Data

Data, don’t make decisions without it. How many of yesterday’s solutions are today’s problems? Overcoming Numerical Literacy

If your measure doesn’t drive a decision, it is probably is not worth much.
If the decision doesn’t make a positive difference, then it is probably not worth much.

We are suffering from chronic Numerical Literacy (coined by Don Wheeler)

The higher they go up the ladder, the more illiterate they get. They fall into the react mode. Journalist and lawyers have the worst case of this, since they gain value by using data to prove their point.
Life is full of variation and random variation is part of every process.
We love data: usually for judgment, but not much for quality.
- There are three things people usually do when they hear bad results:

Leadership response: Good data – good doggie; Bad data – bad doggie.

You can not improve quality in a quarterly format.
- Aggregate data is little good, for it gives no feel of central tendency, and no sense of dispersion. Both are needed to characterize reality.
Three types of data answer which answer three types of questions:

People tend to see trends where there are no trends. Up-good, Down-bad. When you give credit or punish people over what they do not control, you don’t get progress, you get confused fear
- Static data is a snapshot of the process: Tables/pie charts/histograms fall in this category
- Dynamic data is a motion picture of the process: line data shows this
- Six successive data points steadily in an upward or downward constitute a trend
- A pattern is a migration of data up or down over time
Special Cause is non-assignable variation. It is not a part of random process variation
- If arbitrary goals are not grounded in the process, they will create frustration
- If you don’t understand how your data breathes, you will make erroneous judgments.

Some hints on how to improve measurement

Graph data over time
Local collection for local use
Understand when you are ‘tampering’
A smaller set of data with faster feedback is better than larger data with sloooower feedback. Instead of large slow collected data, use small samples and evaluate vigorously
Encourage public posting of data – it makes people ask questions. And questions usually help.

Quick notes taken at Bob Lloyd’s presentation (author of Statistics in Healthcare)

If you would like to provide feedback or discuss the content of this article, please visit the discussion page.

To be automatically notified by email of any changes or additions to the article or discussions about this article, please click here.