The HCR-20 has taken on a life of its own. In forensic services it has been elevated from helpful aide-mémoire into a prophetic tool worthy of Nostradamus himself. Almost every outcome is interpreted through it. Despite the evidence of its limited utility, the difficulties of predicting rare events, the narrative fallacies and other heuristic biases it creates, and the massive opportunity costs it entails, commissioners and services alike mandate its use. Yet in routine practice the problems are not acknowledged, multiple conflicts of interest lie unobserved and other opportunities are neglected.

The HCR-20 is one of the most popular structured clinical judgement tools used in forensic settings; yet, there are no published tools to assess the quality of its use. This study used the CAI-V, a tool to assess the competency of those carrying out risk assessment, to develop a quality tool for the use of HCR-20.

The audit was carried out between July 2012 and July 2013 on all patients resident in St Andrew’s Essex, a low secure unit. The results of the first audit led to an action plan for clinical improvement, subsequently re-audited a year later.

Most of the HCR-20 ratings scored in the competent range in both audits, but the greatest weakness was identified in the treatment planning section. The re-audit showed improvement, but there remained areas for development.

The audit highlighted broad areas of improvement like the need for full multidisciplinary involvement, more attention to formulation, and the need for greater consultation and information gathering from outside professionals and family members. The quality tool developed could be adapted to the requirements of any service, and used accordingly.

There are validated tools for structured professional judgement of risk of violence, but few for risk of suicide. The Suicide Risk Assessment and Management Manual (S-RAMM) is a new structured professional judgement tool closely modelled on the HCR-20. This is the first validation study for the S-RAMM. We measured inter-rater reliability, internal consistency, concurrent validity with another validated risk instrument (HCR-20) and with a measure of psychopathology (PANSS). We tested whether the tool could distinguish between groups of patients clinically assessed as at varying levels of risk of suicide or self harm.

Two researchers jointly interviewed 25 current in-patients for inter-rater reliability (Cohen's kappa) and internal consistency (Cronbach's alpha) and interviewed 81 of 83 current in-patients to assess whether the mean scores for different wards were significantly different (using ANOVA). Two other researchers made independent ratings of the HCR-20 and PANSS.

Inter-rater reliability was acceptable for all items (Cohen's kappa >0.5 for all but three items) and all sub-scale and total scores (Spearman correlations all >0.8). Internal consistency was high, (Cronbach's alpha all sub-scales >0.6). Scores stratified significantly with high scores for admission and intensive care units and progressively lower scores in rehabilitation and predischarge units. The HCR-20 historical and S-RAMM background scores did not correlate but the dynamic sub-scales correlated significantly. PANSS scores also correlated significantly with S-RAMM scores.

The S-RAMM has better than minimum acceptable characteristics for use as a clinical or research tool. Prospective studies of sensitivity and specificity are now required.