Evaluating diagnostic tests: getting the evidence straight!

Tests were in the spotlight during the pandemic like never before. Ensuring evidence about the clinical performance and effectiveness of diagnostic tests is fit for clinical and policy decisions is challenging because the evidence is often lacking or inadequate. The pandemic created immense pressure for rapidly available diagnostic testing solutions and consequently inadequate evaluations in some instances. The Royal Statistical Society (RSS) was especially concerned that many new diagnostic tests for SARS-CoV-2 antigen or antibodies were coming to market for use both in clinical practice and for surveillance without adequate provision for statistical evaluation of their analytical and clinical performance. Motivated by these concerns, the RSS convened a working group of 6 statisticians (including Prof. Takwoingi) to review the statistical evidence needed to assure the performance of new in vitro diagnostic tests for infectious diseases. The RSS Working Group on Diagnostic Tests report published in June 2021 included lessons learned during the pandemic and recommendations that address study design, regulation and transparency issues. Prof. Takwoingi's talk will focus on the study design recommendations by providing a snapshot of test evaluation, and highlighting key challenges and opportunities for getting the evidence ‘straight’, particularly for in vitro diagnostics for infectious diseases.

Using Model-based Geostatistical Methods to Support Neglected Tropical

Disease Control Programmes

Neglected Tropical Disease (NTD) Control Programmes aim to reduce the prevalence of NTDs to a level at which they can be considered to have been eliminated as a public health problem or, more ambitiously, eradicated altogether. The core statistical challenge in this endeavour is to design and analyse prevalence surveys that achieve pre-agreed performance levels (typically, acceptably high positive and negative predictive values for declaration of elimination at a specified level of spatial resolution) at minimum cost.

The NTD community is beginning to recognise that geospatial statistical methods are better suited to this task than are the currently prevailing methods based on classical survey sampling methodology (Diggle et al, 2023). In this talk, Prof. Diggle will review the basic components of model-based geostatistics (Diggle, Moyeed and Tawn, 1998), show how these have recently been applied to a national survey of lymphatic filariasis prevalence in Guyana,  and discuss other applications (some incomplete) where extensions of the basic methodology are required.

Statistical thinking is essential by governments, press and the public: effectiveness of Royal Statistical Society Working Parties, press and public engagement

Fellows can contribute to statistical science through myriad roles in the Royal Statistical Society (RSS). Roles range from volunteering, membership of local groups or sections, elected service on Council, as committee-member, editorship or refereeing, invited membership of RSS Working Parties, service as Honorary Officer or Vice-President; or simply by inspired ideas. During her talk, Prof. Bird will outline the highlights from her RSS roles on Working Parties and as Vice President of External Affairs.