4Journal of Epidemiology and Community Health 1996;50:24-29

Epidemiology and patterns of hospital use afterparasuicide in the south west of England

David J Gunnell, Jane Brooks, Tim J Peters

AbstractObjective - To describe the epidemiology,management, and outcome ofparasuicidein the south west of England.Design - Descriptive analysis using rou-tine information systems.Setting - The former South WesternRegional Health Authority, populationaged 10 and over: 2*9 million.Subjects - These comprised 5080 residentsof the South Western Health Region, ad-mitted on 5770 occasions for parasuicidebetween April 1992 and March 1993 to hos-pitals in the south west.Main outcome measures - Person basedage and sex standardised admission ratiosfor parasuicide; readmission for para-suicide in the subsequent 12 months; ad-mission to psychiatric hospital afterparasuicide; in-hospital mortality forthose admitted after attempted suicide.Results - The overall rate of parasuicideadmission was 174 per 100 000 per year.Rates were highest in males aged 20-24(381 per 100 000) and in females aged 15-19(625 per 100 000). Parasuicide is the thirdmost frequent cause of acute medical ad-mission in the south west. A total of 10-0%of admissions received psychiatric in-patient care as a direct result of theirparasuicide admission and 10-1% were re-admitted in the following 12 months with arepeat episode of parasuicide. Significantvariations in standardised admission ra-tios for parasuicide were observed betweenthe districts. Some of this variation is re-lated to socioeconomic differences be-tween districts, the rest is probably due todifferences in practice between districts.There is no clear evidence that these vari-ations result in differences in readmissionrates. Districts where psychiatric in-patient facilities were located on the samesite as the general hospital tended to admita greater percentage of parasuicidepatients for psychiatric inpatient care. Aquarter ofall suicide deaths from overdoseoccurred in hospital. It is estimated thatthere are 87 000 parasuicide admissions inEngland and Wales annually.Conclusions - Parasuicide is a commoncause of acute hospital admission andthere is evidence that hospital admissionpractices for parasuicide vary across thesouth west. Randomised controlled trialsare needed to evaluate the most ap-propriate form of management for those

patients who do not require admission onmedical grounds.

(J3 Epidemiol Community Health 1996;50:24-29)

After showing a reduction in the 1980s, para-suicide rates in this country are once againrising in some areas."q Reduction in rates ofboth suicide and parasuicide have been iden-tified as targets in health strategy documentsin this country and elsewhere.5-7 Both suicideand parasuicide are used in these documentsas proxy indicators of the mental health of thepopulation. While there is uncertainty aboutthe value of suicide as an outcome indicatorfor mental illness services, parasuicide un-doubtedly remains a common and importanthealth care problem. Although some in-terventions for parasuicide are effective in re-ducing short term repetition rates in high riskgroups, no randomised controlled trials havebeen large enough to show the effectiveness ofinterventions in unselected patient groups.8

In 1984, a Department of Health workinggroup reversed previous recommendations thatall deliberate self harm patients should be as-sessed by psychiatrists.9 The group ac-knowledged that some patients would bemanaged by their general practitioner aloneand that others would not be admitted afterattendance at accident and emergency de-partments. In this context there is evidencein recent years that admission practices varybetween hospitals. In some hospitals less thanhalf of those who attend accident and emer-gency after parasuicide are admitted, in othersover 80% of attenders are admitted.2810 Wehave examined the epidemiology ofparasuicidein the south west region between April 1992and March 1993 using routinely available dataon hospital admissions. In particular, inter-district variations in parasuicide admissionrates and repetition rates are examined togetherwith differences in management.

MethodsDATA SOURCESDetails on each hospital admission for para-suicide were obtained from routine health in-formation systems used in the former SouthWestern Health Region (Patient Admin-istration System (PAS)). This records patientdetails (date of birth, sex, and postcode) anddetails of outcome (length of stay, death, dis-charge to another hospital) for all hospital ad-missions. Information on the methods used in

Department of SocialMedicine,Canynge Hall,Whiteladies Rd,Bristol BS8 2PRD J GunnellT J Peters

South and West RHAResearch andDevelopmentDirectorate,Canynge Hall,Bristol BS8 2PRJ Brooks

Correspondence to:Dr D Gunnell.Accepted for publicationAugust 1995

24

Epidemiology ofparasuicide in SW England

40

35

' 30.0

CDm 25C

uc 20a)

D 15on

° 1

|El ParasuicideAccidental|

30-39 40-49 50-59 60-69 70-79 80-89

Age range (y)30

Females

25pFD ParasuicideAccidental

CDco

.0

thusesrange10%cveaeofalloan

parasuiciepioe(forinsance ofvterdoeight hea-tceratrictsnAlaredicalrcordedrswithi the bregoad

algesicantplyretrinedtanquilliequetl andited.

cuschtopicngisens). Eachmnrasinghealithncarprovieen hositalsin the region.hsaPS oue

thsInurnDE950-E959 and E98-E89 localver,patientnacdnaepisodesfrec oftherugh healthdistictsmallbecoedicalsodr withidntherEgion9Seniotherefrest meet evrhtremonthcmltopicturcusparasingisses thucmincedabothconsstencytbetwreenhourpanalyse.Ifatin theregionth

examiedno-acienaepisisons of7male; poisf-oning mayre coded as accidentalpoiso0ng9iaeandthreoex ditorginuinothemscompee odeduasofcidenasu(icide wetcombiedparasuiciese cat-egorie ina ouibranalyses.Infatpioingth 12smonthmsexamined, 375uioadmissienstrue" episodes6ofeml)wrcoeasaccidental poisoning inordt a hrfr

ageandt sligh distributimatiof toseparauicdea

rates (by a maximum 7%). This is particularlyso in the older groups where accidental self

poisoning is more likely. Individuals aged 10years or less were omitted from all analyses asintentional self harm in children is rare, and ismost likely to be accidental. No attempt wasmade to include accidental injuries in the sameway as accidental poisoning. This decision wasbased on the assumption that wrist lacerationcoded as accidental is more likely to be trulyaccidental than is overdose coded as accidental,and inclusion of the former episodes may there-fore distort the estimate of incidence. Whilethere is no direct evidence for this assumption,ICD categories for accidental injury and pois-oning are less specific for wrist laceration.

Population data were obtained from the 1991census and OPCS mid year population es-timates. Information on suicides and methodsused for suicide in the south west was obtainedfrom OPCS mortality files.A postal questionnaire was sent to the 14

accident and emergency departments attachedto general hospitals in the south west. In-formation was sought on the total number ofannual attendances following drug overdoseand the number of such attenders who wereadmitted.

PERSON BASED ADMISSION AND READMISSIONRATESTo assess person based admission rates andassess the repeat parasuicides rates, separateepisodes of care were linked using district num-ber, date of birth, and sex. Each patient wastracked for a 12 month period after their firstdate of admission, and readmission followingfurther episodes of parasuicide recorded. Res-idents receiving treatment outside the region(less than 2-5% of the total in those districtson the borders of the region) were omitted aswere residents from other regions admitted tohospitals in the south west. In-hospital deathssubsequent to parasuicide admission werecounted as suicides.

PSYCHIATRIC ADMISSIONSInspection of the data suggested that in a num-ber of units, particularly those with on-sitepsychiatric facilities, some parasuicide ad-missions are coded as both medical and psy-chiatric episodes of care. Similarly, thepsychiatric assessment of patients on the med-ical wards after parasuicide may have beencoded as separate consultant episodes of care.To overcome the resulting overestimation ofpsychiatric admission rates we calculated therates using two bases. The first estimate in-cludes all parasuicides who had a psychiatricfinished consultant episode (FCE) during theirperiod of inpatient admission. The second ex-cluded all those patients whose length of psy-chiatric admission was less than two days.

STATISTICAL ANALYSESAge and sex standardised admission ratios(SAR) for each district health authority werecalculated by the indirect method of stand-ardisation using OPCS mid-year 1992 regional

25

Gunnell, Brooks, Peters

700

600 -

o 500 D Females0

400-

C. 300-a)e 200-

100

010-14 15-19 20-24 25-29 30-34 35-39 40+

Age range (y)

Figure 2 Parasuicide admission rates for different agegroups in the South Western Health Region between April1992 and March 1993.

population estimates. An estimate of the num-ber of parasuicide admissions in England andWales was obtained from the South WesternHealth Region age/sex specific rates multipliedby the England and Wales population for eachage/sex population strata. Townsend scoreswere calculated for each of the districts fromthe 1991 census.'1 These scores were used toexamine the relationship between deprivationand SAR for parasuicide, and also to considervariations in admission rates both before andafter adjustment for deprivation.

Standardisation indicates whether a givendistrict's parasuicide experience is comparableto that of the regional population. This doesnot, however, give sufficient evidence of adifferential pattern of admissions across dis-tricts; for this a comparative analysis of districtparasuicide ratios is required. Assuming a Pois-son distribution, it is possible to use an ap-proximate X% test statistic to make a suitablecomparison33 since the standard employed hereis the region itself.

ResultsOVERALL RATESBetween April 1992 and March 1993 therewere 5770 parasuicide admissions in the southwest (admission rate for parasuicide 174 per100 000 per year). Altogether 5415 (93 8%)admissions were for self poisoning and 2763of these were following overdoses using an-algesics, antipyretics, and anti-rheumatics(mainly paracetamol and aspirin). Parasuicidewas the third commonest cause of acute med-ical admission after acute myocardial infarction

(8685 admissions) and heart failure (6257 ad-missions).A total of 2603 (45.1%) of the admissions

involved males; 3167 (54.9%) were female.The highest parasuicide rates in males were inthose aged 20-24 years (381 per 100 000) andin woman rates were highest in those aged15-19 years (625 per 100 000) (fig 2). Reasonsfor admission other than drug overdose in-cluded: injuries by cutting and piercing in-struments (135), self inflicted poisoning bygases (58), injuries by jumping from high places(28), and the remainder (134) were admittedfollowing attempted suicide by drowning,hanging, shooting, or unspecified means. Ex-trapolating these rates onto the population ofEngland and Wales, it is estimated that thereare in the order of 78500 admissions afterepisodes of deliberate self poisoning annually,and a total of87 000 admissions after deliberateself harm of some kind. Of these 39 900 arefor self poisoning with analgesics, antipyretics,and anti-rheumatics (mainly paracetamol andaspirin).The total number of individuals admitted was

5080. Table 1 and figure 3 show the totalnumber of parasuicide admissions and personbased age and sex standardised admission ratiosfor each of the health districts. The stand-ardised admission ratio for parasuicide rangedfrom 78-2 to 130 6 across the district healthauthorities. This variance was highly significantbased on an approximate X2 test for homo-geneity (X2= 107 6 on 7 df, p<0001). Thosedistricts with greater socioeconomic dep-rivation, as measured by the Townsend score,tended to have increased SARs for parasuicide(Spearman's rank correlation coefficient 0-3195% CI -0 43, 0-86, fig 4). Using the rankeddata, about 9% of the variance in parasuicideSAR was due to variation in Townsend score.

MANAGEMENTThe postal questionnaire was returned by 11of the 14 casualty departments attached togeneral hospitals in the South Western HealthRegion. Two of the three departments that didnot respond were within districts with othercasualty departments from which responseswere obtained. The sociodemographic mix ofthe populations served by the non-respondingdepartments is typical of that in the south west.Reported admission rates after attendance at

Table 1 Total parasuicide admissions in the south west April 1992-March 1993. Ratesare indirectly standardised against the population of the former South Western HealthRegionDistrict of residence District Total parasuicide Person based agelsex standardised

population admissions parasuicide admission ratio (95%CI)

1 731 289 1615 105 8 (99 3, 112 5)2 418 197 840 107-4 (99-7, 115-8)3 283 950 490 89-8 (81-8, 98 6)4 122722 310 130-6 (115-2, 148-1)5 294872 778 1224 (113-6, 132-0)6 220476 312 80-1 (71-0, 89-2)7 472 666 881 93 0 (86-7, 99-7)8 366 587 544 78-2 (71 5, 85-4)South Western RHA 2 910 759 5770 100

a 160In=o , 140=>°120

1004- CO 80-

20

°

40 a

i i

1 2 3 4 5 6 7 8District

Figure 3 Person based standardised admission ratios forparasuicide in health districts in the South Western HealthRegion.

26

Epidemiology ofparasuicide in SW England

140 rA

V_ C'Aa)I0a)c

.- occ

U)0

'uE

a-

120 e-

A100 t

80A-

60 F-

40 F

20 e-

-4 -2 0

Townsend sc

Figure 4 Relationship between th

admission ratio for parasuicide anceight health districts in the South I

to admit a greater percentage of parasuicideA

patients for psychiatric inpatient care. Therewere significant variations between districtsusing either definition of psychiatric inpatientepisode. There was a positive relationship be-tween the parasuicide SAR and the proportionofpatients admitted to psychiatric units (Spear-man's 0-48; 95% CI -034, 0-89; partial cor-relation coefficient, adjusted for Townsend:0 46). For the seven districts where informationon the percentage of patients admitted after

2 4 6 parasuicide was obtained, there was no strongcore relationship (Spearman's0 I1; 95% CI -070,

0 80) between the percentage of individualsstandardised admitted and the number of subsequently re-

I Townsend score for ferred for psychiatric care. Adjustment forVester Health Region.

Townsend score did not affect this.

Table 2 Results ofpostal questionnaire to accident andemergency departments (A&E) in the south west regardingattendances and admission for parasuicide

Location ofA&E Admission/A&E attendances (%)

District 1 712/991 (71-8)District 1 259/436 (59 4)District 1 228/295 (77 3)District 1 226/334 (67-6)District 2 245/554 (44-2)District 4 104/128 (81-3)District 5 742/974 (76-2)District 6 231/447 (48 7)District 7 422/636 (66-4)District 7 110/185 (59-5)District 8 300/418 (71-8)

accident and emergency for parasuicide rangedfrom 44-2% (95% CI 40-1, 48 4%) to 81-3%(95% CI 74 5, 88-0%) (table 2). The healthdistrict with the highest SAR for parasuicidealso reported the largest proportion of patientsadmitted after accident and emergency at-tendance (81 -3%); Spearman's rank correlationcoefficient between SAR for parasuicide andthe percentage of accident and emergency at-tenders who were admitted was 0-43 (95% CI-0-48, 0-89). The partial correlation co-efficient (adjusted for Townsend score) was0-55 (95% CI -0-35, 0 92).The median length of inpatient stay was one

day. This did not vary across the districts.Altogether 666 (11-5%) patients were dis-charged to a psychiatric inpatient unit (table3). If those patients whose psychiatric episodelasted less than two days are excluded, 579(10%) of parasuicide patients received sub-sequent inpatient psychiatric care. Districtswhere psychiatric inpatient facilities were loc-ated on the same site as general hospitals tended

OUTCOMEAltogether 512 (10-1%) patients were re-admitted with one or more further episodes ofparasuicide in the 12 months subsequent totheir index admission. No relationship wasfound between SARs for parasuicide and theproportion of patients readmitted after para-suicide in the next 12 months (Spearman'srank correlation coefficient 0-07; 95% CI-0O67, 0 74; partial correlation coefficientadjusted for Townsend: 015). There was apositive though non-significant relationshipbetween the proportion of patients dischargedto psychiatric hospitals and the proportion ofpatients readmitted with a diagnosis of para-suicide in the subsequent 12 months (Spear-man's rank correlation coefficient: 0 55; 95%CI -0O25, 089; partial correlation coefficient,adjusted for deprivation: 052).

Overall there were 42 in-hospital deaths(0-8% of all admissions) after parasuicide ad-mission. Twenty nine (69%) of these were afteroverdose.

DiscussionHospital admission statistics underestimatetotal parasuicide episodes as many of thoseseen in casualty are not admitted to hospital.In addition, many cases may be managed inthe community without the use of hospitalservices.8 012-'143 Our aim, however, was to as-

sess differences in hospital admission rates andthe patterns of care among hospital admittedparasuicides. Our data are based on districtof residence - we were not able to examine

Table 3 Proportion ofpatients undergoing further psychiatric inpatient care following parasuicide admission, andproportion ofpatients readmitted in the subsequent 12 months

District of residence No (%) ofparasuicide No (%) ofparasuicide admissions No of patients (%) readmittedadmissions discharged to where length ofpsychiatric inpatient with parasuicide diagnosis inpsychiatric inpatient units* episode two days or moret the subsequent 12 mtht

1 144 (8 9) 129 (8 0) 94 (6-6)2 128 (15-2) 111 (13-2) 79 (10-5)3 64 (13-1) 55 (11-2) 62 (14-3)4* 61 (19-7) 49 (15-8) 28 (10-4)5* 119 (15-3) 102 (13-1) 95 (14-3)6* 53 (17-0) 44 (14-1) 38 (13-3)7 54 (6-1) 52 (59) 70 (90)8 43 (7-9) 37 (6-8) 46 (9-4)South Western RHA 666 (11-5) 579 (10 0) 512 (10-1)* Districts where psychiatric inpatient beds available on-site in the general hospitalstX2=95-5 (p<0001); t 2=65-9 (p<0-001).

27

Gunnell, Brooks, Peters

admission rates for individual hospitals (as de-nominator population data are not available).Differences between hospitals are likely to havebeen diluted in this aggregate level of analysis.The reliability of hospital admission data on

self poisoning in adolescents has been assessedin Oxford.32 Here over 95% of the episodeswere correctly recorded. In the south westcoders meet regularly to discuss coding issuesand accuracy will probably have increased afterthe recent health service reforms with pur-chasers and providers placing increased import-ance on accurate recording of all episodes ofhospital admission. To ensure complete caseascertainment, and minimise variations re-sulting from differences in coding, we examinedall admissions whether coded accidental (7%of total), uncertain, or deliberate.

Parasuicide is the third commonest reasonfor acute medical admission in the south west.Over 90% of these admissions follow drugoverdose, and of these over half involve drugseasily purchased over the counter in chemistsand other retail outlets (paracetamol and as-pirin). Admission rates were highest amongmen aged 20-24 and in women aged 15-19;similar to the pattern and rates found in Ox-ford.2 The 12 month repeat rate of 101% isslightly lower than that reported in other stud-ies; it is an underestimate of the overall re-petition rate as those patients who are notadmitted after a further parasuicide are notidentifiable.212 Our estimate of 87 000 hospitaladmissions in England and Wales is slightlylower than Hawton and Fagg's in 1992. How-ever, their estimate of 100 000 per year is forhospital referrals (rather than admissions) andis based on figures from Oxford.There were significant variations in the SARs

between districts. These differences are un-likely to be due to coding errors and moreprobably reflect genuine differences in ad-mission rates. There are two possible ex-planations for these differences, both of whichare likely to contribute to the observed vari-ation. Firstly, differences in socioeconomic sta-tus may significantly affect rates of parasuicide.The data show that those districts with higherlevels of socioeconomic deprivation, as meas-ured by the Townsend score, tended to havehigher parasuicide SARs. A similar ecologicallink has recently been found for the 24 localitiesin Bristol.'5 Using ranks, 9% of the variationin parasuicide SAR is explained by deprivation.If the outlying district which has the highestparasuicide SAR and admits the greatest pro-portion of accident and emergency attenders isomitted, 68% of the variation is explained bydeprivation. The other possible explanation forthe observed variation is that hospital admissionpractices for parasuicide vary across the southwest.

Further evidence of variations was derivedfrom the postal questionnaire sent to accidentand emergency departments in the region. Apositive correlation was found between the SARfor parasuicide and the proportion of patientsadmitted from accident and emergency. Theaccuracy of the data reported in the postalquestionnaire is likely to be variable as the

sophistication and completeness of casualty in-formation systems varies across the region (seediscrepancies between tables 1 and 2). How-ever, we feel there is no reason to suspectsystematic bias in the information we have beengiven. In addition, for the calculation of theproportion of parasuicide patients admitted werequested information from the departmentson both the number of attendees and the num-ber of admissions. Any underestimate or over-estimate is likely to affect both the numeratorand denominator in this proportion as bothwere provided by the individual units.

Analysis ofthe relationship between the man-agement ofparasuicide and outcomes is limitedby the number of districts. All reported cor-relations have wide confidence intervals andshould be interpreted with caution. Where re-lationships were found four substantive ex-planations are possible. Firstly, the data maybe incorrect and relationships result from sys-tematic differences in coding practices betweenhospitals. Secondly, in-patient services maydiffer between districts - some may have psy-chiatric units on-site at district general hospitalsand this may alter management strategies.For example, the three districts where the

greatest proportion of attendees were admittedhad psychiatric inpatient facilities on-site inall their district general hospitals. Thirdly, theseverity of casemix may vary between districtsand this rather than clinical practice may resultin the observed variations in outcome. Lastly,management policies and treatment servicesfor parasuicide are likely to differ; this mayresult in differences in outcome. The role ofeach of these factors should be fully exploredusing more in-depth, focused, and preferablyprospective data collection.About 10% of admissions are referred for

psychiatric inpatient treatment emphasising theimportance ofparasuicide as a marker for severepsychiatric illness, highlighted by the long termsuicide risk. This referral rate is similar to thatfound in Oxford and Edinburgh.'2 Surprisingly,our data suggest that there is little relationshipbetween the proportion of patients admittedfrom accident and emergency and the pro-portion subsequently referred for psychiatricinpatient care. If some hospitals only admita population already screened for severity ofpsychiatric illness in accident and emergency,one would expect such hospitals to have agreater proportion of their patients sub-sequently admitted for psychiatric care. Ouranalysis found no evidence to support a policyof admission for all those who self harm. Ifanything those districts referring a greater per-centage of patients for subsequent inpatientpsychiatric care experienced a higher rate ofrepeat parasuicide admission. The basic lackofrelationship here may either be because thosedistricts with on-site psychiatric facilities ad-mitted a greater proportion of parasuicidepatients or because of local differences in theseverity of psychiatric illness among those whodeliberately self harm.A third of suicides have a history of previous

parasuicide, and in the year following para-suicide 1% of individuals succeed in killing

28

Epidemiology ofparasuicide in SW England

themselves, with the eventual figure being 3-10%.25-29 Mortality statistics show a total of104 suicide deaths from poisoning in 1992 inthe south west, thus about a quarter of thesedeaths occur in hospital in the South WesternRegion. Thus at least a quarter of all suicidedeaths by overdose in the south west occurredin hospital and it is possible that some of thesemay be preventable by following best practicemanagement guidelines. Overall, the pro-portion who died after admission for overdosewas 0 5%. These figures underestimate thetotal number of fatalities among those whoreach hospital alive, since some patients withsevere liver damage will have been transferredto specialist liver units elsewhere in the countryfor further management. The figures alsounderestimate total deaths from overdose as

many patients die prior to hospital admission.The variations seen in the south west have

been observed elsewhere and demonstrate un-

certainty concerning the best managementstrategy for these patients.810 Such variationsmay be reduced by consensus statements pro-

duced by authoritative bodies. However, as

attempts to reduce parasuicide repetition haveproved largely unsuccessfull'20 except in highrisk groups21-23 the reasons for variation shouldbe further explored. The effect of differentadmission policies on outcome should also beexamined to clarify best practice guidelines forparasuicide. There remains uncertainty aboutwhether, for most patients who self harm, ad-mission to hospital is beneficial.Randomised controlled trials are required to

determine whether, for those patients wherehospital admission is not indicated on medicalgrounds, admission is beneficial in the man-

agement of parasuicide. Management sub-sequent to acute medical admission alsorequires further controlled evaluation.

We would like to thank: Professor HG Morgan, Dr J Evans,Dr RM Kammerling, and MrW White (regional coding officer)for advice, comments and assistance; casualty consultants andmanagers in the South Western Region and the InformationDepartment Royal United Hospital, Bath for providing us withinformation on attendances at their departments.

1 Stark C, Smith H, Hall D. Increase in parasuicide in Scot-land. BMJ 1994;308:1569-70.

2 Hawton K, Fagg J. Trends in deliberate self poisoning andself injury in Oxford, 1976-90. BMJ 1992;304:1409-1 1.

3 Beck P, Lazarus J, Scorer R, Smith P, Routledge P. Increasingrates of deliberate self harm. BMJ 1994;308:789.

4 Alderson M. National trends in self-poisoning in women.

Lancet 1985;i:974-5.5 Secretary of State for Health. The Health of the nation: a

strategy for health in England. London: HMSO, 1992.6 Faculty of Public Health Medicine of the Royal College of

Physicians UK Levels ofHealth. Second Report, July 1992.

London: Faculty of Public Health Medicine 1992;27-29.7 World Health Organisation. Health for all targets. The health

policy for Europe. Copenhagen: WHO, 1993.8 House A, Owens D, Storer D. Psycho-social intervention

following attempted suicide: is there a case for betterservices? International Revue of Psychiatry 1992;4:15-22.

9 Department of Health and Social Security. The managementof deliberate self harm. Health Notice HN (84)25. London:DHSS, 1984.

10 Owens D. Self-harm patients not admitted to hospital. JRoy Coil Phys 1990;24:281-3.

11 Phillimore P, Beattie A, Townsend P. Widening inequality ofhealth in Northern England. BMJ 1994;308:1125-1128.

12 Platt S, Hawton K, Kreitman N, Fagg J, Foster J. Recentepidemiological trends in parasuicide in Edinburgh andOxford: a tale of two cities. Psychol Med 1988;18:405-18.

13 Cameron I, Renvoize E, Garside G. Deliberate self-pois-oning: the need for a new approach. Health Trends 1990;3:126-8.

14 Kennedy P, Kreitman N. An epidemiological survey ofparasuicide (attempted suicide) in general practice. Br JPsychiatry 1973;123:23-34.

15 Gunnell DJ, Peters TJ, Kammerling RM, Brooks J. Relationbetween parasuicide, suicide, psychiatric admissions andsocio-economic deprivation. BMJ 1995;311:226-30.

16 Hawton K, McKeown S, Day A, Martin P, O'Connor M,Yule J. Evaluation of out-patient counselling comparedwith general practitioner care following overdoses. PsycholMed 1987;17:751-61.

17 Gibbons JS, Butler J, Urwin P, Gibbons JL. Evaluation ofsocial work service for self-poisoning patients. BrJ Psych-iatr 1978;133:111-18.

18 Chowdhury N, Hicks RC, Kreitman N. Evaluation of anafter-care service for parasuicide (attempted suicide)patients. Social Psychiatry 1973;8:67-81.

19 Ettlinger R. Evaluation of suicide prevention after attemptedsuicide. Acta Psychiatr Scand 1975;260(suppl): 1-125.

20 Deykin EY, Hsieh C, Joshi N, McNamarra JJ. Adolescentsuicidal and self-destructive behaviour. Results of an in-tervention study. J Adolesc Health Care 1986;7:88-95.

21 Linehan MM, Armstrong HE, Suarez A, Allmon D, HeardHL. Cognitive-behavioural treatment of chronically para-suicidal borderline patients. Arch Gen Psychiatry 1991;48:1060-4.

22 Montgomery SA, Montgomery D. Pharmacological pre-vention of suicidal behaviour. J Affective Disord 1982;4:291-8.

23 Salkovskis PM, Atha C, Storer D. Cognitive-behaviouralProblem solving in the treatment of patients who re-peatedly attempt suicide. Br J Psychiatr 1990;157:871-876.

24 Morgan HG, Jones EM, Owen JH. Royal College of Psy-chiatrists. The general hospital management of deliberateself harm. A consensus statement on standards for serviceprovision. London: Royal College of Psychiatrists. 1994.

25 Gunnell D, Frankel S. Prevention of suicide: aspirationsand evidence. BMJ7 1994;308:1227-33.

26 Barraclough B, Bunch J, Nelson B, Sainsbury P. A hundredcases of suicide: clinical aspects. Br3tPsychiatry 1974;125:355-73.

27 Monk M. Epidemiology of suicide. Epidemiol Rev 1987;9:51-69.

28 Hawton K, Fagg J. Suicide, and other causes of death,following attempted suicide. Br J Psychiatry 1988;152:359-66.

29 Nordentoft M, Breum L, Munck L, Nordestgaard AG,Hunding A, Laursen Bjaeldager PA. High mortality bynatural and unnatural causes: a 10 year follow up studyof patients admitted to a poisoning treatment centre aftersuicide attempts. BMJ7 1993;306: 1637-41.

30 Hawton K, Fagg J, Simkin S, Mills J. Attempted suicide inOxford 1991 and 1992. Oxford: University Department ofPsychiatry, Warneford Hosptial, and Barnes Unit, 1993.

31 Dennis M, Owens D, Jones S. Epidemiology of deliberateself poisoning: trends in hospital attendences. HealthTrends 1990;22:125-6.

32 Sellar C, Goldacre M, Hawton K. Reliability of routinehospital data on poisoning as measures of deliberate selfpoisoning in adolescents. J Epidemiol Community Health1990;44:313-15.

33 Breslow NE, Day NE. Statistical methods in cancer research.Vol II. The design and analysis ofcohort studies. Lyon:IARC,1987;96.

29