Evaluating spoke facilitation costs of implementing TelePain in the Veterans Health Administration

Abstract

Background

The high prevalence and significant morbidity and mortality associated with chronic pain among veterans has made expansion of pain services within the Veterans Health Administration (VHA) a key priority. TelePain, in which services are provided via telehealth from central “hub” sites to patients at decentralized “spoke” sites, is one such model with positive implementation findings to date. However, the staff effort and cost of implementation have yet to be examined when considering TelePain or similar virtual hub-and spoke models of specialty pain care, information that is critical for expansion of services.

Methods

Using an established time-based activity tracker designed for implementation facilitation, study staff tracked minutes spent on implementation activities at 11 spoke sites. Annual salaries were extracted to calculate an average cost per minute for each personnel type. Costs per personnel minute were combined with activity data to calculate costs of implementation activities at spoke sites. Implementation reach outcomes for the first 36 months of implementation were extracted from the electronic health record. Service utilization data was combined with cost data to calculate cost per patient reached and per clinical encounter achieved at each site.

Results

Total facilitation costs (range: $1,746-$7,978) and unique patients reached (range: 2–46) varied considerably across spoke sites and greater staff implementation efforts (measured in time or cost) were not associated with greater numbers of patients reached. Therefore, costs per patient reached also varied widely (range: $120–2,569) across spoke sites. Key challenges included high rurality and small clinic size; insufficient engagement of frontline providers; lack of referral options for high acuity patients; and lack of existing programming within which to situate pain services.

Conclusions

At spoke sites where patients were consistently referred and reached, costs were relatively modest, particularly compared to the high cost of untreated chronic pain, suggesting the potential impact of this model at scale. However, in spoke sites where referrals and encounters were low during initial implementation, cost per patient was high. Findings highlight the need for better methods for tailoring of facilitation interventions to spoke site’s individual needs to maximize impact.

Contributions to the literature

• Expansion of chronic pain services is much needed among the US adult population, particularly among veterans.

• Cost is an understudied implementation outcome, so here we evaluate costs of implementing TelePain, a Veterans Health Administration virtual care model for expanding access to pain services.

• Costs per patients reached varied widely across spokes and were high at sites with few patients reached.

• Implementation barriers can inform much needed future implementation research and underscore the need to effectively tailor facilitation interventions to site needs

Background

Pain represents one of the most widespread chronic conditions among adults in the United States (US) [1]. Within the general US adult population, the prevalence of chronic pain is estimated at 20.5% [2]. Prevalence rates are substantially higher among military veterans, ranging from 30 to 51% in a recent systematic review and meta-analysis of the general veteran population, and 81.5% in the population of veterans who served in combat settings [3]. In addition to being common, chronic pain is associated with higher morbidity and mortality across all populations. Among veterans, chronic pain is associated with elevated rates of opioid misuse, alcohol misuse, and polysubstance use [4]. In a study of over 30,000 veterans who have attempted suicide, chronic pain conditions were also associated with increased odds of suicide attempt, even after adjusting for mental health and medical comorbidities [5]. Further, chronic pain is associated with higher service utilization and healthcare costs among veterans [6, 7]. For all these reasons, strengthening management of chronic pain and expanding chronic pain services is a key priority within the Veterans Health Administration (VHA) [8, 9].

Unfortunately, comprehensive chronic pain services, like many other specialty services, have historically been limited to large hospitals and academic medical centers within the VHA and similar health systems. Staffing challenges impact pain services; sixty percent of pain management teams across the VHA reported being not staffed or only partially staffed in a 2022 analysis [10]. Access to pain services in rural areas where veterans are served primarily by smaller community-based outpatient clinics (CBOCs) is particularly limited [11]. In addition to geographic and staffing challenges, over the past 15 years, pain services at the VHA, and in other US health systems, have undergone a shift away from physician-led, opioid-focused approaches and towards multidisciplinary, nonpharmacologic approaches that must involve a wider range of providers [12].

Such access, staffing, and opioid safety issues for pain care are a key concern of the VHA and have led to the development of programs aimed at expanding access to nonpharmacologic treatment options [13]. Models can involve expanding training for providers at rural sites or recruiting centralized providers to move to these more rural sites to deliver care [14]. The objective of this manuscript is to evaluate the implementation strategy costs of TelePain, a model that trains generalist providers at rural sites to identify and refer veterans to telehealth programs in which veterans are digitally connected with and treated remotely by specialty providers in more centralized locations (i.e., a hub-and-spoke model [15]). TelePain was developed in a VHA Health Care System in a major West Coast city and covers rural areas of four nearby states [16]. Promising outcomes for this service model have been reported across several areas, including patient reported ease of accessing services, an increase in documented delivery of pain services, and positive provider perceptions of the program [17,18,19,20]. However, better understanding expenditure, in both time and money, needed to achieve good program outcomes will be critical before it can be scaled up and help close the current treatment gap for nonpharmacologic pain services.

In general, cost as an implementation outcome remains under-represented in the implementation research literature [21, 22]. Though there are several published protocols stating authors intend to look at the cost of implementation strategies aiming to expand access to pain services, actual results are extremely limited to date [23, 24]. Understanding the time and cost impact of implementation initiatives on rural sites is important because many initiatives to increase access to care run the risk of increasing demands on clinicians and other frontline workers who must participate in implementation activities [21]. Therefore, this paper helps close the knowledge gap on the cost of a specific implementation package (facilitation) used to expand access to pain services within the VHA that also has important implications for other settings. We specifically focus on costs and challenges at rural spoke sites to identify lessons most relevant to dissemination in lower resource settings, either within the VHA or other health systems, that may deploy a hub-and-spoke approach [25, 26].

Methods

Methods are reported in alignment with the Standards for Reporting Implementation Studies (STARi) checklist (please see Additional File 1) [27].

Study setting

Details about the TelePain hub-and-spoke model [16, 19] and its implementation in the region [28] have been described elsewhere; the current micro-cost analysis study focuses only on time and cost of staff associated with facilitation. Briefly, the TelePain model of care consists of an interdisciplinary pain management team (the “hub”) that offers multimodal, biopsychosocial comprehensive services for chronic pain via telehealth, typically synchronous video visits, to “spoke” sites located in rural areas in the same state or region.

The present project focuses on the initial rollout of TelePain from an interdisciplinary hub located across a large VHA healthcare system in a major West Coast city to spoke sites in rural areas of two additional states within the same Veterans Integrated Services Network (VISN). The hub site had a long history of offering interdisciplinary pain care, although the expansion to telehealth and the hub-and-spoke model was new. The spoke sites were selected based on the need for pain services that could not be met by their local VA clinic. Eleven spoke sites, made up of rural CBOCs (enumerated here as Spoke 1–11) were organized administratively by three smaller healthcare systems (System A; System B; and System C). The 11 spoke sites served patient populations ranging from 1,600 to 9,300 patients (mean = 5,282). For 7 out of the 11 clinics, 98–99% of the patient population were classified as living in rural or highly rural areas; of the remaining 4 clinics, 2 were located in urban/metropolitan areas and 2 had patient populations of 30–40% rurality. Using the VHA classification system for facility complexity, which considers the complexity of the clinical services offered and the level of education and research conducted at the facility, all 11 clinics were classified as the lowest level of complexity (Level 3; [29]). Level 3 facilities offer primary care services and may have circumscribed, brief mental health services, but specialty medicine is usually nonexistent or very limited.

A facilitation team consisting of an external and an internal facilitator based at the hub site supported TelePain implementation across spoke sites from August 2019 to April 2020 (approximately 9 months). The external facilitator was a master’s-level research scientist with training in community-oriented public health practice and community organizing. Their role was to assist in identifying implementation barriers and facilitators, provide recommendations on courses of action, build relationships with stakeholders, monitor implementation progress, and assist with aspects of project management. The internal facilitator was a doctoral-level clinician in the TelePain hub who brought clinical and health system contextual knowledge; their role was to work with the external facilitator on all implementation tasks, and they received protected time to do so. Facilitation was selected as an a priori implementation strategy because of its evidence for supporting implementation of similar team-based innovations in VHA [30].

Activity tracking

Each spoke hospital or clinic was considered its own site for activity tracking purposes, though outcomes are also aggregated at the system level for reporting. We utilized an established format and definitions for time-based activity tracking in implementation facilitation studies [31, 32]. During the pre-implementation (August– November 2019) and implementation phases (November 2019 – April 2020), implementation facilitators used Clockify, a time tracking app on their smartphones to record time expenditure, in minutes, spent on each facilitation activity that day.

Facilitated implementation activities were operationalized as the following using Ritchie et al.’s pre-established definitions: assessment; preparation and planning; education; stakeholder engagement; program adaptation; problem identification and problem solving; data audit and feedback; program marketing; network development; and other [31, 32]. Please see Additional File 2 for definitions.

Additionally, facilitators recorded the type and number of staff involved in each activity at that spoke site. The time that each facilitator spent with spoke site personnel was captured as part of activity tracking; however, for practical purposes, spoke site personnel were not asked to track additional time spent on implementation activities outside of meetings with the facilitation team. Spoke site personnel roles were categorized in the following ways: licensed independent practitioners; nurses; administrative clerks; telehealth clinical technician; facility telehealth coordinators; clinic supervisors; service level leadership; internal facilitator; external facilitator; other (e.g., research staff). Though the internal and external facilitators were based at the hub site, their efforts aside from data collection were also included in time tracking at each site when they were working with that site to capture variation across spoke sites. This is the only cost from staff at the hub site that we include in time and cost analyses as the focus of the analysis is the time demands and costs of implementation for lower resource spoke sites rather than the costs of setting up the hub clinical program. We did not include costs unique to conducting an implementation study (e.g., data collection) that would not generalize to other hub-and-spoke models developed outside of research studies. Lastly, we did not track time expenditure for intervention delivery at either the hub or spokes, as our emphasis was on implementation.

Cost analyses

Our focus is limited to facilitation and spoke site staff costs and our goal was to calculate a cost per unique patient reached via a TelePain encounter (total cost per site divided by number of unique patients seen by TelePain at that site) and cost per TelePain visit (total cost per site divided by number of TelePain encounters at that site) for each site. These outcomes were chosen both based on concept mapping by the project team prior to implementation as well as in alignment with the costing literature [33, 34]. The RE-AIM framework was used to define implementation outcomes that the team expected to be associated with facilitation, as it is a commonly used framework in VA implementation and “reach” in particular is often identified as a key target for improvement by funders and stakeholders [35,36,37,38].

First, an average annual salary cost was identified for each role. Research team members documented the actual names and/or titles of staff within each staff role who were involved in implementation activities in a separate tracker at the time of activity tracking. To derive average annual salary costs, salaries were then extracted from public government databases using names, or from Veterans Affairs pay tables using titles. A flat fringe rate of 35% was then added to all individual salaries. Within each role, the average across individuals was then taken to arrive at an annual average salary. Annual costs were then transformed to cost-per-minute by dividing by the number of working hours per year and then the number of minutes in an hour. Please see Additional File 3 for the cost per minute for each role in each implementation year.

Next, minutes spent on each activity on each day at a site, as documented during activity tracking, were multiplied by the number and cost per minute of all staff involved in that activity on that day at that site. Then, values were summed across days to arrive at a cost for that activity across roles and across days within that site within that year. Lastly, costs across activities and across both implementation years were summed to arrive at a total cost for that site. Formal comparisons of costs using statistical tests were not conducted between spoke sites due to the range of sizes in spoke sites.

Extraction of implementation outcomes

Information on TelePain consults and days with encounters at implementing sites was extracted from the VA Corporate Data Warehouse (CDW). To be included, consults and encounter days needed to have taken place between November 2019 and November 2022 (from the start of implementation to the end of this iteration of the TelePain service). A new electronic health record system was introduced at two spoke sites from June 2022 onwards, but these data were not included as that system was not fully operational at the time. Consults were identified by the inclusion of keywords in their names (e.g. “TelePain,” “tele pain”), while encounter days were identified by combination of relevant primary and secondary stop codes, which indicate delivery of specific clinical services within VHA, and then consolidated by the date of visit. We chose to use encounter days, as opposed to individual encounters, because this allowed us to eliminate duplicated encounters that often occur in the CDW and ensure we were not inflating the volume of care rendered. Qualifying encounters included a “stop code” (VHA service identifier) indicating services rendered within a pain clinic (typically the primary stop code), and another stop code indicating telehealth service delivery (e.g., telephone visit, synchronous video telehealth received at home, synchronous video telehealth received at spoke site). Consults needed to have originated from a TelePain-implementing site. Encounter days were included if they could be linked to a patient who 1) also received a TelePain consult, and 2) had, in the two years prior to their TelePain encounter day(s), been seen most for their primary care in a TelePain-implementing site. “Seen most” was defined as having the largest number of primary care encounter days at that site, based upon observing encounters with specific primary care stop codes. In instances where patients were seen equally frequently across two TelePain-implementing sites, the most recent visit day was used as a tiebreaker. Because we limited qualifying encounter days to patients with an observed TelePain consult, it is possible that not all TelePain encounter days during this time period were captured (e.g., it is hypothetically possible for patients to enter into care via another pathway besides consults, though unlikely). If so, our cost per outcome estimates would be slightly inflated.

Methods for relating costs and outcomes

With complete time and cost data and data on consults and encounters at each site, a cost per unique patients reached via a TelePain encounter and cost per TelePain visit was calculated for each site. We also examined the correlation between total implementation costs at each site and number of unique patients reached.

Results

Please see Table 1 for the costs of each implementation activity and summary costs for implementation activities at each site during the period under study. Table 2 presents time associated with each activity. Costs ranged from $1,746 to $7,978 across spoke sites. The activity associated with the highest average cost was preparation and planning ($1,388). Activities associated with the lowest costs were data auditing and network development, as many sites were recorded as allocating no time to those activities. Sites with the most time spent on preparation and planning had the highest total costs, though higher cost sites also had more time devoted to ongoing processing monitoring, program adaptation, and problem identification and solving, and program marketing. Please see Table 3 for facilitator contacts with each spoke site in support of implementation activities. Facilitation support ranged from 62 to 146 contacts with each spoke.

Table 1 Facilitator and spoke staff costs (in USD) for implementation activities at each site

Table 2 Facilitator and spoke staff time (in minutes) for implementation activities at each site

Table 3 Facilitator support provided to each spoke site

Regarding consults to TelePain during this period, System A had 84 documented consults, System B 123 consults, and System C 51 consults. Across all three systems, the bulk of consults came from one spoke site, always the largest hospital/clinic in that system: Spoke 1, 6, and 9. Please see Table 4 for unique patients reached and encounters provided to each spoke site. The number of unique patients reached at sites ranged from 2 to 46, and the number of encounters at sites ranged from 5 to 449, again with some sites accessing far more services than others. Please see Table 5 for the average cost per unique patient and per encounter day at each site. The lowest cost per patient reached was $120, while the highest was $2,569. Difference in costs per patient reached across spoke sites was driven both by higher facilitation and staff time expenditure on implementation activities at some sites and varying numbers of patients reached across sites. There was no correlation between a site’s overall implementation costs, representative of facilitator and rural staff time spent on implementation activities at a given spoke site, and number of patients reached (r = 0.16).

Table 4 TelePain consults, patients reached, and encounters at each spoke site

Table 5 Costs per services provided (in USD)

Regarding common challenges at higher cost per outcome sites, the three sites with the highest implementation costs per patient and per encounter were small, rural or highly rural clinics where baseline costs for uptake take longer to be distributed across patients due to the restricted size of the patient population. Spoke 9 had the highest total implementation costs and a relatively high cost per patient reached and per encounter. One post-hoc observation from the implementation team was that frontline provider buy-in was not fully addressed during the preparation and planning stage at this spoke, limiting the numbers of patients reached and referred, even though a substantial amount of time was devoted to stakeholder engagement and preparation and planning with leadership. One of the primary concerns at this spoke site was leadership concerns about how opioid prescriptions would be managed by the pain service, and time spent on addressing these concerns drove up facilitation costs without resulting in greater frontline provider engagement.

Discussion

This is among the first studies examining facilitation time and costs related to expansion of access to chronic pain services, providing important information for the VHA and other health systems considering a hub-and-spoke model for pain services. Results demonstrate relatively modest facilitation costs per patient reached at sites with a strong uptake, particularly as compared to the high cost of untreated chronic pain [6, 7], which is promising. However, greater time expenditure for facilitators or staff did not result in better outcomes across sites. This underscores the importance of tailoring facilitation strategies to specific facilities and specific teams even within the same system, iteratively understanding and intervening on implementation barriers at sites with modest or low uptake, and of drawing implementation lessons from sites that do have strong outcomes. Current findings will be helpful not only for decision-makers within the VHA but also help generate important knowledge about the implementation of hub-and-spoke models for pain care more broadly and may also apply to other health systems aiming to expand access to pain services.

Among the smallest and most rural clinics it may be unsurprising that total numbers of patients and encounters were relatively low, given the population sizes of these clinics and known barriers to accessing healthcare, even virtual care, in highly rural areas. Thus, the cost per patient and cost per encounter were high at these sites. There are baseline facilitation and staff time expenditures needed to launch any new program, which may make it difficult to keep implementation costs low at small, remote sites. On the other hand, these are often the sites with the highest need and greatest barriers to care, which is important for decision-makers to weigh as they examine costs and cost per implementation outcome.

Some sites also experienced challenges around the composition of site planning teams. The importance of champions is well documented within the implementation science literature [39,40,41,42,43]. However, even if substantial time is spent on engaging site team members in preparation and planning, as happened at Spoke 9, if the individuals engaged are not the right stakeholders presented with the right information and requests at the right time (e.g., only leadership is engaged rather than frontline providers), efforts to build buy-in will likely fall flat. This speaks to the importance of effective, site-specific situational analysis prior to implementation at each implementation site to guide careful selection of and planning for engagement of champions, a finding not specific to chronic pain services but true of novel program implementation in general. Situational analyses not only allow for the identification of which staff need to be engaged (in what ways and at what stage) but also for implementation team members to adequately understand competing priorities that sites may have [44]. While the assessment and planning activities under facilitation in this implementation package are similar to the activities conducted in a situational analysis—conducting assessments about clinic and organizational context and collaborating with stakeholders—merely spending time on these activities does not guarantee that the correct information was assessed or acted upon. Future research could guide facilitators in how to effectively conduct situational analyses that will best inform downstream implementation outcomes.

In addition to conducting situational analyses to understand context, it is also critical that findings from situational analyses influence program implementation. In this iteration of TelePain, there were challenges in using implementation feedback to change the program itself, which may have impacted uptake. For example, at the time of implementation, patients with a recent positive Columbia Suicide Severity Rating Scale (C-SSRS) [45] were excluded from referral. The primary reason given for this exclusion criteria was clinician concern about how to virtually or remotely manage acute mental health support services in patients' communities. Further, opioid management was not consistently offered at the time due to legal concerns about prescribing controlled substances across state lines and TelePain’s emphasis on expanding nonpharmacologic strategies for pain care. It is possible that in the exclusion of high-risk and more complex patients, such as those with recent elevated suicide risk or who need both nonpharmacologic treatments and opioid medications for pain, concerned spoke site leadership and undermined provider buy-in across spoke sites as these are the exact patients who likely are most in need of specialized services. While facilitators brought this information to program leadership, it was not possible or looked upon favorably at this time to change the program criteria.

It is also important to draw lessons from sites where implementation was successful. Spoke 6, which had the lowest cost per patient reached, had a particularly robust Whole Health program prior to TelePain implementation. Whole Health is a biopsychosocial model of health that situates pain within other domains of health and well-being [46]. At this site, non-specialist providers may have therefore been primed to engage in facilitation activities and more active assessment of a patient’s need for pain services and consider referral to nonpharmacologic pain treatment than at sites without this existing foundation. This example highlights the importance of considering the order in which services are implemented and of leveraging synergies between programs to support implementation.

Overall implementation costs for spoke sites during the 9 months of active facilitation were relatively modest and much lower than facilitation costs seen in some more extensive VA implementation projects [31]. Ritchie et al. conducted a 28-month-long facilitation study in which they were implementing a new, team-based clinical intervention at each site. In contrast, the spoke sites in the present study were tasked only with patient identification and referral to the hub, which provided the clinical service via telehealth. Therefore, these findings also provide some data on how implementation facilitation time and costs vary widely depending on the complexity of the innovation being implemented.

Findings highlight several directions for future research related to cost. In general, our findings underscore the ongoing need for scalable nonpharmacologic pain care models that can accommodate patients with elevated suicidal risk or who also use opioid medications [47], which is the direction TelePain has evolved towards in more recent phases. Better understanding the cost outcomes of extending chronic pain services to these more complex patients would be important in advocating for these services. Of note, a recent VHA analysis found that the cost of expanding novel programming may be higher at sites serving more complex patients even with good implementation [48]. Despite diversity among sites, these data were also all collected within one VISN, which means costs may not generalize to other regions of the VHA. Understanding the costs as they occur across the VHA more broadly would be important before larger scale up of this facilitation package. Lastly, due to our focus on burden on the spokes, we do not include patient, caregiver, productivity or other costs in our calculations, though these would be important to examine in future research.

Study findings should be interpreted considering limitations. First, TelePain is an ongoing care delivery model and findings reported here represent only the first implementation phase. However, examining staff time and cost expenditures during this initial period highlights important findings related to the launch of hub-and-spoke models and pain services. Findings related to service utilization that we report were based on clinical and administrative data and time tracking was done by facilitators rather than site staff themselves; data may naturally have errors or missing data within it and therefore mis-estimate costs per patient at a given site. However, differences between sites are still clearly apparent in results and the use of clinical data for this type of research allows for the best approximation of costs for potential future implementation. Further, we used well-established and clearly operationalized activity tracking tools to maximize the reliability of time-based activity data [30, 31]. We also excluded data from the new electronic health record system rolled out at two sites in the last several months of our follow-up period. However, these systems were not fully implemented at the time, so we anticipate that few instances of care were lost or unobserved. Lastly, the implementation and data collection periods include the early days of the COVID-19 pandemic. However, as the VHA already had a strong telehealth care delivery platform and quickly transitioned routine care online [15] and the TelePain intervention being implemented was itself a telehealth intervention, the impact of the pandemic on results was likely mitigated.

Conclusions

At sites with good TelePain uptake, costs were modest, helping to close the considerable knowledge gap related to the cost of increasing access to pain services. This is promising for both the VHA and other health systems considering a hub-and-spoke model. However, our findings reveal substantial variations in costs and costs per outcome across sites, which related to key implementation challenges at some sites. Such challenges indicate important future directions for research aiming to increase access to chronic pain services, specifically the need for methods for better tailoring of facilitation interventions to individual site needs to ensure less uneven outcomes.