needs of prisoners and at avoiding giving spuriouslabels to large numbers of "normal" criminals.Criminality and substance misuse were recordedindependently and the diagnosis of personalitydisorder reserved for those with other evidence ofdisordered functioning.We did not find the subcategories in the ICD

(ninth revision) useful in describing inmates withpersonality disorder as most of the inmates (over80%) showed pronounced features of two or moresubcategories. Axis II of DSM-III-R suffers froma similar problem, assigning several labels toone patient. This phenomenon has been calledcomorbidity but can best be regarded as reflectingthe unsatisfactory state of existing classifications ofpersonality disorder.

Despair over the diagnosis of personality dis-order has led some professionals to reject thediagnosis' and others to reject patients who havebeen given the diagnosis. We adopted Lewis'sview that the diagnosis is problematic but indis-pensable in referring to a group of patients whoshow profound psychiatric disturbance but do notfit readily into other categories of mental illness.4The inmates we identified stood out from theirpeers by virtue of their mental state or behaviour.Usually the interviewee, other inmates, prisonofficers, and doctors shared our view that theirpersonality problems were of a nature and severitythat warranted psychiatric attention. More time orinformation may have yielded more cases, butwe would claim a degree of face validity. Acomprehensive psychiatric service for prisonerswould have to take these inmates into account.

Deciding which diagnosis is primary depends onthe purpose for which the question is asked. Ourcriterion in compiling table II was the provisionof services: which problem would dictate theimmediate management of the patient? It repre-sents an oversimplification of the reality ofpsychiatric practice.

A MADENDepartment of Forensic Psychiatry,Institute of Psychiatry,Loindon SE5 8AF

I Guze SB. Criminalliv and psychiatric disorders. New York: OxfordUniversity Press, 1976.

2 1)ell S, Robertson G. Sentenced io hospital: offenders in Broadmoor.Oxford: Oxtord University Press, 1988. (Maudslev monographNo 32.)

3 Lewis G, Appleby L. Personality disorder: the patients psy-chiatrists dislike. Brj Psych 1988;153:44-9.

4 Lewis A. Psychopathic personality: a most elusive category.PsycholMfed 1974;4:133-40.

HIV and discriminationSIR,-Although there is much emphasis on thediscrimination carried out by the United Stateswith the travel restrictions on foreigners with HIVinfection,' very little is officially known on thediscrimination carried out by health workersagainst people with HIV infection either in oroutside of the United States. In Italy we conducteda prospective study to evaluate this phenomenon.A coded questionnaire was distributed to all

outpatients and inpatients of the AIDS unit of ourinstitution. Informed consent had been obtainedand questionnaires were anonymous. Between30 May and 7 August 1991, 86 subjects filled in thequestionnaire. Sixty three used intravenous drugs(48 men and 15 women), 11 were heterosexuals(five men and six women), nine were homosexualmen, and three were men without known riskfactors for HIV infection. Among these persons,84 were HIV positive; the two others were HIVnegative but at high risk of HIV infection. Of the34 reporting episodes of discrimination by healthworkers in public or private Italian institutions,eight reported more than one episode. Twentythree of the 42 episodes involved dentists; seveninvolved surgeons; six involved internists; four

involved other specialists; and one involved ageneral practitioner. It must be emphasised,however, that some persons reported the opposite-for example, that general practitioners took careof them in a more heedful way than previously.Different kinds of episodes of discrimination werereported, but particularly common was the refusalto give the requested health service (37 episodes).If these findings are confirmed in other prospectivestudies the health authorities should considerintervening with practitioners who are not follow-ing the ethical rules of their profession.

UMBERTO TIRELLIVINCENZO ACCURSO

MICHELE SPINAEMANUELA VACCHER

AIDS Unit,Centro di Riferimento Oncologico,33081 Aviano,Italy

Delamothe T. America worries about contagion. BMJ 1991302:1418. (15 June.)

General practitioners' access tox ray servicesSIR,-Dr N E Early states, "It might be pertinentto ask radiologists how many referrals they reject(as a proportion of the total) from junior hospitalstaff, consultant hospital staff, non-fundholdinggeneral practitioners, and fundholding generalpractitioners."'

I do not have any figures for the number ofreferrals rejected but can assure him that generalpractitioners are not the only group of doctorsbeing asked to reduce the number of requestsfor examinations. We have achieved a considerablereduction in the numbers of preoperative chestradiographic examinations and of contrast ex-aminations of the urinary tract that we perform.Hospital doctors as well as general practitioners arebeing asked to reduce their requests for x rayexaminations.

T S BROWNBradford Royal Infirmary,Bradford,West Yorkshire BD9 6RJ

I Early NE. General practitioners' access to x ray services. BM_1991;303:122. (13 July.)

Complications ofpregnancy anddelivery and psychosis in adultlifeSIR,-Do obstetric complications constitute a riskfactor for later schizophrenia? From the results oftheir national follow up study Dr D John Done andcolleagues conclude that they do not.' Their studyis impressive, with large numbers and elegantstatistical analyses. But is their conclusion justified?Two aspects of the study give cause for concern:

the statistical power and the clinical factors used todefine risk. With an overall sample size of some16000 the issue of statistical power might seemirrelevent. Yet, as the authors themselves hint, thefinal number of under 50 cases of schizophreniamay well be too small to test adequately thehypothesis in question. Obstetric complications ingeneral probably confer in the order of a twofoldincreased risk of later schizophrenia.2 This is not alarge effect in comparison with that of familial riskfactors, for example. The analyses of the subgroupof patients with high risk showed that all thegroups with psychosis, including schizophrenia,but not those with neurosis had an odds ratio ofstillbirth or neonatal death of between 1-4 and 2 4.The wide confidence intervals on these figurestestify to the small sample sizes and may well

explain the inability to show a significant effect.The second problem is the choice of clinical

variables used to define obstetric risk. A model thatincludes only one variable, the prescription ofdrugs, as indicating the condition of the baby mustbe interpreted with the greatest caution. Under-standing of neonatal physiology at that time waspoor, and drug treatments were empirical-forinstance, the main indication for treatment withnikethamide (Coramine) was impending deathfrom any cause, and most babies dying in hospitalwould have received this as a last resort. It thusmakes little sense to include this as an independentpredictor of neonatal death.

Increased rates of obstetric complications in thehistories of patients with affective psychosis, aswell as schizophrenia, compared with neuroticpatients have been shown before.' One way inwhich the authors might examine further the issueof obstetric risk and later schizophrenia is to lookfor an inverse correlation between calculated riskand age at onset of the illness. Several previousstudies have reported that obstetric complicationspredict an earlier onset, and if this can be shownnot to be the case in the reported sample it willstrengthen the authors' conclusions that no linkbetween obstetric risk and later schizophrenia hasbeen shown,

SHON LEWISDepartment of Psychiatry,Charing Cross and Westminster Medical School,London W6 8RP

ANN STEWARTDepartment in Paediatrics,University College and Mliddlesex School of Medicine,London V'C1E 6JJ

1 Donc DJ, Johnstone EC, Frithi CI), Giolding J, Shepherd PM,Crow TJr. Complications of pregnancy and delivery its relationto psychosis in adult life: data from the British perinatalmortality survey sample. B.MJ 1991;302:1576-80. ( 29 June.:

2 Lewis SW. Congenital risk factors for schizophrenia. I'syvcholMtd 1989;19:5-13.

3 Lewis SW, M\urray RMi. Obstetric complication, neurodevelop-mental deviance, and risk of schizophrenia. .7 Psvchiatr Res1987;21:413-2 1.

Child sexual abuseSIR,-Dr Brendan McCormack's editorial onsexual abuse and learning disabilities drew atten-tion to an important problem affecting the lives ofchildren and adults who are unable to speak forthemselves because of severe communication orcognitive impairments, or both.' The inadequateprotection provided by current law is particularlyworrying.

There is one aspect of the author's discussion ofdiagnosis, however, that needs clarification.Dr McCormack emphasised the difficulty ofrecognising that sexual abuse is occurring andmentioned the presence of "sexualised behaviour,temper tantrums, and challenging behaviour" aspointers. Unfortunately, the latter two features arevery common in conditions in the spectrum ofautistic disorders; they arise from the characteristicsevere impairments of understanding of socialinteraction and the rules governing social be-haviour. "Sexualised behaviour" in the form ofmasturbation in public is also frequently seen. Ifthe author's recommendation that "such behaviourshould always give rise to suspicion of abuse inan adult with learning disabilities" is adopteduncritically then this would involve a large pro-portion of people with autistic conditions. If it isassumed that the parents or other carers are atfault, this would add immeasurably to the stress oflooking after those with communication problemsand socially inappropriate behaviour. I haverecent experience of three such cases in whichparents were unjustly accused solely because ofthe behaviour of their autistic children-classicexamples of the difficulty of proving a negative.

Autistic children and adults are, of course,potentially vulnerable to all kinds of abuse. The

582 BMJ VOLUME 303 7 SEPTEMBER 1991