Regression Analysis and Predictions
Learning Goal: I’m working on a management question and need an explanation and answer to help me learn.
consider the example used in the instructional materials above. You are an administrator whose responsibilities include a cardiac unit. Because congestive heart failure (CHF) is a high volume condition for your unit, you want to have a better understanding of some of the drivers of length of stay (LOS) for these patients. After a brief review of the literature, you find that several of the key factors affecting patient LOS for CHF are: the number of medications the patient is on upon admission, the duration of intravenous diuretics, and the number of comorbid conditions. You are also interested in whether gender affects LOS for your patient population. You ask a colleague whom you know to be familiar with statistics to perform separate, simple regression analyses regarding LOS and each of the factors of interest. Using the framework LOS = intercept + slope*key factor, your colleague presents you with the following results:
LOS as a function of the number of medications upon admission:
- Intercept = 5.2
- Slope = 0.15
LOS as a function of the duration of intravenous diuretics:
- Intercept = 5.1
- Slope = 0.25
LOS as a function of the number of comorbid conditions:
- Intercept = 5.0
- Slope = 0.4
LOS as a function of gender (when the patient is a male):
- Intercept = 5.85
- Slope = 0.05
Prepare an informal cheat sheet for yourself indicating the predicted LOS for each of the following:
- What would LOS be for a patient admitted taking 0 prescription drugs? A patient taking 3 prescription drugs? A patient taking 6 prescription drugs?
- What would LOS be for a patient receiving intravenous diuretics for 0 days? A patient receiving intravenous diuretics for 2 days? A patient receiving intravenous diuretics for 4 days?
- What would LOS be for a patient who has 0 comorbid conditions? A patient who has 3 comorbid conditions? A patient who has 6 comorbid conditions?
- What would the LOS be for a patient who is male? A patient who is female? Use a “dummy variable” to isolate the impact of gender on LOS. To do this, assign males the value of 1 and females the value of 0 when conducting your analysis