Andrew Admon, MD, MPH, MSc is a Clinical Lecturer and Research Fellow in the Department of Internal Medicine Division of Pulmonary & Critical Care Medicine, with advanced training in public health and health care research. Dr. Admon has a total of 37 publications and has been supported by both the Michigan Institute for Data Science (MIDAS) and an NIH/NHLBI NRSA F32 award.
Dr. Admon recently received funding approval (and a perfect score of 10) on a resubmitted career development award proposal for the NIH National Heart, Lung, and Blood Institute (NHLBI) Mentored Clinical Scientist Research Career Development (K08) Award. These grant proposals are reviewed by a large panel of clinicians and scientists, and are evaluated based on the merits of the candidate, career development plan, research plan, mentorship team, and academic environment.
The purpose of the NIH K08 Award is to prepare clinically trained individuals for careers that have a significant impact on the health-related research needs of the Nation. In Dr. Admon's words, "this award provides five years of support to early-career faculty, allowing them to pursue deeply specialized scientific training and engage in mentored research with increasing levels of autonomy. Ultimately, this period of training prepares junior faculty for careers as independent clinician-scientists."
In this Clinical Research Spotlight, Dr. Admon talks about the research project he submitted with his K08 award proposal and what he hopes to achieve.
What is your long-term career goal and career development plan?
My long-term career goal is to become an independent investigator who applies advanced observational research methods attentive to causal inference to improve the organization and delivery of hospital care.
This goal is built upon the premise that modern analytical techniques (e.g., those attentive to complex causal and temporal relationships between variables in an analysis) applied to new observational data resources can inform decisions in care delivery where randomized experiments are often difficult or even impossible.
As part of this long-term goal, I hope to both generate knowledge advancing my own scientific interests in hospital and critical care, while also contributing as a team member to work led by other researchers seeking to address scientific questions using observational data.
Motivated by this goal, my K08 proposal included five years’ worth of deep training in epidemiology, biostatistics, computer science, and econometrics, mentored research, and active engagement in selected seminars and lab groups. Together, I expect that this training will allow me to bring expertise in causal inference back to my own work and to the work of others with whom I collaborate.
What clinical problem are you attempting to address?
Acute respiratory failure leads to 2,000,000 hospitalizations and 400,000 deaths in the U.S. each year. Outcomes after respiratory failure vary widely across hospitals, suggesting that improvements to how care is organized and delivered at some hospitals may improve survival. Yet, an increasingly common feature of hospital care that may be working against patients with respiratory failure is care team fragmentation - discontinuous care spread out across multiple clinicians during a hospitalization. Over the past few decades, broad changes to how we organize and deliver hospital care have encouraged more frequent turnover among inpatient teams, contributing to increasingly fragmented care. While these changes have typically been made with specific positive goals in mind, we’ve not yet evaluated their potentially unintended consequences and what to do about them. Care team fragmentation is also a perfect problem through which to gain new experience applying advanced methods for observational causal inference.
What are the specific aims of your project?
Using methods for observational causal inference applied to new data sources, I plan to:
- compare the causal effects of fragmented care on in-hospital and post-hospital respiratory failure outcomes
- determine whether specific complications occur more often in fragmented care and contribute to poorer outcomes
- test whether specific features of inpatient care are protective against harm when care is fragmented
What do you hope to achieve with this grant?
Care team fragmentation will likely be an increasingly common aspect of hospital care going forward. This research will allow us to identify its potentially unintended consequences and inform strategies to mitigate them. At the same time, the training afforded by this grant will allow me to gain advanced skills in observational research methods attentive to causal inference. In line with my long-term career goal, this expertise will allow me to use novel observational data to address other research challenges in health care delivery where randomized trials are difficult or impossible.
What do you mean by “causal inference”?
'Causal inference’ means thinking about whether and how an action (like taking a medication, performing a procedure, or changing a policy) leads to a consequence (like better disease control, improved survival, or fewer rehospitalizations). Studies employing causal inference seek to establish a few things (we usually combine them, but it can help to split them up for purposes of explanation):
- Causation - E.g., “if I take this drug at this dose for one year, It will cause me to have improved control of my COPD compared to if I do not take this drug.”
- Effect Estimates - E.g., “if I take this drug at this dose for one year, I will enjoy 30 more days out of the hospital and at home compared to if I do not take this drug.”
Through the use of research methods aimed at casual inference, researchers hope to provide accurate information to guide patient, clinician, and policymaker decisions.
A common (and effective) way of establishing causation and effect estimates in medicine is with experiments, like randomized clinical trials. However, randomized clinical trials are expensive and can take years to complete. As a result, we are limited in what we can study at a given time. And, for some important interventions, like transferring a patient with respiratory failure to another hospital, merging two health systems together, or rearranging clinician schedules to promote patient continuity, experiments are very difficult or impossible.
My goal is to apply modern analytical methods to increasingly available observational data to generate causal inferences when clinical trials are not available, are difficult, or are impossible.
Like causal inference using experiments, I hope to do this to provide the best possible (i.e., most accurate) information to guide patient, clinician, and policymaker decisions. Additionally, I hope to help prioritize interventions and guide study designs for later randomized experiments, increasing the likelihood that these experiments provide informative results that improve lives.
Accomplishing this requires the deep training in research methods that I described in my K08 proposal, centered on advanced analytical techniques from epidemiology, econometrics, biostatistics, and computer science. The problem of fragmented care is an ideal example through which to study many of these techniques.
Research proposals take a village to complete. Who was involved in yours?
- My mentor, Colin Cooke. Colin is a leading critical care health services researcher and epidemiologist. I started training with Colin as an intern in 2013, and he has been incredibly generous towards me throughout my career. Colin made himself available to me around the clock as I was putting the proposal together. Because I have a toddler and clinical obligations, I frequently took him up on this.
- Jack Iwashyna is my co-mentor and advises Colin on his mentorship of me (he is a “mentor-to-the-mentor”). He also met with me frequently as I was putting the proposal together. Among his many contributions to my training and development, Jack challenged me to think hard about my passions as a physician and scientist, and about how I could use those passions to inform a longer-term vision for my scientific career. He has been an incredible mentor, sponsor, and role model as I’ve worked towards this vision.
- Sarah Krein and Andrew Ryan also mentored me as I put together the grant proposal. Their input on my career development and research plans pushed me to make the most of my period of training during the proposal. They also provided specific expertise relevant to care coordination, patient safety, causal inference, and survey methods.
- Theodore Standiford, my Division Chief, has always been supportive of me and of our health services research group in pulmonary and critical care medicine. As one of my scientific advisors, he has frequently made the trek over to NCRC to meet with me to discuss a research idea or my career development.
- Others in our critical care health services research and epidemiology group have always been incredibly generous towards me - Hallie Prescott, Michael Sjoding, Thomas Valley, and Elizabeth Viglianti. They’ve always been willing to drop everything to provide feedback, provide examples of their work, share data or analytical resources, and offer opportunities to me as a junior colleague.
- Kristin Poole, our Division’s Pre-Award Grants Specialist, was tremendously supportive with getting the grant proposal together, submitted, and resubmitted.
- Members of the IHPI and National Clinician Scholars Program (NCSP) "DJ Lab" (named after its outstanding leader, Dr. Deena Costa in the School of Nursing), “JackLab” (after Jack Iwashyna), and the Multidisciplinary Intensive Care Research Workgroup (MICReW). Each of these groups provided feedback on early versions of the grant and made it a lot stronger. Group members also generously presented their own work (including similar types of grant proposals) over the past few years, allowing me to incorporate lessons that they learned and strategies that they found effective. This latter input influenced my proposal in unmeasurable ways.
- The Institute for Healthcare Policy and Innovation (IHPI) and Center for Healthcare Outcomes and Policy (CHOP) career development proposal repository and mock study sections (and all people contributing their time and proposals to these resources) gave me useful insight into how to write an effective grant proposal and provided important feedback on my early proposals.