Control charts Significant changes

Performance test results should be plotted on charts for trend analysis. Statistical quality control charting methods are used to detect statistically significant changes in instrument performance. Action limits can be set for test charts based on historical data so that appropriate repairs can be made when necessary. Examples of unacceptable instrument performance may be valuable in setting action limits for future performance tests. 

The SDC plots cover both the actual OD and the processed Pl% values. It is important that both types of plot be examined together. Thus, the charts should be placed in close proximity. This should immediately alert operators to unusual fluctuations from the expected values in both charts as well as deviations in one chart but not in the other. When the Cm OD values are within limits, it is expected that the control serum values for OD and hence Pl% will be within limits. When the Cm OD values do start to increase or decrease significantly (approach limits), it is possible that "alterations" in OD value are not mirrored by a significant change in the control Pl% plots. In other words, when Cm OD values are showing trends, as shown in Fig. 23, the control C++ and C+ PI% charts should be consulted to measure the effect of the OD changes on the PI% values. 

As in statistical control work, points that fall outside the upper or lower control limits may be of statistical significance. For example, two successive out-of-control points may indicate that the population being observed has changed. That is, conditions in the work environment may have changed or process/work improvement been realized. Control charts can show the progressive improvement work, thus triggering the need to change the standards applied. 

To illustrate an example of when it may be necessary to recalculate a control chart due to significant changes the following scenario was constructed ... 

If there is a significant change identified on a control chart, what action can be taken ... 

If entries were made in a control chart for each of 24 months for OSHA incident rates, adding a month as a month was dropped, the depiction would indicate whether performance was in control in relation to computed upper and lower control limits. If performance spiked in either direction, the indication is that a significant change had occurred for which inquiry would be made. 

The control chart has the advantage that changes over time can be seen clearly and, on occasion, significant departures from the acceptable range predicted. The predictive value of control charts should not be overlooked. However, unless they are assessed frequently by a competent analyst, trends suggesting future problems will be missed. 

Is the LTI-rate sensitive to change To put the question in another way, will a change in the risk level at a workplace show up in the control chart as a significant jump in the LTI-rate The answer is most often no . This... 

The third simple chart type is the Control Chart initially developed by Shewhart. Control Charts can be used for both variable and attribute data. The various chart types deal with both some measure of change in an average value over time as well as changes in the variability of the system. In the context of process control, the measurements are frequently of process variables that are recognized from previous efforts with Pareto Charts and C E Diagrams to be significant in the conversion of inputs to outputs that meet customer expectations. 

The key idea about control charts is that if an observation is outside the control limits of the chart or if a significant pattern is seen in the chart the production process has changed and is labeled out of control and should thus be investigated. If the process is found to be out of control , the problem should be identified and corrected before production continues. 

Implementation by Use of the Z and Chi-Square (yf) Tests Charts for mean, range, and standard deviation are seldom found in clinical laboratories because of the time required for calculating the control statistics. These control procedures are probably practical only when computerized data handhng is available. An easy way of implementing these procedures on computers is to employ statistical tests of significance. A Z-test can be used to determine whether the mean has changed from its original value, and a can... 

---