Table 1.

Trials in which conditions were potentially met for negative spillover*

Study	1. Resource shared across study arms	2. Scarce resource present in study setting	3. Study intervention affects resource use	4. Study outcome affected by resource availability
Randomized trial of patient navigation for urban minority women with abnormal mammograms [22]	Patients individually randomized to patient navigation or control groups, which received care at the same breast clinic	Resource: Specialist appointments Concerns over appointment availability and wait times for diagnostic services following abnormal cancer screening results are common [19, 23]	Navigators helped patients schedule appointments. The navigation group obtained care leading to diagnostic resolution earlier than control patients	Outcomes: Time to diagnosis, patient anxiety, patient satisfaction Increased demand for specialist appointments has been linked to increased wait times for cancer screening procedures [23]. Scheduling difficulties can delay breast and colorectal cancer diagnosis [24, 25]
Randomized trial of patient navigation in an underserved population [11]	Patients individually randomized to patient navigation or control groups, which received care in the same health system	Resource: Specialist appointments Concerns over appointment availability and wait times for diagnostic services following abnormal cancer screening results are common [19, 23]	Navigators helped patients schedule appointments. The navigation group obtained care leading to diagnostic resolution earlier than control patients	Outcome: Time to resolution of an abnormal screening finding Increased demand for specialist appointments has been linked to increased wait times for cancer screening procedures [23]. Scheduling difficulties can delay breast and colorectal cancer diagnosis [24, 25]
Cluster randomized trial of EHR trigger-based reminder system to reduce diagnostic delays [26]	Primary care providers randomized to intervention or control groups. Their patients received care in the same health systems	Resource: Specialist appointments Concerns over appointment availability and wait times for diagnostic services following abnormal cancer screening results are common [19, 23]	Intervention involved reminder calls to primary care providers to schedule patients for diagnostic services. The navigation group obtained care leading to diagnostic resolution earlier than control patients	Outcome: Time to diagnostic evaluation Increased demand for specialist appointments has been linked to increased wait times for cancer screening procedures [23]. Scheduling difficulties can delay breast and colorectal cancer diagnosis [24, 25]
Randomized Trial of Induction Versus Expectant Management (ARRIVE) [27]	Patients individually randomized to induction or expectant management groups, which received care in the same hospitals	Resource: Clinicians; labor and delivery rooms Delivery rooms unavailable for a few trial participants. Other studies of perinatal inpatient settings report lower than recommended nurse staffing levels [28] and that some nurses are routinely “overwhelmed” [29]	Elective induction group spent 45% longer in labor and delivery but had fewer cesarean deliveries and shorter postpartum stay Some inductions rescheduled when rooms were unavailable	Outcomes: “Composite of perinatal death or severe neonatal complications,” cesarean delivery, duration of stay Other studies report adverse birth outcomes increase on days with more patients and fewer clinicians [30] and that higher workloads may lead to higher risks to patients [16] or lower rates of other procedures [31]
Randomized trial of machine learning-based sepsis prediction algorithm [32]	Patients individually randomized to usual sepsis detection system or usual system plus alerts from an additional sepsis prediction algorithm. Patients received care in the same ICUs	Resource: Clinician time and attention Concerns over clinician-to-patient ratios, staffing shortages, and care rationing in ICU settings are common [15, 33]. Clinicians’ lack of time is cited as a barrier to providing patient care and following infection-control procedures in the ICU [34, 35]	Patients monitored by the added physiological alarm system received earlier tests and treatment for sepsis No information provided on number of added alarms for intervention patients. Other studies have reported that responding to alarms adds to clinician workload and can disrupt patient care [36, 37]	Outcomes: Duration of ICU stay, in-hospital mortality ICU nurse availability, in terms of workforce size and workload, has been linked to patient mortality [21]. Exposure to excessive numbers of alarms can overload clinicians and lead to missed alarms and potentially patient deaths [38]
Randomized trial of artificial intelligence-enabled electrocardiographic alarm system (AI-ECG) [39]	In-hospital patients individually randomized to usual care or monitoring by AI-ECG alarm system. Patients received care from same clinicians and hospital units	Resource: Clinician time and attention; ICU beds Concerns over clinician-to-patient ratios and staffing shortages in hospital and ICU settings are common [15, 21, 33] ICU beds are not always available to meet the needs of all patients [4, 15, 20, 40]	Intervention effect “may be attributed to the increased attention of the enrolled physicians.” AI-ECG system monitored 8001 patients and generated 709 alarms Patients assigned to AI-ECG alarm system were more likely to receive medical interventions or be admitted to the ICU	Outcomes: All-cause mortality, medical care utilization Clinician availability and workload has been linked to in-hospital mortality [18, 21]. Patients who are not admitted to an ICU due to lack of beds face higher mortality risks [20, 40]
Randomized trial of rapid whole-genome sequencing (NSIGHT1) [41]	Critically ill infants were individually randomized to receive standard genetic tests or standard tests plus a new genomic test. All tests were performed in same laboratory	Resource: Laboratory staff and other resources Turnaround of test results took multiple days. Others have raised concerns that staffing in clinical laboratories is sometimes inadequate [42]	The intervention consisted of performing a new genomic test in addition to standard tests, which increases laboratory workload NICU genetic sequencing and interpretation can take multiple hours or days of laboratory staff time [43, 44]	Outcome: Number of infants receiving genetic diagnosis within 28 days Turnaround time for standard and additional genomic tests directly affects rate of diagnosis within 28 days. Concerns have been raised that inadequate staff resources can affect turnaround time in clinical laboratories [42]. Added workload may lead to delayed test results for all patients
Randomized trial of text-messages to encourage uptake of COVID-19 vaccination [45]	Patients invited for COVID-19 vaccination were randomized to receive text-message reminders or no reminders. Patients scheduled vaccine appointments in the same health system	Resource: Vaccine doses and appointment slots The trial began January 2021. The number of invited patients was limited to ensure sufficient vaccine supply, but appointment availability was not reported. Other vaccination programs in the same county and university system reported limited appointment availability for eligible adults at this time [46, 47]	Patients who received text-message reminders scheduled vaccination appointments earlier and at a higher rate	Outcome: Number of participants scheduling a vaccination appointment within 6 days Availability of vaccination appointments at convenient times could have affected whether an individual scheduled an appointment

*Most trials provided limited or no information on resource availability, utilization, or effect on outcomes. As a result, I use evidence from other studies in similar settings to discuss the potential role that a scarce resource may have played during the trial