It was interesting to read Wing et al's comments (above) on the HoNOS papers which appeared in the May 1999 issue of the Journal. As the creators of HoNOS it is natural that they take an interest in its development, but it is apparent that this six-year-old is now displaying some independence. The emergence of the instrument into the public domain has led to some unexpected results.

In their first sentence, Wing et al indicate that the scale is targeted mainly to clinicians within the National Health Service. In settings where the public/private mix is more equal than in the UK, and with the increasing blurring of the distinction, there is a need for a scale to be equally relevant across diverse settings. Although the HoNOS has been used with adults in in-patient, out-patient, public and private settings it is clear that its original and primary focus is service users in the public sector, and it is not so well targeted to conditions (e.g. eating disorders) encountered more frequently in the private sector.

At several points, Wing et al talk of the importance of training and supervision. They suggest that the Victorian training consisted of a brief international video link with them. That session was for the benefit of a few key staff in order to resolve a few ambiguities in the College Research Unit's training material. After this, trainers and clinical staff were trained in the recommended fashion.

The training materials issued by the College Research Unit, and the article introducing Version 4 (Reference Wing, Beevor and CurtisWing et al, 1998), talk of the importance of training, but supervision is hardly ever mentioned. It may be that as the novelty of the HoNOS wears off, and as clinicians' HoNOS training recedes further into the past, there is progressive loss of fidelity to the rules. Ventura et al (Reference Ventura, Green and Shaner1993), in relation to the Brief Psychiatric Rating Scale, wrote of the need to maintain consistency over time, of interviewer style and interrater reliability. In the real world of mental health care, the availability of quality supervision is not evenly distributed between agencies and professions, and any scale needs to survive and perform in that environment. To the extent that supervision additional to initial training is required, the economy of routine use of the instrument is somewhat diminished, but in return one would hope to gain increased confidence in the ratings.

Once a measure is introduced into routine practice (as is about to happen in Victoria), the question of maintaining data quality arises, and it is presumably in this respect that the idea of supervision has been introduced. Continuous local monitoring is one form of supervision; a centralised system of accreditation is another. It is unclear just what kind of supervision Wing et al have in mind. Now that the instrument is being used in several countries, a system centralised in Britain seems inappropriate. Ultimately, the best guarantee of data quality is meaningful use, feedback and ongoing monitoring.

The presence of a prompt, relevant and user-friendly feedback arrangement is crucial to clinician acceptance and compliance (Reference Callaly, Trauer and HantzCallaly et al, 1998a ). Much of the clinician resistance alleged by Stein (Reference Stein1999) and questioned by Wing et al can be traced to the situation whereby clinical staff fill out forms for some obscure management purpose. If staff do not have the necessary tools to use the data they themselves have collected, is it any wonder that they should be less than enthusiastic? Graphical feedback via computer seems a most suitable medium for returning ratings to raters. The College Research Unit has freely distributed a program called HoNOSSoft which does this. Potential users should be aware that this program cannot discriminate between the missing value rating of nine and the real number nine. Thus, service users with one or more missing ratings attract grossly elevated total scores. A program without this fault and a number of extra features has been developed locally (Reference Callaly, Trauer and HantzCallaly et al, 1998b ).

Wing et al make the good point that an instrument like the HoNOS should not be viewed in isolation, but ideally as part of a wider data set, like a minimum data set. In our article (Reference Trauer, Callaly and HantzTrauer et al, 1999) we were able to analyse HoNOS results against service utilisation data, and showed that certain useful conclusions could be reached. There are now several articles based on Australian in-patient settings examining the changes associated with acute psychiatric hospitalisation (public and private) and HoNOS has been shown to have a key role to play in psychiatric case mix classification (Reference Buckingham, Burgess and SolomonBuckingham et al, 1998). The outcomes information that a scale like the HoNOS can provide lends meaning and relevance to input and process information which are generally routinely collected.

Finally, we may speculate whether the current version (Version 4) of the HoNOS is the final one. Wing et al (Reference Wing, Beevor and Curtis1998) describe it as the final version, but the commentary suggests that further modifications might be needed. There is acknowledgement of the low reliabilities of certain items, and the possibility of a slightly longer instrument is entertained. In Victoria, where we have substantial experience and data on the scale, some of us are considering some modifications which, while retaining its essential features, will overcome some of the uncertainties in glossary descriptions and anchor points. To take just one example, it would be good to agree whether tobacco use is ratable on Scale 3.