DNP Project Statistical Methods Algorithm Use this algorithm to

DNP Project Statistical Methods Algorithm

Use this algorithm to design the approach for data collection and analysis for your DNP project.

Evaluate which option(s) most align with the desired project outcomes. Utilize no more than

two and no less than one statistical analysis tool(s). Use the table below to map out the

statistical test(s) you will use to analyze collected data. *

Project Objectives:

Planned data collection

approach to achieve

objective (if applicable):

Data Analysis (if applicable):

choose option A, B, or C

Objective 1:

Objective 2:

Objective 3:

*Some objectives don’t require data collection to determine if the objective was met. For

example, educating staff would not require data collection. Determining compliance or rates of

infection would require data collection. If no data collection is required, write “n/a” in the

column for data collection/analysis.

Option

Outcome

Considerations

Option A: Paired

samples t-test OR

Wilcoxon signed

rank test

Type of data: Numerical scores or

rates that can be assigned to

participant or provider location

and measured before and after

the intervention. Looking at

paired data.

Examples:

1. Pre/post observations on

the same person for each

data set (examples:

cultural competence

survey, intent to stay

survey, etc.)

2. Compliance rates per

provider, reviewing the

same number of charts

before and after per

provider

Evaluate the assumptions of each

test to determine which test to use.

• Paired samples t-test: use to

compare means, assumption

of approximately normal

distribution for the

differences.

• Wilcoxon signed rank test:

use when assumption of

normality is not met.

Compare ordering of the

data.

Option B: Chi-

Square Test OR

Fisher’s exact test

Type of data:

Counts of participants, providers,

or charts that can be assigned to

Evaluate the assumptions of each

test to determine which test to use.

a particular category, such as

compliant/not compliant, or

injured/not injured.

Examining the difference

between expected and observed

outcomes.

Examples:

1. Compare provider

compliance rates on a

protocol before and after

training (provider not

identified individually,

and compliance is yes/no)

2. Compare rates of hospital

acquired infections after

intervention if the

intervention was not

present before. Compare

rates pre-intervention to

rates post intervention.

• Chi-Square Test: use to

compare observed vs

expected values. At least 5

expected values for each

combination of the two

variables should be present. If

this is not met then use

Fisher’s exact test.

Option C:

Descriptive

statistics with

Confidence interval

(CI)

Descriptive statistics used to

describe phenomenon from a

sample of a population.

Confidence intervals reflect

uncertainty about how the

estimate applies to the

population as a whole.

Examples:

1. Out of 30 telemedicine

patients who screen

positive for depression,

how many of them

received appropriate

referral for psychiatry?

x/30= xx%

Descriptive statistics example: Often

displayed in table and graph to show

rates.

• Mean and standard

deviation/95% CI for each

group (approximately normal

continuous (ordinal or ratio)

data)

• Percent with 95% confidence

interval, making sure to

include the sample size.

• Frequency table, preferably

including counts and

percentages in some format.

Option A

Paired samples t-test: Utilized to compare means between two similar samples. Generally, you

pair the same group of people’s test results before and after an intervention such as pre-

posttest. Example: Pre-test and post-test

Assumptions of Paired samples t-test:

1. Independence: two separate observations are being compared.

2. Normality: Normal distribution between pairs

3. No extreme outliers

If any assumption is violated, then this would be an invalid test.

If a different group of people is examined before and after, see option B.

Wilcoxon signed rank test: Utilized to compare two separate observations between two similar

samples when the assumption of normality is not present. Wilcoxon signed rank test should be

used over a t-test if there will be outliers in the data. Where a t-test examines the means

between two data sets, the Wilcoxon signed rank test examines the ordering of the data

instead of the means of the data. An example where this may be more helpful is if there will be

various disciplines of medicine with widely varied educational background taking the same

survey. Example: Likert scale

Assumptions of Wilcoxon signed rank test:

1. The dependent variable is ordinal and continuous.

2. Independent variable being compared is matched or related, or the same subjects are

examined before and after.

3. Distribution of differences is symmetrical between groups.

Option B

Chi-square test: Examines observed vs expected values.

Example: Implement a new protocol and examine an outcome, such as protocol compliance; **

a specific categorical outcome or nominal outcome that is expected after implementation.

Example: BMI screening, asthma action plan

Assumptions of Chi-square Test:

1. Data should be randomly sampled from the population of interest.

2. Comparing two categorical or nominal variables.

3. At least 5 expected values for each combination of the two variables. (If fewer than

5, consider Fisher’s exact test)

**When evaluating compliance with multiple variables, be sure to define level of compliance

when using YES/NO for consistency. For example, if a protocol requires 5 steps and you

determine the provider is compliant if they achieve 4/5 steps, please define this ahead of time

to collect consistent results from the chart audit.

Fisher’s exact test: Examines observed vs expected values. Use as an alternative to the Chi-

square test if a combination of two variables has less than 5 in expected value.

1. Data should be randomly sampled from the population of interest.

2. Comparing two categorical or nominal variables.

Example of what data might look like:

Compliant

Not Compliant

Before

After

Option C

Descriptive statistics with Confidence Interval testing: Descriptive statistics summarize data

collected in a sample population. Common approaches to descriptive statistics include mean

and standard deviation.

Confidence interval testing is used to determine if a true population mean has been assessed.

Example: descriptive statistics (% compliant before vs after, with 95% confidence interval,

estimable here: http://vassarstats.net/prop1.html)

Assumptions:

1. Random sampling

2. Normal distribution of sample.