Craig White

Craig White

Senior Vice President and General Manager, Concerto HealthAI

Dissertation Title: "Delivering the Right Amount of Care - Sometimes More is Less"New technologies utilized in clinical practice offer improved tools for diagnosing and treating patients. In some cases these improvements cause unintended consequences by allowing diagnosis of indolent or otherwise clinically irrelevant disease. In my first two chapters, I examine technological advances in screening and diagnosis which have led to questions regarding whether it is necessary to treat disease at as early a stage as it is possible to diagnose it. I conduct cost-effectiveness analyses to determine whether application of less-intensive clinical regimens leads to better and more cost-effective outcomes for patients with thyroid and prostate cancer. In these two cases I find that less intensive therapeutic options provide better results. In my third chapter, I use electronic medical records data to examine a case where old technology is sufficient to identify disease at a stage suitable for the application of well-established and highly efficacious treatments. As a corollary to thyroid and prostate cancer, I find that a significant proportion of patients eligible for hypertension treatment are not treated, despite simple criteria for diagnosis, low-cost therapies and clear recommendations for their use within well-established and accepted guidelines.

Chapter 1: Less Intensive Treatment of Low-Risk Prostate Cancer. Which strategy is best?

In the early 1990s, rates of prostate cancer incidence in the United States increased dramatically due to broader use and widespread adoption of the prostate-specific antigen (PSA) test. Although these diagnosis rates peaked in the mid-1990s, today, rates remain elevated by about 50% compared to pre-PSA-testing levels. Despite this, mortality rates have remained fairly constant, suggesting that many patients identified through PSA screening may represent "overdiagnosed" patients. Over the past 15 - 20 years clinicians have developed and tested less intensive, yet highly effective Active Surveillance treatment regimens. These regimens offer patients who satisfy a very specific set of criteria an option to delay immediate radical treatment thereby avoiding the morbidities that are associated with surgery, chemotherapy or radiation therapy. Several variations of Active Surveillance have been practiced in different centers around the world, but there is no clear consensus regarding which is best, and for which patient. In my first chapter I undertake a systematic literature review to identify variants of Active Surveillance, identifying three frequently reported variations representing a high, medium and low-intensity protocol. I undertake a decision analysis to compare these to the existing radical treatment mix that is practiced in the US. I find that Active Surveillance of medium intensity is the most efficient option, allowing men with low-risk prostate cancer to achieve an additional 217 quality-adjusted life days at an incremental cost of just over $2,000, making it a highly cost-effective strategy. I also find that a modified version of this protocol is more efficient but may be less desirable to clinicians due to its longer period between surveillance biopsies and its lack of widespread clinical use.

Chapter 2: Is the Less-intensive 2015 American Thyroid Association Treatment Protocol a cost-effective update for Patients with Papillary Microcarcinoma?

Similarly to prostate cancer, rates of thyroid cancer have increased dramatically in the past 15 years, yet mortality has remained constant. In this chapter I conduct a cost-effectiveness analysis focusing on treatment for patients with papillary thyroid carcinoma, a subtype of thyroid cancer that accounts for approximately 85% of incident cases. In the most recent (2015) update to the American Thyroid Association (ATA) "Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated Thyroid Cancer", the ATA recommended two major changes compared to their 2009 guidelines. First, that patients with specific tumor characteristics could be treated with lobectomy, a less intensive type of surgery than the prevailing surgery, total thyroidectomy, and second; for patients with specific nodule/tumor characteristics, Active Surveillance of their cancer via annual ultrasound imaging was a viable option. For my second chapter, I created a Markov microsimulation model to determine whether these major changes to the guidelines improved outcomes, decreased costs, or both, for patients with papillary thyroid carcinoma. I find that the 2015 guidelines are a dominant treatment strategy compared to the strategy recommended in the 2009 guidelines. Even after taking uncertainty into account via deterministic and probabilistic sensitivity analysis the 2015 strategy remains dominant, or at worst, highly cost-effective.

Chapter 3: Evaluation of Physician Adherence to JNC7 Hypertension Treatment Guidelines using a Large National EMR Database

In this chapter, I utilize a national electronic medical records database to evaluate physicians' behavior with regard to rates of prescription for pharmacologic therapies to treat hypertension. In 2003, the "Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure" issued their 7th revision of the guidelines for the diagnosis and treatment of hypertension (JNC7). Within these guidelines the JNC provided clear blood pressure thresholds of 140mmHg systolic and 90mmHg diastolic for the initiation of hypertension treatment and specified appropriate pharmacologic therapies. I utilized electronic medical records data from approximately 50M patients captured during 2010 - 2014 across the United States. Using these data I determine how often treatment was consistent with that which would be expected if the recommendations of the JNC7 guidelines were followed for those patients. I find that the majority of patients meeting the JNC7 systolic and diastolic blood pressure criteria for treatment do not receive a prescription as recommended by the JNC7 guidelines and that the rates of treatment and guideline concordance vary by age, race, and sex. Through the use of logistic regression analysis I determine that patients satisfying the JNC7 criteria for systolic blood pressure have odds of treatment of 2.98 compared to those who do not, but that for patients with systolic blood pressure readings near the threshold of 140mmHg the odds of treatment for those who satisfy the JNC7 criteria increase to 7.75. From these results, I infer that the JNC7 guidelines stimulate treatment of patients who satisfy the criteria defined in the JNC7, but that there is still significant undertreatment of eligible patients which varies by age, race, and sex.


Graduation Year