Send letter to journal:
Re: Re: Simpson's Paradox

Dear Editor

Thanks to Dr Yen and colleagues for their interest. We agree that the possibility of a case-mix or wound complexity change could account for bias in this before and after study. However, knowing our case-mix we feel this is unlikely. Most of our wounds are simple skin lacerations and anything more complex than this goes to theatre with general or plastic surgery. It would be unusual for a wound to require more than 6 stitches in our department. The study would be more robust if we had graded wound complexity, recorded the number of sutures and length of operation and matched the two groups. We cannot however at this stage go back to retrieve this data as it will not all be recorded. Randomization would have been useful but very difficult to perform. Randomizing patients would require the doctors to use alternate methods ( traditional suturing v safe suturing) on each patient which is biased in itself. Randomizing the doctors would require half the doctors in the department to use the safe method which creates an ethical problem for the others, and as these staff work side by side one group would tell the other what they were doing thus introducing bias.

Randomizing two separate hospital units would introduce case-mix and cultural bias.

Regardless of the results of the study it was disappointing to note that needle-handling with fingers and a slightly casual attitude to needle -handling persisted after the training. This is a concern for us. This is partly related to the low level of blood-borne virus in the local population, < 1% for Hep C and lower than this for Hep B and HIV. It would be interesting to perform the study in a unit where anxiety about blood-borne virus infection was higher and see if compliance and perforation rates were any better.

Send letter to journal:
Re: Simpson's Paradox

Dear Editor

We read with great interest the paper written by McAdam et al [1]. However, we have some doubts regarding their interpretations due to the study was not a randomized study. One of the key assumptions was that they assumed the sutures were performed in a similar or identical environment during pre- and post- intervention periods. If this assumption is not true, Simpson's Paradox should be considered before interpreting the results.

They found the crude rate of glove perforations was higher in the post-intervention stage. However, some potential factors may confound this result. For example, the rate of glove perforations was found higher in complex surgeries [2]. It is possible that there were more complex sutures in the post-intervention stage. Although the non-touch suturing technique effectively reduced the rate of glove perforations, the crude rate of glove perforations in the post-intervention stage was still higher. However, if we compare stratum-specific (simple and complex sutures) rates, it is possible that the rate of glove perforations in simple or complex sutures is lower in the post-intervention stage.

To prevent Simpson's Paradox and to improve the internal validity of the study, we would suggest them to provide some important characteristics of the patients who received sutures in the emergency department during pre- and post- intervention periods. We believe that these essential information will help to clarify our concerns and will also support further interpretations with less bias.

References

1. McAdam TK, McLaughlin RE, McNicholl B. Non-touch suturing technique fails to reduce glove puncture rates in an accident and emergency department. Emerg Med J 2004;21:560-1.