University College Cork Biennial Child Protection & Welfare Conference 2011 Decision Making Workshop Dr Brian J Taylor Contents • • • • • • Assessment in decision making Predicting child homicide & abuse Use of risk factors in decision making Are we achieving anything?! Coordinated & specialist assessment Models to support analysis in assessment & decision making 2 Assessment & Decisions • “… the aim of assessment is to gather and order information for analysis so as to inform professional judgement and decision processes about care [and protection].” Taylor BJ (2010) Professional Decision Making in Social Work, Exeter: Learning Matters Post Qualifying Social Work Series [p.110] 3 Decisions and Assessment • 4.1 factors that contributed to failure: – ineffective assessment processes – faulty decision making • 4.4.1 A notable feature throughout the ... period under examination is the absence of any formal assessment of this case, particularly in relation to risk to the children. Gibbons N (chairperson) (2010) Roscommon Child Care Case: Report of the Inquiry Team to the Health Service Executive (27 October 2010), Dublin: HSE BUT what sort of assessment? 4 Some purposes of SW assessment 1. 2. 3. 4. 5. Appraise risk of specific harms Support analysis of information gathered Engage child & family in problem solving Decide on intervention Decide on eligibility for services FOR SOME OTHER PURPOSES SEE: Taylor BJ (2010) Professional Decision Making in Social Work, Exeter: Learning Matters [ch. 8] 5 Prediction in SW Assessment • We want to predict harm so as to focus scarce resources on those most likely to come to harm • Particularly we want to predict the most serious (& less frequent) harms • BUT predicting rare events is particularly problematic … 6 Intuitive (Clinical) Prediction • My own life experience as a child? • My own life experience as an adult? • Relevant similar cases in my working life? • My understanding of research (?!) • My understanding of theories (e.g. stigma, attachment, bonding, loss & bereavement) & their relevance? 7 Clinical Prediction & Bias • Limited life & work experience • Feedback on success with clients & families is generally limited • My perceptions may be influenced by vivid or more recent events, workload pressures, personal crises etc … • My judgements may be unduly influenced by the media, peers, political pressure … 8 Informed by Evidence? • IF we knew how often the undesirable event occurred in the past • AND the relevant characteristics of the situations where it occurred • THEN we could use that knowledge to predict how often the event might occur in the future • = actuarial prediction • used by insurance companies 9 Knowledge of Risk Factors • Re-offending by serious offenders • Homicide by people with mental health problems • Suicide • Admission to long-term nursing home care • Child abuse & neglect 10 Suicide risk factors (England & Wales) (Gunnell, 1994) < 4 wks after disch. from psych hospital, male *200 < 4 wks after disch. from psych hospital, fem *100 alcohol abuse or drug misuse *20 current or ex-psychiatric patient *10 prisoner *5 doctor *2 farmer *2 unemployed *2 HOW DO WE USE THIS SORT OF KNOWLEDGE? 11 Example of A Scoring Tool • • • • • • • • Age Less than 25 Victim gender Any male Relationship to victim Any non-related Past sex offences 0 1 conviction OR 1 to 2 charges 2 to 3 convictions OR 3 to 5 charges 4+ convictions OR 6+ charges 1 1 1 0 1 2 3 Thornton D (2007) ‘Risk Matrix 2000 (Revised): Assessment and Management of Sex Offenders’, Probation Circular PC17/2007, 1 July 12 Recidivism & predictive validity Score No (%) in group at 5 yrs at 10 yrs 0 1 2 3 4 5 527 (20%) 806 (31%) 742 (29%) 326 (13%) 139 (5%) 52 (2%) 4% 8% 14% 25% 33% 50% 7% 11% 21% 37% 49% 73% 13% 20% • Total 2592 13 Predictive assessment: 4 possibilities 1.correctly predict the harm -true positive 2.indicate that harm will not occur, but it does -false negative 3.correctly predict that harm will not occur -true negative 4.indicate that harm will occur, but it does not -false positive 14 Suicide Prediction • 2006 NI: 227 male & 64 female • 300 in population of 1.5 million • i.e. 2 in 10,000 • Could we use a screening (assessment) tool with 90% accuracy to identify these people so as to target services? 15 TOOL: Yes REALITY: harm No harm Total Yes harm 270 30 300 No harm 149,970 1,349,730 1,499,700 Total 150,240 1,349,760 1,500,000 16 Issues in using risk factors • False positives = a child taken wrongly from a family [happens anyway?!] • Labelling if there is an issue of stigma • Human rights • LIMIT: Better prediction for the original population = less generalisable to others • We can do nothing about static factors; need to focus on dynamic risk factors 17 Actuarial vs Intuitive Prediction • Most of the research shows that (actuarial) prediction using risk factors is more accurate than (intuitive) professional ‘clinical’ prediction • When the base rate (incidence) is low any prediction has poor accuracy • The causal routes to extreme harm are varied, so even good (simple) use of risk factors (as by insurance companies) is difficult 18 Expected to use risk factors? • Inquiry reports seem to assume that risk factors will be taken into account by professionals in safeguarding roles • In social work we need to learn to understand & use this type of data so as to make better informed decisions • Decision model 1 19 Prospects for Risk Factors • screening from a sub-population (eg those referred) is more accurate than predicting from the general population • Predicting RE-abuse or RE-offending from amongst previous abusers or offenders is likely to be more accurate than predicting first time events • USE risk factors to INFORM current modes of unaided professional judgment 20 Predicting harm to individual children in social work is like expecting an insurance company to predict precisely WHICH young drivers will have an accident! 21 Are we achieving anything in terms of protecting children?! • • • • • • Death as the “sharp end” of abuse World Health Organisation data Use 3 year averages as small figures 1980 = average of 1979, 1980 and 1981 2005 = average of 2004, 2005 and 2006 International Classification of Diseases ICD-10 cause of death definitions: http://apps.who.int/classifications/icd10/browse/2010/en 22 Violent Possible ‘Child Abuse Related Deaths’ = Homicide + Other External Causes of Death + Accidents & Adverse Events (rate per million, age U 14) • rpm 1980 2005 %change • RoI 209 30 -86 • UK 183 54 -70 NB death through Ill Defined Signs and Symptoms may include some Sudden Infant Death Syndrome, but these are NOT included here 23 Nos Child (U14) Deaths 1980 vs 2005 • • • • • • • Rep Ireland 1980 All Causes 1217 Homicide 03 OECD 00 AAE 177 IDSS 127 poss ‘CARD’ 307 2005 330 01 01 24 32 58 %change -73% -67% +100% - 86% -75% -81% 24 Improving assessment & decision making 5.3.2 It is recommended that: A national common assessment framework be introduced without delay for all child welfare and protection cases. The framework needs to identify core components while allowing for flexibility. Gibbons N (chairperson) (2010) Roscommon Child Care Case: Report of the Inquiry Team to the Health Service Executive (27 October 2010), Dublin: HSE 25 Integrated Assessment: Aims • E.g. CAF (England & Wales), UNOCINI (NI), GIRFEC (Scotland) • Assessment proportionate to need • Gather information once • Capture all relevant perspectives • Structure information from multiple sources • Simplify access to information • Promote integrated professional working • Standardise data gathered for service development 26 Holistic, integrative assessment • • • • • • • Provide a holistic overview Ensure comprehensive breadth Avoid missing the ‘obvious’ Reduce burden by reducing duplication Enable multi-professional working, inc: Synthesise specialist assessments Support consistent decision making 27 Holistic & Specialist Assessment • Social Workers are normally given the role of coordinating assessment across professionals & organisations • BUT we also need to contribute specialist social work assessment 28 Specialist Social Work Assessment • Family functioning: eg genograms • Engaging parents re neglect: eg GRADED Care Profile • Support for parents: eg Arizona Social Support Scale • Parenting & environment: eg HOME scale • Risk of re-offending: eg Matrix 2000 • Domestic Violence: eg Maddie Bell Tool 29 Munro E (2010) Munro Review of Child Protection - Part One: A Systems Analysis, London: Department for Education • Beware of exclusive focus on technical solutions – increasing rules, more detailed procedures, more use of ICT • We need: – skills to engage with families – expertise to bring about enduring improvements in parenting behaviour – support that enables [us] to manage emotional dimensions of the work without it harming judgment or own well-being 30 Assessment Tools should support: • Effective engagement of clients & families in discussing risks, values, facts, knowledge • Gathering & ordering of appropriate client & family data • Analysis of this data to aid: • Decisions in uncertainty regarding possible services or statutory interventions 31 Analysis in Assessment • Most tools are good at: – Ensuring suitable data gathering – Helping to order data – Supporting communication across professions & organisations • Most of our tools are poor at: – supporting analysis of the data gathered ie relating client data to theory about families, research on what works, etc 32 Models of Decision Making to aid analysis of assessment information Model 2 Image Theory for Client Decision Making - sequenced stages, used only as required 1 fundamental values 2 life goals 3 tactics to achieve my aims e.g. engaging with a case conference Consider Social Work role in this Lee Roy Beach & Terry Connolly (1997) The Psychology of Decision Making: People in Organisations California: Sage 34 Model 3 Balancing Benefits and Harms for Care Planning Decisions 1. Clarify options 2. Give a Value for each outcome 3. IF UNCERTAIN: Clarify Likelihood of each option 4. Value x Likelihood 5. Weigh up = SUBJECTIVE EXPECTED UTILITY 35 Weighing up options for a child leaving state care: - Value V * Likelihood L = Total T - Support: V=6 L = 0.7 T = 4.2 - Likely to be led astray: V = -8 L = 0.3 T = -2.4 Total = 1.8 Value scores -10 to +10 in this example Can develop into a decision tree 36 Model 4: Fuzzy Prediction • Brearley Model (fuzzy set theory) • Conceptualise risk factors without quantifying them • Vulnerabilities (predisposing factors) • Triggers (precipitating factors) • Strengths & Mitigating factors Paul Brearley (1982) Risk in Social Work London: Routledge & Kegan Paul 37 Model 5: Use a Decision Process 1. Clarifying decision context 2. Involving the client 3. Meeting with stakeholders 4. Thinking and feeling about the decision 5. Framing the decision situation 6. Choosing an option 7. Evaluating the decision Terence O’Sullivan (1999) Decision Making in Social Work Basingstoke, Hampshire: Macmillan 38 Model 6: Threshold Decisions • Protection or eligibility for services • Consider situation against a threshold • Model 6A: Satisficing Model – IF limited time or information – SELECT the first option found that is ‘good enough’ • eg placement during out-of-hours Herbert Simon (1956) Rational choice and the structure of environments, Psychological Review, 63, 129-138 39 Model 6B: Bounded Rationality Extends & generalises Satisficing Model • RULE 1: How to search for information • RULE 2: When to stop searching • RULE 3: A simple rule for choosing an option Gerd Gigerenzer & Richard Selten (2002) Bounded Rationality: The Adaptive Toolbox, Cambridge, MA: MIT Press 40 FOR MORE on this topic Decisions, Assessment, Risk and Evidence in Social Work Conference: 02 & 03 July 2012 organised by University of Ulster Email: dare@ulster.ac.uk www.socsci.ulster.ac.uk/irss/dare2012 Templepatrick nr Belfast International Airport 41 University College Cork Child Welfare Conference 2011 Decision Making Workshop THE END Dr Brian J Taylor