0
$\begingroup$

I am a novice in statistics so pardon my rather simple question.

I have a retrospective case-control study, I would like to show the rate of admissions (ie number of admissions per year) of my cases and controls. I have individual level data on a = total number of admissions of a given patient during follow up period y = total number of years of follow up of a given patient

and I have n1 = number of cases and n2 = number of controls.

If I would like to show the mean number of admissions per year of the cases, and the mean number of admissions of controls per year (with SDs), should I be doing

mean= sum(a/y)/n for both groups, and standard deviation? in R I can easily do this by calculating a new column r (admission rate of an individual) = a/y, then do mean(r) and sd(r)

or should it be

mean= sum(a)/sum(y) for both groups (sum of n1 admissions, divided by sum of y1 years of follow up for cases; and sum of n2 admissions, divided by sum of n2 years of follow up for controls) - however, how do I derive standard deviation from this?

$\endgroup$

1 Answer 1

0
$\begingroup$

Edited:

total cases: n1
total control: n2

each case/control has $a$ admission numbers in $y$ period of time.

If $a$ and $y$ are varying/different for each patient $i$

Mean number of admissions per year of the cases:

The mean, $\mu$ is $\frac{\sum_{i}^{n_1}(a_i)}{\sum_{i}^{n_1}(y_i)} = \frac{a_1 + a_2 + ... + a_{n1}}{y_1 + y_2 + ... + y_{n1}}$

basically, add all the individual patient admissions together and divide by adding all the individual patient's periods.

special case: If $a$ and $y$ is the same for all patients then mean number of admissions per year of the cases:= is total number of cases $(n1 \cdot a)$ divided by total time $(n1 \cdot y)$

mean, $\mu$ = $\frac{(n_1 \cdot a)}{(n_1 \cdot y)}$

replace n1 by n2 above and you can get the mean number of admissions of controls per year.

for standard deviation you just need to apply the formula.

$\frac{1}{\sum_{i}^{n_1}(y_i)} \cdot \sum_{i}^{n_1}(a_i -\mu)^2 = \frac{(a_1 - \mu)^2 + (a_2 - \mu)^2 + .... + (a_{n_1} - \mu)^2} {y_1 + y_2 + ... + y_{n1}}$

change n1 to n2 for standard deviation in controls.

$\endgroup$
2
  • $\begingroup$ no, a = total number of admissions per patient for the entire follow up period (y) of this patient. And n is number of patients, therefore there are n1 cases and n2 controls. I will edit my question to be more clear. $\endgroup$
    – Hong
    Jul 27, 2023 at 10:45
  • $\begingroup$ Please check now. $\endgroup$ Jul 28, 2023 at 7:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.