Skip over navigation

Predicting Health Outcomes among the Elderly: Racial
Differences in Long-Term Care Transitions and Mortality

Adviser: Scott M. Lynch

Matthew L. Edwards Jr.


“The ability to approach novel situations critically is essential for any long-term goal I can imagine, whether in academia, medicine, law, or industry.”


After reflecting on my experiences the summer prior to my junior year of college, I found it remarkable that I learned as much about legal issues as medical issues while interning with residents and attendings at an obstetrics/gynecology clinic. For example, I learned that the need to carefully document a patient’s treatment, whether a patient understood their diagnosis and treatment options, and whether a patient consented to treatment was as much about establishing (or relinquishing) liability as it was a concern of practicing good medicine. I was always prone to asking “sociological questions,” despite knowing very little about sociological theory and methodology before enrolling at Princeton; this summer experience proved no different. I had well-developed interests in science and medicine, but as this experience demonstrated, I had only the faintest grasp of the enormous complexity of the health care industry, an industry shaped by culture, history, politics, law, and socioeconomic factors. I actively sought enrollment in broad survey courses such as “The Sociological Perspective” and “The Rise of Modern Biomedicine” in order to gain a greater appreciation of the ways in which society shapes medicine and health care, and it seemed quite natural that I would explore these interests in health care in my independent work in sociology.

Not surprisingly, my junior paper examined how individuals perceive the medical malpractice crisis, the sources that shape perception, and resulting implications for policy and legal reform. I read and scored 100 media articles based on similar objective criteria and interviewed seven physicians from the Princeton alumni network in order to compare physicians’ subjective experiences with malpractice to more objective trends from media coverage. Although my junior paper seemed to pose as many questions as it did answers, it gave me an appreciation for the enormous potential of social science research on health care and broadened my interests in the sociology of medicine and health.

The summer prior to my senior year was equally instrumental in influencing what I researched the following year. That summer, I noticed a growing number of retirement residences near my home and commercials for “long-term care” insurance plans, and I was generally struck by the growing political debates concerned with providing for the nation’s elderly, especially in the context of Medicare and comprehensive health care reform. After meeting with my adviser, Professor Scott Lynch, early in the fall, I began reading academic literature on aging in gerontology journals and noticed a trend in nursing home research: The vast majority of research focused on predictors of nursing home admission or discharge. This niche in the literature was well researched and there were a number of significant studies addressing nursing home predictors. Quite naturally, I began to wonder, if these factors influence who is admitted to a nursing home, what determines who stays? This question seemed largely unanswered as I continued to read more nursing home research articles. I was determined to find out what happens when an individual is admitted to a nursing home. How long is a typical stay? How likely are individuals to exit, and what factors contribute to discharge?

My senior thesis sought to address these questions by expanding the scope of previous studies, accounting for various factors (namely racial differences) that influence not only entries into, but also exits from, nursing homes. The data used in my study came from a longitudinal survey of community-dwelling elderly from New Haven, Connecticut, and North Central, North Carolina. The study included demographic, social, and medical histories of the individuals, and I hoped to examine the factors that shaped individuals’ trajectories through formal and informal long-term care services. By focusing merely on initial predictors of admission or discharge, previous studies inherently assumed that the probability of leaving a nursing home is zero; these studies failed to recognize what I saw as a bidirectional mobility of some individuals’ long-term care patterns. My goal was to demonstrate that (a) elderly Americans engage in a number of different long-term care trajectories, moving in and out of the community with different propensities, (b) these trajectories depend on an individual’s relative health and social status, and (c) that race, social, and health factors play significant and independent roles in explaining long-term care transitions as they compete with mortality.

I soon realized that there were practical explanations as to why this question was largely unanswered. There was a lack of surveys that would be useful in addressing these questions, and where they existed, they were largely inconsistent over time. This is partly explained by the fact that it is inherently difficult to track individuals over time as they move into and out of nursing homes. As a result, I was forced to sift through a number of data sets before finding a suitable study, a process that consumed a significant amount of my time and energy. I grew more and more frustrated as each data set appeared incapable of answering my question, and I met with my adviser with frequent attempts to change my question to whatever that data set seemed capable of answering. I made countless trips from my adviser’s office to the Data and Statistical Services lab, often finding a new complication or what might be a possible solution. Professor Lynch, however, was persistent at keeping me focused and on track, never failing to suggest other avenues to pursue in order to explore my original question. He was patient and understanding of the problems with nursing home data, but still stressed the importance of a well-defined question firmly grounded in a theoretical framework.

My research question became well developed despite the initial setbacks, and I came across a theoretical model that was directly relevant to my question. Furthermore, it had been applied to previous research in nursing home predictors. My study used multinomial logistic regression methods and a discrete-time hazard model that took into account multiple trajectories of individuals by tracking their residential living/mortality status at every stage of the survey. With these methods, I was able to predict the probability that individuals would enter and exit nursing homes, remain in nursing homes or the community, and die in nursing homes or the community. The results showed that blacks are significantly less likely than whites to reside in a nursing home or the community at the beginning of a given period. This also confirmed the fact that blacks were much less likely than whites to ever transition into a nursing home even across comparable levels of health.

It took a while for me to fully understand the significance of the finding, but my adviser was particularly effective at explaining the mathematical basis of the models I used, especially given the fact that multinomial and discrete-time models were inherently difficult to use. I generated elaborate matrix calculations of probabilities and had a number of multinomial regression outputs for my models. I was overwhelmed by the results, as it seemed that I had to analyze more probabilities than I could grasp. I met with my adviser a number of times after getting my results to simply talk through them, and he suggested using a path diagram to model the results. It was not until I worked on this part of the project that the “big picture” seemed to emerge. The path diagram clearly illustrated the significance of my findings: While blacks have more need for services on average, they are less enabled to act on that need. Thus, they lack the very thing that empowers them to act on their perceived need and diagnosed illnesses by transitioning into nursing homes. After I realized the “real-life” significance of my results, I was able to focus on more practical policy concerns, given the complicated nature of long-term care in the United States. My results seemed to suggest that race-driven as well as socioeconomically driven policies were needed concurrently to combat differences and potential disparities. This is because (a) there are different levels in health status among blacks and whites, and (b) perhaps more importantly, there also are different transition patterns across comparable levels of health.

I was enthusiastic about pursuing this project throughout my senior year, but I admit that it was difficult to balance the demands of the thesis (with its often unexpected turns) with the demands of my regular academic coursework. Fortunately, I was organized enough that I was able to work on my thesis at regular intervals. Thankfully, meetings with my adviser were not always solely concerned with the thesis. We often discussed my career interests in academic medicine, the nature of academia, and many nonacademic concerns. I always thought the claim that the thesis is the “capstone” of a Princeton education was merely a testament to the fact that it is a Princetonian’s last significant commitment as a student. My experience writing my thesis, however, changed the way I viewed my Princeton education. The thesis taught me how to be persistent in my search for understanding, but also how to reroute my course of action when circumstances demand it. More than anything, the thesis was an act in balancing significant demands—balancing the demands of original thought against the need for theoretically sound investigation, the demands of being enrolled in a full course load against the rigor of conducting an original study, and the demands of finding something about which you are truly passionate against the fear of immersing yourself in unfamiliar territory.

The last point helped make the thesis the “capstone” of my Princeton education, and it was one that I envision shaping my long-term career goals. The ability to approach novel situations critically is essential for any long-term goal I can imagine, whether in academia, medicine, law, or industry. I had a passion for health care and desired to understand the complexity of health care more fully. If I were a senior with the foresight for what I have just mentioned, I would invest a significant amount of time exploring what really interested me, take cues from whatever grasps my attention and imagination at the moment, and think critically about what most interests me about those issues. I think this is the essence of what it means to be passionate about an issue—whether that issue is engineering a new type of pacemaker, exploring the classics, or understanding nursing home behavior. Passion allows you to pursue understanding of an issue not merely as a means to an end, but as an end in and of itself.

Predicting Health Outcomes among the Elderly: Racial
Differences in Long-Term Care Transitions and Mortality

Matthew L. Edwards Jr.

Scott M. Lynch

Associate Professor of Sociology

“... The senior thesis must break new ground—it must involve learning something that one cannot simply discover by reading someone else’s writing.”

At its broadest, sociology is the study of human variability within and between societies. A consequence of variability, or perhaps another way to think about variability, is in terms of inequality. In every society, a social structure emerges that stratifies statuses and roles (positions) and ultimately distributes resources unevenly so that some persons end up “on top,” while others end up “on bottom.” As a sociologist, my job is to understand the process of distribution of resources, the extent of inequality in those resources, and the consequences of inequitable distribution. The better we understand these processes, the better position we are in to create and modify policy to improve society.

In my research, I am interested in health disparities as a consequence of other forms of inequality like socioeconomic and racial inequality. Matthew Edwards shared a similar interest when he first met with me to discuss my advising his senior thesis. In particular, Matthew was interested in racial differences in nursing home experiences in the United States. This interest is an excellent topic for a senior thesis. Nursing homes play an important role in the living arrangements and health status of elders in the U.S., both at a societal level and at an individual level. At a societal level, mortality rates have fallen steadily over the last century, and in the last half of the 20th century, the rates fell faster at older ages (say above 65) than younger ages. At the same time, fertility rates have fallen dramatically over the last half-century. The result has been that, while in 1960 about 9 percent of the population was over the age of 65, by 2000, about 12 percent of the population was over age 65, and when the retirement of the baby boomers is complete, about 20 percent of the population will be over 65. Despite the improving health of older persons, somewhere between one-quarter and one-half will spend some time in a nursing home, and nursing home costs are increasing rapidly, just as overall health care costs are. A key result is growing inequality in the rates of nursing home admissions, discharges, and lengths of stays. At the individual level, this implies that many of us will spend some time in a nursing home, and that our experiences will vary greatly. Matthew was interested in understanding the extent of racial inequality in this long-term care process, including the ultimate reason for nursing home discharges (death versus recovery).

After a decade of advising senior theses at Princeton, I have discovered that there are basically three stages students go through in the process of thesis completion: (1) developing an answerable research question, (2) finding (or possibly collecting) appropriate data to answer the question, and (3) analyzing the data and writing up the results of analyses. Invariably, the first two stages are the most difficult and take up the majority of the year. The difficulty arises because these stages must, to some extent, be undertaken simultaneously. In my view, and based on our department’s standard, the senior thesis must break new ground—it must involve learning something that one cannot simply discover by reading someone else’s writing. Thus, the research question must be unique. At the same time, the question must be answerable, meaning that data must either exist or be collectable to address it. These two stages often seem overwhelming to students. Most students have a general interest in a topic, but once they “dive” into the extant literature, they find that many of their initial questions have already been answered by others: In sociology, there are huge literatures on almost every topic imaginable! Further, if the initial question hasn’t been addressed in the literature, it is often because there are no data on the topic. As a result, the student usually ends up having to balance developing an unanswered question with finding data that exist that can help answer it. It is at this point that the initial research question tends to get refined and the scope of the eventual thesis narrows.

Matthew’s thesis process began more auspiciously than most, because he came to our first meeting with an important and seemingly manageable initial question. Yet, as he delved into the literature, he rapidly found that racial inequalities in long-term care have been investigated extensively. Worse still, finding an existing survey data set with enough detail to answer a refined question proved to be incredibly difficult—more so than I expected. Seemingly undeterred, Matthew spent hours upon hours searching for data and refining his research question until he obtained the right “combination.” Ultimately, his review of the literature found that, while racial differences in rates of nursing home entry had been investigated extensively in the literature, no one had investigated racial inequalities in the entire process of nursing home entry, length of stay, and exit. Most studies had simply found that blacks and whites, in particular, differed in their rates of admission, with whites entering nursing homes at greater rates. Some studies suggested that this disparity was the result of nursing home cost as well as difference in family members available to take care of elders; it is not a function of differential need.

Part of the explanation for the extant literature’s limited focus on nursing home entry is that most surveys only interview noninstitutionalized persons (i.e., persons not in nursing homes). It is difficult to find survey data that follows individuals beyond entry. The data that Matthew ended up using for his investigation was a long-term survey of several thousand elders who were residing in the community in the early 1980s. These persons were followed up on at least four occasions for more than a decade, allowing Matthew to statistically analyze transitions into and out of nursing homes across their later years. Matthew excelled with the analyses. Statistically modeling transitions between different living arrangements is difficult; yet Matthew learned the types of methods required to do so, and he did so amazingly quickly.

In the end, Matthew’s thesis was excellent. His key finding, which makes an important contribution to the literature, is that blacks and whites differ not only with respect to their nursing home admission rates, but also with respect to their length of stay in nursing homes and their method of exit. In brief, black Americans are less likely to be admitted to nursing homes, they experience shorter stays in nursing homes—despite having poorer health—and are more likely to exit nursing homes both via death and via returning to the community. Since he completed his thesis, Matthew and I have discussed publishing his thesis in a sociology or gerontology journal, and I expect the published version will be well received. Overall, working with Matthew on his thesis was one of the most rewarding experiences I’ve had at Princeton.