Abbott Laboratories - Global Services Transfusion Abbott Park, Illinois, United States
Background/Case Studies: Laboratory staffing shortages1,2 highlight the need for service operational efficiency and reduced labor cost overages. Our transfusion market instruments transmit >2TB of data daily from 800 hardware components, >3600 failure modes. Detailed troubleshooting resources also increase searching challenges. We hypothesize that this data can be used to develop advanced information technology applications (AITA) that improve field service engineer (FSE) efficiency. Here we present quantifiable real world and modeled improvement in key performance indicator (KPI) metrics with the use of Prognostic Health Notifications (PHN – predictive alerts of needed service) and AI-powered Diagnostic Tool (AIDT – FSE bidirectional troubleshooting guide).
Study
Design/Methods: Mean Time To Repair (MTTR) for service tickets (ST) calculated using total ST hrs over total # ST for periods Jan21-Dec21 & Jan22-Apr22. PHN MTTR was normalized over non-PHN MTTR. MTTR delta and MTTR % difference calculated. Current PHN annualized projected instrument downtime (ID) savings modeled by calculating ST/instrument during Jan22-Apr22 and multiplying by MTTR against non-PHN annualized ID. Predictive warning time from Q1‘22–Q4‘23 was evaluated by Machine Learning (ML)-model to identify earliest notification of known events. AIDT assessed by total submissions of diagnostic queries and search per user. This was correlated against MTTR, Same Day Fix Rate (SDFR), Part Usage per instrument from Q4’23-Q1'24. Users were categorized and assessed by various criteria: time since instrument trained, AI tool use frequency, AI tool live training. High frequency (HF) AIDT users (90th%ile) and newly trained FSE are compared to other users.
Results/Findings: Summary in Table 1. 19354 total scenarios (14197 Non-PHN, 5167 PHN) & 1780 instruments analyzed. Normalized PHN labor hours were 0.84 and 0.77 with instrument downtime reduced 41% and 36% analyzed periods chronologically. Each period saw an improvement with PHN supported events of 0.52hr(31min)-17.8% reduction and 0.78hr(47min)-26.2% reduction. The ML-model identified a PHN alert generation Max of 2 months prior to actual instrument downtime. 3,515 queries from 170 AIDT users showed HF users improve rates 4.3% and decreased time to repair of 36 min over others. Part usage also decreased 15% for newly trained HF users over others. Conclusions: Data supports integration value of AITA into service programs and suggests a direct positive impact to laboratory instrumentation service management. PHN predicts events to avoid workflow interruptions while reducing servicing time. AIDT demonstrates significant FSE KPI enhancements. This trend towards improved KPI metrics was maintained through eh studied period. Ultimately, this suggests improved instrument and customer support particularly laboratory staff and instrument up-time thus greater availability of safe blood and plasma. Additional studies required.
