Among critically ill older adults receiving care in an intensive care unit (ICU), delirium is very common (>70%) and strongly associated with long-term cognitive impairment similar to Alzheimer's Disease and Related Dementias (ADRD). With growing recognition of this major problem, there is an increasing number of randomized controlled trials (RCTs) evaluating preventative and therapeutic interventions for delirium among critically ill older adults. Efforts are underway to improve the scientific rigor of such trials; specifically, a NIH/NIA NIDUS collaborative (R24AG054259) and a Canadian CIHR-funded project that aim to harmonize use of outcome measurement instruments to improve comparability and consistency across delirium RCTs. However, despite these efforts, guidance regarding delirium endpoint definition and appropriate statistical methods for evaluating treatment effects in delirium RCTs is lacking. Defining and evaluating a delirium endpoint is especially challenging for older adults in the ICU as delirium status can vary during the follow-up period (e.g. 28 days) and measurement may be terminated by patient discharge or death, referred to as ?competing events? in the statistical literature. Both discharge and death are correlated with the risk of delirium, but likely in opposite directions, and may be affected by the intervention, further complicating evaluation of delirium endpoints. Thus, our overall objective is to improve the design and statistical analysis of preventative and therapeutic delirium RCTs in critically ill older adults. To achieve this objective, we will conduct a systematic review of delirium endpoint definitions and statistical analysis methods, followed by a rigorous evaluation of their statistical performance (e.g., false positive rate (Type I error) and statistical power), using simulation studies based on real data from 6 diverse RCT exemplars and 2 large clinical cohorts (Aim 1). Joint models, that include survival models for both recurring delirium and a single competing event, have recently been applied in two prominent RCTs of pharmacological interventions for delirium in the ICU (Aim 2). In this proposal, novel extensions of these joint models will be developed that better mimic key features of delirium among critically ill older adults by including a recurrent event survival model for delirium, as well as allowing two survival models, one for each competing event. Thereafter, current and novel extensions of the joint models will be compared to the endpoint definitions and statistical analysis methods identified in the preceding systematic review, using simulation studies (Aim 3). Based on these findings, we will make recommendations for the design and statistical analysis of delirium RCTs, accounting for the possibility that discharge and death may be affected by the study intervention. We will disseminate these novel joint models, in addition to our findings and recommendations, through publications, open source statistical software, a stand-alone statistical application, an established ICU research methodology dissemination infrastructure, and aging, delirium, and statistical professional societies.
There is a growing number of randomized controlled trials (RCTs) evaluating interventions to prevent or treat delirium among patients in the intensive care unit, whom are increasingly older and have significant risk for long-term cognitive impairment, that is similar to Alzheimer's Disease and Related Dementias. This proposal will systematically evaluate delirium endpoint definitions and statistical analysis approaches for detecting a treatment effect by conducting a systematic review of the literature, performing empirical analyses and statistical simulation studies, along with the development of novel extensions to existing advanced statistical methods. Based on this work, we will develop recommendations for delirium endpoint definition and statistical analysis approaches that will minimize the rate of false positives (i.e., Type I error) and maximize the chance of identifying a true treatment effect (i.e., statistical power) in delirium RCTs, with the long-term goal of reducing delirium and long-term cognitive impairment in critically ill older adults.