Skip to content

Sam Watson’s journal round-up for 21st March 2016

Every Monday our authors provide a round-up of some of the most recently published peer reviewed articles from the field. We don’t cover everything, or even what’s most important – just a few papers that have interested the author. Visit our Resources page for links to more journals or follow the HealthEconBot. If you’d like to write one of our weekly journal round-ups, get in touch.

Firearm legislation and firearm mortality in the USA: a cross-sectional, state-level study. The Lancet [PubMedPublished 10th March 2016

A good example of not letting bad statistics get in the way of a good story? Possibly. The authors of this study attempt to analyse the effect of different firearm laws on firearm mortality across states in the US. The authors set up a cross-sectional regression of state level gun deaths on indicators for a wide range of different gun legislation such as background checks and identification requirements. Those coefficients coming back as statistically significant are reported to indicate successful legislation. Some criticism of the study has arisen: there were only 50 data points for 25 regressors and there was no correction for multiple testing, among other things. The authors have recently penned a reply. A study from two years ago conducted a similar exercise and reported that the strength of gun legislation was associated with fewer gun deaths, but on closer inspection it does not seem to explain much of the variation in gun deaths. I’ll let you be the judge of the quality of the study, but it’s a good exercise in trying to interpret what may well be noise.

Multilevel structural equation models for longitudinal data where predictors are measured more frequently than outcomes: an application to the effects of stress on the cognitive function of nurses. JRSSA Published 10th March 2016

This study provides both an interesting method and application. The authors explicate a multilevel structural equation model for when the predictors are measured more frequently than the outcomes. Such a scenario may occur in studies of, for example, the effect of childhood factors on adult outcomes. Birth cohorts often have various aspects of their lives measured at multiple time points but outcomes in adulthood are measured less frequently or even only once. The method is applied to look at how stress affects the cognitive function of nurses working on a telephone helpline. Stress is found to affect reaction times and the risk of making an error and some decompositions into within nurse and between shift effects are provided.

Regression discontinuity designs in healthcare research. BMJ [PubMedPublished 14th March 2016

There is often a delay between the development of statistical methods and their wider acceptance in applied research. Regression discontinuity designs have been a mainstay in economics and other quantitative research for a while. The discontinuity in question is an arbitrary threshold above or below which a treatment is applied, people close to the threshold on either side are likely to be very similar to one another but only those on one side get the treatment. The publication of methods papers such as this for the general reader facilitate the use and reporting of novel methods in non-specialised journals. This paper provides a good overview of the methods and a few examples from published studies including: patients need to have a CD4 count <200 cells/mm3 to be eligible for HIV treatment in South Africa, and babies born at less that 1,500g in the US are recommended more treatment than those born heavier.

Florence Nightingale, William Farr and competing risks. JRSSA Published 10th March 2016

And finally an interesting piece of statistical and health services research history. Florence Nightingale was both a pioneering nurse and reformer of hospitals and was the first female member of the Royal Statistical Society after being recommended by her long time collaborator William Farr. Farr and Nightingale analysed hospital mortality rates using some techniques more complete than many used today. In this paper, the authors discuss the relationship between competing risks (where a patient may exit into one of a number of mutually exclusive states like death or discharge), incidence rates (number of new cases with respect to time at risk), and incidence proportions (number of new cases with respect to number of patients at risk). In a competing risks framework the incidence proportion, which is often calculated for hospital epidemiology studies, is biased in the presence of right censoring, but the incidence rate of one event is not interpretable without the incidence rates of the competing outcomes being calculated. This papers argues that Nightingale and Farr were well aware of this problem unlike many of their modern day contemporaries.

We now have a newsletter!

Sign up to receive updates about the blog and the wider health economics world.

0 0 votes
Article Rating
Notify of

Inline Feedbacks
View all comments
Join the conversation, add a commentx