|
Complete Article from Edition 44 APG eNewsletter
Article written by James Hoban , Quality Control Supervisor, APG
APG's customers cover a wide range of commercial, industrial, and government laboratories around the country and across the world. APG is dedicated to committing resources aimed at informing and educating current and potential customers about the value of Proficiency Testing and Quality Control monitoring programs. Opportunities, such as this newsletter, conference presentations, and personal visits, are used to demonstrate how APG's products and experience can help improve processes or aid laboratories in attaining regulated or voluntary measures, as well as providing laboratories a benchmark for displaying their own expertise.
The Quality Control and Proficiency Testing samples available at APG allow for point-in-time assessment of performance. Analysis of these samples can highlight an immediate need for corrective action. However, routine use of Quality Control samples combined with effective data harvesting can provide a laboratory with powerful tools to indicate need for preventative action, thereby increasing continuous throughput and reducing the chance of a Proficiency Testing remediation or process failure.
Control Charting begins with the collection of percent recovery data for known samples. This data is compiled electronically using a commercially available or custom provided statistical program. Twenty data points are typically collected before the Mean and Standard Deviation are calculated, along with the values for twice and three times the Standard Deviation. The latter two calculations are approximations of the 95% and 99% Confidence Intervals respectively.
Data |
Mean |
Standard Deviation |
2s |
3s |
|
Data |
Mean |
Standard Deviation |
2s |
3s |
99.0 |
98.5 |
6.4 |
12.9 |
19.3 |
|
96.0 |
100.1 |
6.5 |
12.9 |
19.4 |
94.0 |
98.4 |
6.3 |
12.7 |
19.0 |
|
96.0 |
99.8 |
6.4 |
12.8 |
19.2 |
100.0 |
98.9 |
6.2 |
12.4 |
18.6 |
|
104.0 |
100.1 |
6.3 |
12.7 |
19.0 |
104.0 |
99.2 |
6.2 |
12.4 |
18.6 |
|
99.0 |
100.5 |
6.3 |
12.5 |
18.8 |
105.0 |
99.5 |
6.2 |
12.4 |
18.6 |
|
105.0 |
101.2 |
6.2 |
12.5 |
18.7 |
105.0 |
100.0 |
6.2 |
12.4 |
18.6 |
|
97.0 |
101.5 |
6.2 |
12.3 |
18.5 |
90.0 |
100.0 |
6.3 |
12.6 |
19.0 |
|
101.0 |
101.2 |
6.1 |
12.2 |
18.2 |
103.0 |
99.6 |
6.3 |
12.5 |
18.8 |
|
101.0 |
101.0 |
6.0 |
12.0 |
18.0 |
107.0 |
100.0 |
6.3 |
12.7 |
19.0 |
|
98.0 |
100.5 |
5.9 |
11.9 |
17.8 |
110.0 |
100.7 |
6.6 |
13.1 |
19.7 |
|
96.0 |
100.5 |
5.1 |
10.1 |
15.2 |
For example, the tables show a subset of twenty data points extracted from a larger record. The values begin with 99.0 in the leftmost portion and end with the data point 96.0 on the right. The twenty data points in this subset have a calculated mean of 100.5 and a standard deviation of 5.1, with the 2s and 3s values included. From this data, a Control Chart is created by plotting points for the data along with points for the mean plus and minus each of the final standard deviation calculations.
These charts present a graphical visualization of process performance. The widest limits at 3s, or the Control Limit, represent a one percent chance the next sample tested will exceed those limits. As an initial application of the Control Chart, when the process is adjusted to ensure the Control Limits remain within allowable maximum ranges, there is only a one percent chance the process will fail on or before the next Quality Control check execution.
Further review of the Control Chart can provide valuable information on the need for preventative action. There are a number of trending situations identified by their statistical unlikelihood that can indicate a shift occurring in the process. These trends include:
- A run of eight points on one side of the center line
- Two of three consecutive points outside the 2s Warning Limits
- Four of five consecutive points beyond the 1s limits
- An unusual or nonrandom pattern in the data
- Several points near a warning limit
Control Charting provides the laboratory with the ability to recognize a need for and execute necessary investigation or maintenance procedures to continue. When operating with good confidence, failures will not happen. Additionally, the information collected can be used to estimate the uncertainty value associated with all results reported by the laboratory.
A combination of the standard deviation and the number of data points collected expresses the uncertainty. When this data is collected and stored, the laboratory has presentable records showing reported results are accurate within a calculated precision. An acceptable calculation of the uncertainty is twice the current standard deviation divided by the square root of the number of data points.
This table shows a running assessment of the laboratory's uncertainty values. This value, divided by the current average, translates to a percentage value that is included with any laboratory result. For instance, a reported value of 2 obtained at the end of the table provided is expressed as 2 +/- 0.08.
Control Charting gives the laboratory confidence in both the ongoing process and the precision of any values reported. The cultivation of available data allows early warning of possible issues and action taken before out-of-control results are received. This method of proactive review and preventative action creates an environment where corrections are made by scheduling when they are convenient instead of as a process halting necessity.
|
