PROGRESS Partnership’s recent PROGNOSIS RESEARCH: CONCEPTS, METHODS AND CLINICAL APPLICATION, International Summer School, June 30th – July 2nd 2014 at Keele University, UK.
I undertook this course as my PhD centres around developing the basis for a future clinical decision aid for use with patients with musculoskeletal shoulder disorders. In 2013, I noted with great interest that the PROGnosis RESearch Strategy (PROGRESS) Partnership (a UK Medical Research Council (MRC) funded, international, interdisciplinary collaboration developing understanding in research into quality of care outcomes, prognostic factors, risk prediction models, and predictors of differential treatment response) had published four key publications in the BMJ and PLOS Medicine on: (i) A Framework for Researching Clinical Outcome, (ii) Prognostic Factor Research, (iii) Prognostic Model Research and (iv) Stratified Medicine Research. Having really enjoyed these four publications, I wondered what other pearls of wisdom were held by this group of notable academics and researchers and I therefore enrolled on this Summer School.
The Summer School consisted of a series of lectures with linked interactive practical sessions. Course attendees were most certainly a keen bunch with questions, queries and debates coming thick and fast after each lecture. Given the highly varied mix and range of researchers, clinicians, statisticians, epidemiologists, public health, actuaries and PhD students present, this course constituted a collective journey guided by the brilliant PROGRESS partnership team but steered by keen individuals with varying interests in the field of prognosis research. The final practical session involved a number of small groups generating and presenting a bid for a prognosis study to a fictional funding body where the panel consisted entirely of PROGRESS partnership members – oh my, a tough crowd! The bids, ideas and rebuttals were innovative and entertaining in equal measure, a fitting close to a highly interactive and ‘progress’-ive course.
6 Top Tips on Prognosis Research from the Summer School
- Avoid using cut-off points in prognostic models (Sun et al., 1996). Too many examples of data fishing until a significant cut-off is found. Also important to consider statistical versus clinical significance. Better to keep variables continuous (or use more than 4 categories if you must)
- Know the pros & cons of prospective and retrospective cohort studies. Both have distinct advantages but are often inappropriately employed.
- Model building – avoid use of univariable analysis to select variables for inclusion in the multivariable model (Sun et al., 1996). This increases risk of excluding variables that are really important when in the presence of other variables.
- Model validation – Pre-plan how you will do this when planning your study. Consider the setting, country etc., of the validation cohort. Avoid splitting the original sample into model generation and model validation samples (Steyerberg et al., 2001). Identify or design and collect separate samples and consider bootstrapping if sample size is an issue.
- So your model tells a tall tale, but does it improve patient outcome or save any money? Plan health economic evaluation of models.
- If you are lucky enough that your model turns out to be wonderful, expect the long road to clinical implementation.
In spite of being a ‘Keele-ite’ or student at Keele University, I thoroughly enjoyed becoming resident on campus for 3 days whilst the ‘familiar became unfamiliar’. Course conversations flowed from breakfast until sunset as the international cohort enjoyed impromptu explorations of the Keele Arboretum, dining in Keele Hall, and sampling the selection of wines, ciders & real ales on offer in the Keele Post-graduate Association (KPA) pub.
If prognosis research and/or stratified care are of interest to you, I recommend this Summer School. Information on future summer schools may be found here.
Cliona McRobert is a SPCR PhD Student at Keele University