STUMP » Articles » Mortality with Meep: Student Deaths at a Large University » 13 December 2019, 17:42

I started this post back in November, when the following story came out: USC student found dead off campus is university’s 9th death this semester, report says

The University of Southern California is investigating the death of its ninth student in three months, a phenomenon that’s sending alarm throughout the campus community.

Three of the deaths have occurred since Friday.

A 28-year-old student was found dead in off-campus housing on Monday by USC public safety personnel, who were responding to a welfare check. The cause of death has not been released, but the LAPD says no foul play is suspected after the death was reportedly considered suspicious on Monday, according to CBS Los Angeles.

Just one day before, the university sent a letter to students and parents, acknowledging the then-eight deaths among the student body since late August. The university’s president, Carol Folt, said that many students had been assuming most of the deaths were the result of suicide, which was “not correct.”

See, I would assume most of the deaths were drug overdoses, but that’s me.

“These tragic losses have resulted from a number of different causes. In some cases the cause of death is still undetermined, and in others loved ones do not want details disclosed,” Folt said in the letter. The causes of death vary and some have not yet been determined. At least three were suicides.

In the first death of the semester, freshman Matthew Olson, 18, was hit and killed on the freeway while walking between lanes.

Or accidental death. That’s a top cause of death for people under age 35. (Especially males between ages 15 and 25, who do a lot of dumbass things.

But let’s get to the bottom line:

The university said it typically sees between four and 15 deaths per year in the community of about 47,000 undergraduate and graduate students, according to KTLA.

Okay. Let us see how suspicious this is. The raw death rate is about, let’s say, 10 per 47,000, or .0002128. That’s 21.28 per 100,000. And here I’m assuming they’re including the whole year, but if we assume they cover only 9 out of 12 months, or 75%, I will scale this up to 28.37 per 100,000.

Let’s check what death rates for people age 18-35 are in the U.S.

10-14 18.6 12.3
15-19 72.7 29.4
20-24 137.9 50.9
25-29 171.3 68.8
30-34 196.3 93.5

On the left are male mortality rates, per 100,000. On the right are female.

Okay, let’s try 9 people in 3 months. I will get to Poisson point processes in a moment, but let’s say this scales up to 36 people in a 12-month period. Out of 47,000 people. So that gives me 76.60 per 100,000 per year.

Now, that may be a little high for college students at Harvard, say, but this is not that far out in rate compared to the general population.

But I think this may be somewhat misleading. The problem is that when you’ve got a very low rate of something occurring, you often get “clumping” every so often.

POISSON POINT PROCESSES

Here’s a wiki article on Poisson processes, but hey, what you really want to read about is the Poisson distribution.

When there’s a very low rate of death, in general, it’s best to simulate using a Poisson process.

So let us go with an annual average of 12 deaths in a year, or 1 death per month, using what the college folks said above (4 to 15 deaths per year…but, we’ll come back to that).

If we use a Poisson model, then in 3 months we expect 3 deaths, and the probability of getting 9 deaths in 3 months equals:

Okay, 0.3% is pretty unlikely.

Well, let’s take a look at what if those expected 12 deaths are actually over 9 months, so you would expect 4 deaths in 3 months.

Here’s the probability with that new assumption:

That’s almost 5 times more likely.

TAKEAWAYS: IT’S NOT UNREASONABLE TO WORRY BUT MAYBE IT’S NOTHING

So while it could just be the roll of the dice that this is an anomalous year for USC itself, that they opened a new psychiatric clinic is not necessarily a bad response. Some of the students did die by suicide, and it could be a worsening situation.

For what I can tell, in my calculations above, this is not necessarily a high rate compared to some other universities.

In particular, I know MIT has a particularly bad rate. Cornell got a reputation, but the actual numbers are within normal ranges (yes, it would be best to reduce that rate. It’s just that there’s nothing special about Cornell.) Similarly, at NYU, there was an attractive nuisance, as the insurance folks say.

So, while this may be a bad year for mortality at USC, it’s not clear to me that this is (yet) signs of a deteriorating situation. Sometimes bad events do clump up just by coincidence and sometimes the rates themselves are not anomalous, but people simply paying attention to the story when in prior years they ignored it.

If I don’t hear any more about this situation, I would not be surprised. Like the Dominican Republic deaths, one can get a bunch of “suspicious” deaths in quick succession… and then you don’t.