Introduction
“Cash can’t purchase happiness.” “You possibly can’t decide a ebook by its cowl.” “An apple a day retains the physician away.”
You’ve in all probability heard these sayings a number of occasions, however do they really maintain up after we have a look at the information? On this article sequence, I need to take widespread myths/sayings and put them to the check utilizing real-world knowledge.
We would affirm some surprising truths, or debunk some widespread beliefs. Hopefully, in both case we’ll acquire new insights into the world round us.
The speculation
“An apple a day retains the physician away”: is there any actual proof to help this?
If the parable is true, we should always anticipate a detrimental correlation between apple consumption per capita and physician visits per capita . So, the extra apples a rustic consumes, the less physician visits folks ought to want.
Let’s look into the information and see what the numbers actually say.
Testing the connection between apple consumption and physician visits
Let’s begin with a easy correlation verify between apple consumption per capita and physician visits per capita.
Information sources
The info comes from:
Since knowledge availability varies by yr, 2017 was chosen because it offered essentially the most full when it comes to variety of international locations. Nonetheless, the outcomes are constant throughout different years.
Visualizing the connection
To visualise whether or not larger apple consumption is related to fewer physician visits, we begin by taking a look at a scatter plot with a regression line.
The regression plot exhibits a very slim detrimental correlation, that means that in international locations the place folks eat extra apples, there’s a barely noticeable tendency to have decrease physician visits.
Sadly, the development is so weak that it can’t be thought-about significant.
OLS regression
To check this relationship statistically, we run a linear regression (OLS), the place physician visits per capita is the dependent variable and apple consumption per capita is the unbiased variable.
The outcomes affirm what the scatterplot steered:
- The coefficient for apple consumption is -0.0107, that means that even when there may be an impact, it’s very small.
- The p-value is 0.860 (86%), way over the usual significance threshold of 5%.
- The R² worth is sort of zero, that means apple consumption explains just about none of the variation in physician visits.
This doesn’t strictly imply that there isn’t a relationship, however moderately that we can’t show one with the accessible knowledge. It’s doable that any actual impact is just too small to detect, that different elements we didn’t embody play a bigger function, or that the information merely doesn’t replicate the connection effectively.
Controlling for confounders
Are we accomplished? Not fairly. Up to now, we’ve solely checked for a direct relationship between apple consumption and physician visits.
As already talked about, many different elements could possibly be influencing each variables, doubtlessly hiding a real relationship or creating a man-made one.
If we take into account this causal graph:
We’re assuming that apple consumption straight impacts physician visits. Nonetheless, different hidden elements may be at play. If we don’t account for them, we threat failing to detect an actual relationship if one exists.
A well known instance the place confounder variables are on show comes from a research by Messerli (2012), which discovered an fascinating correlation between chocolate consumption per capita and the variety of Nobel laureates.
So, would beginning to eat numerous chocolate assist us win a Nobel Prize? In all probability not. The possible rationalization was that GDP per capita was a confounder. That implies that richer international locations are likely to have each larger chocolate consumption and extra Nobel Prize winners. The noticed relationship wasn’t causal however moderately as a consequence of a hidden (confounding) issue.
The identical factor could possibly be occurring in our case. There may be confounding variables that affect each apple consumption and physician visits, making it troublesome to see an actual relationship if one exists.
Two key confounders to contemplate are GDP per capita and median age. Wealthier international locations have higher healthcare programs and totally different dietary patterns, and older populations have a tendency to go to docs extra typically and will have totally different consuming habits.
To regulate for this, we modify our mannequin by introducing these confounders:
Information sources
The info comes from:
OLS regression (with confounders)
After controlling for GDP per capita and median age, we run a a number of regression to check whether or not apple consumption has any significant impact on physician visits.
The outcomes affirm what we noticed earlier:
- The coefficient for apple consumption stays very small(-0.0100), that means any potential impact is negligible.
- The p-value (85.5%) continues to be extraordinarily excessive, removed from statistical significance.
- We nonetheless can’t reject the null speculation, that means now we have no sturdy proof to help the concept that consuming extra apples results in fewer physician visits.
Similar as earlier than, this doesn’t essentially imply that no relationship exists, however moderately that we can’t show one utilizing the accessible knowledge. It may nonetheless be doable that the true impact is just too small to detect or that there are but different elements we didn’t embody.
One fascinating statement, nonetheless, is that GDP per capita additionally exhibits no important relationship with physician visits, as its p-value is 0.668 (66.8%), indicating that we couldn’t discover within the knowledge that wealth explains variations in healthcare utilization.
Then again, median age seems to be strongly related to physician visits, with a p-value of 0.001 (0.1%) and a constructive coefficient (0.4952). This implies that older populations have a tendency to go to docs extra regularly, which is definitely probably not shocking if we give it some thought!
So whereas we discover no help for the apple fable, the information does reveal an fascinating relationship between getting older and healthcare utilization.
Median age → Physician visits
The outcomes from the OLS regression confirmed a robust relationship between median age and physician visits, and the visualization beneath confirms this development.
There’s a clear upward development, indicating that international locations with older populations are likely to have extra physician visits per capita.
Since we’re solely taking a look at median age and physician visits right here, one may argue that GDP per capita may be a confounder, influencing each. Nonetheless, the earlier OLS regression demonstrated that even when GDP was included within the mannequin, this relationship remained sturdy and statistically important.
This implies that median age is a key consider explaining variations in physician visits throughout international locations, unbiased of GDP.
GDP → Apple consumption
Whereas indirectly associated to physician visits, an fascinating secondary discovering emerges when trying on the relationship between GDP per capita and apple consumption.
One doable rationalization is that wealthier international locations have higher entry to recent merchandise. One other chance is that local weather and geography play a job, so it could possibly be that many high-GDP international locations are situated in areas with sturdy apple manufacturing, making apples extra accessible and inexpensive.
After all, different elements could possibly be influencing this relationship, however we gained’t dig deeper right here.
The scatterplot exhibits a constructive correlation: as GDP per capita will increase, apple consumption additionally tends to rise. Nonetheless, in comparison with median age and physician visits, this development is weaker, with extra variation within the knowledge.
The OLS confirms the connection: with a 0.2257 coefficient for GDP per capita, we are able to estimate a rise of round 0.23 kg in apple consumption per capita for every improve of $1,000 in GDP per capita.
The three.8% p-value permits us to reject the null speculation. So the connection is statistically important. Nonetheless, the R² worth (0.145) is comparatively low, so whereas GDP explains some variation in apple consumption, many different elements possible contribute.
Conclusion
The saying goes:
“An apple a day retains the physician away,”
However after placing this fable to the check with real-world knowledge, the outcomes appear not according to this saying. Throughout a number of years, the outcomes had been constant: no significant relationship between apple consumption and physician visits emerged, even after controlling for confounders. Plainly apples alone aren’t sufficient to maintain the physician away.
Nonetheless, this doesn’t fully disprove the concept that consuming extra apples may cut back physician visits. Observational knowledge, irrespective of how effectively we management for confounders, can by no means absolutely show or disprove causality.
To get a extra statistically correct reply, and to rule out all doable confounders at a degree of granularity that could possibly be actionable for a person, we would want to conduct an A/B check.
In such an experiment, individuals could be randomly assigned to 2 teams, for instance one consuming a hard and fast quantity of apples day by day and the opposite avoiding apples. By evaluating physician visits over time amongst these two teams, we may decide if any distinction between them come up, offering stronger proof of a causal impact.
For apparent causes, I selected to not go that route. Hiring a bunch of individuals could be costly, and ethically forcing folks to keep away from apples for science is certainly questionable.
Nonetheless, we did discover some fascinating patterns. The strongest predictor of physician visits wasn’t apple consumption, however median age: the older a rustic’s inhabitants, the extra typically folks see a health care provider.
In the meantime, GDP confirmed a light connection to apple consumption, probably as a result of wealthier international locations have higher entry to recent produce, or as a result of apple-growing areas are usually extra developed.
So, whereas we are able to’t affirm the unique fable, we are able to provide a much less poetic, however data-backed model:
“A younger age retains the physician away.”
For those who loved this evaluation and need to join, you could find me on LinkedIn.
The total evaluation is on the market on this pocket book on GitHub.
Information Sources
Fruit Consumption: Meals and Agriculture Group of the United Nations (2023) — with main processing by Our World in Information. “Per capita consumption of apples — FAO” [dataset]. Meals and Agriculture Group of the United Nations, “Meals Balances: Meals Balances (-2013, previous methodology and inhabitants)”; Meals and Agriculture Group of the United Nations, “Meals Balances: Meals Balances (2010-)” [original data]. Licensed underneath CC BY 4.0.
Physician Visits: OECD (2024), Consultations, URL (accessed on January 22, 2025). Licensed underneath CC BY 4.0.
GDP per Capita: World Financial institution (2025) — with minor processing by Our World in Information. “GDP per capita — World Financial institution — In fixed 2021 worldwide $” [dataset]. World Financial institution, “World Financial institution World Growth Indicators” [original data]. Retrieved January 31, 2025 from https://ourworldindata.org/grapher/gdp-per-capita-worldbank. Licensed underneath CC BY 4.0.
Median Age: UN, World Inhabitants Prospects (2024) — processed by Our World in Information. “Median age, medium projection — UN WPP” [dataset]. United Nations, “World Inhabitants Prospects” [original data]. Licensed underneath CC BY 4.0.
All photographs, until in any other case famous, are by the creator.
