Baljash Cheema, MD, MSCI – The Early Career Voice

Baljash Cheema is a second-year cardiology fellow at Northwestern University in Chicago, Illinois. While he has loved his experience in all aspects of cardiovascular medicine, he plans to specialize in advanced heart failure and transplant and care for our sickest. He has a particular interest in improving patient care using machine learning and artificial intelligence. When he isn't in the hospital, he enjoys spending time with his family and friends and watching sports with his wife. Follow him on twitter: @JCheemaMD

Data Science and Coding for Clinicians – Where to Start

August 6, 2021July 29, 2021 Baljash Cheema, MD, MSCI

Medicine is seeing an explosion of data science tools in clinical practice and in the research space. Many academic centers have created institutions tailored to integrating machine learning (ML) and artificial intelligence (AI) into medicine, and major associations including the AHA have created funding opportunities and software tools for clinicians interested in harnessing the promise of big data for their research.

While knowledge on the underlying algorithms and writing code is not necessary to lead a multidisciplinary team working in this space, there are those that want a working knowledge of what is happening under the hood. Thankfully, the computer science (CS) and AI communities have numerous free, online resources to help with this. As I embark on a Masters in Artificial Intelligence, I have used these courses as prep work and found them to be highly educational.

Python for Everybody – By Dr. Charles R. Severance, University of Michigan

This course is meant to get those with no programming background up and running with Python. It focuses on understanding the underlying syntax of the language and the various data structures that come standard in Python. It also touches on web applications, SQL, and data visualization. Thorough, but approachable, this is a great place to start.

CS50x – By Dr. David Malan, Harvard University

One of the most popular courses at Harvard, this course is an intensive introduction to computer science, focusing on key concepts and using various programming languages to illustrate them. The first half or so of the course teaches you to program in C, a low-level language that illustrates how a computer really functions, before moving on to Python (and various Python frameworks), SQL, and web programming. While the juice is definitely worth the squeeze, this course is a commitment and takes significant mental energy to get through.

Machine Learning – By Dr. Andrew Ng, Stanford

One of the courses that popularized the massive open online course (MOOC) revolution, here AI visionary Dr. Ng takes you through a survey of ML/AI algorithms with real world examples and problem sets to work through. The main programming language is MATLAB. This course is enough to give you a basic overview of how these algorithms run and the types of data they are best at handling, serving as a solid introduction to the field.

Machine Learning for Healthcare – By Peter Szolovits and David Sontag, MIT

Healthcare in general and the data it generates is unique, posing challenges distinct from other fields where ML/AI are commonly employed. This course highlights these points through a thorough investigation of healthcare data, common questions clinicians ask in routine patient care, and the clinical integration of ML. It touches on many different topics, including ML for cardiac imaging, natural language processing and clinical notes, and reinforcement learning. No coding is required for this course.

While these courses are just a start, they provide the groundwork for further investigation. in many cases, they are enough to develop an intuition of more complex material including deep learning. If these are topics that interest you, I encourage you to jump on in!

“The views, opinions and positions expressed within this blog are those of the author(s) alone and do not represent those of the American Heart Association. The accuracy, completeness and validity of any statements made within this article are not guaranteed. We accept no liability for any errors, omissions or representations. The copyright of this content belongs to the author and any liability with regards to infringement of intellectual property rights remains with them. The Early Career Voice blog is not intended to provide medical advice or treatment. Only your healthcare provider can provide that. The American Heart Association recommends that you consult your healthcare provider regarding your personal health matters. If you think you are having a heart attack, stroke or another emergency, please call 911 immediately.”

Stage D Heart Failure – Who and When to Refer (#ACC21)

May 28, 2021May 21, 2021 Baljash Cheema, MD, MSCI

#ACC21 came and went, bringing the usual flurry of practice-changing clinical trials, new scientific theories and inquiries, and a wealth of creative ideas showcased through poster presentations. While the virtual format is quite the departure from the in-person atmosphere, it allows flexibility in viewing sessions on-demand and allows individuals that may have an otherwise challenging time traveling to join the discussion. Aside from the trials and presentations that got the most headlines, I wanted to highlight a talk within the advanced heart failure space that expanded on a challenging clinical scenario we encounter routinely. This blog contains screenshots that are directly from the talk Moving Beyond NYHA Class: Risk Stratification and Prognosis in Advanced Heart Failure (within Session 603 The Advanced Heart Failure Therapies of LVAD and Transplant: Who, What, When, Where, Why, and How?) by Dr. Garrick Stewart from Brigham and Women’s Hospital.

Dr. Stewart starts with an overview of how we think about and classifies patients who have heart failure, starting with the history of the New York Heart Association Class grading schema. While it is simple to use and universally known, its limited in its ability to discriminate how sick those with heart failure truly are. Specifically, it cannot tell you who is at the highest risk for morbidity and mortality. To try and address those specifically with advanced heart failure, the INTERMACS Profiles were created. He outlines in his talk how these schemes are related

While this helps us think more critically about this patient population, there remains the issue of knowing who to refer for advanced heart failure therapies. Timing is anything from trivial, as those that are referred too late have worse outcomes, and those that are referred too early are placed at the risks associated with the therapies before they may actually need them. A commonly used mnemonic to remind clinicians of red flags for patients with advanced heart failure is “I NEED HELP.” If your patients has any of these criteria, and certainly if they have several, it may be time to refer.

Despite this, improvements in the referral process are still needed. Thankfully this is an area of active investigation! Congrats to Dr. Stewart on creating this excellent review.

Reference:

Stewart, Garrick. Moving Beyond NYHA Class: Risk Stratification and Prognosis in Advanced Heart Failure. Session 603 The Advanced Heart Failure Therapies of LVAD and Transplant: Who, What, When, Where, Why, and How? ACC 2021 Scientific Session. May 15, 2021.

Top 10 Tips for Incoming Cardiology Fellows

April 30, 2021April 30, 2021 Baljash Cheema, MD, MSCI

Cardiology fellowship comes at you fast. Within the first day, you realize how much nuance exists within the field that you hadn’t been exposed to in internal medicine, and there are lots of patients whose care depends on those details. At the same time, you quickly come to appreciate how much of an impact you can make on a patient’s health and just how rewarding the field is. it’s a beautiful journey! In thinking back on the last two years, I wanted to share my top 10 tips and insights on fellowship aimed at incoming fellows.

Learn from everyone. Nurses, techs, the cath lab team, sonographers, any staff with any clinical experience. There is often a sense you get when things aren’t right, and it takes a while to learn. These folks have developed it.
It will take you at least 6 months to start to feel somewhat comfortable, a year before you think you got a hang of things, and that’s normal.
Ask for help often. Key resources: senior fellows. They know everything, or they know who does.
When you are on call, you are never alone. Get in the habit of communicating with your attending, even in the middle of the night. It’s the best thing for patient safety and your learning, and the attendings want it too.
“Don’t guess when you can know.” The credit for the quote goes to Dr. Neil Stone, but the point is to get all the information you need (safely) and double-check the basics. The stakes are high in cardiology, so do the little things that prevent errors.
Stay well outside of work. Family, friends, exercise, sleep, hobbies, whatever makes you you. These things are never more important than now. Burnout is real, prevalent, and painful.
Meaningful learning happens through rote repetition in cardiology. Whether it’s in the cath lab or on echo, you will make insights through monotonous reps of seemingly routine studies/cases that you can’t make through any other means. Show up and dive in.
It may take you a while to have the bandwidth to delve into academic pursuits outside of “just being a fellow” – that’s okay. Fellowship is hard.
Recognize sick from not sick, and if someone is sick, move fast.
When you consent a patient for a procedure, know the facts, and tell them. There is no such thing as a no-risk procedure. I will never forget this, after being involved in a case of a PA rupture during a straightforward right heart cath leading to a cardiac arrest. Make sure they know what they are signing up and consent is truly informed.

I would recommend going into cardiology to those who are interested 10 times out of 10. Congrats to those just starting out! I hope this list gives some pointers that help as you embark on your own journey in the field.

Coronary Allograft Vasculopathy – The Achilles’ Heel of Heart Transplant

March 31, 2021March 29, 2021 Baljash Cheema, MD, MSCI

As a 3^rd year medical student in the coronary care unit (CCU), I helped care for a patient whose story I will never forget. She had developed advanced heart failure due to peripartum cardiomyopathy in her 20s giving birth to her only child and required a heart transplant. She did well for a number of years, but I met her in her 30s when she was admitted post-MI in cardiogenic shock. Her coronary atherosclerosis was due to severe coronary allograft vasculopathy (CAV), an aggressive form of CAD transplant patients may develop. She got a LAD stent and was supported with a balloon pump but was tenuous at best. Some days after her PCI, in a moment seared into my memory, she let out an ear-piercing yell and suddenly arrested and died, her daughter at the bedside. I’ll always remember the pain on her child’s face when she passed, and I will always have a sincere appreciation for the misery CAV can cause. This blog is meant to provide some historical context to heart transplantation and the issue of CAV, as well as to discuss ways we can prevent it.

Since the first heart transplant in Cape Town, South Africa, there have been tremendous advances in cardiac transplantation with median survival now around 12 years. It didn’t always appear that this would be the case, with mortality so high in the early days that many felt heart transplant wasn’t worth it. The advent of calcineurin inhibitors with cyclosporine in the 1980s and tacrolimus in the 1990s were key (Figure 1). Steady improvements in infection prophylaxis, screening for and treating rejection, and surgical technique and expertise further helped the cause.

But as we addressed one set of problems, we found another. CAV is an aggressive form of coronary artery disease (CAD) present in 30% of heart transplant recipients at 5 years and 50% at 10 years. Those with it have worse survival. It shares some risk factors with classic CAD but has several of its own, and there are key pathophysiologic differences (Figures 2 and 3). Our patient was unique in that she had a true plaque rupture MI, typically occurring less often with CAV relative to classic CAD, but this may have been related to a donor transmitted lesion acting more as typical CAD would.

Figure 2. Pathophysiologic Differences

Figure 3. Risk Factors

So how do we prevent CAV? Our best data comes from statin trials in the 1990s-2000s (pravastatin, simvastatin, and atorvastatin studied), showing lower rates of rejection and CAV with improved survival in transplant patients treated with statins. This makes intuitive sense, as dyslipidemia is a rock-solid risk factor for classic CAD and nearly universally seen post solid organ transplantation due to the metabolic consequences of common immunosuppressives. These immunosuppressives, while life-saving in their own right, also lead to worsening glucose control, hypertension, obesity, and kidney disease. Addressing each of these while encouraging a heart-healthy diet and routine exercise is of paramount importance in keeping our transplant patients healthy. Finally, a reminder that there are many drug-drug interactions with transplant medications. Figure 4 is adapted from Warden et al and shows the relative degree of interactions between immunosuppressives and common lipid-lowering drugs.

Figure 4. Drug-Drug Interactions

While this story was tragic for the patient and her family, it’s given me a profound respect for CAV that I will carry forward when I eventually care for heart transplant patients in my career. Below are the references for this article from which parts of the figures were taken. Each of these is a fantastic resource for further learning.

References:

Stehlik, J., et al. (2018). “Honoring 50 Years of Clinical Heart Transplantation in Circulation: In-Depth State-of-the-Art Review.” Circulation 137(1): 71-87.
Warden, B. A. and P. B. Duell (2019). “Management of dyslipidemia in adult solid organ transplant recipients.” J Clin Lipidol 13(2): 231-245.
Costanzo, M. R., et al. (2010). “The International Society of Heart and Lung Transplantation Guidelines for the care of heart transplant recipients.” J Heart Lung Transplant 29(8): 914-956.

Evaluating ML/AI Models in Clinical Research

March 1, 2021February 26, 2021 Baljash Cheema, MD, MSCI

The number of machine learning (ML) and artificial intelligence (AI) models published in clinical research is increasing yearly. Whether clinicians choose to dive deep into the mathematical and computer science underpinnings of these algorithms or simply want to be conscientious consumers of new and relevant research to their line of work, it is important to become familiar with reading literature in this field.

To that end, Quer et al. recently wrote a State-of-the-Art Review in the Journal of The American College of Cardiology detailing the research landscape for ML and AI within cardiology including concrete tips on how a non-ML expert can interpret these studies. At its core, ML is about prediction, and models are created to make accurate predictions on new or unseen data. Inspired by their work and incorporating many of their recommendations, below is a list of considerations for when you are critically evaluating an ML/AI model in clinical research:

What question is addressed and what problem tackled? How important is it? Regardless of a model’s performance or the accuracy, its usefulness is determined by its clinical application. Everything must go back to the patient.
How does the ML/AI model compare to traditional models for the given task? Many studies have shown little additional benefit when comparing ML/AI models to standard statistical approaches including logistic regression for clinical questions that have been extensively researched in the past with key predictors of the outcome of interest identified. The promise of ML/AI really exists in incorporating novel data sources and data structures, including time-series information and continuous input from wearable sensors, raw images and signals such as that from common studies including echos and ECGs, and harmonizing unique data types together.
To which broad category does the model fall into? Most machine learning models fall into buckets of supervised learning algorithms, unsupervised learning algorithms, or reinforcement learning. Each approach is slightly different with a unique end product. Supervised learning algorithms learn patterns in the data that allow them to predict whether a specific observation falls within a specific class or category, for example determining if a photo is a cat or a dog. This requires data that is labeled for the algorithm to learn from, i.e. someone or something has provided data that is correctly tagged as a dog or cat. Unsupervised learning does not require observations with labels but instead combs through the observations to look for those that are similar to each other. Reinforcement learning a separate task in which an agent is trained to optimize choices made to attain a stated goal. All of these have been used clinically in recent literature.
How were the data and labels generated? Garbage in = garbage out. Your model is only as good as the data it was trained on and the accuracy of the labels. It’s important to know where this information came from.
Model training, validation/performance, generalizability. A common approach to training models is to split the data into a training set with unique observations left for the test set to validate the model. It is critical to train and test on different data with no overlap. Model performance is tracked with metrics similar to those used to evaluate clinical models, including sensitivity, specificity, positive predictive value, negative predictive value, and AUC, although the names associated with those measures may be different. Additional measures such as an F-score may be used. Arguably more important, however, is generalizability. This is how well the model performs in an entirely unique cohort, often from another center, although many of the currently published studies do not include this step.
How clinically useful are these findings, and is the model interpretable? Basically, is the juice worth the squeeze? And can a human understand why the model made its conclusion? A common knock against deep learning neural networks for example is that although they are incredibly skilled at learning from data and making accurate predictions on new data, how they do so is a “black box,” although new ML/AI methods have started to account for this.
How reproducible are the results? Did the authors share their code or dataset? If they used an EHR phenotype to generate their cohort, can you do the same thing at your institution?

These points are meant to summarize and add to some important aspects of this recently published article, but it is an excellent read and I encourage everyone to review it in its entirety.

Reference:

Quer, G., et al. (2021). “Machine Learning and the Future of Cardiovascular Care.” Journal of the American College of Cardiology 77(3): 300-313.

Non-Fiction Books as a Learning Tool in Training

February 1, 2021January 29, 2021 Baljash Cheema, MD, MSCI

The early days of cardiology fellowship are mentally draining. You flip from being an experienced medicine resident to a newly minted cardiologist-to-be, and it feels like you know nothing. Your brain is fried at the end of each day, but there’s always that nagging voice in the back of your head telling you to read more and tackle all of that new material.

At those times, I found non-fiction books by physician-writers to be a fantastic resource. These incredible authors have an uncanny ability to walk the line between medical experts providing state-of-the-art care and storytellers able to convey a message that anyone can understand.

These books gave me a way to put my clinical experiences into context, to learn the history of cardiology, and to develop a vocabulary I could use when discussing heart disease with patients. They were also a pleasure to read.

Here are a few of the books I loved, recommended to me by mentors/experts in our field:

Heart: A History, by Dr. Sandeep Jauhar

In this book, Dr. Jauhar takes us through the story of cardiology as a medical discipline by putting advances in the field into the context of individual patients, discussing his own personal encounter with cardiovascular disease and how it has led to tragedy in his own family.

The Heart Healers: The Misfits, Mavericks, and Rebels Who Created the Greatest Medical Breakthrough of Our Lives, by Dr. James S. Forrester

It turns out cardiac surgeons and cardiologists are quite the characters! Here, Dr. Forrester highlights the effervescent personalities in the field while walking us through his professional goal of stomping out coronary artery disease and its downstream consequences.

Deep Medicine, by Dr. Eric Topol

The AI revolution is touching every aspect of our lives, and here Dr. Topol lays out a plethora of examples of its impact in medicine while offering insights on how things will look in the future.

Digital Doctor, by Dr. Robert M. Wachter

The implementation of the EHR and its unintended consequence, medical errors, virtual patient encounters, Silicon Valley’s impact on healthcare, and diagnostics through smartphone algorithms are all discussed in Dr. Wachter’s work on the digital transformation of American medicine.

Grit, by Angela Duckworth

This is a bonus, as Dr. Duckworth is the only non-physician doctor on this list, but she is a brilliant researcher who lays out her evidence for grit, passion, and persistence being key factors in individuals achieving remarkable accomplishments. Excellent motivation for trainee spending hours in the hospital!

There are many, many more books that fall into this category. Happy reading!

AHA2020 – The Next Steps in Treating Heart Failure

December 29, 2020December 28, 2020 Baljash Cheema, MD, MSCI

AHA 2020 came and went, and now is the time to put into context the scientific advances presented. While all areas of cardiology saw therapeutic innovations, the ever-evolving landscape for heart failure (HF) therapies stood out in particular.

These were among the key discoveries shared at AHA20 in the HF space:

GALACTIC-HF: In patients with chronic heart failure with a reduced ejection fraction (HFrEF), the cardiac-specific myosin activator omecamtiv mecarbil reduced the primary composite endpoint of time to HF event or cardiovascular death, driven by a reduction in hospitalizations and ED visits. Importantly, the therapy appeared to be hemodynamically neutral, and subgroup analysis showed those with lowest ejection fraction (EF) may benefit in particular.

AFFIRM-AHF: In patients with HFrEF and iron deficiency stabilized from an acute HF event, IV iron repletion reduced the risk of subsequent hospitalization for HF but not death.

SOLOIST-WHF: In patients with worsening HF, the SGLT1/2 inhibitor sotagliflozin significantly reduced the risk of death and hospitalization for HF subgroup analysis showed the results persisted regardless of EF.

SCORED: In patients with diabetes and chronic kidney disease, sotagliflozin reduced the risk of cardiovascular death and subsequent hospitalization and/or urgent visits for HF. Similarly, the effect was seen regardless of EF.

These results not only add to the proven therapies for HFrEF including the cornerstones of ARNI, MRA, BB, and SGLT2 inhibitors, they add therapies for worsening heart failure and strongly suggest therapy for heart failure with a preserved ejection fraction. They may even hint at therapy for those with very low EF. With the VICTORIA trial showing benefit for vericiguat at ACC 2020, and additional therapies already indicated for subsets of patients including ivabradine and fixed-dose isosorbide dinitrate and hydralazine, we now find ourselves with a number of medications our patients should be receiving.

The path forward will be deciphering how best to implement these therapies at doses with proven benefit. Dealing with the issue of cost will be key. Sequencing trials, collating datasets with prescription fill data, machine learning tools to support clinical decision making, and personalized medicine through “omics” technologies may all play a role, as recently discussed by the HF Collaboratory (1).

While there is much to be seen, it’s certainly a very exciting time for heart failure!

Reference

Bhatt AS, Abraham WT, Lindenfeld J et al. Treatment of HF in an Era of Multiple Therapies: Statement From the HF Collaboratory. JACC: Heart Failure 2020.

Buzzword Alert! Artificial Intelligence – Just the Hype Man or a Genuine Showstopper?

November 17, 2020November 17, 2020 Baljash Cheema, MD, MSCI

Conversations of the utility and promise of machine learning (ML) and artificial intelligence (AI) permeate all fields of medicine, and cardiology is no exception. A quick search shows that 69 posters containing the keywords “machine learning” made it into AHA’s Scientific Sessions 2020. But is it for real? Will we really see a future in with ML/AI factors into all aspects of clinical care and in fact, re-write the script on how we care for patients?

Below is some of the discussion points and imperatives that stood out to me today from the “Hope or Hype? Artificial intelligence and Machine Learning in Imaging.” session at #AHA20 featuring thought leaders Drs. Marielle Scherrer-Crosbie, Alex Bratt, David Ouyang, Tessa Cook, Damini Dey, David Playford, and Geoffrey Rubin.

While awe-inspiring in its ability to make inferences and predictions human beings often cannot themselves, we must be aware that ML/AI algorithms can recreate and reinforce the bias pre-existing in our society. We must fight this by knowing it is a possibility, screening for it, and training algorithms on datasets that are truly representative. As much of the political landscape and national conversation right now centers on structural racism and bias in America, it’s is prudent to understand how the models we create can perpetuate this.

Separate low hanging fruit from the unrealistic (at the moment) and consider the unrealistic tasks in the realm of discovery science. A quick rule of thumb provided by Dr. Ouyang, summarizing the words of Dr. Andrew Ng, first determines if it is possible for humans to do a task relatively quickly. If it is, we can probably automate it with AI now or in the near future.

Scrutinize our data. How much do we trust it? High-quality data for ML/AI means broad, accurate, and plentiful. We need robust training labels, as free from subjectively as possible.
How open is our data for inspection? Fields in computer science are far further along than medicine in deploying and improving ML/AI models because of open data sets and shared code, allowing groups to verify, tinker, and re-create to move the needle forward. Medical AI has not been so forthcoming.

As new technology is rapidly evolving and making it into the clinical space, we need to be responsible for mistakes. This means we need to assess not only our model performance before deploying but also the consequences of using the model in real life. This may require RCTs and to consider ML/AI algorithms like we consider new therapeutics.

What we really want is the AI running in the background saying “Hey, this task was automated and is now solved for you. Proceed as you see fit.” Humans and machines are in this shared space. The more we can integrate ML/AI to help us with tasks we are already doing, the better our results will be.

So where does this leave us? Most in our field believe ML/AI will play an important role in our future. Ideally, we will do it in a way that will make sure human intelligence is always paired with artificial intelligence to create a product neither of the two could be alone, will ensure our algorithms are free of bias and openly shared to allow for continuous improvement.

Highlights from Day 3 of #AHA20

November 15, 2020November 17, 2020 Baljash Cheema, MD, MSCI

What an end to the weekend! As Day 3 of AHA 2020 continues, there is already much to digest and discuss. From Late-Breaking Trials to specific programming for Fellows in Training and Early Career individuals, there was something for us all. My time was spent tuning in to informative discussions from leaders in the advanced heart failure world, and below I reflect on some of what stood out to me today.

Implementation science matters as Dr. Saurer shares with us on Twitter. As our armamentarium of GDMT grows, it will be key to figure out how best to get these medications prescribed to our patients in order to maximize therapy while minimizing side effects and managing cost. But how do we do that? What order should be prescribed our drugs in? What about devices? Is the same approach applies to all? There is an area primed for more research. What an exciting time to be in the heart failure space.
Take note of the trajectory. It is important to be aware of and routinely reassess the trajectory of our patients with heart failure, both in the inpatient and outpatient setting, as discussed by Dr. Hollenberg. In our sickest patients, those with Stage D heart failure, we have the option of considering VAD or transplant as well as palliative care approaches including home inotropes, but patients are often flagged too late and no longer eligible for certain therapies. Use the “I NEED HELP” mnemonic to try and identify these patients early, as Dr. Breathett shares with us.
Know when to escalate care. While it is not always crystal clear who needs a higher level of support, Dr. Cogswell gives us a clinical pearl: if your patient is hypotensive with heart failure, start to think about what is next, whether that is temporary support or a durable device or ultimately both. She gives an example of pausing an IABP and seeing if her patient becomes hypotensive in order to consider an Impella or LVAD, as opposed to adding more drips which may not ultimately be enough.
#ReviveTheSwan! As eloquently and definitively stated by Dr. Hall, all patients with cardiogenic shock need a Swan-Ganz catheter. Our physical exam “just is not that good”. Care is improved by using invasive hemodynamics
A civilized debate is possible! Despite the politics in our country as of late, today’s sessions clearly showed the civil, intellectual, informative debates are still possible in our society, and they are for the betterment of all involved! Kudos to all who gave us their time and wisdom today.

While the day is early and there is much yet to be seen, the teaching on Day 3 of AHA20 has already been fantastic.