Family Relationships Quarterly No. 19
- Screening for family violence
- Current trauma: The impact of adoption practices up till the early 1970s
- Responding to the children of women in prison: Making the invisible visible
- Trends in couple dissolution: An update
- Walking the talk: Facilitating evaluation in a service environment
- Parenting efficacy: How can service providers help?
Screening for family violence
Some comments relating to Family Violence: Towards a Holistic Approach to Screening and Risk Assessment in Family Support Services by Elly Robinson and Lawrie Moloney
by Bryan Rodgers
Robinson and Moloney (2010) have raised and discussed several important issues around screening for and assessment of family violence. Family violence is widely acknowledged as an important matter for research, policy, intervention and prevention, but many unanswered questions remain regarding its identification and measurement. It is helpful to see what lessons can be learned from other areas where screening has been problematic but which have longer histories of trying to meet challenges. The following comments are structured around five themes that have emerged in the measurement of other constructs of a psychosocial nature:
- Screening for what?
- The continuum of severity and where to draw the line.
- How screening tools are applied.
- Measures of signs and symptoms versus measures of functioning.
- The comparative nature of reliability and validity.
Screening for what?
Robinson and Moloney (2010) are clear from the outset that there is no single accepted definition of family violence and that this construct can mean different things to different people. This can be an especially pertinent issue if screening tools are developed by professionals who hold a different view of violence from the individuals who will be assessed using those tools. This is not a matter of "correct" and "incorrect" views but is a reflection of a fundamental feature of screening for many psychosocial constructs where there is no gold standard against which measures can be assessed. In many instances, it is more appropriate to acknowledge from the outset that we are not dealing with a single discreet entity and that the task of screening may involve the identification of two or more aspects of family violence. This, in turn, may or may not require more than one screening tool, and formal enquiry is needed to find that out.
To draw a parallel in the field of mental health assessment, the distinction between depression and anxiety has long been an area of conceptual debate and empirical investigation (Eaton & Ritter, 1988; Feldman, 1993). There are established criteria for identifying anxiety disorders and depressive disorders and many measures are available that could serve as screening instruments for one or other category of disorder (Goldberg, Bridges, Duncan-Jones, & Grayson, 1987). In practice, however, screening tools that identify depression are fairly good for identifying anxiety disorders and vice versa (Andrews & Slade, 2001; Gill, Butterworth, Rodgers, & Mackinnon, 2007). The co-existence of different disorders in the same individuals is part of the reason behind this (Andrews, Henderson, & Hall, 2001). We might expect something similar if we considered sexual violence and physical violence as separate constructs, and we need research evidence to determine whether separate tools are needed to identify each.
Overall, the absence of a gold standard in any field signals the need to consider multiple definitions and constructs, and to develop and test alternative approaches to screening for different outcomes.
Continuum of severity
Even when we hold a clear view of a unitary concept, it is unusual for psychosocial attributes to have just two states of being "present" or being "absent"; rather we are typically dealing with a continuum where something ranges from being trivial through to being prominent and serious. Reflecting this continuum, measurement tools typically provide levels of severity. This can pose a problem in the context of service delivery where choices are required between discrete service responses, such as taking no action, conducting further follow-up, or referral to a specialist service. Often, there is an arbitrary point at which such decisions are made. Where this is formalised in relation to a screening assessment, decision-making inevitably represents a trade-off between sensitivity and specificity. Sensitivity is the likelihood of identifying a problem when it is known to exist and specificity is the likelihood of correctly categorising that a problem does not exist. We can increase sensitivity very simply by applying a lower threshold for decision making, but the trade-off is such that this will necessarily lead to a decrease in specificity. In a service context, providers would be flooded with clients, including many who had little need for help. The dilemma in constraining the number of referrals (i.e., increasing specificity) is that this would decrease sensitivity and some families needing help would not get it.
The crunch for the sensitivity - specificity trade-off is where to draw the line. Even if the accuracy of screening is known precisely, this decision depends on the costs of incorrect decisions, that is, the false positive findings (that violence occurred when in fact it did not) and the false negative findings (that violence did not occure when in fact it did) highlighted by Robinson and Moloney (2010). Where to draw the line is further influenced by the benefits arising from true positives (correctly identifying people who need help) and from true negatives (correctly identifying people who don't need help), although the latter is typically of no consequence. In many settings, we do not know the actual value of the relevant costs and benefits but it could be instructive to use a range of guestimates and observe how this informs the decision as to where to draw a line for the purpose of screening.
How screening tools are applied
There is a tendency to view the reliability and validity of a measurement instrument as inherent attributes of the tool itself, but there is ample evidence to show that the way in which a tool is applied can change its measurement attributes. The same measure (ostensibly) can yield different information if used as part of face-to-face interaction, if administered in a self-completion format, or if presented in an online or computer scripted format. Many research tools that were once considered to require sensitive face-to-face interviewing are now delivered by computer administration as the preferred mode. Estimates of substance use (legal and illicit) differ substantially according to the method by which information is collected (Booth-Kewley, Larson, & Miyoshi, 2007; Richman, Weisband, Kiesler, & Drasgow, 1999). This is an issue that requires fuller investigation in the field of family violence.
As well as the costs and benefits for clients relating to the sensitivity and specificity of screening, there are important economic costs to services arising from the screening process. A similar trade-off may operate in this respect; organisations utilising less staff time for screening an individual client/family may then be able to screen more frequently and, therefore, better monitor the ongoing risk pointed out by Robinson and Moloney (2010). Of course, it is essential to consider any distress to clients during assessment and how support is provided where this occurs (and not just the accuracy of elicited information), in keeping with the concerns for safety emphasised by Robinson and Moloney (2010). This may involve the collection of information in a private way but with appropriate monitoring and follow up of responses.
Measures of signs and symptoms versus measures of functioning
Drawing a parallel with the measurement of mental health problems is again illuminating in relation to the issue of using functioning in personal and social roles as a means of assessment. Most mental health measures are based on the clinical signs and symptoms of specific disorders. However, there has been increasing recognition of the significance of mental health problems in terms of the impact on social functioning and quality of life (and, indeed, this is reflected in the current diagnostic criteria of most disorders). Assessment of functioning provides an alternative or complementary approach to the basis of assessment in that the impact on various roles (parenting, employment etc.) and the level of personal distress (including fear) can be as useful a guide to the need for support as is information on events and features of violent behaviour.
The comparative nature of reliability and validity
One notable weakness in the literature surrounding many areas of psychosocial measurement is the tendency to fall back on such statements as "measure X is a reliable and valid measure of construct A". There is no pre-determined point at which an individual measure can be determined as "reliable" or "valid" and even if that were possible it is probably not the information we are really after for either research or service-delivery uses. Usually, what we need to know is the relative performance of the different instruments available to us; that is, whether measure X is a more valid indicator than measure Y for a particular purpose. Across a wide range of subject areas, it is remarkable how few studies compare the usefulness of different measures, and most published studies report on a single instrument (Rodgers, Pickles, Power, Collishaw, & Maughan, 1999). Again, drawing on the mental health field as an example, there are just sporadic reports of studies that directly compare the performance of different measures in similar circumstances (Furukawa, Kessler, Slade, & Andrews, 2003; Gill et al., 2007; Schmitz , Kruse, Heckrath, Alberti, & Tress, 1999). This is something that should be encouraged in the family violence field as, ultimately, it is through comparative studies that informed choices can be made of which instruments are better suited to particular needs.
Screening for and assessment of family violence is in a stage of early development. "Further research is needed ..." is something of a cliché in these circumstances. Hopefully, we can add to that time-worn statement some pointers as to the direction in which future research should progress. The five themes covered above are not exhaustive; there are other themes to be explored and there are other points that could have been raised within the five mentioned. However, they serve as a rough guide, based on previous experience in comparable fields. If there is a unifying message across the five themes, it is that screening and assessment in psychosocial fields is not about the search for abstract and perfect measures of things well defined, but is essentially about the development of tools that are useful for expressed purposes. Professionals in the field of family violence have key roles in stipulating their requirements for such tools and working with those who have the technical skills to construct and refine the necessary measures.
- Andrews, G., Henderson, S., & Hall, W. (2001). Prevalence, comorbidity, disability and service utilisation: Overview of the Australian National Mental Health Survey. British Journal of Psychiatry, 178, 145-153.
- Andrews, G., & Slade, T. (2001). Interpreting scores on the Kessler Psychological Distress Scale (K10). Australian and New Zealand Journal of Public Health, 25, 494-497.
- Booth-Kewley, S., Larson, G. E., & Miyoshi, D. K. (2007). Social desirability effects on computerized and paper-and-pencil questionnaires. Computers in Human Behavior, 23, 463-477.
- Eaton, W. W., & Ritter, C. (1988). Distinguishing anxiety and depression with field survey data. Psychological Medicine, 18, 155-166.
- Feldman, L. A. (1993). Distinguishing depression and anxiety in self-report: Evidence from confirmatory factor analysis on nonclinical and clinical samples. Journal of Consulting and Clinical Psychology, 61, 631-638.
- Furukawa, T. A., Kessler, R. C., Slade, T., & Andrews, G. (2003). The performance of the K6 and K10 screening scales for psychological distress in the Australian National Survey of Mental Health and Well-Being. Psychological Medicine, 33, 357-362.
- Gill, S. C., Butterworth, P., Rodgers, B., & Mackinnon, A. (2007). Validity of the Mental Health Component Scale of the 12-item Short-Form Health Survey (MCS-12) as measure of common mental disorders in the general population. Psychiatry Research, 152, 63-71.
- Goldberg, D. P., Bridges, K., Duncan-Jones, P., & Grayson, D. (1987). Dimensions of neuroses seen in primary-care settings. Psychological Medicine, 17, 461-470.
- Richman, W. L., Weisband, S., Kiesler, S., & Drasgow, F. (1999). A meta-analytic study of social desirability distortion in computer-administered questionnaires, traditional questionnaires, and interviews. Journal of Applied Psychology, 84, 754-775.
- Robinson, E., & Moloney, L. (2010). Family violence: Towards a holistic approach to screening and risk assessment in family support services (AFRC Briefing Paper No. 17). Melbourne: Australian Family Relationships Clearinghouse, Australian Institute of Family Studies. <www.aifs.gov.au/afrc/pubs/briefing/b017/index.html>
- Rodgers, B., Pickles, A., Power, C., Collishaw, S., & Maughan, B. (1999). Validity of the Malaise Inventory in general population samples. Social Psychiatry and Psychiatric Epidemiology, 34, 333-341.
- Schmitz, N., Kruse, J., Heckrath, C., Alberti, L., & Tress, W. (1999). Diagnosing mental disorders in primary care: The General Health Questionnaire (GHQ) and the Symptom Check List (SCL-90-R) as screening instruments. Social Psychiatry and Psychiatric Epidemiology, 34, 360-366.
Bryan Rodgers is a Professor of Family Health & Wellbeing at the Australian Demographic & Social Research Institute, ANU, and an NHMRC Principal Research Fellow.