Can psychiatrists predict which new referrals will fail to attend?

on REHAB: the mean community/self-care score (Pearson's r= -0.85, P<O.OOOI); staff opinion of problems relevant to community living (Spearman's rho = 0.26, P < 0.0001); and the levels of support recommended by staff for community accommodation (Spearman's rho=0.52, P<O.OOOI). A X analysis showed a significant association between the hard-to-place category on the CPQ and a score of greater than 80 on the REHAB General Behaviour score X = 41.2, d.f. = I, P<O.OOOI, with Yates' correction.

A X 2 analysis showed a significant association between the hard-to-place category on the CPQ and a score of greater than 80 on the REHAB General Behaviour score X 2 = 41.2, d.f. = I, P<O.OOOI, with Yates' correction.

Comment
The validity of the CPQ is supported by the high levels of agreement between some of its measures and those of the REHAB. REHAB is shorter, with less than half the number of items of the CPQ, but the CPQ asks for more information regarding the specific requirements of patients, and may therefore be more useful in service planning. Further research is required to assess directly the predictive validity of both scales.
However, although there was a significant overall agreement between the CPQ's hard-to-place category and a score of greater than 80 on the REHAB General Behaviour, this was largely due to the substantial Psychiatric Bulletin (1992),16, 18-19 Nelson et al numbers of patients identified as not presenting marked difficulties. The measures of potentially hard to place patients did not identify the same individuals. While this may not affect the usefulness ofthese scales as planning instruments for the hospital population as a whole, it does suggest that one should not rely exclusively on a single measure to identify those individual patients who may prove difficult to place in the community. Can psychiatrists predict which new referrals will fail to attend?

References
JONATHAN P. WOODS, Senior Registrar, Gartnavel Royal Hospital, 1055 Great Western Road, Glasgow, Gl2 OXH It is not uncommon to hear a psychiatrist claim to be able to judge from general practitioners' letters which new referrals will attend and which will fail to turn up. However operational research has failed to define clear characteristics of patients who do not keep first appointments (Hillis and Alexander, 1990;Skuse, 1975;Zegleman, 1988). Also the standard of referral letters has been criticised in the past and shown more recently to omit key items of information (Pullen & Yellowlees, 1985). The aim of this study is to test the ability of psychiatrists of varying experience to predict non-attenders.

The study
Photocopies of ten non-urgent new referral letters were sent to 48 psychiatrists. Three of the ten patients had in fact failed to attend. The letters were selected at random from 100 referrals to the Victoria Infirmary in Glasgow. A new referral was defined as a patient never previously referred to the services or not seen in the preceding year. The patient's name and address and the identity of the general practitioner were omitted.
The psychiatrist indicated how likely attendance was by choosing one of five responses: Yes; Probably; Don't Know; Unlikely; No. If the patient had attended marks were awarded as follows: Yes (2 marks); Probably (1); Don't Know (0); Unlikely ( -1); No (-2). The scoring was reversed if the patient did not attend.

Findings
The questionnaires were completed by 40 doctors: 15 consultants, 11 senior registrars and 14 trainees. The trainee group was made up of 11 trainees in psychiatry and 3 GP trainees.
The pattern of responses to the question "Will this patient attend?" was similar in the different groups of doctors. Overall the response rates were as follows: Yes (200/0), Probably (39%), Don't Know (12%), Unlikely (23%) and No (5%). The actual nonattendance rate was 30% and this rate was well anticipated by the different groups of psychiatrists. This is shown by the following figures which indicate how often the doctors thought attendance would not or was unlikely to occur; all doctors (28%); consultants (26%); senior registrars (300/0) and trainees (310/0). However, despite demonstrating an awareness of the general non-attendance rate the doctors were poor at predicting individual cases. The mean score for all doctors was only 3.2 out of a possible 20. The mean scores for the different grades of psychiatrists were as follows; the consultants 4.5 (range 8-1), the senior registrars 3.1 (range l{}-minus 2) and the trainees 1.9 (range 8-minus 7). As can clearly be seen, the individual doctors varied considerably in their powers of prediction. There was a trend for more experienced doctors to score more highly. The consultants' superiority over the trainees was statistically significant (at the 5% level). The Wilcoxon two-sample tests for unpaired data was used.

Comment
There are issues to be addressed before concluding that psychiatrists are unable to predict nonattendance by new referrals.
Some doctors felt they could have improved their score if they had known each patient's address. Social class may be implied from postal address and 19 Zegleman found lower social class to be associated with non-attending (Zegleman, 1988), although others have not (Hillis and Alexander, 1990;Skuse, 1975). Would the identity of the GP have helped? GPs vary in their referral rates and, interestingly, Creed et al found a significant negative correlation between referral rate and amount of detail in referral letters, i.e. low referrers wrote very detailed letters (Creed et al, 1990). However the attendance rates of patients from high and low referrers does not appear to have been examined.
The Victoria Infirmary is a district general hospital serving a mainly urban area. The out-patients seen most commonly fall into the following diagnostic groups: depression, anxiety states, personality disorder, drug or alcohol abuse, adjustment reactions. New patients are rarely suffering from a psychotic illness. The non-attendance rate in the out-patient clinic is 250/0 for new referrals. It seems unlikely that the patients studied were atypical in any important way.
When a new patient fails to attend it is wasteful of time and resources as well as being frustrating for the psychiatrist. Experience seems to help psychiatrist's judgement to a certain extent. However, it would appear that colleagues who claim to be able to spot non-attenders from referral letters are deluding themselves. This study also suggests, and it should cause concern, that as psychiatrists we make incorrect assumptions about patients before even seeing them.