Physics-style Repeated Measures of multiple predictor variables in survival analysis?

I have inherited a longitudinal dataset where 5 different continuous independent variables were measured multiple times daily for 28 days in individual subjects. Measurement intervals were irregular, both within and between subjects. These independent variables have a relatively high variance, and produce quite chaotic plots over time, although trends are visible over days. These data are needed as predictor variables for a (continuous or discrete) time to event outcome. There are also binomial confounding variables.

The approach i used in my (veterinary-related) masters dissertation was calculate the within-subject rolling means of four consecutive measurements and standardised. For each day of the study, I then used the first of these standardised rolling means for each IV in a multivariate Cox Proportional Hazards model. Unfortunately, this meant 28 models - one for each day of the study period. Not ideal, but it was good enough for the dissertation.

Now I'm back to clinical work, I really want to do justice to this dataset. I would be very grateful if anybody could suggest a technique which I could use to analyse this type of data, preferably retaining the (continuous or discrete) survival outcome, but managing the repeated measures of the independent variables in a more elegant way.

Not enough is known about the way these IVs change with time, so I am trying to retain as much of the temporal resolution of the IV as possible.

Many thanks,