Statistics in Nursing – Patient Falls and Multiple Regression

Statistics in Nursing – Patient Falls and Multiple Regression

Statistics in Nursing – Patient Falls and Multiple Regression

I am convinced that we can predict patient falls, who will fall and, and the extent of their injuries. This will require considerable testing and modeling with many variables and the use of multiple regression. Consider this scenario: We have been asked to form a national task force to eliminate patient falls in the inpatient setting. We will create a list of variables and evaluate these in an extensive research study. Our first task, in this forum, is to brainstorm and create a list of variables which may predict patient falls. Each of your are to provide a creative list of ten variables with a precise method of measuring each. There are no \”bad\” answers. I will provide an initial example list to help get us started and I will use a modified – Delphi technique to come to a consensus on a list of variables. I will upload info! Use US references only.

Multiple Regression

The Basics

1. Purpose

Multiple regression is used to analyze data when the research question asks if two or more variables (predictor or independent variables) (in any type of data) – predict one dependent variable (which is in scale data). If there

is only one independent variable we use Simple Linear Regression and if the dependent variable is not in scale

data we use Logistic regression. The outcome of any regression analysis is the regression equation, which can be

used to input values of the predictor (independent) variable(s) to predict the value of the dependent variable.

The process for all regressions, and the interpretation of the SPSS output is very similar. In this presentation we cover the basics of the multiple regression: purpose, process, Data Output, and reporting the results.

2. Process: Running a Multiple Regression in SPSS
3. From the top toolbar choose Analyze > Regression > Linear
4. Move the Independent (predictor) Variables of interest to the Independent Window (under “Block 1”)
5. Move the Dependent Variable (the variable being predicted) to the Dependent Window
6. From the tab below the Independent Variable list, labeled “Method”, select Stepwise in the dropdown box
   e. Select Statistics in the top right corner, and make sure these are all checked:
   Estimates, Model fit, Descriptives, and Part and Partial correlation; then select “Continue”
7. To keep this simple for this demonstration, then select OK.

 

3. The Data Output (Results) in APA…
4. The first two tables include descriptive statistics and correlations tables. You can review these….

b. Depending on the number of variables included you will get a table entitled “Variables Entered and Removed”, which is part of the stepwise modeling process to determine the best model, good information, but the information in these first three tables is not significant in reporting the results of the regression.

c. Next will be a “Model Summary” table (See Table 1.). This is important. The R is similar to a correlation, but it is not necessary to report this – the important number here is “R Square” (written also as R² This is the percentage of variability in the dependent variable that is explained in the regression equation. Use the  “Adjusted R Square” figure, as this takes sample size into account. Report the R² of the model having the highest value if there is more than one model, here we only have one model.

Statistics in Nursing – Patient Falls and Multiple Regression

Table 1.

*In this table the important information is the “Adjusted R-Square” value R²=.044. (This means that only 4.4% of the variance in the dependent variable is explained in the equation. That is not going to be significant…

1. Next is the “ANOVA” table. For this only use the data from the “best fit” model chosen in the prior table, the one which had the highest R² in step 3c above. What is important here is the regression and residual dfs, the F statistic, and the Sig. (See Table 2). (In the old days the Sig was determined by going to a table with the df and F statistic. We do not do this any more, but we do report these numbers…)

2

Table 2.

*In this table we want to note the regression and residual df, the F statistic “F”, and the Sig (the p-value).

1. Next is the “Coefficients” table. Again only use the results from the same Model selected in the prior tables.

Here there is only one Model, use this one. The important information is the Unstandardized Coefficient column

“B”. Note the numbers for each variable and the number for the constant. These values are used for the regression equation (See Table 3).

KINDLY ORDER NOW FOR A CUSTOM-WRITTEN, PLAGIARISM-FREE PAPER WITH COMPLETE ANSWERS

Table 3.  *The important information is what is in column B.

4. Presenting Results in APA

This is how these results would be written in an APA paper. Only include the regression equation if p <.05.

 

Phrasing Insignificant Results

A multiple linear regression equation was calculated to predict a patient’s LOS based on their Hematocrit, Estimated Blood Loss, POCobe (misspelled (cell saver)), Packed RBCs received, and body weight. The regression equation was not significant (F(5, 43) = 1.437, p =.230); Hematorcrit, estimated blood loss, POCobe, packed RBCs received, and body weight are not predictors of patient LOS.

 

Phrasing Significant Results

Assume that for this output we had the same results, except we had R² = .334 and p = .001. This would indicate significant results and so we would write this up more like this:
A significant regression equation was found (F(4, 43) = 1.437, p = .001), R² = .334. Participant’s predicted LOS is equal to: .558(H) – .002(EBL) + .029(POC) + .006(PRBCs) + .416(Wt) – 17.311 = LOS, where Hematocrit (H) is in percent; Estimated Blood Loss (EBL), POCobe (POC), and Packed Red Blood Cells (PRBCs) are all in ccs; Weight (Wt) is in pounds; and Length of Stay (LOS) is in days. LOS increases with increases in Hematorcrit, POCobe, PRBCs, and Weight, but decreases with increases in Estimated Blood Loss.
(The R² of .334 indicates that this equation accounts for 33.4% of the variance in the dependent variable!)