Hi everyone,
I am conducting a study for association of cold season temperature with hospital admissions. I want that for a decreasing temperature and hospital admission if
RR > 1 it means less risk
RR < 1 means more risk...
Does RR works in reverse for decreasing temperature.
Let' say we have RR 0.87 for a GLM model of cold season temperature with total hospital admission.
Can we say that " with each 1 degree celsius decrease in temperature there is an increasing risk of 0.87 times? How does it workds? Can someone please explain about it more?
Dear Hasan Bin Sohail , I would interprete it this way; the risk of hospital admission is (1-0.87)*100= 13% among those exposed to cold temperature compared to those not exposed to cold temperature. Meaning, individuals who are exposed to cold temperature are less likely to be admitted in hospital compared to those who are not exposed to cold temperature. There could be factors confounding the result so it is good to take into consideration other factors for hospital admission apart from cold temperature through multivariate analysis.
When RR= 1 risk in exposed group is same as risk in non-exposed group,
When RR>1 risk in exposed group is greater than risk in non-exposed group,
When RR
Dear @brenda ayugi
Thank you for the detailed answer. However, I assume that the RR works in a different way for decreasing temperature. In the regression model, we say that a one-unit increase in the predictor variable is equal .... times outcome risk.
In my case i.e. cold temperature, I have to explain as 1 unit decrease in temperature (as it's a cold-weather) (not 1 unit increase). So, keeping that in mind, I assume the RR would also be interpreted in a reverse way. For example, if we get an RR of 0.87. We can say that a 1-degree decrease in cold temperature would increase the risk by 0.87 times?
I am just assuming so. Please correct me or help me to explain how would RR be explained for a 1-degree decrease in temperature?
It doesn't matter whether you have a cold/hot temperature as the independent variable. Say, your independent variable is x. The interpretation is for each unit increase in the x, there is (0.87-1)*100 = -13 = 13% less risk of hospital admission on average.
Say, the binomial model is log(p) = b0 + b1x = -2.5-0.14x. Then the RR for one unit increase in x is exp(-2.5-0.14*1) / exp(-2.5-0.14*0) = 0.87. You can check by putting other values as well. On the other hand, you can say, for one unit decrease in x results in 1/0.87 = 1.15 times (or 15% higher) risk of hospital admission on average. Usually, the interpretation in terms of 'an increase in x' is used in the literature.
Thank you, Md. Belal Hossain . Very interesting explanation.
You said " You can check by putting other values as well. On the other hand, you can say, for one unit decrease in x results in 1/0.87 = 1.15 times (or 15% higher) risk of hospital admission on average ".
Why you divided 1by 0.87? Why we can't just say that "1-degree decrease in x results in 13% increased risk for hospital admissions"?
Dear Hasan Bin Sohail , 1/0.87 was done to give the risk of hospital admission for those not exposed to cold.
Their risk is higher (1.15 times) than the risk of those exposed to cold(0.87) for every unit increase in temperature.
This is an alternative calculation. You can put the value and check. For example, for one unit decrease in x is associated with exp(-2.5-0.14*1) / exp(-2.5-0.14*2) = 1.15 times risk of hospital admission on average.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis