Missing data and Mixed effects model analysis

[Son Tran]

Hi,
I had an issue with my project. I would like to compare the effect on heart rate and blood pressure between treatments during surgery. These vital signs had been collected each 5 minutes. However, it depended on the operative time of each case. The shortest time was recorded at 20 minutes and the longest time was at 100 minutes. Therefore, I have a lot of missing data of these variables between two timepoints. The below table shows the percentage of missing data at different timepoints.

Timepoint	Missing	Total	Percentage
20	4	82	4.88
25	9	82	10.98
30	14	82	17.07
35	24	82	29.27
40	30	82	36.59
45	36	82	43.9
50	38	82	46.34
55	42	82	51.22
60	46	82	56.11
65	55	82	67.07
70	58	82	70.73
75	61	82	74.39
80	67	82	81.71
85	72	82	87.80
90	74	82	90.24
95	77	82	93.90
100	77	82	93.90

I think these are missing at random because they depend on the complexity of lesion operated and the experience of surgeon.
I'm wavering between multiple imputation and removal of data at timepoints with high percentages of missing. Which one will be suitable for this issue?
I found that the linear mixed effects model can handle with missing at random because it's based on Maximum likelihood Estimation.

I look forward to your worthy recommendations to solve my obstacle.
Best regards

[Carlo Lazzaro]

Son:
welcome to this forum.
It's recommended to deal with missing data (diagnosis of the underlying missingness mechanism and related therapy, say multiple imputation if the missingness mechanism is MAR) instead of getting rid of them.
As far as your second statement is concerned, Stata omits observations with missing values in any variables.

[Son Tran]

Thank Mr Carlo for your recommendations.

Missing data in my study may be MAR. Therefore, I will implement multiple imputations. How many imputations should be conducted? Some authors recommend 5 imputations, while stata's guidelines are 20.

[Carlo Lazzaro]

Son:
given the great support that Stata can give you with -mi- I would try 20.
As an aside, Carlo is enough ! Thank you.

[Son Tran]

Carlo:
I'm gratitude you recommend for my issue.
I will try with 20 imputations.