Communication between GPs and psychiatrists : the long and short of it

We examined GP new patient referrals and the corresponding psychiatric assessment letters on consecutive attenders in the general adult psychiatry out-patient department in Central Manchester between August 1994 and January 1995. to obtain 100 pairs. The referrals were made by 61 GPs and seen by a doctor from one of 13 clinical teams (trainees and consultant/ locum consultant). Only those patients new to the Central Manchester service, who were not admitted, and where there was no documenta tion other than the GP letter itself, were included. We recorded basic data concerning the case together with the length (in words and number of pages), number of subheadings and score for both the GP and psychiatrists' letters. The letter


Good communication promotes effective patient management
and the clinical letter remains central to this process, especially in GP referrals for out-patient assessment.Williams & Wallace (1974) used a survey of GPs and psychiatrists to identify "key items" which each felt to be important elements in an effective communica tion from the other, from which they constructed a simple rating scale to measure the quality of such letters.This methodology was developed further (Yellowlees & Pullen 1984: Pullen & Yellowlees, 1985) to include an assessment of both the length and the format of the psychia trist's report, while also identifying five key items for each type of letter which were almost identical to those previously established.These criteria have been used to assess the characteristics of GPs' and psychiatrists' letters in out-patient settings (Pullen & Yellowlees. 1985;Prasher et cd, 1992).
We carried out a local audit of new referral letters by GPs and the consequent psychiatric assessment letter in reply, to investigate any differences between training grades and to identify areas for improvement.

The study
We examined GP new patient referrals and the corresponding psychiatric assessment letters on consecutive attenders in the general adult psychiatry out-patient department in Central Manchester between August 1994 and January 1995. to obtain 100 pairs.The referrals were made by 61 GPs and seen by a doctor from one of 13 clinical teams (trainees and consultant/ locum consultant).Only those patients new to the Central Manchester service, who were not admitted, and where there was no documenta tion other than the GP letter itself, were included.We recorded basic data concerning the case together with the length (in words and number of pages), number of subheadings and score for both the GP and psychiatrists' letters.The letter scores (0=item absent, l=item present, max imum score=5) were derived from the 5 'key items' for letters writen by GPs (medication, family history, main symptoms/problems, rea son for referral and past psychiatric history) and by psychiatrists (diagnosis, treatment, follow-up, prognosis and a concise explanation of the condition) (Pullen & Yellowlees, 1985).Inspec tion of the data showed that the length of GP letters was highly positively skewed, so the data were log,0 transformed to facilitate analysis and provide a better estimate of the group average.Group differences were tested by one-way ana lysis of variance or unpaired f-tests, and correla tional analysis was performed using Pearson's test.The data are presented as mean (s.d.) for normally distributed data, and geometric mean and range for log,0 transformed data.

GP letters
All but four GP letters (96%) were contained on a single sheet of paper and only 2% made use of subheadings.The mean score (average number of key items per letter) was 2.5 (0.9).There was no difference in score whether the referring letter was addressed to a named consultant or to the department in general.Ninety-seven per cent of letters described the main problem, 69% the medication, 38% the reason for referral, 35% the psychiatric history, and only 7% any details of family history.The GP letters were generally short (geometric mean 106 words, range 26 to 686) and there was a positive correlation between the score and the number of words per letter (r=OA4, P< 0.001).

Psychiatrists' letters
Psychiatric assessment letters were between one and four pages in length (mean 2.5 (0.6) pages).The majority (63%) did not use subheadings, with some using up to 12.The mean score for psychiatrists' letters overall was 3.9 (0.6) with no significant difference between grades (junior trainees 4.0 (0.7), senior registrars 3.8 (0.6), consultants 3.9 (0.5)).Diagnosis and treatment was described in all letters, and follow-up in 98%.A brief explanation of the condition was given In 57%, with prognosis the least described item (21%).The psychiatrists' letters were markedly longer than those from GPs (849 (360) words, ranging from 130 to 2244), with the senior registrars' letters significantly longer than those writen by either junior trainees or consultants (P<0.05).In contrast to GP letters, there was a negative correlation between the score and the number of words in letters written by psychiatrists (r=-0.47,P<0.001).

Comment
The GP letters scored lower in our study (mean 2.5) than in previous work which reported a mean number of items between 3.4 and 3.6 over a period of a decade (Pullen & Yellowlees, 1985).This was mainly because of less reporting of family history (7 v. 35% of letters) or the reason for referral (38 u. 88%).In agreement with a previous study which found that GP referral letters were shorter than psychiatrists would like (Blaney & Pullen, 1989), the GP letters In our study were on average very short.The positive relationship between length and key-item score suggests that an improvement in the quality of referral letter could be achieved at the cost of only a modest increase in length.The best way to achieve this is not clear, and a previous attempt to improve GP referrals to a child psychiatry clinic through an educational approach and requests for standard referral details was un successful (Naik & Lee, 1993).
Opinions regarding the appropriate length and format of psychiatrists' assessment letters have varied widely (Margo, 1982;Pether et al 1993).Key issues include the amount of time 'wasted' preparing (and reading) long letters and the often conflicting role of the letter as both a commu nication to the GP and a typed summary for the case notes.One study which addressed these questions through a survey of GPs' opinions concluded that a one-or two-page letter contain ing two or three subheadings and including the five key items was the preferred compromise (Yellowlees & Pullen, 1984).
In the present study psychiatrists' letters scored well with a mean of 3.9 compared with previous findings of 3.2 and 3.6 (Pullen & Yellowlees, 1985).The profile of key items in the psychiatrists' letters resembled that of Prasher et al (1992) apart from a lower score (57 u. 100%) for the explanation of the condition.Our finding for this item was similar to that of Pullen & Yellowlees (1985) which probably reflected strict scoring with a mark being given only for a "concise" explanation, defined as less than one paragraph in their study and less than 200 words in ours.Prognosis has been consis tently poorly described (between 14 and 27%) in studies to date.A reasonable assumption is that this is not information generally valued by psychiatrists.In contrast, GPs rank prognosis higher than hospital doctors, probably because they are more often asked about it by patients (Solomon et al, 1995).
The length of the psychiatrists' letters was broadly comparable with previous reports in varying from less than one to four pages of A4 (Pullen & Yellowlees, 1985;Prasher et al 1992).This is a relatively crude measure which is further complicated since the format of letters and size of print was not consistent.We therefore measured letter length in number of words.In marked contrast to earlier findings of an inverse relationship between seniority and length of letter, with junior trainees writing twice as much as consultants (Pullen & Yellowlees, 1985), we found that senior registrars' letters were longer than either junior trainees or consultant letters which were of equal length.One possible ex planation for this discrepancy is that Edinburgh consultants do indeed write much shorter letters than those in Manchester; certainly the 'ideal' specimen letter provided by Yellowlees & Pullen (1984) contained only some 300 words.Alter natively it may reflect closer supervision of junior trainees than senior registrars by consultants In this aspect of practice.
The negative correlation between score and length for psychiatrists' letters suggests that a relatively brief assessment letter (300-500 words) does not result in loss of relevant information and hence is more efficient given the savings in medical and secretarial time.Previous research suggests that it is preferred by GPs and therefore is also likely to be more effective in communicat ing information (Yellowlees & Pullen, 1984).
In conclusion, we would suggest that adoption of the Yellowlees & Pullen (1984) preferred format and its incorporation in junior staff training by their educational supervisor would result in more efficient communication between psychiatrists and GPs.How to improve the information in GP referral letters is unclear.