Use this URL to cite or link to this record in EThOS:
Title: Identifying the elements of effective hospital management of self-harm : instrumental variable and meta-regression approaches
Author: Carroll, Robert
Awarding Body: University of Bristol
Current Institution: University of Bristol
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
Background: People presenting to hospital following an episode of self-harm have an estimated risk of suicide 50 times that of the general population. Providing effective clinical care during this period is therefore critical to prevent further self-harming behaviour and subsequent death by suicide. A number of interventions are commonly implemented when treating these patients, such as psychosocial assessment, but the evidence supporting their effectiveness is largely based on expert opinion. Observational analyses evaluating the impact of these interventions have produced mixed results and are likely to be limited by the effects of confounding by indication. Aims: To estimate the association of routine elements of self-harm patient care with risk of repeat self-harm and suicide. These aspects of care will include admission to a hospital bed, psychosocial assessment, and referral to outpatient care. Methods: Meta-regression was used in a systematic review to investigate whether heterogeneity in study estimates of the risk of repeat self-harm and suicide were explained by variation in the prevalence of different treatments. As routine aspects of self-harm patient care are commonly reserved for patients at increased risk of poor outcomes, the ability of traditional epidemiological approaches to assess their effectiveness is limited. Instrumental variable methods were implemented to overcome these problems of confounding by indication which are often encountered when assessing treatment effects in observational data. The different instruments tested included those based on temporal/seasonal variation in treatment and those based on variation in treatment across institutions or preference based instruments. These potential instruments were implemented using large datasets from prospective registries (Bristol Self-harm Surveillance Register, n=2676; Manchester Self-harm project, n=15,113; National Registry of Deliberate Self-harm, Ireland, n=43,595). Results: Pooled meta-analysis of available (n=177) study estimates suggested that within one year of an initial presentation 16.3% (95% Cl 15.1 % to 17.7% ) of self-harm patients will have had a repeat self-harm episode and 1.6% (95% Cl 1.2% to 2.1 %) will have died by suicide. Study estimates of these outcomes were not strongly related to reported levels of clinical care. In a subgroup analysis, studies reporting higher levels of medical admission had lower estimates of suicide within one year (change in mean odds: -24.0%, 95% Cl -39.9% to -3.8%, p=0.034). Of the various instruments investigated, time of day of hospital presentation was most robustly associated with the likelihood of treatment. In contrast to conventional ordinary least squares regression, instrumental variable analysis suggested psychosocial assessment reduced the risk of repeat self-harm (Risk difference (RD) -0.18,95% Cl -0.32 to -0.03, p=0.017), but the possibility of residual confounding by indication, especial through the effects of alcohol use, could not be ruled out. A preference based instrument utilising between hospital variations in admission rates was used to investigate the effect of medical and psychiatric inpatient admission on risk of repeat self-harm. Medical admission was found to have little impact on risk of repeat self-harm using this instrument (RD -0.02, 95% Cl -0.04 to 0.01, p=0.153). Psychiatric inpatient admission was associated with an increased risk of repeat self-harm (RD 0.12, 95% Cl 0.05 to 0.19, p=0.00l). Conclusions: The findings from this thesis add to a converging body of evidence of the effectiveness of psychosocial assessment in reducing repeat self-harm. These data support NICE guidelines recommending such assessments for all self-harm patients. The lack of evidence describing a therapeutic benefit of medical admission justifies the policy shift away from an emphasis on this intervention. Psychiatric inpatient admission may be associated with poor outcomes for a small subgroup of patients but this finding needs further investigation owing to the considerable effects of confounding by indication associated with this intervention.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available