Simultaneous Recumbent Cycling and Cognitive Training

Investigators will conduct a 2-year RCT, single-masked (participants), pilot feasibility and acceptability study, with repeated measures to compare SRCCT to usual care on cognition and QOL in MA, ICU survivors who experienced at least one episode of delirium in an ICU and have mild to moderate cognitive impairment. The SRCCT will include 12 weeks of simultaneous recumbent bicycling and cognitive training delivered at the same time in a rehabilitation center followed by a 6-month maintenance phase when no intervention will be delivered. Cognitive training will be administered via the POSIT Science (TM) Brain HQ computer-based program. At 12, 24, and 36 weeks post-baseline, cognition will be measured using Montreal Cognitive Assessment (MoCA) and PROMIS Cognitive Function Abilities SF8a (PRO Cog), and QOL will be assessed using PROMIS Global-10 Health Short Form (PGH-10).

Setting and Study Population: The study will take place in two large, urban health care systems in Kansas City, Missouri: St. Luke's Hospital (SLH) and Truman Medical Center (TMC). St. Luke's Hospital is located in downtown Kansas City with 600 staff physicians practicing in 60 specialty areas. This centrally located hospital has 77 ICU beds. Truman Medical Center is the safety net hospital for the Kansas City area and has 41 ICU beds. The rehabilitation centers employ healthcare staff who work with outpatients who are deconditioned and recovering from hospitalization. Data provided by SLH and TMC indicate an adequate pool for study recruitment. In 2018, 87 MA patients were admitted to TMC ICUs; 1366 were admitted to SLH ICUs. Combined ICU admissions age 45-64 were 1453 and 12% (174) experienced delirium (using CAM-ICU). Given 79% (138) of patients who have ICU delirium experience cognitive impairment and recruitment rate of 75%, there are 103 ICU survivors possible and therefore the recruitment goal of 50 participants for this study is achievable. Fifty patients meeting inclusion criteria will be recruited and enrolled.The final patient sample will comprise 50 MA ICU survivors who have had at least one episode of delirium in the ICU and have mild to moderate cognitive impairment (using MoCA). Registered Nurses (RN) from SLH and TM CICUs will identify and HIPAA-consent patients who meet study inclusion and exclusion criteria. Once consented, the RN will notify the study Research Assistant (RA). After discharge from the ICU to a hospital room, the RA will meet with the patient to explain the study and obtain consent. Once consented, the RA will contact the PI's who will determine random assignment to either treatment (SRCCT) or usual care. Block randomization will be created using a computer-generated list of random numbers blocked into balanced groups of four. Before recruitment begins, the RAs will practice the consent process in simulated situations with PIs. Rapport will be maintained by pairing RAs and participants for the duration of the intervention thereby enhancing retention. Investigators anticipate low attrition but recognize participants may withdraw 1) on the advice of their healthcare provider or for other health reasons, 2) due to changes in life circumstances, or 3) loss of interest in the study. In such cases, the RAs will document the event and notify the PIs who will track all withdrawals. Patients who are randomized to the SRCCT intervention treatment group will be screened and consented after transition from ICU into their hospital room. The SRCCT will begin approximately two weeks after discharge to home. The intervention was developed by Dr. Lasiter (Co-PI and Critical Care nurse), Dr. Chrisman (Co-PI and physical activity interventionist), and Dr. Russell (Co-I and feasibility/acceptability/RCT expert) and will be administered by Drs. Lasiter and Chrisman, and two trained RAs. A rehabilitation center RN will monitor the participant's condition while engaged in the SRCCT or usual care. Dr. Lasiter and Dr. Chrisman will oversee the intervention protocol. Research assistants who are baccalaureate and master's-prepared RNs experienced in RCT studies will undergo training by Drs. Lasiter, Chrisman, and Russell, including the following components: 1) project rationale and overview; 2) detailed information on the SRCCT intervention and usual care; 3) scripts for use when consenting and collecting data; 4) how to use Research Electronic Data Capture (REDCap) database (redcap.umkc.edu); 5) detailed training on administering study instruments (MoCA requires training before permission is granted to use the instrument), and 6) using a tablet. Simulation and role-play will be used until the RAs are consistently applying the protocol as judged by investigators. To ensure the highest level of RA protocol knowledge and skills, the training sessions will also include role-playing of disruptive situations for the intervention. To monitor RA intervention fidelity, a fidelity protocol checklist will be used during all participant encounters to document key elements of the protocol, including number of intervention sessions, session duration, length of time between sessions, and intervention steps. Each element will be rated as completed, partially completed, not completed, or N/A. Field notes will be kept for every encounter. Participants randomized into the usual care arm will receive scheduled follow-up office appointments with their provider post-discharge at which time a member of the study team will administer the scheduled study measures at specific time points. The usual care group will not receive SRCCT.

Data analyses: Drs. Lasiter, Chrisman and Staggs (biostatistician) will conduct analyses. Feasibility will be tracked using checklists for 1) educating nursing staff about inclusion criteria and HIPAA consent, 2) educating RAs about interpersonal communication, recruitment, consent, intervention procedures and fidelity and, 3) guiding the intervention to insure fidelity between sites and study staff/investigators. Acceptability will be evaluated by assessing four domains: participant satisfaction, participant preferences, participant burden, and suggestions for improvement. Limitations: One limitation is that the study will not include exercise-only or cognitive-only training groups. Literature supports that each is effective independently and, rather than isolate their effects, investigators will examine combined effects which has the potential to be synergistic in efficacy. Potential difficulty may be recruitment; if recruitment difficulty occurs, investigators will consult with the study team and study site contacts to potentially widen the scope of recruiting participants, add a third study site, or hire an RA who closely matches the ages of the study sample. Another potential difficulty is low retention, and/or patient mortality. To minimize attrition, an honoraria will be provided for each data collection time point. Although mortality in MA ICU survivors is higher than in the general MA population, investigators do not expect mortality to have a major effect on retention in this 36-week study.

Data Analyses Plan: Descriptive statistics (e.g., standardized mean difference) will be used to compare the intervention and control groups on selected patient characteristics and adjust for meaningful imbalances on any variables by including them as covariates in statistical modeling. In the focal analysis investigators will assess the effect of the SRCCT intervention on each of the three Aim 1 outcome measures at T3 by fitting three ANCOVA models with the T3 measurement as the dependent variable, the baseline measurement (T2) as a covariate, and group (intervention vs. control) as an explanatory variable. In secondary analyses the intervention effect at T4 and T5 will be assessed analogously. If the usual linear modeling assumption of Normal residuals is not met, investigators will either transform the dependent variable, fit an appropriate generalized linear model, or obtain confidence limits and p-values by non-parametric bootstrapping. Power and precision of estimation for the ANCOVA model described above were assessed using a simulation study. For each of 1,000 simulated data sets, investigators randomly generated Normal T2 scores and then used these to generate Normal T3 scores, assuming a strong intervention effect of 0.5 SDs and a correlation of 0.7 between T2 and T3 scores within each arm (implying roughly 50% of variability in T3 scores can be explained by T2 scores). In keeping with the exploratory nature of this aim, power was estimated to detect a non-zero intervention effect with a one-sided test at α = 0.10. The average length of the 90% confidence interval for the intervention effect across the simulated data sets was computed. Estimated power was 86%, and confidence intervals averaged 0.68 SDs in length (equivalent to an average margin of error of 0.34 SDs). Actual power will depend on the true intervention effect size and true correlation between T2 and T3 scores, both of which are unknown; however, in this preliminary effort to demonstrate efficacy the focus will be on estimating the intervention effect size, not making a yes/no decision in a null hypothesis test based on an arbitrary α cutoff. Analyses will be run using an intent-to-treat approach, with missing data addressed either by multiple imputation or nonresponse weighting.