Search

BACKGROUND: Validating new algorithms, such as methods to disentangle intrinsic treatment risk from risk associated with experiential learning of novel treatments, often requires knowing the ground truth for data characteristics under investigation. Since the ground truth is inaccessible in real world data, simulation studies using synthetic datasets that mimic complex clinical environments are essential. We describe and evaluate a generalizable framework for injecting hierarchical learning effects within a robust data generation process that incorporates the magnitude of intrinsic risk and accounts for known critical elements in clinical data relationships. METHODS: We present a multi-step data generating process with customizable options and flexible modules to support a variety of simulation requirements. Synthetic patients with nonlinear and correlated features are assigned to provider and institution case series. The probability of treatment and outcome assignment are associated with patient features based on user definitions. Risk due to experiential learning by providers and/or institutions when novel treatments are introduced is injected at various speeds and magnitudes. To further reflect real-world complexity, users can request missing values and omitted variables. We illustrate an implementation of our method in a case study using MIMIC-III data for reference patient feature distributions. RESULTS: Realized data characteristics in the simulated data reflected specified values. Apparent deviations in treatment effects and feature distributions, though not statistically significant, were most common in small datasets (n

Non-randomised studies of the effects of interventions are critical to many areas of healthcare evaluation, but their results may be biased. It is therefore important to understand and appraise their strengths and weaknesses. We developed ROBINS-I (“Risk Of Bias In Non-randomised Studies - of Interventions”), a new tool for evaluating risk of bias in estimates of the comparative effectiveness (harm or benefit) of interventions from studies that did not use randomisation to allocate units (individuals or clusters of individuals) to comparison groups. The tool will be particularly useful to those undertaking systematic reviews that include non-randomised studies.

Background: Citizen science is a way to democratise science by involving groups of citizens in the research process. Clinical guidelines are used to improve practice, but their implementation can be limited. Involving patients and the public can enhance guideline implementation, but there is uncertainty about the best approaches to achieve this. Citizen science is a potential way to involve patients and the public in improving clinical guideline implementation. We aimed to explore the application of citizen science methods to involve patients and the public in the dissemination and implementation of clinical guidelines in oral health and dentistry. Methods: We developed GUIDE (GUideline Implementation in oral health and DEntistry), a citizen science online platform, using a participatory approach with researchers, oral health professionals, guideline developers and citizens. Recruitment was conducted exclusively online. The platform focused on prespecified challenges related to oral health assessment guidelines, and asked citizens to generate ideas, as well as vote and comment on other citizens' ideas to improve those challenges. Citizens also shared their views via surveys and two online synchronous group meetings. Data were collected on participant's demographics, platform engagement and experience of taking part. The most promising idea category was identified by an advisory group based on engagement, feasibility and relevance. We presented quantitative data using descriptive statistics and analysed qualitative data using inductive and deductive thematic analysis. Results: The platform was open for 6 months and we recruited 189 citizens, from which over 90 citizens actively engaged with the platform. Most citizens were over 34 years (64%), female (58%) and had a university degree (50%). They generated 128 ideas, 146 comments and 248 votes. The challenge that led to most engagement was related to prevention and oral health self‐care. To take this challenge forward, citizens generated a further 36 ideas to improve a pre‐existing National Health Service oral care prevention leaflet. Citizens discussed motivations to take part in the platform (understanding, values, self‐care), reasons to stay engaged (communication and feedback, outputs and impact, and relevance of topics discussed) and suggestions to improve future platforms. Conclusion: Citizen science is an effective approach to generate and prioritise ideas from a group of citizens to improve oral health and dental services. Prevention and oral health self‐care were of particular interest to citizens. More research is needed to ensure recruitment of a diverse group of citizens and to improve retention in citizen science projects. Patient or Public Contribution: This project was inherently conducted with the input of public partners (citizen scientists) in all key aspects of its conduct and interpretation. In addition, two public partners were part of the research team and contributed to the design of the project, as well as key decisions related to its conduct, analysis, interpretation and dissemination and are co‐authors of this manuscript. [ABSTRACT FROM AUTHOR]

Background/aims To assess baseline ocular parameters in the prediction of long-term intraocular pressure (IOP) control after clear lens extraction (CLE) or laser peripheral iridotomy (LPI) in patients with primary angle closure (PAC) disease using data from the Effectiveness of Early Lens Extraction for the treatment of primary angle-closure glaucoma (EAGLE) tria. Methods This study is a secondary analysis of EAGLE data where we define the primary outcome of 'good responders' as those with IOP<21 mm Hg without requiring additional surgery and 'optimal responders' as those who in addition were medication free, at 36-month follow-up. Primary analysis was conducted using a multivariate logistic regression model to assess how randomised interventions and ocular parameters predict treatment response. Results A total of 369 patients (182 in CLE arm and 187 in LPI arm) completed the 36-month follow-up examination. After CLE, 90% met our predefined 'good response' criterion compared with 67% in the LPI arm, and 66% met 'optimal response' criterion compared with 18% in the LPI arm, with significantly longer drops/surgery-free survival time (p<0.05 for all). Patients randomised to CLE (OR=10.1 (6.1 to 16.8)), Chinese (OR=2.3 (1.3 to 3.9)), and those who had not previously used glaucoma drops (OR=2.8 (1.6 to 4.8)) were more likely to maintain long-term optimal IOP response over 36 months. Conclusion Patients with primary angle closure glaucoma/PAC are 10 times more likely to maintain drop-free good IOP control with initial CLE surgery than LPI. Non-Chinese ethnicity, higher baseline IOP and using glaucoma drops prior to randomisation are predictors of worse long-term IOP response. [ABSTRACT FROM AUTHOR]

We describe our experience of using a modified version of the Cochrane risk of bias (RoB) tool for randomised and non-randomised comparative studies. Objectives: (1) To assess time to complete RoB assessment; (2) To assess inter-rater agreement; and (3) To explore the association between RoB and treatment effect size. Methods: Cochrane risk of bias assessment was performed on a sample of full text primary reports included in a systematic review comparing operative techniques for radical prostatectomy. Inter-rater agreement was assessed using the kappa statistic. Results: Twenty-four studies were judged as high overall RoB, 13 were judged as low RoB and 11 were unclear. The weighted Kappa value was 0.35 indicating fair agreement. The median (range) time taken to rate each study was 30 min (10-49). The effect estimate for all studies was 0.61 (95% credible interval (CrI) 0.46-0.83) and 0.73 (95% CrI 0.29-1.75) for low risk studies. Conclusions: Although the process was time consuming, using a modified version of the RoB tool proved useful for demonstrating conservative effect estimates. That we only achieved a fair agreement between reviewers demonstrates the urgent need for further validation to improve inter-rater agreement. We suggest additional RoB levels could improve inter-rater reliability. [This article was written with the assistance of the UK Robotic Laparoscopic Prostatectomy HTA Study Group.]

Background: Methods need to be further developed to include non-randomised studies (NRS) in systematic reviews of the effects of health care interventions. NRS are often required to answer questions about harms and interventions for which evidence from randomised controlled trials (RCTs) is not available. Methods used to review randomised controlled trials may be inappropriate or insufficient for NRS. Aim and methods: A workshop was convened to discuss relevant methodological issues. Participants were invited from important stakeholder constituencies, including methods and review groups of the Cochrane and Campbell Collaborations, the Cochrane Editorial Unit and organisations that commission reviews and make health policy decisions. The aim was to discuss methods for reviewing evidence when including NRS and to formulate methodological guidance for review authors. Workshop format: The workshop was structured around four sessions on topics considered in advance to be most critical: (i) study designs and bias; (ii) confounding and meta-analysis; (iii) selective reporting; and (iv) applicability. These sessions were scheduled between introductory and concluding sessions. Summary: This is the first of six papers and provides an overview. Subsequent papers describe the discussions and conclusions from the four main sessions (papers 2 to 5) and summarise the proposed guidance into lists of issues for review authors to consider (paper 6).