Paul Duncan-Jones

Detecting anxiety and depression in general medical settings.

To aid general practitioners and other non-psychiatrists in the better recognition of mental illness short scales measuring anxiety and depression were derived by latent trait analysis from a standardised psychiatric research interview. Designed to be used by non-psychiatrists, they provide dimensional measures of the severity of each disorder. The full set of nine questions need to be administered only if there are positive answers to the first four. When assessed against the full set of 60 questions contained in the psychiatric assessment schedule they had a specificity of 91% and a sensitivity of 86%. The scales would be used by non-psychiatrists in clinical investigations and possibly also by medical students to familiarise them with the common forms of psychiatric illness, which are often unrecognised in general medical settings.

Measuring social relationships The Interview Schedule for Social Interaction

It is generally recognized that social relationships occupy a central position in psychiatry. To examine the role of social relationships in the onset of minor psychiatric morbidity it is necessary to construct an instrument which will meet the need for a valid, reliable and comprehensive index of social relationships. Such an instrument has now been developed and used in a sample of the general population. The Interview Schedule for Social Interaction arose from a research need, and was constructed to assess the availability and perceived adequacy for any individual of a number of facets of social relationships. These consist both of persons and of the provisions obtained through them. Data from a general population sample suggest this instrument to be sufficiently valid and reliable, and also sensitive to predictable variations between sociodemographic groups, to justify its use in clinical and epidemiological studies, both in psychiatry and general medicine.

Psychiatric disorder in Canberra. A standardised study of prevalence.

A standardised survey of prevalence has been carried out in a general population. The epidemiological method employed is innovative in its use of two established instruments, the GHQ and the PSE, harnessed together in a two‐phase design. 756 persons were interviewed, giving a response rate of 85 % in phase 1. 157 were then interviewed with the PSE in phase 2 within a few days, giving a response rate of 92 % in this weighted subsample.

The Parental Bonding Instrument (PBI): an epidemiological study in a general population sample.

The Parental Bonding Instrument (PBI) was used in a 2-wave community survey of 386 persons. The two factor structure of the instrument was confirmed, as well as the high stability of its scales over time. Respondents who had had children differed significantly from others. No association was found between social desirability, neuroticism or extraversion and either of the scales. No association was observed between the scales and psychiatric disorder, despite using covariance structure analysis to remove the effects of age and attenuation due to measurement error. Estimates of relative risk for affectionless control were low in contrast to estimates calculated from samples with psychiatric disorders. These findings may be characteristic of symptomatic persons in a general population. Because of the importance of the affectionless control construct, the findings invite further investigation in other community and treated samples.

The use of a two-phase design in a prevalence survey.

SummaryA method is described for the use of a two-phase design in a prevalence survey of non-psychotic psychiatric morbidity. The method consists of screening a population sample, then giving a standardised psychiatric interview to varying proportions of those screened, stratified according to the probability of each respondents being a case. Prevalence rates can then be calculated by weighting back to the original population. By harnessing together two well-proven instruments, the General Health Questionnaire and the Present State Examination, one can obtain a rigorous and internationally comparable description of morbidity while at the same time efficiently deploying scientific staff in the field.

The utility of latent trait models in psychiatric epidemiology

Latent trait modelling is a recent psychometric technique with great potential for the construction and refinement of psychiatric instruments. It provides a greater insight into the nature of measurement in psychiatry and the statistical machinery for improving it. This expository paper starts with a non-technical outline of the latent trait model, gives a detailed analysis of the 12-item General Health Questionnaire (GHQ) and examines points raised by the empirical analysis through computer stimulation. It is shown that the latent trait model can give a good representation of empirical data and uncover new aspects of a familiar instrument. It provides a precise methodology for evaluating the functioning of a questionnaire and for developing better short instruments. It highlights the need, and provides the means, to tailor instruments for different tasks, such as (a) screening, and (b) measuring over the whole range of the population. We examine scoring in the light of the model, and show that simple scoring is often adequate. While points for further methodological development are noted, it is argued that the method is already sufficiently developed for general application.

The reliability of reporting adverse experiences

The reliability of reporting of life-events was examined in 52 subjects attending clinics. An inventory of events and longer-standing difficulties was administered on 2 occasions, 7-14 days apart. High levels of reliability were found for the number of events, the mean score for distress or change over all events, and for the single event with highest score. The reporting of individual events was less reliable: only 70% of those events reported at either interview were reported under the same heading at both interviews. Subjective reactions to events differ in reliability according to the type of response, and they are less reliable for single events than overall. Lastly, the reliability of highly distressing events in fact lower than for the less distressing. These findings point to some of the shortcomings of inventory methods in life-event research.

The reliability of the Geriatric Mental State Examination. Community survey version.

ABSTRACT ‐ A version of the Geriatric Mental State Examination has been prepared for use in community surveys. Its reliability has been investigated on a sample of geriatric day‐patients (n= 52). Two psychiatrists separately examined each patient and audiotaped the interviews. It is argued that unless certain requirements in the data from such a study are fulfilled, interpretable reliability statistics can‐not be calculated for individual items. Where the data were sufficient in this study, the mean phi coefficient was 0.84 within interviews and 0.56 between interviews. The reliability of individual items has been assessed as a basis for further improvement in the instrument.

The path to prescription: sex differences in psychotropic drug prescribing for general practice patients.

SummaryThe excess prescribing of mood-modifying drugs to women has been observed for some time, but the explanation for the sex difference has resisted analysis. In this paper, we develop five distinct hypotheses which summarise several of the most popular explanations and subject them to empirical test using data collected independently from doctors and patients in 1301 general practice consultations. The survey was conducted in Sydney, Australia in 1976, and includes information on psychiatric morbidity, presenting complaint, diagnosis, and prescription. Cross-tabular and regression analyses suggest that higher rates of consulting by women and perhaps a higher incidence of female morbidity account for much of the prescribing differential. That is, the main sources of the sex difference in prescribing appear to lieoutside the consultation rather than arising from doctor/patient interaction. The one exception to this is that doctors tend to “underdiagnose” psychiatric disturbance in males.

The Reliability of Reports of Early Separations

A random sample of adult community residents (n 244) was asked about separations in their childhood. Eight months later, the group was asked the same questions again. Only the death of a parent and the break-up of the parental marriage were reported with high reliability. These events made up one third of all reported separations.