emory Institutional Review Board (IRB00102971).

An overview of the features of the pilot version of DialysisConnect is presented in Figure 1. The home page included an overall hospitalization report, listings of patients who were hospitalized at the time of the study and were previously hospitalized, and a message center (Multimedia Appendix 3). The admission feature allowed hospital providers to search for and select patients and review medical information about the patients and identify reasons for hospitalization; instant automated messages informed dialysis providers of admissions (Multimedia Appendix 3). The message feature allowed providers to view automated messages (at admission, discharge, and document upload), send and view user-initiated messages, check read status for messages, and request and upload documents (Multimedia Appendix 3). To prompt log-ins, users also received automated messages externally, by email (default) or SMS text messages, with links to the system. Hospital providers could use the discharge feature to complete a brief, instantaneously delivered discharge report, including discharge date, status, and diagnosis; antibiotics to order; medication changes; and dialysis prescription or dry weight changes (Multimedia Appendix 3). The system was flexible and allowed modifications during the pilot based on user requests or inputs (Textbox 1).

In this study, we developed and piloted a flexible web-based communication portal between providers at dialysis facilities and hospitals, the design and content of which were driven by potential users in both settings, suggesting that DialysisConnect is a potentially feasible solution to the gap in communication between dialysis facilities and hospitals. Although most of the potential users with access to the system were physicians, APPs in both settings and dialysis nurses were the most active users during the pilot. Our data also suggest that DialysisConnect is a usable solution. In midpilot and end-of-pilot surveys, both users and nonusers of the system rated DialysisConnect positively, and users also stated that the system was easy to use and did not disrupt their workflow. Survey participants identified multiple motivators for using DialysisConnect, predominantly the potential of the system to reduce time and energy spent in gathering information from other settings. However, future work in this area would need to address several reported barriers to the use of DialysisConnect, including a lack of time in addition to usual clinical duties, perceived lack of a system champion, and insufficient training and documentation.

Provider perceptions of lack of time in addition to usual clinical duties was an expected barrier in this pilot, given the new, untested nature of the system and the general challenges of affecting provider behavior change. Including multiple, active, and contextually relevant behavior change strategies, such as continuing education incentives, might help overcome this barrier [20,21]. In fact, we found that the barrier of perceived duplicated work effort was no longer identified as a barrier by participants at the end of the pilot study, suggesting that users may have dropped processes made obsolete by DialysisConnect (eg, phone calls).

Providers also commonly reported that a lack of training and supporting documentation was a barrier to using DialysisConnect. However, our research team provided multiple group and individual training sessions and detailed user guides for potential users in both settings at the start of the pilot, as well as additional, as-needed group or individual training sessions and updated user guides with every DialysisConnect user communication. The perception of lack of training and documentation despite these efforts may have been partially because of COVID-19–related disruptions; the research team was unable to provide formal and informal in-person training, as originally planned, and many potential users may not have had web-based meetings and email communications in their usual workflow and, thus, may have missed training opportunities. Importantly, the perception of lack of training is also likely partially because of another commonly reported barrier: the reported lack of support by management and key people driving DialysisConnect forward. This lack of support may have driven poor attendance in the web-based group training sessions.

In addition to the abovementioned barriers, several potential users noted in the surveys that the lack of integration, or even an icon to enter DialysisConnect from the desktop rather than navigating to the website, was a barrier. The lack of EHR integration also meant that automated information on patients had to be uploaded via patient census files that were provided by the dialysis facilities (but not hospitals), causing some lags in the information, particularly for new patients. The feasibility of including this information for other dialysis facilities or hospitals in future implementation would depend on the willingness of health care administration and information technology to provide such files or facilitate EHR integration of the system. However, the lack of EHR integration provided some advantages as well, as the Health Insurance Portability and Accountability Act–compliant system was web-based and, therefore, EHR-agnostic. It was implementable in both settings with fewer institution-specific, often burdensome information technology compliance requirements than EHR-integrated systems. Thus, we were able to offer the system to potential users across both settings and make changes to the system in real time.

Despite these barriers, 81.3% (296/364) of hospital admissions were captured over the course of the pilot study. Furthermore, this capture rate was accomplished by a single APP in the hospital setting and suggests that fidelity to the intervention can be high, even with few users: overall, a small number of users, who were primarily APPs and nurses, were responsible for most DialysisConnect activities. APPs were far more active than physicians at either site, which may reflect the APPs’ greater time available for clinical care, site champion encouragement of APPs specifically, and the existing roles of APPs in institutions as independent physician extenders. In fact, our APPs were already performing most of the intersetting communication at the start of the pilot. Although social workers expressed initial enthusiasm for DialysisConnect in early focus groups [17], none of the social workers, who were given full access to the system and who facilitated critical nonclinical support (eg, transportation and financial assistance) to ensure a successful care transition, logged into the system during the pilot. This may reflect a lack of site champion encouragement of social workers; the perception of DialysisConnect as a platform to exchange medical information only; or particular disengagement of social workers, given pandemic-related increased workloads (and, for dialysis social workers, adjustments because of working from home during our pilot). Other essential care transition roles, such as care coordinators and vascular access team members, were also not represented among the users.

Finally, there is evidence that DialysisConnect could be a sustainable intervention. Most survey participants who were users of DialysisConnect stated that they would continue using the system at the end of the pilot. Furthermore, we found that many hospitalization events continued to be entered into DialysisConnect after the conclusion of the pilot when no communication encouraging its use was being sent by the research team. Site champions who identify and assign users at defined points in the workflow are likely to increase their sustainability. Sustainability is a key attribute for interventions such as DialysisConnect, which would be used in pragmatic settings where research teams are usually not embedded.

The reporting of interventions to reduce readmissions, specifically in the population of patients receiving dialysis, is limited to a few quality improvement projects and is generally focused on effectiveness rather than implementation. For example, the study by Wingard et al [22] reported the results of a pre-post study of the effectiveness of a phased, multicomponent intervention aimed at reducing hospital readmissions among patients at 26 US dialysis facilities (with patients from 18 nonrandomized facilities serving as controls). Readmissions were reduced after the introduction of the intervention (0.88-0.66 per patient-year) in the intervention facilities; however, the decline was not statistically significant compared with the decline in the control facilities (P=.26). Sparse information about implementation was provided, except that they found that only 42% of the patients were successfully contacted by assigned case managers within 30 days of discharge for the posthospitalization call or visit intervention component. Similarly, in another pre-post quality improvement study with a single-component intervention (postdischarge telephone follow-up) aimed at reducing readmissions in a dialysis population admitted to a single hospital, readmissions were reduced (from 28.4% to 24.6% in the 3-month project). However, in an audit, investigators found that only 71% of visits were assigned to a manager; of these, only 80% recorded a patient call attempt, and of these, 38% of patients completed the interview [23]. Although our intervention did not include a direct patient contact component, these prior results highlight the challenges for staff, particularly nurses, in dedicating time to such tasks, even when the tasks are supported by health care administration and assigned as part of the clinical workflow. This difficulty is exacerbated by inadequate dialysis nurse staffing levels, which have been shown to be predictive of hospital readmission [24,25].

Limitations other than those noted previously deserve mention. First, the design of DialysisConnect precluded the inclusion of nephrologists who attended Emory Dialysis and EUHM as users in both settings; thus, they were only able to log in as dialysis facility users. Second, the user surveys had low response rates, which may have biased our results in either direction; however, we were able to collect data from users versus nonusers and from both settings. Third, some suggested changes to the system may have increased its use but were very labor intensive for our initial, limited pilot, including adding the ability for dialysis providers to start a hospitalization event when they send the patient to the hospital and incorporate information from the emergency department. Fourth, DialysisConnect was developed without patient or surrogate input and was unlikely to address the sociodemographic factors associated with hospital readmission in previous studies [26,27]. Fifth, system use logs are temporary and potentially helpful information, such as pages visited more frequently, SMS text messages versus email preferences for alerts, and access via mobile versus web browsers, which cannot be determined; for future studies, these and similar systems could be modified to create permanent logs of system use to better inform implementation. Sixth, the number of users trained was difficult to track in the virtual environment, and it was even more difficult to track the level of engagement with training; future studies could include brief surveys about the level of confidence in using the system to inform training efforts. Finally, the COVID-19 pandemic both delayed and shortened our pilot and altered our approach to rollout and training. Although few providers named lack of time because of additional duties related to COVID-19 in our surveys (none by the end of the pilot), it is likely that the pandemic also had some impact on provider behavior and willingness to use DialysisConnect, especially given that a surge of patients receiving dialysis and hospitalized at EUHM for COVID-19 and related complications occurred during our pilot (January 2021 to February 2021).

Future work using DialysisConnect or similar systems could leverage the most commonly reported motivator: the potential to save time over current processes. Testimonials from current users, who were generally satisfied and reported that DialysisConnect not only provided informed care but also saved time, may be helpful in convincing reluctant users to invest initial time to learn the system and incorporate it into work processes. These users, who are peers rather than members of the research team and are current and frequent users of the system, would be ideal system champions and essential for successful implementation. When targeting potential users and champions, investigators should consider that APPs (if available) may be more likely to engage with the system than physicians. Dialysis nurses, who are critical to the postdischarge processes, could be encouraged and potentially incentivized to learn and use the system. In addition, social workers should be engaged directly to ensure that they understand their essential role in care transitions, as well as the ability of the system to inform timely recognition of the need for services. The addition of content highly relevant to social workers (such as transportation needs for follow-up appointments or requests for referrals for mental health services) might also improve engagement. As much as possible, involving hospital or dialysis facility administration to encourage providers to use the system and facilitate the workflow changes required could increase provider engagement. EHR integration, if possible, is ideal; however, the inclusion of automated file feeds where available, the use of checkboxes and drop-down lists to limit typing, and the inclusion of only the most critical elements for admission and discharge considerably minimizes the time spent in the system, which is a message that can be prioritized by site champions.

DialysisConnect is a flexible and adaptable intervention for enhancing care coordination between dialysis facilities and hospitals that demonstrates the feasibility, usability, and sustainability. In addition, this pilot implementation study informs strategies to overcome multiple potential barriers to the use of DialysisConnect and similar interventions. Future work in this area should consider appropriate system champions, identification of key users and tasks to be incorporated into the workflow, and site-specific methods to encourage provider behavior changes.