Despite the shared interests that led researchers and healthcare systems to collaborate in the first place, health system leaders’ end-of-trial decisions may be guided by shifting priorities as the trial unfolds. Such decisions can be highly contextual and localized, driven by different sets of conditions across clinical settings where the trial took place.

Timelines for trial results and health system decision-making may diverge, creating conditions in which site-specific circumstances dictate clinical partners’ choice to continue or discontinue the intervention before trial results are known. When this happens, key questions are raised about commitment, resources, relationships, and other kinds of obligations that trial investigators must be ready to address. This chapter offers a framework to guide researchers as they approach end-of-trial decision-making.

To introduce this topic, we present 3 case studies of concluded trials in which the context at each participating site drove health system decisions about the intervention.

Case Studies

All 3 case studies discussed here are NIH Collaboratory Trials.

Lumbar Imaging with Reporting of Epidemiology (LIRE)

The Lumbar Imaging with Reporting of Epidemiology (LIRE) trial evaluated the impact of including benchmark prevalence data in routine spinal imaging reports on subsequent spine-related healthcare utilization and opioid prescriptions. Using a stepped-wedge, cluster randomized design, the trial took place in 98 clinics across four large health care systems. Its primary objective was measuring 12-month spine-related relative value units. Secondary objectives included evaluating subsequent X-sectional imaging and subsequent opioid prescriptions.

This proved to be a relatively straightforward and inexpensive intervention to implement. In the interim period between the end of active intervention and the time when trial results would become available, the 4 participating health systems considered factors that led each one to make a decision regarding maintaining or discontinuing the inclusion of epidemiological data with spinal imaging reports.

In one of the participating systems, the intervention had been in place before the trial. With no new resources needed to continue and no anticipation of potential harms, this system chose to continue the intervention while awaiting trial results. A second participating site made a different choice. In this system, the intervention was not present before the study and new resources were needed to continue. Their decision was to discontinue the intervention while waiting for results. Similarly, a third system was experiencing a transition to a new EHR when the trial ended. In addition, continuing the intervention would have required new resources. Under these circumstances, this system elected to cease the intervention and wait for results to see if intervention benefits justified effort and cost. For the fourth system, cost was not an issue as no additional resources were needed to continue the intervention. The intervention had also garnered positive reviews from clinicians. Considering these factors the system decided to continue the intervention.

Results revealed no reduction in spine-related healthcare utilization in the intervention group, and only a slight reduction in subsequent opioid prescriptions. The diversity of system-level decisions before these outcomes were known underscores how local resources, preferences, and competing needs may have a significant impact on healthcare system leaders’ choices independent of trial results.

Improving Chronic Disease Management with Pieces (ICD-Pieces™)

The Improving Chronic Disease Management with Pieces (ICD-Pieces™) trial aimed to help primary care physicians treat patients with co-existing chronic kidney disease (CKD), type 2 diabetes, and hypertension in more effective ways to reduce hospitalizations, emergency department visits, cardiovascular complications, and deaths. The intervention included an informatics component (to identify people eligible for the intervention and generate registry lists to prompt the intervention) and practice facilitation (to encourage and support treating clinicians to address gaps in care).

This was a cluster randomized trial (randomized by primary care practice) in 4 large health systems. The systems differed greatly in their organizational structures, EHR systems, and patient populations. The primary outcome was 1-year documented hospitalization rates. Secondary outcomes included cumulative incidence for death, and cumulative incidence for dialysis, emergency department visits, 30-day readmissions after the first inpatient hospitalization, cardiovascular procedures, cardiovascular events, and transplantation.

Trial results were not available for almost a year after the intervention ended. Although practice facilitation was observed to support delivery of evidence-based care, all 4 sites chose to discontinue the intervention in the interim period.

These decisions were mostly due to lack of funds to support the practice facilitators. Practice facilitators were local employees who were paid by the trial for its duration but had to be redeployed afterwards. Other factors also played a role in each system. Lack of IT support to maintain the informatics tool was an issue in one system, while loss of interest in the research prevailed at two other systems.

The trial did not achieve its intended outcomes as the intervention did not reduce hospitalization rates compared to usual care. This example illustrates that the question of how personnel essential to delivering the intervention could be funded posttrial can arise and determine health system decisions while data analysis is still underway.

Advance Care Planning: Promoting Effective and Aligned Communication in the Elderly (ACP PEACE)

The Advance Care Planning: Promoting Effective and Aligned Communication in the Elderly (ACP PEACE) trial tested a comprehensive intervention to help patients with advanced cancer and their providers communicate about serious illness care that is concordant with the patient’s wishes, values, and goals. The intervention consisted of an advance care planning program including clinician communication skills training and patient decision aid videos. The stepped-wedge cluster randomized trial involved 29 oncology clinics in 3 large health systems (13,800 patients total) and relied on EHR data to study 4 outcomes: advance care plans completion, medical orders for resuscitation preferences, palliative care consultations, and hospice use.

Contextual factors affected uptake of the intervention during the trial and may influence final health system decisions. Implementation efforts and commitments varied across sites due to organizational buy-in, personnel turnover, and impact of the COVID-19 pandemic on maintaining in-person contacts between researchers and health systems. Initial uptake also varied from high (with integration of the intervention into standard work) to very low (with only partial implementation of the intervention).

While awaiting the results, all intervention materials remain available, though there is no additional trial funding for intervention continuation. Future contact between interested sites and the research team will determine training needs and available funding for new personnel to learn the intervention.

Although different in many respects, these studies illustrate how end-of-trial decision-making is highly context dependent. Trial sites weigh differently the pros and cons of continuing the intervention and reach conclusions driven by local conditions. This potential variability challenges trial investigators to maintain an ongoing dialogue with health system partners to provide needed or wanted input (including potentially helpful implementation data) while upholding rigorous equipoise until results are available.