Friday, March 30, 2007

Bad bets in health care

I hope this link stays active long enough for anyone interested to read it. It makes an interesting point about public policies that are "bad bets" that have a strong human factors component.

Consider this example: There are two people who are very sick and require medical care that costs $25,000 to treat. Person A has a 50% chance of getting better with treatment. Person B has a 5% chance of getting better with treatment. You can either:
1) give Person B the treatment
2) give Person A the treatment
3) split the money and half treat them both, which reduces their chance of improving by 90% (so it would be 5% and 0.5% respectively.

Most people would select option 2. It gives the best bang for the buck. But what if there is no Person A. We can either:
1) treat person B and have a 5% chance of having an effect
2) distribute the $25,000 in medical care to other people in general, who have an average chance of improving of 50%.
3) distribute the $25,000 to the entire population to spend however they want.

Now which one should we do? Option 2) in the second situation is identical to option 2) in the first example in its effect. So if you preferred option 2) in the first example, you can't choose option 1) in the second. But the problem faced in our health care system is that when we are faced with a real Person B on one hand and a vague, unseen population on the other, we find it very hard not to treat person B and hope for the best. And the when we have a lucky 5%-er, it makes the news and makes it even harder to choose option 2) the next time.

This is why our health care system is so wasteful. We keep funding the bad bets because we can't look at a sick person and say no.

Tuesday, March 20, 2007

after a big error

When is the best time to be a patient? Ironically, it is immediately after your doctor makes a salient error. The reason is simple - that is when he/she is going to be paying the closest attention and double checking everything. If the error was a slip (insufficient attention) then the doctor will be paying extra attention. If the error was a mistake (incorrect situation schema) then the doctor will be double checking and studying the situation. If the doctor doesn't know why the error was made, he/she will probably be doing both.

Ironically, the way most organizations deal with big errors is to suspend the person from doing the activity. Quarterbacks are taken out of the game, employees are suspended, children are grounded, etc.

The better thing to do is to send them right back out with maximum repetitions so that the extra carefulness gets cemented into their normal routine. At the same time, the system needs to be evaluated to identify the root cause of the error. Are employees not sufficiently trained to recognize situation schema? Are there distractions that hide important information? Is the information not readily available?

Of course, this is when it was an error. If the person intentionally violated a rule, then the correct response is punishment. But people violate rules for many reasons. It may be that the rule was not presented with sufficient importance and consequence. Or the person was trying to satisfy one requirement by violating another. Even intention does not 100% indicate that the root cause is the violator him/herself.

This is a subject for a different post. Or you can take my upcoming summer course at FIU on performance management. It is available by Internet if you don't live in Miami.

Monday, March 19, 2007

speech fire alarms

I heard today about a new product that claims to be a more effective smoke/fire alarm. Instead of a siren, the alarm is a recorded voice. The idea is that when little kids wake up to an alarm, they are disoriented and confused. To hear specific instructions coming from their mom's voice is supposed to be helpful. The recording can be more than just "wake up." It can have detailed instructions like "There is a fire!! Get up and go out the back door right away!!!"

But I am conflicted on whether this is really better. I think that from a pure sensory salience perspective, there is not way that a voice is as loud or alarming as a siren. Also, the mom's voice is a common thing, so they won't instinctively associate it with an emergency. If they are sleeping, they may not hear enough of the voice to know it is the fire alarm. Also, as early as 10 years old, I am sure kids are trained to ignore their mom's voice waking them up: "Mom, I don't want to go to school today."

So the tradeoff is that a siren is better to wake them up and the voice is better to remind them of where to go. So which is more likely, that they fail to wake up or that they panic and forget where to go? I don't know this answer, but I suspect that the designers did not do the study to find out either. They guessed that this would be better and the lives of thousands of children will depend on them being correct. I hope they are.

airline pilot pay

The salaries of airline pilots is a great example of a way to structure pay that is misaligned with performance. The hardest part of flying is takeoff and landing. So if salaries is based on the task difficulty, then short haul pilots would get paid more. But in fact it is the opposite.

I suspect that these salaries are based on the size of the plane. Long haul flights generally have bigger planes. This could make a little sense if it is based on responsibility - more people in the plane = more people relying on the pilot's skills. But if takeoff and landing are really the hard parts, then many smaller flights still has more responsibility.

Or, they are influenced by seniority. Since longer flights are easier (the plane mostly flies itself), more senior pilots get them. Basing pay on seniority is somewhat controversial. It is common in unionized organizations. It makes sense to reward employees for staying at the company, especially in industries where there is tacit knowledge that makes the employee better at his/her job. But I personally think it is better to link the salary to this tacit knowledge directly - not the predictor of job tenure.

But this is a good example of the complexity of the real world. If pay were structured based on what really motivates performance (long term and short term), seniority would have some impact, but not nearly as much as it does. Politics and other non-engineering factors still play a large role.