How much does health drive healthcare costs anyway? Entry 20 – 2009


When medical and disability costs are high, conventional wisdom assumes there must be more illness driving up costs, right? But how much of total cost can we actually attribute to health status versus other things?

Four Parts
There are actually four driving components of health and absence cost, the first three of which we’ll cover here, and the fourth in the next blog. To give away just a little of the secret in advance, it may surprise some readers to learn that health status is not as powerful a predictor of cost as one might expect.

Believe it or not, our research on nearly 2 million employees and their families across the US finds that a surprisingly small amount of the variation in healthcare costs can be attributed to health status.

We’ve studied how each of four components independently influences cost when all the others are held constant. (We won’t get technical about this other than to say that our analysis included medical and disability data from tens-to-hundreds of thousands of people, running two-part regression models to estimate independent effects.)

The first two components involve “non-modifiable” costs that cannot easily be influenced or changed, while the second two involve costs we consider to be “modifiable.”

Part I: Basic needs and bad luck
Could healthcare and disability costs actually go to zero if we had a very young, generally healthy population? Clearly not. To explore the possibility though, we constructed a model that would approximate such a population. We selected characteristics that correlate with lower costs. We took a young (late 20s), mostly male, single (e.g., not having children), highly-educated, highly-paid workforce, in a region known for low-cost care, with all benefits policies and business practices aligned for optimal use of benefits.

Can you guess what it would cost to cover the healthcare spending of this virtually-risk free group? Our data says it is somewhere near $1,300 on average per year. Some costs would be associated with basic needs, and some would be the result of misfortune due to genetics or accidents. As you might expect, the majority in this population would have very small expenditures, with a few high outliers.

(My colleagues and I debated about this number because it is virtually impossible to have a population this young, highly-paid, in a specific region, and with a specific gender and marital-status profile. However, this was never intended to be an achievable situation, just the lowest imaginable.) So, the lowest imaginable total for Part I: $1,300 / adult person / year.

Part II: Demographics and labor market
Face it, age matters. Our bodies wear out. Other factors such as where we live and the type of work we do matter, too.

To explain how demographics and labor market affect cost, figure on AVERAGE:

  • Older workers spend more on health than younger workers;
  • Women cost more than men (at least up to a certain age);
  • Lower education and lower salary correlates with higher medical and absence costs; and
  • Workers in some regions spend more (North East) than others (Rocky Mountain).

Companies naturally hire a workforce with the skills and characteristics needed for the services and products they produce. One company might attract an older, mostly female, less-educated workforce who will earn minimum wage in Minnesota. Another might attract highly-skilled, younger, male engineers in Boston. Because most companies tend to have a consistent labor market to choose from, and because the demographics of those hired rarely involve drastic changes in type of workers, level of pay, or location, we consider the “Demographics and Labor Market” part of cost to be largely non-modifiable. *

To illustrate the impact of this component, a company in New England with an average employee age of 40 and hourly workers making $40,000 per year would be expected (OTHER COMPONENTS KEPT EQUAL) to add another $1,500-$1,800 per employee above the basic ($1,300) amount from Part I. The same group aged 50 or 60 years would add almost $2,200 or over $3,400 per employee, respectively. (click on the table to enlarge)

Non-modifiable total for Parts I & II: These two components can vary, as we see, from under $1,300 (in our “lowest cost” situation) to $4,700 in the extreme. For the companies we work with, the non-modifiable portion often sits in the $2,500 to $3,500 range. However, total costs for these companies often range from under $4,000 to almost $6,000 per person per year! So if basic costs, bad luck, labor pool and demographics only account for about 60%, where does the rest of the cost come from?

Now for the modifiable parts. By modifiable, we mean something that can be altered by the individual, and/or influenced by the employer. Above, we categorized demographics as non-modifiable because you cannot change them unless you change who you hire. Modifiable factors are those you can theoretically change in the people you already have.

Part III: Health status

When health declines, costs go up. Naturally, we put this component in the ‘modifiable’ section of cost, because each of us can decide to what degree we avoid risk and protect our health.

Once again, to isolate the influence of health, we hold constant the basic needs, bad luck, demographics and labor pool factors described above. In the end, our research finds that a 10% improvement in health will influence and reduce costs by about the same amount, between 7-11%.

Here is what I mean:

  • A 10% decrease in the number of diagnoses people have correlates to a medical and disability cost difference of 11%.
  • A 10% decrease in the number of medications people receive results in a medical and absence cost difference of 7%.
  • A 10% improvement in self-reported health status (a 10% average group score on a scale from poor to excellent, e.g. from 3.0 to 3.3) correlates with combined medical and disability cost decrease of approximately the same amount, 9%.

For those who want a more technical explanation…basically these analytic models tell us that when health-related metrics indicate that when the same population (same demographics, jobs, work environment, location) is 10% healthier, they will be about 10% less costly. If the population is 20% healthier, we would expect them to be 20% less costly.

So let’s do the math. If a group has non-modifiable costs (from Part I & II) of around $4,500 per employee, their total costs could be = $4,050 if they had 10% better health status than average people of that age/gender/location, etc.

What catches my attention is the magnitude of change in health status required to produce a cost savings. On the one hand, it validates what we all know: if we live healthy lifestyles and avoid many of the preventable illnesses we develop as we age, we will feel better and cost less.

On the other hand, after evaluating the broad range of health programs, disease management, case management, and wellness efforts available to employers across the US, I don’t recall a workplace intervention that produced a full 10% improvement in actual health status across all employees.

My colleagues in the wellness industry will likely object to this number, because published studies have shown larger dollar figures regarding the cost of health risks. We often quote studies saying that a risk factor equates to thousands of dollars. While this is a topic for a different blog, it reminds us to look carefully: have we ever seen a case where every member of the entire workforce successfully eliminated a full risk factor? If we convince only part of the population to accomplish this, the dollar value will be a fraction of the full value.**

So, if improving health status by 10% would reduce costs by about 10%, what else is there?

To isolate the independent effects of the first three parts above, we held everything else constant: not only characteristics of the workers, but also the employee policies and business practices in their place of employment. This set of predictors is usually left out of health-related modeling, but you’ll soon agree, it shouldn’t be.

Which leads us to Part IV.

Part IV: Business Practices

In the next blog, we will cover a critical and often overlooked driver of healthcare and absence costs: business practices. All of the underlying incentives inside an employer culture matter—how benefits are designed, how people are paid, how they are managed—each influences utilization of benefits. More often than not, these factors have a stronger influence on cost than health status.

Why this matters: In some instances, we overestimate the influence of health status on healthcare and absence costs. As a result, companies presume that ever-increased spending on health programs will reduce their overall healthcare costs. Being realistic about what really drives benefits costs will help organizations make investments in their human capital that more effectively produce return-on-investment.



*In this brief discussion we are really only looking at “demand-side” components of cost; we are not going to address how the supply-side (meaning differences across providers) affects costs, although we acknowledge this phenomenon is very real. To some degree this is included in regional differences.

**Certainly, some of the differences in different regions of the country, which we are attributing to demographics here, reflect differences in health habits. We know from national statistics that obesity rates and smoking rates differ in different parts of the country. In this discussion we are taking the point of view that the employer has a specific labor pool to choose from. From here we ask: how would a change in health status in this labor pool affect costs?


Leave a reply