A predictive framework for modeling healthcare data with evolving clinical interventions

Medical interventions critically determine clinical outcomes. But prediction models either ignore interventions or dilute impact by building a single prediction rule by amalgamating interventions with other features. One rule across all interventions may not capture differential effects. Also, interventions change with time as innovations are made, requiring prediction models to evolve over time. To address these gaps, we propose a prediction framework that explicitly models interventions by extracting a set of latent intervention groups through a Hierarchical Dirichlet Process (HDP) mixture. Data are split in temporal windows and for each window, a separate distribution over the intervention groups is learnt. This ensures that the model evolves with changing interventions. The outcome is modeled as conditional, on both the latent grouping and the patients' condition, through a Bayesian logistic regression. Learning distributions for each time-window result in an over-complex model when interventions do not change in every time-window. We show that by replacing HDP with a dynamic HDP prior, a more compact set of distributions can be learnt. Experiments performed on two hospital datasets demonstrate the superiority of our framework over many existing clinical and traditional prediction frameworks.

Letters, green cards, telephone calls and postcards: systematic and meta-analytic review of brief contact interventions for reducing self-harm, suicide attempts and suicide

BACKGROUND: There is growing interest in brief contact interventions for self-harm and suicide attempt. AIMS: To synthesise the evidence regarding the effectiveness of brief contact interventions for reducing self-harm, suicide attempt and suicide. METHOD: A systematic review and random-effects meta-analyses were conducted of randomised controlled trials using brief contact interventions (telephone contacts; emergency or crisis cards; and postcard or letter contacts). Several sensitivity analyses were conducted to examine study quality and subgroup effects. RESULTS: We found 14 eligible studies overall, of which 12 were amenable to meta-analyses. For any subsequent episode of self-harm or suicide attempt, there was a non-significant reduction in the overall pooled odds ratio (OR) of 0.87 (95% CI 0.74-1.04, P = 0119) for intervention compared with control. The number of repetitions per person was significantly reduced in intervention v. control (incidence rate ratio IRR = 066, 95% CI 0.54-0.80, P<0001). There was no significant reduction in the odds of suicide in intervention compared with control (OR = 0.58, 95% CI 0.24-1.38). CONCLUSIONS: A non-significant positive effect on repeated self-harm, suicide attempt and suicide and a significant effect on the number of episodes of repeated self-harm or suicide attempts per person (based on only three studies) means that brief contact interventions cannot yet be recommended for widespread clinical implementation. We recommend further assessment of possible benefits in well-designed trials in clinical populations.

Keeping up with innovation: a predictive framework for modeling healthcare data with evolving clinical interventions

Medical outcomes are inexorably linked to patient illness and clinical interventions. Interventions change the course of disease, crucially determining outcome. Traditional outcome prediction models build a single classifier by augmenting interventions with disease information. Interventions, however, differentially affect prognosis, thus a single prediction rule may not suffice to capture variations. Interventions also evolve over time as more advanced interventions replace older ones. To this end, we propose a Bayesian nonparametric, supervised framework that models a set of intervention groups through a mixture distribution building a separate prediction rule for each group, and allows the mixture distribution to change with time. This is achieved by using a hierarchical Dirichlet process mixture model over the interventions. The outcome is then modeled as conditional on both the latent grouping and the disease information through a Bayesian logistic regression. Experiments on synthetic and medical cohorts for 30-day readmission prediction demonstrate the superiority of the proposed model over clinical and data mining baselines.

Interventions for treating obesity in children (Review)

Childhood obesity affects both the physical and psychosocial health of children and may put them at risk of ill health as adults. The number of obese children in developed countries is increasing dramatically. There are many different lifestyle change programmes to help parents and children control their weight. Although 18 research studies were found, most of these were very small studies and so evidence from them is limited. In conclusion, there is a limited amount of quality data on the effects of programs to treat childhood obesity, and as such no conclusions can be drawn with confidence.

Diabetes prevention research: a systematic review of external validity in lifestyle interventions

Context

Type 2 diabetes is a major contributor to disease burden globally. A number of systematic reviews support the efficacy of lifestyle interventions in preventing Type 2 diabetes in adults; however, relatively little attention has been paid to the generalizability of study findings. This study systematically reviews the reporting of external validity components and generalizability of diabetes prevention studies.

Evidence acquisition

Lifestyle intervention studies for the prevention of Type 2 diabetes in adults with at least 6 months' follow-up, published between 1990 and 2011, were identified through searches of major electronic databases. External validity reporting was rated using an assessment tool, and all analysis was undertaken in 2011.

Evidence synthesis

A total of 31 primary studies (n=95 papers) met the selection criteria. All studies lacked full reporting on external validity elements. Description of the study sample, intervention, delivery agents, and participant attrition rates were reported by most studies. However, few studies reported on the representativeness of individuals and settings, methods for recruiting settings and delivery agents, costs, and how interventions could be institutionalized into routine service delivery. It is uncertain to what extent the findings of diabetes prevention studies apply to men, socioeconomically disadvantaged individuals, those living in rural and remote communities, and to low- and middle-income countries.

Conclusions

Reporting of external validity components in diabetes prevention studies needs to be enhanced to improve the evidence base for the translation and dissemination of these programs into policy and practice.