Intended for healthcare professionals

Letters

Effects of drug overdose in television drama on presentations for self poisoning

BMJ 1999; 319 doi: https://doi.org/10.1136/bmj.319.7217.1131 (Published 23 October 1999) Cite this as: BMJ 1999;319:1131

Antifreeze poisonings give more insight into copycat behaviour

  1. Martin J Veysey, specialist registrar,
  2. Robie Kamanyire, information officer,
  3. Glyn N Volans, director
  1. Greenwich District Hospital, London SE10 9HE
  2. Medical Toxicology Unit, National Poisons Information Service (London), Guy's and St Thomas's NHS Trust, London SE14 5ER
  3. Greater Glasgow Health Board, Dalian House, Glasgow G3 8YU
  4. Department of Mental Health, Gloucester House, Southmead Hospital, Bristol BS10 5NB
  5. ICRF/NHS Centre for Statistics in Medicine, Institute of Health Sciences, Oxford OX3 7LF
  6. Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 7JX

    EDITOR—Hawton et al highlight the effect of the media on influencing the incidence of deliberate self poisoning.1 However, they and other authors suggest that the changes noted are the result of spontaneous variation in the patterns of particular overdoses rather than a direct effect of the specific televised incident.2 3 One of the limitations of previous studies has been that the investigators have monitored the total numbers of deliberate self poisoning and, specifically, paracetamol overdoses, which are comparatively common. A clearer picture emerges for agents used less commonly for deliberate self harm, such as antifreeze, which commonly contains ethylene glycol or methanol.

    The figure shows the numbers of intentional and accidental cases of poisoning by ethylene glycol reported to the National Poisons Information Service (London) during two specific months and, for comparison, from January 1996 to January 1997. In April 1995 the Independent reported an inquest into an antifreeze poisoning,4 which subsequently received further media coverage. On 15 February 1997 an episode of the BBC television drama Casualty depicted an incident of self harm with ingestion of antifreeze.

    Figure1

    Numbers of intentional and accidental cases of ethylene glycol posoning reported to the National Poisons Information Service (London)

    The mean number of intentional antifreeze poisonings for 1996 was 2.0 per month (range 1-3 per month). Moreover, the mean number of cases reported during 1995 and 1997, excluding the incident months, was 1.9 and 1.8 respectively. For April 1995 and February 1997 the number of reported cases was 9 and 6—a significant increase (P=0.016) Interestingly, all the cases of intentional ingestion of antifreeze during April 1995 and February 1997 occurred after the announcements in the media. Furthermore, in one specific case not only the agent but also the manner in which the antifreeze was taken (mixed with lemonade and drunk in a field) was identical with that reported.

    These data further support the concept that media portrayal of self poisoning influences subsequent self harm behaviour.

    References

    1. 1.
    2. 2.
    3. 3.
    4. 4.

    A causal association cannot yet be inferred

    1. Jill Pell, consultant in public health medicine,
    2. Robert Murdoch, information support manager
    1. Greenwich District Hospital, London SE10 9HE
    2. Medical Toxicology Unit, National Poisons Information Service (London), Guy's and St Thomas's NHS Trust, London SE14 5ER
    3. Greater Glasgow Health Board, Dalian House, Glasgow G3 8YU
    4. Department of Mental Health, Gloucester House, Southmead Hospital, Bristol BS10 5NB
    5. ICRF/NHS Centre for Statistics in Medicine, Institute of Health Sciences, Oxford OX3 7LF
    6. Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 7JX

      EDITOR—Hawton et al compared hospital presentations for self poisoning during the three week periods before and after the screening of a medical television drama.1 Their aim was to ascertain whether the programme altered the incidence of self poisoning. They attributed the increased frequency during the second period to a possible short term influence.

      The Scottish morbidity record system collects information on all admissions to Scottish NHS hospitals. These data are collated by the Information and Statistics Division of the Common Services Agency and linked to the general registrar for Scotland's database, which provides supplementary information on deaths outside hospital. Data were obtained on admissions and deaths attributed to all self poisoning and paracetamol overdoses between January 1995 and December 1997 inclusive.

      During this time the number of admissions attributed to deliberate poisoning rose steadily and showed substantial monthly variation. The number attributed to accidental poisoning remained comparatively constant, implying that the trend in deliberate poisoning was unlikely to be an artefact due to changes in coding practice. Deaths attributed to all self poisoning and paracetamol poisoning varied widely by month.

      If the conclusions of Hawton et al were correct, a significant increase would be expected in the numbers of admissions to hospital and deaths in November 1996. In fact, the number of admissions for all self poisonings fell in November 1996 and continued to fall until March 1997. Admissions for paracetamol poisoning rose in November 1996, but the rate of increase was lower than that for the rise from September 1996 to October 1996 and was well within the parameters of chance variation. The number of deaths from all poisonings rose slightly but was also within the range of chance variation.

      Although the episode of Casualty may have had an impact on hospital presentations, a causal association cannot be inferred without showing among other things that the increase shown by Hawton et al could not be explained simply by an existing upwards trend or normal chance variation.

      References

      1. 1.

      Study is impressive but raises methodological concerns

      1. Simon J C Davies, senior house officer,
      2. Mary Atherton, senior house officer,
      3. Timothy Williams, senior house officer,
      4. Jethro Purkis, senior house officer,
      5. Gillian Combe, senior house officer,
      6. Andrew J Brindley, senior house officer,
      7. Geraldine McCarthy, specialist registrar,
      8. Roger Denny, specialist registrar (roger.denny{at}lineone.net)
      1. Greenwich District Hospital, London SE10 9HE
      2. Medical Toxicology Unit, National Poisons Information Service (London), Guy's and St Thomas's NHS Trust, London SE14 5ER
      3. Greater Glasgow Health Board, Dalian House, Glasgow G3 8YU
      4. Department of Mental Health, Gloucester House, Southmead Hospital, Bristol BS10 5NB
      5. ICRF/NHS Centre for Statistics in Medicine, Institute of Health Sciences, Oxford OX3 7LF
      6. Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 7JX

        EDITOR—The study by Hawton et al examining the effects of a medical television drama on the incidence and nature of general hospital presentations for deliberate self poisoning1 has provoked considerable interest in medical and popular media. The role of imitation in suicide attempts is not a new phenomenon. It was named the Werther effect after an epidemic of copycat suicides in young men who read Goethe's Die Leiden des Jungen Werthers. The effect has been described after television films depicting fictional suicide.2 Television is certainly a powerful medium in contemporary society.

        We have a methodological concern about the study of Hawton et al which was not addressed in the discussion: the study was not well controlled They compared overdose rates between a baseline period before the drama was broadcast and each of the three weeks after the index episode.

        The authors acknowledge the possibility that an overdose depicted in a second soap opera six days after the index episode may have had a minor influence on results. We note that the end of British summer time also coincided with the observed trend of increasing rates of overdoses that began in the week before the drama was broadcast. The consequent reduction in evening daylight represents an external stressor that could be associated with parasuicide. Although Shapiro et al could not confirm such an association,3 the temporal relation of the return to winter time with the rise in overdose rates raises the possibility of this being a confounding factor that cannot be excluded given the methods of the study. Other potential confounding factors include seasonal changes in weather and temperature and the impact of unrelated items in national and international news.

        A more robust study design would entail a comparison of the change in overdose rates before and after the November 1996 screening to the change over a similar period one year earlier or later. An alternative would be an investigation of any change in overdose rates in a region or state where the index episode was broadcast compared with an area where it was not shown.

        The study of Hawton et al was certainly impressive in its coordination of data from many sources. It will continue to stimulate valuable debate on the responsibilities of media producers in portraying self harming behaviour. We believe that the results must be interpreted with caution. The limitations of the study design should be recognised.

        References

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

        Authors' reply

        1. Jonathan J Deeks, medical statistician,
        2. Keith Hawton, professor of psychiatry,
        3. Sue Simkin, senior researcher
        1. Greenwich District Hospital, London SE10 9HE
        2. Medical Toxicology Unit, National Poisons Information Service (London), Guy's and St Thomas's NHS Trust, London SE14 5ER
        3. Greater Glasgow Health Board, Dalian House, Glasgow G3 8YU
        4. Department of Mental Health, Gloucester House, Southmead Hospital, Bristol BS10 5NB
        5. ICRF/NHS Centre for Statistics in Medicine, Institute of Health Sciences, Oxford OX3 7LF
        6. Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 7JX

          EDITOR—These correspondents express concerns about the design and conclusions of our study of presentations for self poisoning after the broadcast of an episode of Casualty. They contest the robustness of our comparison and suggest that seasonal trends, random variability, or other confounding factors could explain our findings.

          Pell and Murdoch also report that there was no increase in monthly admission rates for self poisoning in Scotland after the broadcast, but we are concerned that aggregation at a monthly level may obscure an increase that at most lasted two weeks.

          To estimate random variability in weekly self poisoning rates we recorded presentations in 49 centres for three weeks immediately before the broadcast. Our statistical analysis showed the observed 17% increase in presentation rates after the broadcast to be substantially larger than could reasonably be explained by natural variation. Random variability is not a tenable explanation.

          Interrupted time series studies cannot provide totally convincing evidence of effect as it is impossible to exclude the possibility that other unknown factors and trends may confound the comparison. However, we do not think that the suggested mechanisms are of concern. Studies of seasonal effects on suicide attempts across Europe consistently show peaks in the spring and nadirs in December, with a maximum variation of 26% between the two, the variation being stronger in women than men.1 This pattern is contrary to the suggestion of Davies et al but in line with the observations of Pell and Murdoch. Changes in suicide rates with weather do exist but are small,2 as is the ongoing upward trend in admissions for self harm. These factors are unlikely to have caused the increase in self poisoning presentations of 17% in one week across 49 centres.

          Although alternative comparison groups may seem preferable, the practicalities of undertaking this project make many suggestions impossible, and others are not as reliable as one might first think. Studies based on comparisons between places or between disjoint time periods may be seriously biased, not only because of confounding but also because of the difficulties in ensuring consistent data collection. We aimed at minimising these problems by collecting data from adjacent time periods and making comparisons within centres.

          We anticipated the weaknesses of a time series design and additionally surveyed a sample of patients admitted with self poisoning throughout the study, asking about their motivations and television viewing. We found that those among them who watched Casualty changed their choice of substance after the broadcast, the proportion using paracetamol increasing by 106% (95% confidence interval 28% to 232%) This is the crucial supporting evidence on which judgments of causality should be made.

          References

          1. 1.
          2. 2.
          View Abstract