Intended for healthcare professionals


Preventing injuries in children

BMJ 1999; 319 doi: (Published 28 August 1999) Cite this as: BMJ 1999;319:574

Elements of trial's design and analysis might have biased results

  1. Elizabeth Waters, coordinator (,
  2. Patricia Priest, Nuffield fellow,
  3. Charlie Foste, research officer, British Heart Foundation Health Promotion in Primary Care Research Group,
  4. Anne Andermann, DPhil student, CRC Primary Care Education Research Group
  1. Child Public Health Unit, Centre for Community Child Health, University of Melbourne, Australia
  2. Department of Primary Care and Public Health, University of Oxford, Oxford OX3 9DU
  3. Department of Public Health, University of Oxford, Oxford OX3 7LF
  4. Division of General Practice, University of Nottingham, Nottingham NG7 2UH
  5. Roundwood Surgery, Mansfield NG18 1RH
  6. Centers for Disease Control and Prevention, Atlanta, Georgia, USA
  7. Trent Institute for Health Services Research, University of Nottingham, Nottingham NG7 2UH

    EDITOR—We overwhelmingly believe that trials should be carried out in primary care to assess child health promotion. Kendrick et al report the lack of effectiveness of a package of health injury prevention strategies delivered by health visitors.1 Before this lack of effectiveness is accepted, we recommend critical analysis of elements of the design and analysis of the trial that might bias results towards the null hypothesis. Further information is needed to enable the generalisability of interventions to be assessed.

    Factors possibly leading to bias

    • The control group was potentially contaminated by health visitors and parents in geographically close areas.

    • Health visitors were selected because of their interest in research into injury prevention; control health visitors may have increased their injury activities in concurrent health promotion initiatives.

    • Were other injury prevention activities occurring at the time?

    • A high proportion of the intervention group (22%) did not receive any intervention.

    • It is unclear whether the time sequence of the study was important as primary outcomes were measured both …

    View Full Text

    Log in

    Log in through your institution


    * For online subscription