ANZBMS - Australian & New Zealand Bone & Mineral Society 

Resouces INDEX
Patient Information
Bone Densitometry
Refs. & Reviews
Policies
Grants/Awards
Journal Subscription
ASM Resources
Links
 
Bone Densitometry
Accreditation Guidelines for Bone Densitometry[ TABLE OF CONTENTS ]

APPENDIX 5 - Detection of "out of control" behaviour: preparation of a multi-rule Shewart chart

The QC parameters obtained from serial phantom measurements can be analysed in various ways to determine drifts, changes in precision and other factors which may affect machine performance over time. Well-established statistical techniques are available for this purpose, usually involving the generation of a "control chart" with appropriate control limits. The actual techniques chosen at a particular centre will depend on its resources, and to some extent its particular preferences. Choice of control limits is a critical issue, requiring that sensitivity to "out of control limits" situations is balanced against the "false alarm" rate. Advice should be sought from a statistician.

One simple method for detecting changes in precision is to examine the cumulative precision value for the densitometer daily QC phantom, as calculated by the densitometry software. Any form of drift, systematic (ie. in one direction) or otherwise, will cause this value (usually <0.5% when expressed as CV [see Appendix 3) to increase. Usually a centre will not have the resources to perform regular precision measurements in-vivo on young volunteers (other than at the time of commissioning of the machine), but ideally in-vivo precisional errors would be charted successfully if this exercise was repeated once every 6 months.

Another method, which can be set to acceptably balance the "false alarm" rate against the "out of control limits" rate, and which is statistically straightforward is multi-rule Shewart charting. Depending on the rule or rules adopted (see below) Shewart charting can identify the growth of both random and systematic errors.

The Construction of a Shewart chart begins with the assembly of around 20 consecutive phantom QC measurements of BMD, or another chosen QC parameter to be followed over time. These data are used to derive the Mean and Standard Deviation (SD) for further evaluation. One or more of the following rules may then be applied, to determine the significance of variations in the QC parameter over time. It is clear that depending on the choice of rules, either systematic or random errors will be highlighted (or a combination of both).:

  1. Warning rule: A control (phantom) measure exceeds the mean ± 2SD of the baseline phantom measures. This occurrence should prompt additional inspection of control data with the following rules.
  2. Three SD rule: A control measure, which exceeds the baseline mean ± 3SD indicates the need for instrument evaluation.*
  3. Two SD twice rule: Two consecutive control measures that exceed the mean + 2SD, or mean - 2SD, dictate instrument evaluation.a
  4. Range of 4SD rule: When the difference between two consecutive control measures exceeds 4SD (specifically when one measure exceeds + 2SD and another exceeds - 2SD) the instrument requires evaluation.a
  5. Four ± 1SD rule: When four consecutive measures exceed the same limit (+ 1SD or - 1SD) instrument evaluation a is required.
  6. Mean x 10 rule: When 10 consecutive control measures fall on the same side of the mean, instrument evaluation a is necessary.

* Instrument evaluation shall involve repeated (five to ten) control measures. With repeated failure, patient measures shall be suspended until the instrument is more thoroughly evaluated.


<< Appendix 4 - Measurement of short
      term precision in-vivo
Appendix 6 - Construction of a Cusum plot >>
[ TABLE OF CONTENTS ]
Jump to top
 
    HOME  |  ABOUT US  |  MEMBERSHIP  |  NEWS  |  EVENTS  |  RESOURCES  |  CONTACT US  
Content copyright 2005 © Australian & New Zealand Bone & Mineral Society
Design copyright 2005 © Anne Kovach, eRadiate
 
-- This page last edited: 06 Dec 2002 --