Patient puzzlers for healthcare | Capability, stability, predictability

NEW! Generate SPC charts with the new release of CHARTrunner Lean.

Have questions? Click to chat with one of our service and support staff.

PQ Systems provides customers like Cleveland Clinic Health with user-friendly SPC software (statistical process control) for their data analysis.

Patient puzzlers for healthcare

Capability, stability, predictability

Lewis N. Clark considers himself an explorer when it comes to statistical process control. His job as deputy quality manager for a major health care equipment manufacturer puts him in touch with a great deal of data, and he likes to pursue his own version of data mining, which to him means making a variety of different visual expressions of the same data.

In the data that has come to his attention is a breakdown of end users of the firm’s adhesive patches used for a variety of medical tests, including electocardiogram screenings (EKGs). Clark decides to sort this data by age, since he believes that the information produced by a frequency distribution will be useful in the company’s marketing efforts.

Lewis makes the following frequency distribution chart:

Percentages of users of adhesive patches, by age:

Class Percentage (%) size

85 and older
75 - 84
70 - 74
65 - 69
60 - 64
55 - 59
50 - 54
45 - 49
35 - 39
30 - 34
25 - 29
20 - 24
15 - 19
10 - 14
5 - 9
under 5

5.7
15.8
7.0
7.4
12.4
11.2
9.2
8.1
5.2
3.1
2.7
2.2
1.1
0.9 0.7
---

The data suggests some interesting conclusions that Clark has already begun to make, based on his analysis of frequency. For example, the largest group of users lies in the 75-84 age group, so more attention should be given to that customer base, he surmises. What has he done in creating this distribution that will render some of his conclusions inaccurate?

A. overlooked rules for creating class intervals

B. created groups that are too nearly equal in size

C. rounded off data

D. aggregated data so it has no meaning

A is the correct answer.

Conclusions based on this frequency distribution will be flawed because Lewis N. Clark has not bothered to determine appropriate intervals for his data, and has indeed created two different intervals, one of 5 years (most categories, including 20-24, for example), and one of 10 years (e.g., 75-84). The number of users in the 10-year interval groups would of course be larger than if that had been distributed into 5-year groups. His conclusion is therefore not correct.

Setting up frequency distributions and histograms so that data can be analyzed accurately involves several steps. To learn these steps, view a sample chapter from Sandra and Byron Murray's book, Practical Tools for Healtchare Quality.