Identifying opportunities to reduce emergency service calls in hematology manufacturing using statistical methods

Abstract

The main goal of this project is to identify opportunities to improve the reliability of the DxHTM product line, an automated hematology instrument for analyzing patient blood samples. The product was developed by Beckman Coulter Diagnostics, a division of a Danaher operating company with principal manufacturing and support operations based near Miami, Florida. A critical business metric used to reflect reliability is the Emergency Service Call (ESC) rate. An ESC for an instrument is defined as the number of unscheduled, on-site technician visits during the one year warranty period. Though Beckman Coulter already deploys an extremely robust quality control system, ESCs can still occur for a wide variety of other reasons resulting in an impact to reliability. Any tools that support the reduction of ESCs may help generate positive perceptions among customers since their instruments will have greater up-time. This project entails an evaluation of a new initiative called "Reliability Statistical Process Control" (R-SPC). R-SPC is a form of manufacturing process control developed internally consisting of an electronic tool that collects raw instrument data during manufacturing. Unusual measurements are automatically sent to a cross functional team, which examines the potential trend in more detail. If an abnormal trend is identified, the examination could generate a lasting improvement in the manufacturing process. Currently, the success of R-SPC is measured by the extent to which it reduces ESCs. Because an unusual measurement engenders further actions to investigate an instrument, it is desirable to show with empirical evidence that the measurement is linked to reliability. To assess whether particular measurements were systematically related to the ESC rate, relevant data were analyzed via the Pearson Chi Squared statistical test. The tests revealed that some of the variables now monitored do not appear to affect the ESC rate for the range of values studied. In contrast, several proposed "derived" parameters may serve as better indicators of an instrument's ESC rate. Moreover, the Chi Squared methodology described can be used to investigate the relationships between other variables and the ESC rate. The thesis concludes by offering several specific recommendations to help refine the R-SPC initiative.