Conectar
Clinical practice guidelines are designed to guide clinical practice and often make causal claims when making recommendations. Sometimes, guidelines make or require stronger causal claims than supplied in the original studies, a phenomenon we call ‘causal language jump’. We aimed to evaluate the strength of expressed causation in guidelines and the evidence they reference to assess the pattern of jumps, taking diabetes as an illustrative example.
This is a systematic evaluation of guidelines and original studies cited by them, using scoping review design with deviations.
Randomly sampled 300 guideline statements (narrative sentences describing evidence to support recommendations) from four selected diabetes guidelines.
The eligible guidelines should focus on non-pharmacological treatments or preventive strategies for adult type 2 diabetes mellitus management and related complications. The eligible action recommendations and guideline statements should intend to support non-pharmacological treatments or preventive strategies of type 2 diabetes or in a general diabetic context.
We rated the causation strength in the statements and the dependence on causation in recommendations supported by these statements using existing scales. Among the causal statements, the cited original studies were similarly assessed. We then evaluated jumps by checking if the causal claims in guideline statements were stronger than in original studies, and if the causation-dependence in guideline recommendations was stronger than supplied in guideline statements. We also assessed how well they report target trial emulation (TTE) components as a proxy for reliability.
Of the 300 statements, 114 (38.0%) were causal, and 76 (66.7%) expressed strong causation. 27.2% (31/114) of causal guideline statements stated stronger causation than any of their references and demonstrated ‘causal language jump’; 34.9% (29/83) of guideline recommendations required stronger causation than provided in statements. Of the 53 eligible studies for TTE rating, most did not report treatment assignment and causal contrast in detail. The prevalence of these jumps could be partially attributed to the suboptimal use of causal and associational words.
Causal language jumps were common among diabetes guidelines. While these jumps are sometimes inevitable, they should always be justified by good causal inference practices.
Due to the growing use of high-dimensional data and methodological advances in medical research, reproducibility of research is increasingly dependent on the availability of reproducible code. However, code is rarely made available and too often only partly reproducible. Here, we aim to provide practical and easily implementable recommendations for medical researchers to improve the reproducibility of their code. We reviewed current coding practices in the population-based Rotterdam Study cohort. Based on this review, we formulated the following five recommendations to improve the reproducibility of code used in data analysis: (1) make reproducibility a priority and allocate time and resources; (2) implement systematic code review by peers, as it further strengthens reproducibility. We provide a code review checklist, which serves as a practical tool to facilitate structured code review; (3) write comprehensible code that is well-structured; (4) report decisions transparently, for instance by providing the annotated workflow code for data cleaning, formatting and sample selection; and (5) focus on accessibility of code and data and share both, when possible, via an open repository to foster accessibility. Ideally, this repository should be managed by the institution and should be accessible to everyone. Based on these five recommendations, medical researchers can take actionable steps to improve the reproducibility of their research. Importantly, these recommendations are thought to provide a practical starting point for enhancing reproducibility rather than mandatory guidelines.
FreshRSS 