Article Text

Download PDFPDF

Identification of strategies to prevent death after pesticide self-poisoning using a Haddon matrix
  1. M Eddleston1,2,
  2. N A Buckley2,4,
  3. D Gunnell5,
  4. A H Dawson2,6,
  5. F Konradsen2,7
  1. 1Centre for Tropical Medicine, Nuffield Department of Clinical Medicine, University of Oxford, Oxford, UK
  2. 2South Asian Clinical Toxicology Research Collaboration
  3. 3Ox-Col Collaboration, Department of Clinical Medicine, University of Colombo, Sri Lanka
  4. 4Department of Clinical Pharmacology and Toxicology, ANU Medical School, ACT, Australia
  5. 5Department of Social Medicine, University of Bristol, Bristol, UK
  6. 6Department of Clinical Medicine, University of Peradeniya, Sri Lanka
  7. 7Department of International Health, University of Copenhagen, Copenhagen, Denmark
  1. Correspondence to:
 Dr M Eddleston
 Scottish Poisons Information Bureau, Royal Infirmary, Little France, Edinburgh EH16 4SA, UK; eddlestonm{at}


Despite pesticide self-poisoning causing around 300 000 deaths each year in the rural Asia Pacific region, no comprehensive public health response has yet been formulated. The authors have developed a Haddon matrix to identify factors that increase the risk of fatal rather than non-fatal pesticide self-poisoning in Sri Lanka. Many important host factors such as age, gender, and genetics are not alterable; factors that could be changed—alcohol use and mental health—have previously proved difficult to change. Interventions affecting agent or environmental factors may be easier to implement and more effective, in particular those limiting the human toxicity and accessibility of the pesticides, and the quality, affordability, and accessibility of health care in the community. Controlled studies are required to identify effective strategies for prevention and harm minimization and to garner political support for making the changes necessary to reduce this waste of life. Lessons learnt from Sri Lanka are likely to be highly relevant for much of rural Asia.

  • ICU, intensive care unit
  • OP, organophosphorus
  • WHO, World Health Organization
  • Haddon matrix
  • pesticide
  • prevention
  • self-poisoning
  • suicide

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Since the 1960s, pesticides have been increasingly used across the Asia Pacific region. Their widespread use in poor rural communities has inevitably resulted in their ready availability for acts of self-harm and a rapid rise in the number of deaths from pesticide self-poisoning.1,2 Pesticide self-poisoning currently causes an estimated 300 000 deaths every year and is responsible for around two thirds of all self-harm deaths in the region.3,4 The World Health Organization (WHO) now considers pesticides to be the most common method of suicide globally.5

Unfortunately, the problem has been largely ignored.6 Despite its recognition in the late 1980s,7 no international convention or forum on pesticides has ever acknowledged that self-poisoning is the major cause of mortality from pesticides, instead preferring to pay attention only to unintentional poisoning.1,6 As a result, no public health approach has been formulated and no country seriously affected by pesticide self-poisoning has yet developed a comprehensive strategy for reducing deaths.

Recent WHO reports on mental health and violence have recommended primary prevention as the best way to reduce the number of deaths globally.8,9 This may be true in Western countries where most deaths occur from firearm injuries, hanging, and gassing—acts that are usually rapidly fatal.4 However, the situation is different in rural Asia where most deaths occur hours to days after an act of self-poisoning. Strategies to prevent deaths by restricting access to toxic pesticides may be effective; however, there are considerable economic and regulatory barriers to their implementation. Instead, the substantial time delay from the act of self-harm to death offers multiple opportunities for harm minimization.

The current situation with pesticide self-poisoning seems analogous to the problem of road traffic deaths in the US during the middle of the last century. Similarly, most effort then went into trying to stop cars crashing, with little success. Then, in 1965, William Haddon radically changed how people thought about preventing road traffic deaths.10,11 Frustrated by the solitary focus on primary prevention, he argued that harm minimization would be more effective and that its adoption earlier in the century would have saved hundreds of thousands of lives on US roads.10

To review possible strategies for prevention and harm minimization, he developed a matrix that organized the factors influencing the likelihood of fatal injury from road traffic crashes into pre-event, event, and post-event factors, associated with the person, host, and environment.10 This matrix served as an impetus for the introduction over the next decade of a wide range of innovative interventions.


We have developed a Haddon matrix to identify factors that affect the likelihood of fatal rather than non-fatal pesticide self-poisoning in Sri Lanka (table 1). We hope that it will initiate the development of a comprehensive public health approach to pesticide self-poisoning by serving as a framework for a program of research to find the most effective methods to reduce the number of deaths, with involvement of all stakeholders and combining both primary and secondary prevention.

Table 1

 Host, agent, and environmental factors affecting the likelihood that pesticide ingestion will result in death


The matrix has been developed after reviewing the literature on pesticide self-harm and data from our studies of pesticide poisoned patients in southern and north central Sri Lanka.2,12–18 We carried out a systematic search for relevant studies by searching PubMed, EMBASE, and the Cochrane databases with the search terms “pesticide” and “poisoning”. We did not limit the search by language. We hand searched Sri Lankan and other South Asian journals (years 1980 to 2005) for information on pesticide poisoning and used information from our studies that have recruited over 6000 pesticide poisoned patients.

We did not address general risk factors for self-harm because we wanted to examine factors specifically involved in fatal pesticide poisoning. Many workers have reported such risk factors and they are as relevant for prevention of pesticide self-poisoning as they are for other forms of self-harm.8,19,20 Both the literature and our own studies in Sri Lanka are likely to systematically miss less seriously injured people. However, as this paper is about preventing fatal self-poisoning, not all self-poisoning, cases that do not come into contact with health services or coroner are unlikely to be relevant and are not further discussed.


Acute alcohol use and chronic alcohol abuse are important issues in self-poisoning.13,15,17 Acute use of alcohol is associated with poor impulse control, impaired judgment, and possibly altered taste that would increase the likelihood of the person ingesting pesticides and of ingesting large amounts. Co-ingestion of alcohol increases the risk of aspiration, coma, and respiratory failure—all common and often fatal complications of pesticide poisoning.21,22

After ingestion of poison, alcohol withdrawal complicates the medical management of pesticide poisoning,23,24 especially in hospitals where it is not possible to routinely sedate patients due to the lack of intensive care unit (ICU) beds and ventilators. It lowers the threshold for seizures and impairs the immune response to infection. Importantly, agitation in inadequately sedated patients results in heat generation that is dangerous to patients being given atropine (and therefore not sweating) in hot humid Sri Lankan wards.

Acute alcohol use impairs myocardial contractility and can induce dysrhythmias; chronic alcohol abuse causes alcoholic cardiomyopathy.25 Both reduce the ability of patients to withstand the tachycardia that may occur from the organophosphorus (OP) poisoning and its therapy. Other comorbidities, such as ischemic heart disease or liver disease, will affect the likelihood of a fatal outcome after poisoning. This may partly explain the higher case fatality seen in elderly people after poisoning,26,27 independent of their higher intent and use of more dangerous methods.28

Older people and men have a higher case fatality from self-poisoning in Sri Lanka. We have found in a logistic regression model, controlling for sex and type of poison taken, that the risk of death increased by 62% (95% CI 45% to 81%) per 10 year increase in age and was 52% (95% CI 4% to 124%) higher in males than females.15

Women are less likely than men to use a pesticide for an act of self-harm,15 and men commonly co-ingest alcohol.13 Bearing in mind the interaction between highly toxic pesticides and alcohol, this may partly explain the excess of deaths in middle aged men.15,29 The majority of deaths in Sri Lankan females occur in young women, yet few take pesticides or co-ingest alcohol. These deaths are due to the large number of young women harming themselves, rather than a particularly high case fatality in this age group. Women are at a disadvantage from their smaller size and the smaller amounts of pesticide therefore required for a severe poisoning.

The time since the last meal affects outcome. The presence of full stomach at the time of paraquat ingestion has been shown to reduce pesticide absorption and improve outcome.30 The same effect may occur with other pesticides.

Studies from industrialized countries show that the case fatality for men is higher than for women across all methods of self-harm.28,31 This suggests that intent may be generally higher in men than women and the outcome of male self-harm worse. A study of oleander poisoning in Sri Lanka found that men ingested more seeds than women, supporting the hypothesis that men in Sri Lanka have higher intent than women.27

Knowledge about the differential lethality of pesticides would allow a person with high intent to select a more toxic pesticide. This, however, appears to be uncommon at present.17,18 Instead, it appears to be more common for people to impulsively ingest the nearest pesticide without considering its toxicity.4,17 Levels of impulsivity may affect how much pesticide is ingested.

Help-seeking behavior affects outcome. People who ingest pesticides where they will not be found are more likely to die. Ingesting pesticides in front of others increases the chance of being taken to healthcare facilities before the onset of symptoms.

Host genetic factors may be important. OP pesticides are activated and metabolized after ingestion.32 Genetic polymorphisms in the relevant enzymes, for example paraoxonase or cytochrome P450s, might affect outcome.33


The acute toxicity of pesticides is currently classified according to their lethal dose in rats after oral administration. Using this information, WHO has divided pesticides into five categories, from extremely hazardous (class Ia) to slightly hazardous (class III), and then pesticides that are “unlikely to present acute hazard in normal use”.34 This classification system is not ideal.35 There are obviously difficulties with extrapolating toxicity classifications from untreated rats to humans treated in ICUs.22 Furthermore, the classification was developed for occupational exposure, by dermal or respiratory routes, rather than intentional ingestion of pesticides.36

However, studies of self-poisoning in Asia indicate that the WHO classification might be crudely useful: class Ia, Ib, and II pesticides are all relatively dangerous pesticides that kill at least 5% of patients who ingest them and present to hospital.13,37,38 A significant proportion of these deaths occur among people who ingest only a very small amount; the case fatality in patients who take mouthfuls of the pesticide is likely to be much higher. In contrast, class III pesticides and those “unlikely to present acute hazard in normal use” cause few severe effects and few deaths irrespective of how much is ingested. Ease of access to class Ia to II pesticides will markedly increase the risk of death after pesticide ingestion.

The amount ingested is an important issue affecting outcome for toxic pesticides, but not for the less toxic pesticides that are inherently safe, whatever the amount. The size and concentration of the pesticide preparations available will affect how much pesticide can be drunk. Bottles containing up to five litres are available in Sri Lanka but few patients are admitted after drinking from containers larger than 400 ml, probably due to the physical difficulty of handling such large containers (Eddleston, unpublished observations). Liquid pesticides are available in concentrations between 5% and 60% with little relation between toxicity and highest permitted concentration. Other factors that will affect the amount ingested include the level of intent and alcohol co-ingestion.

Chemicals added to the pesticide may affect how much of the ingested dose is absorbed into the body. Propriety paraquat has had an emetogenic compound added in the hope of rapidly inducing vomiting.39 More recently, a pH dependent gelling agent that slows paraquat’s passage from the stomach to the small bowel has been added, to give the emetic more time to work.40 Initial studies indicate that the gelling agent results in reduced case fatality.41 If proven to be effective, these approaches could be considered for other toxic pesticides.

Within the dangerous classes of pesticides, there are more subtle issues. The speed of poisoning onset varies between pesticides: for example, patients taking large amounts of parathion can become ill within minutes and unconscious within half an hour;42 patients taking fenthion often do not show signs for many hours.43,44 Ingestion of a fast onset OP risks respiratory arrest in the community with the attendant risks of aspiration, hypoxia, and death before reaching medical care.

Not all pesticides can be treated, in part because effective antidotes exist for only a small proportion of products. Paraquat is classified as a class II (moderately hazardous) pesticide yet it is essentially untreatable, with a case fatality reaching 70% after ingestion.45 Endosulfan is less toxic to animals than the class I OP insecticides it replaced in Sri Lanka during the second half of the 1990s.35 However, there is no effective therapy and many deaths occurred before it too was banned in 1998.14,35

Within pesticide classes, there is marked variation in effectiveness of therapy. OP poisoning is treated with oximes.46,47 Most OPs have either two methyl groups attached to the phosphate or two ethyl groups, and are classified accordingly.48 Oximes appear to be effective for diethyl OPs (for example, chlorpyrifos, parathion) but less so for dimethyl OPs (for example, dimethoate, fenthion).22 This may account in part for the higher case fatality with dimethyl OPs compared to diethyl OPs seen in Sri Lanka.22

Embedded Image

Farmer applying pesticide to rice plants, Sri Lanka (photo: International Water Management Institute).


The storage and distribution methods of pesticides in the community and the range and quantity of pesticides used for farming, affected by agricultural practice, economic reasons, or regulatory actions, determine which pesticides are ingested.19 Similarly the ease of obtaining pesticides affects the availability. Legislation that encourages the use of less dangerous pesticides, limits sales, and requires safe storage of pesticides will reduce the risk of death after self-harm.19

The proximity of other people at the time of pesticide ingestion will often determine how quickly the person is brought to medical attention.17 Their first aid actions may also affect outcome. Inducing vomiting within a few minutes in a fully conscious patient may improve outcome; any act that increases vomiting in an unconscious or drowsy patient is likely to worsen it. People sometimes give liquids to the patient soon after the event—large amounts may well push the pesticide into the small bowel, increasing the speed of absorption and poisoning onset, with harmful consequences.

A delay in reaching medical attention adversely affects outcome. The proximity of healthcare workers with the ability and facilities to treat pesticide poisoning is important, as is the availability of transport. The small hospitals scattered across rural Sri Lanka allow most patients to reach medical attention and to be transferred as necessary to a secondary hospital within an hour or two. A community survey of injury deaths suggests that few poisoning deaths (less than 5%) occur before reaching primary health care.49 However, in communities without the same level of healthcare provision, or with difficult terrain that makes transfer to hospital difficult, more deaths will occur out of hospital.

The cost of health care will affect whether people are sent to medical attention after self-harm. Pesticide poisoning often results in long term ventilation and extensive use of antidotes, which proves expensive where health care is paid for by the patient or their family. Despite this care, severely poisoned patients often die. A negative community view of the likely outcome of treatment, or of people who harm themselves, or of socially/economically marginalized groups (perhaps women or old people), may reduce the likelihood of certain people being taken to hospital in regions where payment is required. Stigma, legal issues, and religious beliefs related to self-harm and the courtesy with which healthcare workers treat self-harming patients50 also probably affect the likelihood of presentation to hospital.

The quality of health care (which includes the availability and affordability of antidotes) in the local hospitals will affect outcome, particularly after OP poisoning. Atropine is occasionally unavailable in some Asian rural hospitals (L Karalliedde, unpublished observations), making management of OP poisoning almost impossible. Oximes are expensive and often unaffordable, particularly where patients have to pay for their medical care. Oxygen and ventilators are essential for OP poisoning but frequently unavailable in rural district hospitals seeing most cases. Finally, the evidence base for guiding management is weak.3,51


The Haddon matrix identifies multiple factors that may affect the likelihood of dying from pesticide poisoning. Prevention strategies designed to affect these factors should be able to reduce the number of deaths. However, we currently have few data with which to compare the cost effectiveness of interventions and prioritize their introduction.52 Research is required to comprehensively determine the effectiveness and practicality of different approaches.

The major alterable host factor in Sri Lanka is alcohol use. Community efforts to reduce alcohol use should reduce the number of self-harm deaths and produce additional major public health benefits. However, such attempts require immense political will, have elsewhere proved difficult to sustain, and will be difficult in Sri Lanka because of the political power of the drinks industry and widespread illicit distilling of alcohol. Alongside alcohol reduction, expansion of community mental health care and counseling may be effective in primary prevention, but there are strikingly few examples worldwide of improved mental health care producing reductions in self-harming behavior.8

Prevention based around the agent may prove more effective. Previous experience in Sri Lanka suggests that legislation to enforce the use of pesticides with low human toxicity and slow onset of action will reduce deaths.14,35 Similarly, reducing the bottle size and pesticide concentration will make it more difficult for a person to ingest a lethal dose.

Altering the environment in which agriculture is practiced may be effective.19 Increased uptake of the Food and Agriculture Organization’s integrated pest and vector management practice will generally reduce pesticide use in rural communities and therefore reduce their availability for self-harm. Provision of more dilute preparations of pesticides should improve safety but this is likely to be unpopular with manufacturers. Safer storage so that they are not accessible at moments of stress may reduce the number of acts of self-poisoning.

Improved affordability and availability of antidotes and ventilators, and improved evidence upon which to base their use, will likely be useful.3 A sustained effort to make hospital care more locally available and affordable—ideally free at point of access—will increase its appropriate use in poor rural areas and improve outcomes.

More research is required to clarify the role community education and raised awareness might have on preventing deaths from pesticide poisoning. There seems little point making people aware about the dangers of pesticides—while there is little appreciation of how toxic particular pesticides are, rural people clearly know that pesticides are generally dangerous.12,18 Education around appropriate first aid, destigmitization of self-harming people, and appropriate use of the healthcare services may reduce harm after the event.


Attempts to prevent deaths from self-poisoning by controlling the availability of highly toxic pesticides will not result in substantial and sustained reductions if the majority of patients who currently die have high levels of intent. There is a risk that patients who survive in the future because only low toxicity pesticides are accessible might simply look for another lethal method. However, current evidence indicates that method substitution is unlikely to be important.53,54 Furthermore, cases in Sri Lanka show that many deaths occur after patients impulsively ingest pesticides of unknown toxicity.4,17,18 Research from rural China similarly suggests that many patients surviving ingestions of highly toxic pesticides had low intent to die.55,56 This indicates that method substitution should be relatively unimportant when pesticides are made less dangerous rather than completely removed.


This Haddon matrix identifies multiple targets for public health intervention that should reduce the number of deaths from pesticide self-poisoning in the developing world. The large number of deaths in impoverished rural communities each year—around 300 000 in the Asia Pacific region alone—offers a moral imperative for rapid and effective action.


We thank Michael Phillips for incisive comments on an earlier draft.



  • See linked article, p 285

  • Contributors

    ME had the idea of developing a Haddon’s Matrix for pesticide self-poisoning and wrote the first draft of this paper. ME had multiple discussions with the other authors both before and after preparing the manuscript and made revisions accordingly. All authors saw and approved the final manuscript.

  • Funding: ME is a Wellcome Trust Career Development Fellow, funded by grant GR063560. The South Asian Clinical Toxicology Research Collaboration is funded by a Wellcome Trust/National Health and Medical Research Council International Collaborative Research Grant GR071669.

  • Competing interests: none.

Linked Articles