Changing Clocks In Fall Will Stroke You Out, Fatty: Research

(My enemies changed the title. I have changed it back. They will pay for this insolence.)

Daylight savings time is something that angers many performatively, more than in truth. If you believe the chatter, wars don’t us hit as hard as falling back an hour once a year.

I grew up with changing the clocks, and now I even sort of look forward to it. Signals the change of seasons and all that. But I can’t say I devote much time to pondering the great question of Time Switching. Because I have never felt the change of only a hour. Except to notice it’s darker a bit faster in fall.

Frankly, I don’t care what rulers decide on this burning question. As long as they make their decision not based on “research”.

Here’s the breathless headline: “Ditching daylight saving time could prevent millions of obesity cases and thousands of strokes, study says“.

Millions. Millions. Millions. Which is to say, millions. The article clarifies that headline’s “thousands” to mean “hundreds of thousands”. Of strokes. Which is a lot of strokes, and seems greater than millions of new fat people. Seeing as we have so many already, a few more wouldn’t hurt.

The far-left political magazine Scientific American is on the story, because of course all things are political in an Our Democracy.

The only reason I’m pestering you with this, dear reader, is they have a quote from one of the study authors that raised my suspicions: “‘Our work reveals that there may be greater health benefits on a population level if we switch to a permanent standard time,’ says study co-author Lara Weed, a bioengineering Ph.D. candidate at Stanford University.”

Why that odd phrase “population level”? You can’t get benefits at a population level if you don’t get them in individuals. Are we looking at some kind of model here? And not observations? (Recalling all models only say what they are told to say.)

The peer-reviewed paper is “Circadian-informed modeling predicts regional variation in obesity and stroke outcomes under different permanent US time policies” by Lara Weed and Jamie M Zeitzer, both from Stanford, in PNAS.

After the Introduction promising this is a most-important problem, rarely researched, we come to the Materials and Methods. Which opens:

Light Simulations. Light exposure patterns (i.e., patterns of daily light exposure) under SDT, DST, and BAS were simulated over one year. We assumed an idealized light exposure pattern (Fig. 1A), with individuals having a regular sleep (10 pm to 7 am, daily, 0 lx) and work schedule (9 am to 5 pm, Monday–Friday,well-lit office of 500 lx), and sunlight exposure in the mornings, afterwork, and on the weekends. [Much of the same and an] evening indoor light brightness of 120 lx was selected to reflect typical values for indoor light at night, such as in a family room or bedroom.

In other words, everything was a simulation. Including how light affects people:

Circadian Models. Light exposure patterns for each county and time policy were used as inputs into mathematical models of the human circadian rhythm (Fig. 2), specifically the St. Hilaire (28) adaptation of the Jewett-Forger-Kronauer model…

The good old St. Hilaire adaptation of the Jewett-Forger-Kronauer model. Well, every field has its jargon. And models.

They did grab some county-level prevalence data of various maladies, including smoking status which ought not to be in the mix because smoking isn’t an outcome. To further muddy the bad health outcome they added another model on top of the implied model (smoothing) of the county-level data:

To prepare the health determinant data for use in statistical modeling, and reduce the number and complexity of overlapping variables, we calculated principal component analysis (PCA) and corresponding scores for all counties. K-nearest neighbor imputation was used to fill missing data before computing z-scores for input to PCA. The first 12 principal components were included in statistical modeling of prevalence…

So not individual health outcomes, but linear combinations of smoothed county-level health outcomes. What is that in terms of real health of real people? Who knows. The thing is some kind of model is all we need know.

Next comes linking this weird (let us call it) Health Mixture model with the simulations of light model and the good old St. Hilaire adaptation of the Jewett-Forger-Kronauer model. And how did they do this linking? With a—stop me if you’ve heard this one before—with a model.

Estimation of the Health Impact of Additional Circadian Phase Shifting Under Each Time Policy. We used Gaussian kernel local polynomial regression to estimate the partial derivative of health outcome prevalence with respect to yearly circadian shifting time. We then integrated the partial derivative to estimate change in health outcome prevalence between time policies for all counties.

Results?

Significant correlations (P < 0.01) between shifting time and east/west location were found for all chronotype and time policy combinations…As compared to the current time policy, the prevalence of obesity and stroke are predicted to decrease under both permanent SDT and DST, while changes in prevalence of arthritis, cancer, CHD, COPD, depression, diabetes are small and not significantly different from current time policy. Obesity is predicted to have the greatest decrease under permanent SDT…

P-values? I weep.

Does changing times twice yearly produce more or fewer bad changes in health? I have no idea, and neither do they. All they have done is string a bunch of models together, which demonstrates great facility with the toys scientists have built for themselves, but not much else. These models build in the conclusions the eventual model of models reaches.

What about the known seasonality—a huge signal—of maladies increasing winter and peaking every mid January, and descending and bottoming out around July? Could that once-known-to-all effect account for why smoothed county level grouped and mixed health outcomes are worse when it’s slightly darker earlier in the fall? At just that time, that is, that maladies are on their way up, Up, UP!

There was no reason to do this research. It adds nothing. It only subtracts from knowledge. It does show the authors know how to work their computers, which is all we are asking of our well remunerated scientists these days. You will thus be happy to know you paid for this work: “Research reported in this publication was supported by the National Heart, Lung, And Blood Institute of the NIH under Award No. F31HL170715.”

When I said government has to get out of the science grant business, I meant it.

Subscribe or donate to support this site and its wholly independent host using credit card click here. Or use PayPal. Or use the paid subscription at Substack. Cash App: $WilliamMBriggs. For Zelle, use my email: matt@wmbriggs.com, and please include yours so I know who to thank. BUY ME A COFFEE.