DELTA
Coping Systems, focused on patient safety and patient-involved medical decision making, has created DELTA, the web-based Data Extraction and Longitudinal Trend Analysis system.
DELTA, designed for patient safety analysis, is able to use large, distributed, heterogeneous clinical data sets to inform patients, doctors, administrators, researchers, and officials - in real time - about higher than expected adverse events, particularly those with a low signal-to-noise ratio.
Figure 1. DELTA Outcome Trending graph showing alert status
In a medical setting it is extremely important to provide risk-adjusted models of the particular expected outcome relative to the patient population. DELTA uses numerous statistical methods and benchmarking approaches to assure its analyses properly reflect the expected outcomes of the population under study.
DELTA delves deeply into data to find issues related to devices, medications, patient history, providers, or procedural settings. Its initial utilization has been in post-market analysis (Phase IV) of recently released medications and cardiac devices, providing the data to warrant further investigation of certain products within specific patient populations.
DELTA continuously monitors patient clinical data, creating online graphs and reports along with sending real-time alerts to key individuals when thresholds have been surpassed. More than just providing observed rates of occurrences or scoring one hospital relative to others, or even providing observed to expected ratios, DELTA provides a subsequent level of statistical scoring (including Logistic Regression, Cohort & Propensity Analysis, SPRT, Bayesian, and others) so that using the results to effect policy change are justifiable by statistical and health policy norms.
DELTA can also accumulate data from more than one site. Not only can it handle heterogeneous data, it understands the confidentiality requirements of certain operational data points (not just patient privacy but hospital confidentiality as well). A DELTA analysis can be run at a collecting node with data from multiple sites, detecting alerts on data that is sharable among sites (such as adverse events for a device), while locally providing alerts to a site regarding data that is confidential to that site (such as adverse events per clinician).