Every Monday our authors provide a round-up of the latest peer-reviewed journal publications. We cover all issues of major health economics journals as well as some other notable releases. If you’d like to write one of our weekly journal round-ups, get in touch.
Medical Decision Making
Volume 41, Issue 1
There are quite a few interesting studies in this month’s issue of Medical Decision Making. There is a study showing that offering COVID-19 testing in US post offices reduces travel distance and improves accessibility to testing; a study reporting a simulation model for obesity in Norway; three studies on shared decision making (one on its measurement in the context of general practice, another in end-of-life care, and one on risk communication); a study on bolt-on dimensions for the EQ-5D; and a study on how to collect and analyse data from clinical trials to inform the costings of a new behavioural intervention.
My highlight is a study on parametric models for survival extrapolation, by Daniel Gallacher and colleagues. In quite the exercise of data simulation and model fitting, Gallacher et al compared the performance of statistic measures of goodness of fit in assessing parametric models for extrapolation. If I understood correctly, the authors started from the published Kaplan-Meier curves, which they digitised, to then recreate the individual-level data (ILD). The next step was fitting parametric models to the recreated ILD, and simulating new ILD to simulate the recruitment of the trial. The new ILD was then used to create a complete dataset and a censored dataset, similar to what was observed in the trial, to which parametric models were fitted. Finally, they compared the predictions of the parametric models fitted to the complete and censored datasets to the source dataset.
In the comparison between log-likelihood, AIC, and BIC, BIC emerged as generally better in selecting the least biased parametric models, as long as the data were mature. When data were immature, none of the measures of goodness of fit selected the least biased models consistently. Also, and as we would expect, the authors found that the parametric distribution needs to match the ‘true’ distribution of the hazard so that the estimated survival is similar to the observed survival times.
My take-home message is that statistic measures of goodness of fit cannot be used in isolation to select parametric models for extrapolation. Qualitative assessments, such as visual inspection, comparison to historical data or data from similar drugs, and feedback from clinicians on the plausibility of predictions, are all valuable tools in selecting parametric models for extrapolation.
The European Journal of Health Economics
Volume 22, Issue 1
This month’s issue of the European Journal of Health Economics covers a wide range of topics. We start with the editorial, on the future of the pharmacy profession. This is followed by studies on policy, including presumed consent for organ donation, managed entry agreements, risk-sharing approaches to health insurance, providers’ response to reimbursement schedules, the effect of accreditation of health providers on health outcomes, and the effect of taxing sugar-sweetened beverages.
Three studies are on cost-effectiveness topics, from the cost-effectiveness of screening for hepatitis C, to modelling approaches for glioblastoma multiforme, and costs of chronic graft versus host disease. Additionally, there is a study investigating protocols for time trade-off surveys.
Being Portuguese, I couldn’t help myself but focus on a Portuguese study about the relationship between multilayer health coverage, healthcare utilisation and health. In Portugal, people can have three layers of coverage: an NHS-type coverage, which is available to all; an additional social insurance type coverage, available to some employees such as civil servants; and voluntary private health insurance. One advantage of multilayer coverage is faster and ungated access to specialist care. Aida Isabel Tavares and Inês Marques investigated whether people with multilayer coverage access health care more often and whether this multilayer coverage is related to health.
They found that people with multilayer coverage use more specialist care but similar quantities of GP care, and they have better health status than people who only have access to the Portuguese NHS. This means that people who only have access to the Portuguese NHS, who are also more likely to be poorer and in poorer health, use less health care. As the authors point out, there is an obvious equity issue as well as an efficiency question. But there is also a political question, in that any changes would need to have buy-in from those who may be faced with losing privileged access to care. Great paper – hopefully it will kick start discussions in Portugal!
Credits
