Kelly Reeve

Biostats PhD student, recovering runner, plant and Labrador mother

[PIPRA @PipraAG]

📢 We are excited to share groundbreaking news from our important publication:
Our latest paper, published today in @Age_and_Ageing, the leading journal in geriatrics, showcases the transformative impact of our clinical decision support tool.
While we’ve already demonstrated our ability to identify patients at risk of delirium, this publication goes a step further. It highlights the significant effect our tool has on clinical outcomes, including a 29% reduction in delirium cases.
Additionally, our solution has led to a reduction in both length of stay by 1.36 days and nursing time by 19.3 hours per medium risk patient following prevention recommendations.

We are incredibly proud of this achievement and its potential to enhance patient care. https://t.co/QqrHcWIiKH

Dodsworth et al. Benefits of an automated postoperative delirium risk prediction tool combined with non-pharmacological delirium prevention on delirium incidence and length of stay: A before-after analysis based on a quality improvement project
@KelDoesBiostats

[Kelly Reeve @KelDoesBiostats]

New work on clinical prediction models: https://t.co/gYY5vJVPLP
Some take aways: (1) Models are not ready for clinical use, (2) Most do not appear to be meant for use, (3) There is a disconnect between methods research and clinical research @ulrike_held @HeidiBaya

[Frank Harrell @f2harrell]

Improving R Workflow e-book to take advantage of cool new code annotation feature in Quarto #rstats @vandy_biostat @VUDataScience #Statistics #stattwitter https://t.co/kj5Isp4lTh

[Kelly Reeve @KelDoesBiostats]

@Kook_Lucas @marmolejo_ramos Oh man, we are all done :)

[Kelly Reeve @KelDoesBiostats]

I wish people would annotate their calibration plots with "under-" and "over-prediction" so I don't have to just remember.

I wish people would provide calibration plots at all.
#statstwitter #epitwitter

[Kelly Reeve @KelDoesBiostats]

@numbersman77 Thanks!

[Kelly Reeve @KelDoesBiostats]

@numbersman77 I see the focus is on data structure and analysis reporting. What about analysis software itself? Does pharma need to 'prove' to the FDA in some way that analysis libraries (like survival, lme4, or metafor) work as intended?

[Kelly Reeve @KelDoesBiostats]

@MaartenvSmeden The new regulations have expanded quite a bit. Any good tips on summaries of the parts relevant to statisticians/developers?

[Kelly Reeve @KelDoesBiostats]

Well aren't those the nicest things anyone said to me all week:
"The applicants describe the planned analysis for the primary outcome measure and model adjustment clearly and this model is appropriate."
and
"I am able to replicate the sample size calculation"
#statstwitter

[Kelly Reeve @KelDoesBiostats]

@JakobDambon Ich hab diesen Brief auch bekommen! :)

[Kelly Reeve @KelDoesBiostats]

Adventures in prognostic model reviews:

In response to "Please provide the coefficients and their measures of uncertainty for your model"

"Please note that sklearn Logistic Regression does not provide measures of uncertainty by default." Coefficients part or request ignored..

[Kelly Reeve @KelDoesBiostats]

Adventures in prognostic model reviews:

"[The AUC is the most important validation measure because it quantifies the ability of the model to estimate well calibrated probabilities]"

[Kelly Reeve @KelDoesBiostats]

@BenVanCalster @Richard_D_Riley @TheStatsGeek Also, maybe due to the heavy focus on using regression models to produce odds/hazards and thinking of the intercept/baseline hazard as nuisances

[Kelly Reeve @KelDoesBiostats]

Pet peeve of the day: journals that put the entire methods section at the end of the paper, as if it's a side note.

What if I only want to read the results if the analysis makes sense?

[Kelly Reeve @KelDoesBiostats]

@laure_wynants Might take time to get used to. A friend tried them for a few minutes and didn't like the feel. And I'm not a sound expert, so I would check internet reviews if that is super important for you.

[Kelly Reeve @KelDoesBiostats]

@laure_wynants I usually only connect to 1 device (so no experience switching between devices). May be something to pay attention to if you are looking to connect wirelessly to several devices.

[Kelly Reeve @KelDoesBiostats]

@laure_wynants I have the Jabra Sport Elite 65t. They are small ear buds. I haven't had a problem with size (they come with 3 different sized rubber pieces) and they stay in well (I wore them for slow long runs without losing them). I've used them for Zoom calls without a problem.

[Kelly Reeve @KelDoesBiostats]

@MaartenvSmeden Out of curiosity, who did they attribute the first remark too (reference 1 and 6)?

[Kelly Reeve @KelDoesBiostats]

Never will I ever write a for loop without a progress indicator again.

[Kelly Reeve @KelDoesBiostats]

@lucasmorin_eolc The bulldozer would have to be an ant