How precise does stratified medicine need to be?

Introduction

This model is designed to help researchers understand how prescribing-algorithms to guide treatment with biologics in rheumatoid arthritis (RA) are valued by users (patients or potential patients). Prescribing-algorithms for targeting safe and effective healthcare are characterised by their accuracy to predict who will, and who will not, safely and effectively respond (positive predictive value and negative predictive value). The predictive value of a prescribing algorithm may be improved by incorporating more clinical or non-clinical variables; however, this could involve additional tests (with an associated financial cost) which could potentially delay the start of treatment. Researchers must therefore decide when a prescribing-algorithm is ‘sufficient’ in terms of whether the marginal benefit of adding additional information is outweighed or equal to the marginal cost of collecting and processing such information. Likewise, patients (current and future) must compare the prescribing algorithm with conventional approaches to selecting treatment and dosage, and balance the associated benefits and risks.

Model design

The model was developed using 3,798 choices from 211 patients and members of the public in a discrete choice experiment (DCE) survey. The survey respondents made choices by trading-off the attributes of prescribing in hypothetical scnearios. This model uses the choice data to create an estimate of the probability of a person choosing one approach to prescribing over another. As all data were collected in hypothetical scenarios, the model produces a forecast but this is not necessarily what will happen in the real world. The results used in the model also reflect the average views of the sample and do not illustrate heterogeneity in the respondents' preferences.

Further information about the study design and analysis are reported in Vass, Barton & Payne (2019) Quantifying preferences for a stratified approach to treatment of rheumatoid arthritis with biologics: patients’ and the public’s perspective.

Using the model

The model is easy to use. The user specifies levels for the attributes (delay to starting treatment; positive preedictive value (PPV); negative predictive value (NPV); risk of infection; and cost saving to the NHS) for their prescribing-algorithm using drop down menus. The initial values for the prescribing algorithm are: 30 day delay to treatment; 50% PPV; 50% NPV; 5% risk of serious infection and £300 per annum per patient cost-saving to the NHS. The expected value of each approach is then estimated and presented as the probabilitiy of an individual choosing that approach to prescribing. These probabilities sum to 100% as the model assumes every indivudal will be prescribed a biologic. The probabilities are also illustrated in a bar chart.

Spreadsheet produced by Dr Caroline Vass, Research Fellow in Health Economics, Manchester Centre for Health Economics, The University of Manchester. Date: 25th April 2019