Help with Regression and Graphing in SPSS. Cubic splines?

#1
Hi Guys,

I am looking at time to surgery and infection. Infection is dichotomous. I ran a graph looking at cumulative sum of infection and time to surgery and it looked somewhat like a logistic curve. Linear at about 4x in the first 6 hours to surgery and then tapers off to about y=x roughly. Now, this would suggest to me there's an association.
There's 4 infections an hour for the first 6 hours in 296 patients. That's like 1.3% an hour increase. However, in SPSS if I do a logistic regression, no effect. beta is below 1.
Then I restrict the data for time to surgery <6 only. beta is 0.86. Don't have a clue why this is happening. If I restrict the regression (logistic) for data with people to surgery >6 hours the beta goes to 1.05 and then 1.07 (both nonsig) after 12 hours.

Why is the cumulative sum going up rapidly if logistic can't find an association during that time period? Makes no sense to me.