If we only consider one possible cause, we will be left with only one type of solution. Health status is not the only predictor of medical costs.
Here is a comparison of medical costs for two different companies. Company #2 workers are 84% more costly. Can you guess why?
Does Company #2 have older workers? Do they have more chronic illness? Do they live in a geographic region with higher rates of obesity or smoking?No, no, and no.
In fact, these two groups are as identical as we could make them in our modeling: same type of workers, same age, same gender, same job, same salary. So, why is there a difference in costs? (We give you the answer later.)
Costs and Health
More than once per week I discuss medical or disability costs with someone who works in health promotion or human resources. Invariably, I’m told that the explanation for medical costs is health status. If we are discussing a group with low costs, they credit health improvement programs and their employees’ dedication to health. If we are discussing a group with high costs, they blame illness, and believe that the demographic just happens to be unhealthy.
At one time, I too believed that medical costs and disability rates were driven primarily by illness in a population. Now I know better.
It’s not that I am any smarter than I was before. It’s just that I now have access to better information. In our Health as Human Capital Research Group, we are able to integrate many sources of data to tease apart the connections between health and costs as well as work and policies. As a result, my thinking has changed. Here are some highlights of what I have learned.
1. Yes, health status correlates with health-related outcomes, but the association is much smaller than is often presumed.
At the extremes, costs do correlate strongly with level of illness. A 60-year-old person who has five chronic illnesses will almost always have higher costs than a 20-year-old with no diagnoses. However, as we look closer the association soon gets murky. Ten people with similar levels of illness can vary dramatically in medical costs and absences from work. Often, a person with seemingly serious illness has lower costs than a person with seemingly minor illness.
Statistically speaking, general health status explains less than 20% of the variation in medical spending.(a) Except in the most extreme examples comparing terrible illness to perfect health, medical costs are not a very good proxy for health status. Two people with the same level of illness are likely to have very different costs.
The other 80% of variation in spending reflects people’s individual preferences and incentives, such as: Do I understand my treatment options? Do I trust doctors? Do I have to pay for this service myself or does insurance cover it? Do I believe that protecting my own health will help me advance my career? How many days off do I have left? Do I find my work highly rewarding? With so many other considerations in the equation, it is accurate to say that health status and medical costs do not at all reflect the same thing.
I find that many of us in the health professions resist this reality. We resort to believing that high medical costs mean that more health improvement investments are needed. We are adamant that someday, improving health will result in lower costs. And, not surprisingly, vendors of health interventions do not want to hear that their products and services may not be the most effective way to manage costs.
2. Policies (health plan design and time-off policies) have dramatic effects on benefit utilization, without much impact on health status.
There is overwhelming evidence, in our data and in large randomized trials, that humans respond to incentives like those built into insurance policies. We have discussed the best study-- by RAND-- in other blogs, showing that free medical care results in much higher (40%) utilization, with little or no change in health status. When individuals pay more for medical care, they use less—again with little effect on health status.
Translation: much of what we spend on medicine has little or no impact on health.
It’s what we’ve been told.
If the only data available demonstrate a correlation between health and medical costs, then it’s the only conversation anyone is having. We have all seen hundreds of examples of studies documenting the costs associated with health risks and health conditions, a few of which are referenced at the bottom of this blog (1-4; I admit I have also co-authored a few myself). If illness is the only CAUSE we have seen, it makes sense that we think improving health status is the only way to lower medical costs.
Am I saying that these studies are wrong? No. But they are not the whole story. And because they omit other factors, the effect gets over-emphasized and other explanations get lost. Until I was able to see more complete information, I focused on this association too.
In our own data, we certainly see the same general relationship—IF that is the only factor we isolate in our analysis. Each additional risk factor or disease increases average medical costs by a measurable degree. So, one might assume that health improvement should be the primary cost reduction strategy, right?
In fact, these two groups are as identical as we could make them in our modeling: same type of workers, same age, same gender, same job, same salary. So, why is there a difference in costs? (We give you the answer later.)
Costs and Health
More than once per week I discuss medical or disability costs with someone who works in health promotion or human resources. Invariably, I’m told that the explanation for medical costs is health status. If we are discussing a group with low costs, they credit health improvement programs and their employees’ dedication to health. If we are discussing a group with high costs, they blame illness, and believe that the demographic just happens to be unhealthy.
At one time, I too believed that medical costs and disability rates were driven primarily by illness in a population. Now I know better.
It’s not that I am any smarter than I was before. It’s just that I now have access to better information. In our Health as Human Capital Research Group, we are able to integrate many sources of data to tease apart the connections between health and costs as well as work and policies. As a result, my thinking has changed. Here are some highlights of what I have learned.
1. Yes, health status correlates with health-related outcomes, but the association is much smaller than is often presumed.
At the extremes, costs do correlate strongly with level of illness. A 60-year-old person who has five chronic illnesses will almost always have higher costs than a 20-year-old with no diagnoses. However, as we look closer the association soon gets murky. Ten people with similar levels of illness can vary dramatically in medical costs and absences from work. Often, a person with seemingly serious illness has lower costs than a person with seemingly minor illness.
Statistically speaking, general health status explains less than 20% of the variation in medical spending.(a) Except in the most extreme examples comparing terrible illness to perfect health, medical costs are not a very good proxy for health status. Two people with the same level of illness are likely to have very different costs.
The other 80% of variation in spending reflects people’s individual preferences and incentives, such as: Do I understand my treatment options? Do I trust doctors? Do I have to pay for this service myself or does insurance cover it? Do I believe that protecting my own health will help me advance my career? How many days off do I have left? Do I find my work highly rewarding? With so many other considerations in the equation, it is accurate to say that health status and medical costs do not at all reflect the same thing.
I find that many of us in the health professions resist this reality. We resort to believing that high medical costs mean that more health improvement investments are needed. We are adamant that someday, improving health will result in lower costs. And, not surprisingly, vendors of health interventions do not want to hear that their products and services may not be the most effective way to manage costs.
2. Policies (health plan design and time-off policies) have dramatic effects on benefit utilization, without much impact on health status.
There is overwhelming evidence, in our data and in large randomized trials, that humans respond to incentives like those built into insurance policies. We have discussed the best study-- by RAND-- in other blogs, showing that free medical care results in much higher (40%) utilization, with little or no change in health status. When individuals pay more for medical care, they use less—again with little effect on health status.
Translation: much of what we spend on medicine has little or no impact on health.
It’s what we’ve been told.
If the only data available demonstrate a correlation between health and medical costs, then it’s the only conversation anyone is having. We have all seen hundreds of examples of studies documenting the costs associated with health risks and health conditions, a few of which are referenced at the bottom of this blog (1-4; I admit I have also co-authored a few myself). If illness is the only CAUSE we have seen, it makes sense that we think improving health status is the only way to lower medical costs.
Am I saying that these studies are wrong? No. But they are not the whole story. And because they omit other factors, the effect gets over-emphasized and other explanations get lost. Until I was able to see more complete information, I focused on this association too.
In our own data, we certainly see the same general relationship—IF that is the only factor we isolate in our analysis. Each additional risk factor or disease increases average medical costs by a measurable degree. So, one might assume that health improvement should be the primary cost reduction strategy, right?
Well, let’s look back at our original graph. This is a cost difference that is NOT attributable to health.
Using a single model, we took medical costs for all people in two companies with a similar job. Then, we accounted for a list of predictors of costs (demographics, health factors, geographic location, company, etc.). We then used the model to predict what expected costs would be for a group with a specified set of characteristics—the same type of people in both companies. Essentially our projection study asked, for a specific type of person, what would we expect the costs to be in each company?
The graph shows the result. For a group of administrative workers who are two-thirds female, average age 46, employed for eight years, making $43,000, living in similar regions, having four diagnoses in a year, and a specified level of illness according to a co-morbidity index,(b)these are the expected costs in these two companies. By controlling for so many things, we are ruling out health differences in these predictions. But for those of you who are still skeptical, here is some other information: Company 2 has more wellness programs, and has a more highly-educated general population.
The reasons the first company has $2,200 lower costs can be attributed to the policies of the organization, which include a high deductible, a consumer-directed account, greater shared responsibility for costs of extended time-off, and general culture of communicating the cost of benefits.
To further demonstrate the power of these policies compared to health status---to overcome this $2,200 difference and make the companies’ costs equivalent, our model predicts that employees at Company #1 (now LESS expensive) would have to average four more risk factors (smoking, obesity, etc.) per person than employees at Company #2.
Why this matters:
Our point here is that while health status is part of the cost issue, if we ignore incentives and benefit design while investing in health, we overlook powerful and sometimes predominant determinants of cost. There are enormous variations in cost that result from the structure of a company’s incentives.
The next time someone tells you his company’s medical costs are high because their people are sicker, tell him he may be missing the other 80% of the problem.
The graph shows the result. For a group of administrative workers who are two-thirds female, average age 46, employed for eight years, making $43,000, living in similar regions, having four diagnoses in a year, and a specified level of illness according to a co-morbidity index,(b)these are the expected costs in these two companies. By controlling for so many things, we are ruling out health differences in these predictions. But for those of you who are still skeptical, here is some other information: Company 2 has more wellness programs, and has a more highly-educated general population.
The reasons the first company has $2,200 lower costs can be attributed to the policies of the organization, which include a high deductible, a consumer-directed account, greater shared responsibility for costs of extended time-off, and general culture of communicating the cost of benefits.
To further demonstrate the power of these policies compared to health status---to overcome this $2,200 difference and make the companies’ costs equivalent, our model predicts that employees at Company #1 (now LESS expensive) would have to average four more risk factors (smoking, obesity, etc.) per person than employees at Company #2.
Why this matters:
Our point here is that while health status is part of the cost issue, if we ignore incentives and benefit design while investing in health, we overlook powerful and sometimes predominant determinants of cost. There are enormous variations in cost that result from the structure of a company’s incentives.
The next time someone tells you his company’s medical costs are high because their people are sicker, tell him he may be missing the other 80% of the problem.
___________________________________________________________________
a) Recently we predicted medical costs in a large population from age, gender, and a rating of general health. The R-square statistic was .11, or 11%. That means that if we take all of the variability in medical costs, we can only predict 11% of it by knowing a person’s age, gender and health rating. The remaining 89% of variability is related to something else.
b) Charlson Comorbidity Index.
References
1. Anderson, D., S. Brink, and T. D. Courtney. 1995. Health risks and their impact on medical costs, Milliman & Robertson, Inc., Brookfield, WI.
2. Goetzel, R. Z., K. Hawkins, R. J. Ozminkowski, and S. Wang. 2003. The health and productivity cost burden of the "top 10" physical and mental health conditions affecting six large u.s. Employers in 1999. J Occup Environ Med 45, no. 1: 5-14.
3. Thorpe, K. E. , C. S. Florence, and P. Joski. 2004. Which medical conditions account for the rise in health care spending? Health Aff (Project Hope) : 10.1377/hlthaff.w4.437 (accessed February 29, 2008).
4. Weaver, M. T., B. G. Forrester, K. C. Brown, J. A. Phillips, J. C. Hilyer, and E. I. Capilouto. 1998. Health risk influence on medical care costs and utilization among 2,898 municipal employees. Am J Prev Med 15, no. 3: 250-3.
posted by Wendy Lynch, Ph.D. at
9:58 AM
0 Comments:
Post a Comment
<< Home