The operational diagnostic systems, such as RDC [1], DSMIV [2] and ICD10 [3], that have been developed over the last 30 years are widely used by clinicians and researchers and have been an important methodological advance over earlier, non-structured approaches [4]. Although open to a variety of criticisms and unlikely to map directly onto the pathophysiology of the disorders, the operational approach is relatively simple, provides acceptable levels of reliability and is useful for communication and decision making regarding management, research and service provision. In most cases the categories defined are informed by a broad range of research data and are revised at regular intervals to take account of new findings and concepts as they emerge.

However, for clinicians and researchers, such as ourselves, interested in disorders lying within the Bipolar spectrum there are several problems in using the current systems for lifetime diagnosis. These include:

1) It is difficult to operationalize the integration of diverse episodes of psychopathology – The operational systems perform best for categorizing a discrete, well-delineated single episode of psychopathology – for example, there are clear-cut criteria to define episodes of mania and major depression. However, a lifetime diagnosis requires integration of the lifetime experience of psychopathology and the criteria used are much less easy to operationalize, typically requiring judgements about the balance between different types of episode [4].

2) Hierarchies lead to loss of information – The existence of explicit, or implicit, hierarchies creates the situation in which certain symptoms "trump" others. For example, an individual can have a diagnosis of DSMIV Schizophrenia despite having had more episodes of mania during his or her lifetime than another individual with a diagnosis of Bipolar I Disorder. For those interested in Bipolarity, this results in a serious loss of important information.

3) Boundaries between diagnostic categories are often arbitrary – Although there is usually a plausible evidence base and/or conceptual basis to support the separation into distinct diagnostic categories, the criteria used to define the boundaries between categories is almost always arbitrary. For example, the level of impairment defines the boundary between Bipolar I and II Disorders – it is most implausible that any specific level of impairment could neatly carve the boundary between distinct disorders. Similarly, in DSMIV the boundary between Bipolar I Disorder and Schizoaffective Disorder, Bipolar Type is defined by the precise timing of occurrence of psychotic symptoms outside of an affective episode. Table ​Table11 lists some key diagnostic boundaries relating to bipolar spectrum disorders and the criteria used in making diagnostic decisions at these boundaries.

4) Boundaries between categories usually require a major element of subjective interpretation – Most of the key boundaries for diagnoses within the Bipolar spectrum require judgements about severity and/or the balance of symptomatology. This substantial subjective element reduces reliability and, as commonsense suggests and everyone who has ever participated in formal reliability exercises knows, cases lying near diagnostic boundaries contribute most of the diagnostic disagreements.

5) Available diagnostic categories are relatively unhelpful in distinguishing severity – No distinction is made between an individual that just meets the threshold for a specific category and another individual that has had multiple severe episodes.

6) Subclinical cases are not accommodated usefully within the current diagnostic categories – An individual who has a DSMIV Diagnosis of Major Depressive Disorder may have experienced multiple mild sub-hypomanic episodes. Indeed, it is being increasingly recognized that the bipolar spectrum extends well beyond the traditional Bipolar I and II categories and includes many individuals with formal diagnoses of unipolar major depression under current operational systems [5,6]. For those interested in bipolarity this is wasteful of information.

7) "Not Otherwise Specified (NOS)" categories are highly heterogeneous – Because of the need for a catch-all category that can accommodate the set of cases not fitting the other operational criteria, the NOS categories in operational systems may include a wide range of types of case ranging between the mild and the severe. Thus, such categories provide little information about the lifetime psychopathology of individuals having the diagnosis.

Dimensional classifications offer an alternative to the conventional categorical approach and have the potential to address many of the issues listed above [7]. Dimensional classifications are not a new idea but have not in the past been widely adopted by either the clinical or research community. Some of the disadvantages, which have impeded widespread use in clinical and research settings in psychiatry, stem from their relative complexity – leading to difficulty in use and interpretation in areas such as communication and decisions regarding management and services.

1) The presence and severity of manic syndromes.

2) The presence and severity of depressive syndromes.

3) The presence of psychotic symptoms and the balance of mood and psychotic symptomatology.

4) The mood congruence of psychotic symptoms and the temporal relationship between affective and psychotic symptomatology.

Our group and our collaborators have extensive experience of use of BADDS as an adjunct to conventional operational diagnosis and it has been part of our standard assessment approach for over 5 years. We have found that it is straightforward to use and adds little to the time taken to complete the consensus diagnostic process.

For researchers, such as ourselves, wishing to establish a measure of "caseness" BADDS can easily be used to define thresholds – for example, a study of mania might require that cases be included only for M > 64. This would allow inclusion of all cases with the equivalent of 3 of more episodes of mania, irrespective of diagnosis. In a study of psychotic Bipolar spectrum illness it might be important to distinguish between cases in which psychotic features were a prominent, recurrent feature of illness (rather than an occasional relatively minor feature). Such individuals could be selected using BADDS as having P > 50, together with M > 60. BADDS can also easily be used in conjunction with categorical diagnoses for case selection.

BADDS was developed within the context of family studies and it lends itself to providing a substantially more useful description of the milder ("sub-clinical") end of the Bipolar spectrum which is frequently encountered within members of families of probands with full-blown Bipolar illness. Conventional categorical approaches often lead to unsatisfactory diagnoses such as "Never ill", "Major Depressive Disorder" or some form of mild "Not Otherwise Specified" category when it is clear that there is some definite, albeit mild, degree of bipolarity. Within the context of family studies it is extremely wasteful to discard such quantitative information about the presence and extent of bipolar features and BADDS provides a simple approach to making simple but efficient use of such data.

Directly related to this issue, there is currently great interest in delineating the breadth and frequency of expression of the bipolar illness spectrum in the population. Recent research, championed by Akiskal and Angst, provides evidence that many cases that have been regarded as being "unipolar major depression" actually have subtle (or not so subtle) bipolar features [5,6] and classifications have been suggested that recognize several categories of milder bipolarity in addition to the conventional DSMIV categories of Bipolar I and II Disorders [eg. [5]]. BADDS provides the capability to capture information about this milder degree of bipolarity – a substantial part of the M dimensions (the range 0 – 39) is available for rating sub-clinical hypomanic features.

The dimensional approach is particularly beneficial for cases close to diagnostic boundaries. As any researcher or clinician knows who has undertaken formal diagnostic assignment using operational classifications such cases may be associated with a substantial investment of time in order to make a finely balanced decision between two (or occasionally more) discrete diagnostic groups. It is common that different raters come down on different sides of the finely balanced decision process leading to a split of diagnoses with eventual agreement on a consensus but often with further agreement that it is a "difficult case" and that the single category chosen does not quite do justice to the complexity of the case. In contrast, the dimensional approach of BADDS provides a scheme which can reflect that different ratings of such cases are relatively close on the quantitative scale. An example is provided by a case considered in the formal reliability exercise in which the subject experienced several severe (but not incapacitating) major depressive episodes and also mild recurrent sub-hypomanic episodes. Of the 7 raters, 4 made a diagnosis of DSMIV Bipolar Disorder, Not Otherwise Specified (the consensus) and 3 a diagnosis of Recurrent major Depressive Disorder. In contrast the dimensional ratings were very similar across all raters (means for those raters making diagnosis of Bipolar Disorder Not Otherwise Specified: M 27.3; D 70.3; P 0; I blank; means for those raters making diagnosis of Recurrent Major Depression: M 23; D 74.7; P 0; I blank).

The primary purpose in developing BADDS was to use it as an adjunct to better describe some key features of cases and provide a simple mechanism for case selection on the basis of these features. BADDS has already been used within family-based studies to investigate intra-familial resemblance for lifetime experience of mania and psychosis [11] as well as investigating the relationship between smoking and psychosis in Bipolar Disorder [12]. We are currently using BADDS to explore genotype-phenotype correlations within the context of both classical and molecular genetic studies of large samples of patients with functional psychosis and mood disorder.

There are several limitations in use of BADDS, most of which are common to other lifetime diagnostic procedures. First, and most obvious, is that the ratings are entirely dependent upon the quality of the data. Poor data will inevitably lead to poor dimensional ratings as well as poor categorical diagnoses. It is essential that multiple data sources are used whenever possible that provide adequate description of an individual's lifetime experience of psychopathology (not just one or two representative episodes). As for any type of rating, poor data would be expected to affect both the validity and reliability of ratings. Second, ratings can only reflect what is known of the lifetime experience of psychopathology up to the time the ratings are made. In the light of new episodes of illness scores on the M and D dimensions may increase; those on the P and I dimensions may increase or decrease. Third, subjective judgments are required in integrating multiple data sources and matching data to the criteria within the guidelines. Within the context of our current approaches to psychiatric classification this is inevitable. Judgements must still be made about the range for a rating – this can be equivalent to making a categorical judgement, except that the different categories lie contiguous with one another on an ordered dimension. Fourth, there are features of Bipolar spectrum illness that BADDS was not designed to capture – examples include the presence and extent of rapid cycling and the extent of mixed episodes (although if all manic episodes are mixed this is denoted in BADDS by adding an "m" qualifier to the M dimension – see rating guidelines in Appendix A). It is possible for additional dimensions to be added to capture additional features. Fifth, BADDS was not developed for use in the general population. It was designed for use in clinical populations likely to contain patients with Bipolar spectrum diagnoses. The dimensions have meaning in providing an ordered measure of specific domains of psychopathology. The distributions remain to be tested in non-clinical populations but will certainly not conform to normal distribution. Sixth, for the M and D dimensions there is a ceiling effect in that these dimensions do not allow discrimination between individuals having more than 11 episodes of incapacitating mania, or depression, respectively. In practice, however, for the populations of patients that we have studied relatively few patients score M = 100 or D = 100. Seventh, BADDS is relatively poor at characterizing cases where the majority of episodes are at a lower level of severity than the most severe.

Our justification for developing BADSS was that no dimensional scale was already available that adequately addressed the issues (1) – (7) discussed in the background section. However, several researchers have described approaches relevant to dimensional ratings of psychopathology including bipolar features. Depue has described a quantitative scale for screening for Bipolar and Unipolar disorders within a non-clinical university population [13]. This derived a bipolar and a unipolar dimension from a modified version of the General behaviour Inventory [14] and focused on screening for affective psychopathology at the milder end of the spectrum. Brockington and colleagues have described a complex procedure for lifetime psychopathological assessment that includes a detailed interview schedule and case note review (taking 9 hours per patient) and produces lifetime summary scores on 30 scales covering a wide range of psychopathology [15]. One popular approach to lifetime rating of psychopathology for functional psychosis is provided by OPCRIT, a computer-based 92 item checklist that includes symptoms over a range of domains including positive, negative and disorganized psychotic symptoms, course variables, depressive symptoms and manic symptoms [16]. OPCRIT can be used in a variety of ways but was primarily developed as a diagnostic system. It performs best for schizophrenia spectrum disorders, although it can be used satisfactorily in diagnosis of Bipolar Disorder [17]. However, OPCRIT does not provide a dimensional measure of severity or frequency/duration of the domains of psychopathology and, in its unmodified form, is much less satisfactory for use with disorders having a predominantly episodic course. However, within these constraints, OPCRIT has been used by several groups for investigating factor structures of patient sets with functional psychotic illness [eg. [18,19]]. Several groups working on the genetics of psychosis have described dimensional approaches that focus on the psychotic domains of psychopathology. Maziade et al have examined lifetime ratings of psychotic symptom dimensions in patients with schizophrenia and Bipolar Disorder [20]. There was no coverage of mood symptoms and assessments were confined to predominant symptoms in acute episodes and predominant symptoms "between" episodes. Kendler et al [21] have used clinical judgement to make ratings on a 4 point scale that reflected severity and duration for each of 9 symptom and 2 course variables which included depressive symptoms and manic symptoms. Levinson and colleagues [22] have recently described a lifetime dimension scale for use in psychosis research, the Lifetime Dimensions of Psychosis Scale (LDPS). This was developed within the context of family-genetic studies of schizophrenia and motivated by several of the same concerns and aims that motivated us in developing BADDS. Ratings are made on a 39 item scale that reflect severity (on a 5 point scale) and duration (on a 5 point scale) for lifetime occurrence of a range of psychotic features encompassing positive, bizarre, negative and disorganized domains plus depressive and manic syndromes. As with the approach taken by Maziade and Kendler, the focus of LDPS is schizophrenia spectrum disorders and a chronic course. There is relatively little attention to the milder mood psychopathology, episodic course and to the relationship between mood and psychotic symptomatology. These are all issues of key importance to study of bipolar spectrum illness and are a focus of BADDS.

A further useful approach in characterization of episodic disorders such as Bipolar Disorder is the life chart method [23] which provides a visual schematic representation of illness using a time-line on which is charted key events, illness episodes and treatments during an individual's life. We find this an invaluable component of our own assessments but it in general it is necessary for quantitative and qualitative information about illness type, frequency and severity to abstracted from the life chart for use in research or clinical settings. BADDS clearly does not provide the full richness of individual description of the life chart method but is designed to capture some of the important features of an individual's lifetime experience of illness.