Yeah, I know what you are thinking–health spending statistics … “booooring”! Please bear with me–I will try to make it as painless as possible. The national health spending per capita ($9,255 in 2013) that I have described in a past blog post is an average (mean) value. Your knowledge of statistics probably dates back to high school where you were given equations to memorize but never fully appreciated how the information may become important in your life. This lack of basic understanding is constantly being used to sell you something–usually something you don’t need.
First of all, the words average and mean are the same thing. I absolutely hate when this happens (almost as much as I hate acronyms). The average (mean) health spending per capita (person) is obtained by simply adding up the health spending (all components) from every person in the United States and then dividing by the total number of people (roughly 315 million in 2013). This average (mean) spending number is not a measure of your specific health spending. You might have spent zero dollars for health care in 2013, but you are still one data point for the statistic.
Another statistical measure that is often used is called the median. It is simply obtained by putting all 315 million health spending data points in numerical order and reporting the number that has an equal number above and below it. The average (mean) and the median values are often different numbers and tell us different information about health spending.
The difference between the median and the average (mean) is shown in the figure below for a hypothetical risk pool of 20: where 18 people spend $500/year (low spender category), 1 person spends $25,000/year (medium spender category), and 1 person spends $150,000/year (high spender category”). The average ($9200) and median ($500) for this hypothetical health spending population of 20 people are very different values.
When the average (mean) is higher than the median, it is an indication that a greater share of the health spending is being spent by a small subgroup of individuals who are heavy spenders. In the case above, one person (in red) spent more than five times the combined spending for all the other 19 people in the risk pool! The heavy spender’s health costs made it seem like the health spending for the majority of the population was higher than it really was if you looked at the average (mean) statistic. In any given year, the majority of Americans are low spenders (i.e., relatively healthy) and therefore the median health spending value is a better measure if you want to know what the majority of the population is spending.
Health Spending Statistics: Composition (Distribution)
In addition to average (mean) and median health spending statistics, the distribution of health spending is the third part of the spending story. Information about the composition (distribution) of health spending can be used in a number of ways:
- Insurance companies use the health spending distribution data (nationally and in their own risk pools) to help set insurance premium prices
- Government health policymakers assess spending distributions for cost containment and health care need. If health spending is found to be concentrated with the over 65 year old Medicare crowd or with low income, uninsured Americans, or even for the treatment of specific conditions or illnesses, then government health reform could target those areas first (most bang for the buck).
- BB’s Brigade can use this distribution data to ask important questions that affect Affordable Health Care and Beyond for All Americans
In the hypothetical example above, you can “see” the distribution of the health spending. For the health spending of our nation, we are talking about more than 315 million people and therefore we need to present the distribution of the data in some other way (315 million stick figures is a lot of stick figures!).
Let’s look at some examples of how the spending distribution can be presented to accommodate larger risk pools . Before we begin, I need to take the low spender (green) category and be more specific about each individual person’s spending. In real life, every person in the low spender category will not spend exactly $500/year. My low spenders could have the following 18 health spending values:
$0,$0,$0, $100, $150, $200, $250, $275, $300, $400, $450, $500, $600, $800, $900, $1000, $1175, and $1900. (average for low spenders = $500)
The average (mean) spending for my hypothetical risk pool of 20 has not changed (still $9200), but the median is now $425 (the average between $400 (person 10) and $450 (person 11)).
Presenting Health Spending Distribution by Dollar
I can take the spending data for the 20 person risk pool and group it into health spending dollar ranges. For example, if I chose 0-$2000 (= low spenders), $2001-$75,000 (= medium spenders), and $75,001-above (= high spenders), then the data could be presented as shown in the bar graph below. Within a given category, I take the average (mean) health spending and report the number at the top of each bar.
Grouping health spending into smaller dollar increments (e.g., $0-$1000, $1001-$2000, $2001-$3000, etc.) would be better for large populations because the distribution evaluation would be more detailed. Take the same data above and present it with these new tighter increments gives a different health spending picture. I get a better picture of what is going at the low end of the spending. The data is the same, just how I presented it is different!
Presenting Health Spending Distribution by Population Percentages
I could have also divided up the data by population instead of by dollar increments. In the figure below, I divided the data evenly by population; namely, the lowest spending third of the population, the next spending third of the population, and finally the last (and highest) spending third of the population. In this case, the raw data would be presented below.
Wow does this presentation of the raw data look different! Dividing the population by thirds hides the fact that two out of twenty people in my hypothetical risk pool have unusually high spending. The graph also tells me that dividing up the populations by thirds is not the best since the average spending for two thirds of the population are not too far apart.
As you can see, how I divide up the data for health spending presentation is very important to the “story” I am trying to tell. All my presentations are valid for the given raw data. I am not lying about the health spending in any of the presentations; I am simply taking the raw data and telling the story I have chosen to tell in each one of them.
The Bottom Line
Health spending statistics are important to know. When average (mean) health spending is much higher than the median, then health spending is concentrated in a small subset of high spenders. Median health spending then becomes a better measure of what the American in the middle of the pack is actually spending. Together with the average (mean) and median health spending statistics, the distribution (composition) of health spending completes the story that needs to be told.
The distribution (composition) of health spending can be presented in many ways depending on how the raw data is categorized and averaged. Be aware that you are probably getting one piece of the story in any presentation. The data is usually “packaged” for a specific audience and to convey a specific message. Insurance companies and government officials “package” data one way for internal consumption and another way for consumer consumption.