Intended for healthcare professionals

Letters

Ethnicity and analgesia in accident departments

BMJ 2000; 321 doi: https://doi.org/10.1136/bmj.321.7259.513 (Published 19 August 2000) Cite this as: BMJ 2000;321:513

Authors did not exclude type II error or perform power calculation

  1. Peter Leman (peter.leman{at}gstt.sthames.nhs.uk), accident and emergency consultant
  1. St Thomas's Hospital, London SE1 7EH
  2. Barts and the London NHS Trust, Department of Anaesthetics, Royal London Hospital, London E1 1BB
  3. Clinical Effectiveness Unit, Royal London Hospital
  4. St Bartholomew's and the Royal London School of Medicine and Dentistry, Department of Environmental and Preventive Medicine, London E1 2AD

    EDITOR—Choi et al investigated whether ethnicity had any bearing on the prescription of analgesia in their accident and emergency department.1 Unfortunately, they have failed to answer the question adequately. By failing to reject the null hypothesis (that ethnicity has no effect on prescribing of analgesia in accident and emergency departments) they allow the possibility of a type II error. They did not perform a power calculation, and hence their conclusion that ethnicity is not a risk factor lacks validity.

    To detect a difference of 10% (say 80% v 70%) in prescribing rates between two groups, with a ratio of 5:1 recruitment to the study, the significance level set at 5%, and power of 90%, they would need 255 patients in the Bangladeshi group to show such a difference if it existed. The published study would seem to have a 90% power to detect a significant difference (at the 5% level) in proportions in the Bangladeshi group compared with the white group only if the Bangladeshi group had a prescribing rate of <55.3% or >94.8%.2 The published study, with an expected prescribing rate of 78.5% in the white group, has less than 30% power to detect a clinically important variation in prescribing of 10%; thus the type II error rate is too high (>70%) to allow any valid conclusions to be drawn.

    The possibility of selection bias must also be considered. The small number of patients in the Bangladeshi group does not reflect well the local population distribution (14% of the study population v >25% of the local population). Additionally, no explanation is offered for the lack of age comparability between the two groups (mean age 33.8 years for the white patients v 25.9 years for the Bangladeshi patients, P<0.05)).

    Thus these groups may have other differences besides ethnicity. The authors state that no attempt was made to adjust for potential confounding factors (although some were measured), and a more robust logistical regression model may allow greater interpretation of the data.

    In summary, the authors' contention that ethnicity does not affect analgesic prescribing in their hospital has yet to be proved. Communication with patients in accident and emergency departments, particularly those who do not speak English, can be difficult.3 I would like to see the authors show more robustly how communication with all ethnic groups in their department has no impact on patient care as this is an important and growing issue and may affect our prescribing habits.

    References

    1. 1.
    2. 2.
    3. 3.

    Authors' reply

    1. Paul Yate, consultant,
    2. Paul Kalinda, manager,
    3. Elizabeth A Paul (E.A.Paul{at}mds.qmw.ac.uk), lecturer in medical statistics
    1. St Thomas's Hospital, London SE1 7EH
    2. Barts and the London NHS Trust, Department of Anaesthetics, Royal London Hospital, London E1 1BB
    3. Clinical Effectiveness Unit, Royal London Hospital
    4. St Bartholomew's and the Royal London School of Medicine and Dentistry, Department of Environmental and Preventive Medicine, London E1 2AD

      EDITOR—We were motivated in our study by the report of Todd et al, who found that, compared with non-Hispanic white patients, Hispanic patients had a relative risk of 2.12 (P=0.003) of not receiving analgesia for long bone fracture.1 They investigated 139 patients, of whom about a quarter were Hispanic. We investigated 307 patients to see if there was a similar association in Bangladeshi patients in our hospital.

      Given that our sample was about double that of Todd et al, it seemed reasonable that there would be adequate power to detect a similar risk in our population. The relative risk in our study was 0.87 (95% confidence interval 0.45 to 1.70), so it seems unlikely that the relative risk for Bangladeshi patients in our population would be as high as 2.12. We agree with Leman that larger studies are needed to estimate the population risk with more confidence.

      The mean age was almost identical in the patients who had analgesia and those who did not (32.6 v 32.3 years, P=0.46), and the proportions of male and female patients who had analgesia were similar (79.2% male v 79.5% female patients). We therefore did not consider age and sex to be confounders in the relation between ethnicity and analgesia. With respect to other potential confounders, there were no significant differences between the white and Bangladeshi patients in bone fractures, reduction needed, admissions, or sex.

      References

      1. 1.
      View Abstract