EVIDENCE REQUIRED BY HEALTH TECHNOLGY ASSESSMENT AND REIMBURSEMENT BODIES EVALUATING DIAGNOSTIC OR PROGNOSTIC ALGORITHMS THAT INCLUDE OMICS DATA.

OBJECTIVES: Multi-analyte assays with algorithmic analyses (MAAAs) use combinations of circulating and clinical markers including omics-based sources for diagnostic and/or prognostic purposes. Assessing MAAAs is challenging under existing health technology assessment (HTA) methods or practices. We undertook a scoping review to explore the HTA methods used for MAAAs to identify the criteria used for clinical research and reimbursement purposes. METHODS: This review included only non-companion (stand-alone) tests that are actionable and that have been evaluated by leading HTA or insurer/reimbursement bodies up to September 2017. RESULTS: Twenty-five reports and articles evaluating seventeen MAAAs were examined, most of which have been developed in oncology. The two main models used were the EUnetHTA Core model and the Evaluation of Genomic Applications in Practice and Prevention ACCE framework. Clinical validity and utility criteria were used, as were economic, ethical, legal, and social aspects. Economic evidence on MAAAs was scarce, and there is no consensus on whether the perspectives used are sufficiently broad to include all relevant stakeholders. CONCLUSIONS: Clinical utility and efficiency were the most used criteria, with stronger evidence needed linking the use of the algorithm with the clinical outcomes in real-life practice. HTA bodies must as well consider questions related to the analytical validity of MAAAs or with organizational aspects. The two main models, the EUnetHTA Core model and the ACCE framework, could be adapted to the assessment of MAAAs.

A quality indicator can be biased by intra-hospital heterogeneity: the case for quality of patient record keeping in France.

BACKGROUND: Since 2008, French health institutions providing medical, surgical and obstetrical care are assessed on the basis of a set of quality indicators. The French National Authority for Health developed a survey design in which 80 records are randomly selected from each institution. The main aim was to assess the effects of internal heterogeneity of a hospital that comprises several units. The survey method is based on the hypothesis of intra-institution homogeneity, which overlooks the fact that in wide hospitals homogeneity is related to departments and thus leads to overall intra-hospital heterogeneity. METHODS: Simulated databases were created to modelise the heterogeneity of our hospital and computed to assess the reliance of indicator measurement. We used real data from a large teaching hospital having internal heterogeneity related to each department. RESULTS: Variance under heterogeneity was greater than under homogeneity (3- to 18-fold) leading to an increased size of the confidence interval (CI) (at 95%) from 9 (given Haute Autorité de Santé sources) to 22 (for greatest internal heterogeneity). CONCLUSIONS: The variations in a quality indicator can be explained by intra-institution heterogeneity and are not related to changes in the quality policy of the hospitals and may lead to errors in terms of pay for performance.