THE OPTIMAL TIME TO PREPARE A FISTULA FOR HEMODIALYSIS PATIENTS

Purpose:  

   The gold standard for hemodialysis delivery is via an arteriovenous (AV) fistula.  Two types of uncertainty make it difficult to know the optimal time to create a fistula: 1) when a CKD patient may need to start dialysis, and 2) when a fistula will mature.  We developed a decision model to evaluate two key timing decisions: 1) when to request laboratory tests to estimate glomerular filtration rate (eGFR), and 2) when to start fistula preparation based on the results.

Methods:  

   Based on observed data, we assume eGFR declines linearly, and that observed values fluctuate around this line according to y(t) = 45  -.375t + ε,  where ε ~ N(0,4²) and t is in months.  We assume that patients start dialysis when their eGFR = 15 and that there is a Uniform[3,9] month distribution of time from when fistula preparation commences until it is ready for use.  We assume the clinician obtains noisy eGFR readings upon each test, fits an updated linear regression, estimates when the line crosses 15, and then subtracts a “backtrack time” to factor in the various uncertainties (random observations and time for fistula to mature).  If the resulting time has already passed or is imminent, fistula preparation commences; otherwise one waits until the next scheduled lab test.

   We simulate the observation process and fistula maturation time via Monte Carlo simulation, and use it to evaluate expected costs associated with various lab testing and fistula start time policies.  We consider three types of cost parameters: 1) c_L—cost per day fistula is ready later than ideal dialysis start time, 2)  c_E—cost per day fistula is ready early, and 3) c_T—cost for each lab test patient undergoes.

Results:

    Among fixed-spacing testing policies, simulation experiments suggest an optimal (inter-test time, backtrack time) of (3 months, 13 months).  In sensitivity analyses, we changed the slope parameter to -.25 (“slower progressor”) and -.5 (“faster progressor”), and obtained optimal solutions of (4 ,14 ) and (2 , 11), respectively.  Results were not very sensitive to the distribution of fistula maturation time.

Conclusions:  

   Monte Carlo simulation is a useful tool for evaluating the interaction between testing frequencies and fistula initiation policies when establishing guidelines for dialysis preparation.