Samir Soneji: Forecasting Population Health

As a demographer, Samir Soneji, PhD, an assistant professor at The Dartmouth Institute for Health Policy and Clinical Practice at Geisel School of Medicine, studies the long-term effect of cancer-related health policy at the population level. Through forecasting future disease and mortality rates, demographers are able to predict how changes in population affect public policy.


Could you describe what you are working on now?

Part of my current work focuses on tobacco regulatory control. In 2009, the Tobacco Control Act granted regulatory authority to the Food and Drug Administration (FDA) over the manufacture, distribution, and marketing of tobacco products. Although the FDA exerts its regulatory authority over cigarettes and has proposed to extend its authority over other tobacco products, we find that significant gaps remain. For example, in Soneji et al. (JAMA Pediatrics, 2015) we find initial use of hookah—a currently unregulated tobacco product—increases the risk of subsequent cigarette smoking. We’re now studying whether electronic cigarettes—another unregulated product—increase the risk of subsequent cigarette smoking.

Another part of my current work focuses on lung cancer screening. Established adult smokers face an ever-increasing risk of developing and dying from lung cancer as they age. For the first time, lung cancer screening was found to be effective in a national randomized clinical trial. We found that lung cancer screening was also cost effective in the trial. While the results from the trial are promising, we do not yet know if they will translate into effective reductions in lung cancer mortality when screening occurs in everyday clinical practice. We’re now studying the clinical and structural barriers patients face as they begin screening for lung cancer.

One of the most exciting fields in demographic research is forecasting. We know a lot about the current health of the population. And we can forecast how our health today might affect our longevity in the future. We consider the decline in smoking and rise in obesity and forecast that US life expectancy will still increase over the next quarter century. The gains in life expectancy we forecast are higher than official government projections. If our forecasts are correct, then the Social Security and Medicare Trust Funds may become insolvent sooner than expected because retirees will be living longer than anticipated.


What about the margin of error—how does that affect forecasting?

One of the joys of forecasting is that it would take 25 years to prove me wrong! However, one of the responsibilities of forecasting is to have enough uncertainty about my forecasts in order to capture the truth with some certainty.

Let’s look at obesity for example. The rise in obesity is much more recent than the decline in smoking. And at a population level, we still don’t see an effect of today’s obesity rates on tomorrow’s mortality rates. So we have to ask: Does obesity not affect mortality at the population level? At this point it’s uncertain. We want to produce forecasts that allow for a range of likely relationships between obesity and mortality while still producing forecasts that maintain well-known demographic properties observed over centuries in dozens of countries.

With smoking it’s a different story—the relationship between smoking and mortality has been well known for more than 50 years now.


Have you uncovered any surprises along the way?

We recently studied the value of US cancer care by comparing spending and mortality in the US and Western Europe. The US and Europe both achieved steady reductions in mortality rates of many leading cancers. And the US spent a lot more money to achieve this progress.

But I was surprised that US lung cancer mortality rates were higher than the corresponding rates in Western Europe over the past thirty years even though the US spent significantly more on lung cancer care.


How are forecasts used?

The information may be used by private organizations or government agencies focused on prevention efforts.

Suppose we have $100 million allocated to reducing cancer mortality in the US and you asked me where we should we spend the money. I’d say we ought to spend it—from a population health perspective—on cancers that have largely declined because of prevention, on cancers that have declined because of screening, and on cancers that have declined because of treatment.

In fact, I wouldn’t hesitate to recommend we spend nearly all of the money on smoking cessation efforts. Smoking is the most important risk factor for lung cancer, of course, and a leading risk factor for many other cancers, too.


What brought you to Dartmouth?

I love the idea that you can use statistics to answer interesting and important questions about human populations and I’m fortunate to be able to work with incredibly generous and accomplished senior faculty here who help me advance my research. The faculty position at TDI and the Cancer Center has been a great move for me—and for my wife, Valerie Lewis, who is also an assistant professor at TDI.

Valerie and I enjoy getting outdoors to explore New Hampshire and Vermont with our 9-month old daughter.