How to Calculate Mean Follow Up Time in Stata

Mean follow-up time is a key metric in longitudinal studies and clinical research. It measures the average duration between the initial event and subsequent follow-up assessments. Calculating this in Stata requires understanding the data structure and applying appropriate statistical commands.

What is Mean Follow Up Time?

Mean follow-up time represents the average duration between the baseline event and subsequent follow-up measurements. In research studies, this metric helps assess the completeness of data collection and the efficiency of the study design.

Key characteristics of follow-up time:

Measured in days, weeks, or months depending on study design
Calculated for each subject from baseline to last follow-up
Useful for assessing study completeness and data quality

How to Calculate Mean Follow Up Time

The basic formula for mean follow-up time is:

Mean Follow-Up Time = Σ (Follow-Up Time for each subject) / Number of subjects

Steps to Calculate

Identify the baseline date for each subject
Identify the last follow-up date for each subject
Calculate the time difference between these dates
Sum all individual follow-up times
Divide by the total number of subjects

Example Calculation

For a study with 5 subjects:

Subject	Baseline Date	Last Follow-Up Date	Follow-Up Time (days)
1	Jan 1, 2023	Mar 15, 2023	64
2	Jan 10, 2023	Apr 1, 2023	83
3	Jan 5, 2023	Mar 20, 2023	67
4	Jan 1, 2023	Feb 28, 2023	57
5	Jan 15, 2023	Mar 30, 2023	76

Mean follow-up time = (64 + 83 + 67 + 57 + 76) / 5 = 341 / 5 = 68.2 days

Stata Implementation

To calculate mean follow-up time in Stata, you'll need to:

gen followup_time = (last_followup_date - baseline_date)

summarize followup_time, meanonly

Step-by-Step Guide

Load your dataset with subject identifiers, baseline dates, and follow-up dates
Create a new variable for follow-up time using the date difference function
Use the summarize command to calculate the mean
Consider handling missing data appropriately

Example Stata Code

// Assuming your dataset has variables: id, baseline_date, last_followup_date
gen followup_days = (last_followup_date - baseline_date)
summarize followup_days, meanonly
tabstat followup_days, stats(mean sd min max)

Note: Stata automatically handles date calculations when using date variables. Ensure your date variables are properly formatted as dates in Stata.

Interpretation and Practical Use

The mean follow-up time provides several important insights:

Assesses the completeness of your data collection
Helps evaluate the efficiency of your study design
Can identify potential biases in follow-up patterns
Useful for comparing different study arms or groups

When to Use This Metric

Mean follow-up time is particularly valuable in:

Longitudinal studies
Clinical trials
Observational studies
Any research requiring time-to-event analysis

Limitations to Consider

While useful, mean follow-up time has some limitations:

Does not account for irregular follow-up patterns
May be skewed by subjects with very long or short follow-up
Does not provide information about follow-up completeness

FAQ

What is the difference between mean and median follow-up time?: The mean is affected by extreme values, while the median represents the middle value. Median is often more robust when there are outliers in follow-up times.
How do I handle missing follow-up data in Stata?: You can use the egen command with the rowmax() function to find the last available follow-up date for each subject, then calculate follow-up time based on that.
Can I calculate follow-up time in weeks or months instead of days?: Yes, you can convert the date difference to weeks or months by dividing by 7 or 30.4 (average days in a month), respectively.
What if some subjects never had a follow-up?: You should either exclude those subjects from the calculation or treat their follow-up time as zero, depending on your research question.
How does follow-up time relate to study power?: Longer follow-up times generally increase study power by providing more data points, but this depends on the specific research question and event rates.