Medical data can be mined for patterns, which may be used to predict candidate diagnoses according to symptoms and other parameters of care. Our hypothesis is that the admission (initial) patient assessment, when combined with seasonal information can provide more accurate insights for the patient diagnosis. For instance, when cough is the symptom, the probability for flu could be higher during the winter (flu season). We hereby present a method to estimate the temporal variation of the probability for a diagnosis, when the initial patient assessment is known. In order to develop the model, we utilized a large synthetic medical claims dataset from the Centers for Medicare and Medicaid Services. We used the Apriori algorithm to calculate the support and confidence for each 'admission_diagnosis → final_diagnosis' itemset. For each itemset, 52 rules were generated, one for each week of a calendar year. The Apriori output was filtered so that only itemsets with the 'admission diagnosis' on the Left Hand Side(LHS) are extracted. We furthermore smoothened, using the Exponentially Weighted Moving Average (EWMA) algorithm, and then visualized the week-by-week variability of confidence, for any 'admission_diagnosis → final_diagnosis' pair of interest. With our approach, researchers can observe seasonal variations of the diagnosis element, and further study these variations for causal knowledge discovery.