Abstract The objective of our study is to derive reference intervals (RI) for total protein (TP), albumin and globulin that are appropriate for both plasma and serum samples by applying statistical analyses to retrospective patient data. Both plasma and serum are acceptable matrices for these tests at our institution, with a single RI that applies to all specimens. The widespread shortage of lithium heparin (LH) tubes that occurred as a consequence of the pandemic prompted our lab to switch the primary tube type for comprehensive metabolic panels (CMPs) from LH to serum separator tubes (SST) in October 2021. The primary tube type reverted to LH in October 2023, and no adjustments were made to RIs at either time. To determine how well our RIs align with the accepted matrices, we used R Statistical Programing version 4.3.1 to apply an indirect, nonparametric statistical approach to data extracted from the laboratory information system (LIS) between July 2018 and December 2023. All orders for TP, albumin and globulin were queried to include the following components in the analysis: TP, albumin, globulin, serum TP, serum albumin and serum globulin. The records were divided into 3 epochs based on the primary tube type for CMP on the order date, with epochs 1 and 3 reflecting LH and epoch 2 reflecting SST. We attempted to identify a subset of serum and plasma samples in healthy individuals by applying exclusion criteria based on inpatient status, age, and draw station at which the specimen was collected. For each component, a resampling bootstrap method was used to calculate the RIs and corresponding 95% confidence intervals by epoch. Our results show that the RI of TP, albumin and globulin changed significantly between the epochs, reflecting the difference in primary matrix. The components serum TP, serum albumin, and serum globulin did not see a significant change in values over the epochs. A verification study looking at plasma and serum samples in 20 healthy individuals support these results. Based on these findings, our reference intervals were updated from 6.6-8.7 g/dL to 5.9-8.3 g/dL for TP, from 3.6-4.9 g/dL to 3.6-5.1 g/dL for albumin and from 2.3-3.5 g/dL to 1.5-3.9 g/dL for globulin. We show that there are statistically and potentially clinically significant differences in the reference intervals of TP, albumin and globulin in serum compared to plasma. This study demonstrates a big data analytics approach to derivation of RIs and emphasizes the importance of RI assessment and adjustment for patient population and acceptable specimen type.