Researcher practices on returning genetic research results.

BACKGROUND/AIMS: as genetic and genomic research proliferates, debate has ensued about returning results to participants. In addition to consideration of the benefits and harms to participants, researchers must also consider the logistical and financial feasibility of returning research results. However, little data exist of actual researcher practices. METHODS: we conducted an online survey of 446 corresponding authors of genetic/genomic studies conducted in the United States and published in 2006-2007 to assess the frequency with which they considered, offered to, or actually returned research results, what factors influenced these decisions, and the method of communicating results. RESULTS: the response rate was 24% (105/446). Fifty-four percent of respondents considered the issue of returning research results to participants, 28% offered to return individual research results, and 24% actually returned individual research results. Of those who considered the issue of returning research results during the study planning phase, the most common factors considered were whether research results were deemed clinically useful (18%) and respect for participants (13%). Researchers who had a medical degree and conducted studies on children were significantly more likely to offer to return or actually return individual results compared to those with a Ph.D. only. CONCLUSIONS: we speculate that issues associated with clinical validity and respect for participants dominated concerns of time and expense given the prominent and continuing ethical debates surrounding genetics and genomics research. The substantial number of researchers who did not consider returning research results suggests that researchers and institutional review boards need to devote more attention to a topic about which research participants are interested.

Increasing Motorcycle Taxi Driver Conspicuity in Tanzania

Motorcycle crash related injuries and deaths are increasing rapidly in many African nations. Utilizing conspicuity measures, such as wearing reflective, fluorescent safety vests, are effective as crash prevention strategies. Furthermore, use of some conspicuity measures is mandated by law among motorcycle-taxi drivers in Tanzania. Nonetheless, uptake remains low. Locally appropriate strategies to improve crash preventative behaviors are needed.

To explore whether use of conspicuity measures could be improved through eliminating cost-barriers, we tested a distribution strategy involving the provision of free motorcycle safety vests among a population of motorcycle-taxi drivers in Moshi, Tanzania. We conducted a cluster randomized controlled trial among 180 motorcycle-taxi drivers in which half of the participants (90) were randomized to the intervention arm and received a free reflective vest. The other half of the participants (90) were randomized to the control arm and did not receive free vests. Whether motorcycle taxi drivers used the reflective vest was then unobtrusively observed on city streets over a period of three months.

Mixed-effects logistic regression was used to estimate differential uptake of the vests between trial arms. At baseline, 3.3% of individuals in both arms used a reflective vest. In three months of follow-up, 79 drivers in the intervention arm and 82 drivers in the control arm were able to be observed. In the intervention arm the average proportion of observations during which drivers were observed to be using a reflective vest was 9.5%, compared to 2.0% in the control arm (odds ratio: 5.5, 95% confidence interval: 1.1-26.9, p-value: 0.04). Distribution of free reflective vests did lead to an increase in vest usage, however, the increase was minimal. Removing economic barriers alone appears insufficient to adequately improve adherence to conspicuity measures.

Finite mixture distributions, sequential likelihood and the EM algorithm

A popular way to account for unobserved heterogeneity is to assume that the data are drawn from a finite mixture distribution. A barrier to using finite mixture models is that parameters that could previously be estimated in stages must now be estimated jointly: using mixture distributions destroys any additive separability of the log-likelihood function. We show, however, that an extension of the EM algorithm reintroduces additive separability, thus allowing one to estimate parameters sequentially during each maximization step. In establishing this result, we develop a broad class of estimators for mixture models. Returning to the likelihood problem, we show that, relative to full information maximum likelihood, our sequential estimator can generate large computational savings with little loss of efficiency.