What is Survival Analysis? Kaplan-Meier and Cox Regression

Power analysis is a method used to determine the probability of a study identifying a statistically significant result. It is used to calculate the necessary sample size to determine a specific effect size within a defined confidence level and margin of error in a planned study. This analysis allows for the prediction of how powerful the research is and prevents working with an insufficient or excessive sample size.

Research Question:

Does providing home care training according to guidelines after patient discharge reduce the rate of readmission within the first 30 days compared to standard training?

Group 1: We have a group that received home care training according to guidelines.

Group 2: We have a group that received home care training according to standard training.

How many people should be included in this study?

We can find this answer using G*Power analysis.

The characteristic of a good sample is, firstly, that it accurately represents the population from which it is drawn, and secondly, that the sample size is sufficient.

If there are not enough people in the study, even if our research has a statistically significant difference, its power to prove this will be insufficient.

In summary; Power analysis, performed at the beginning of a study to find statistically significant differences, helps us determine the minimum number of participants required for the applications.

Power analysis can be performed at the beginning or end of the study. This type of power analysis is called Post-Hoc or Posterior Power Analysis.

Post-Hoc power analysis is performed after the study is completed to determine the power of any statistically significant difference found. Or, if no difference is found, it investigates whether the sample size was sufficient to detect that difference.

What determines sample size?

-It can vary depending on the dependent variable.

-It can be a ratio. For example, readmission rate, treatment success rate, side effect rate, relapse rate, etc…

-It can be an average. The average of blood values, the average of any scale, etc…

-The outcome variable (dependent variable) should be determined at the beginning of the research.

A study may have more than one outcome variable. A separate power analysis is performed for each outcome variable.

STEP 1: DETERMINING THE OUTCOME VARIABLE AND FORM OF STATEMENT

Outcome variable: Re-hospitalization within the first 30 days due to home care training;

Form of statement: Rate

STEP 2: ASSESSING THE SITUATION BY REVIEWING THE LITERATURE

We need to identify at least one of the groups by reviewing the literature. This is an important step in the power analysis of the study.

According to research results found in the literature, the re-hospitalization rate in the last 30 days was found to be 15%. The study sample could be from abroad, but if it is from Turkey, it would be better because it would be more realistic and have a higher similarity rate.

This shows where we are currently.

STEP 3: CALCULATING THE EFFECT SIZE

Effect size is the expected difference between two groups. It will show where we are going from where we are. Effect Size: Found to be 15% for the standard treatment group.

How much will my application reduce this rate?

5%?

10%?

This can be found from pilot studies or literature based on researchers' experiences.

STEP 4: DETERMINING THE LEVEL OF ERRORS THAT CAN BE MADE DURING HYPOTHESIS TESTING

Type I Error (alpha):

This is the probability of detecting a difference that does not exist in the hypothesis test result. Like telling a man he is pregnant. The internationally accepted limit is 5%. The p-value we obtain from the statistical significance result indicates the level of Type I error for this comparison. If we base our predictions on a 5% Type I error, we can say that we made our predictions with 95% confidence.

Type II Error (beta):

This is the probability of failing to detect a difference that exists. Like telling a pregnant woman she is not pregnant. The internationally accepted limit is 20%. The power of our test is considered to be 80%.