Outlier Tracking

Real-Time Signal Detection

There are a multitude of ways your clinical study can get derailed with bad data - but what about problems with good data? How can good data be problematic? When data falls within the bounds of the edit checks, but in aggregate signals that something isn't quite right. How can you detect that signal? Two choices - conduct a lot of high burden manual analyses on an ongoing basis, or use MedNet's fully automated Outlier Tracking Solution.

Outlier Tracking is built specifically to highlight data that requires data management review. Here's how it works: fields in the ENLIGHTEN^TM eClinical Solution have two types of built-in logic based edit checks to ensure clean data is collected. A "hard" edit check is one where the value is not physiologically possible - for example, a Systolic Blood Pressure of >300 mmHg. If a value of more than 300 is entered, the system will not allow that value to be saved. A "soft" edit check, by contrast, is one which is possible, but unusual. In this example, perhaps a Systolic BP of 200-300. In this case, the system warns the user with a message like "this value seems high - are you sure?" If the user indicates the value is correct, the system will accept it.

While an individual instance of a systolic BP of 220 might be acceptable, what if all the patients at that investigative site had BP's in the "are you sure" range? How would you detect this? In the ordinary course of events, you may only find out when the co-morbidities of hypertension start affecting your study outcomes and AE's - and by then your whole dataset might be at risk.

Outlier Tracking automatically captures all data that fall within the "soft" edit check ranges, allowing data mangers to quickly review, on demand with real-time data, trends that may be emerging within the study, and take corrective action. Actions might include retraining on a particular question if values are consistently dubious, examination of edit check parameters for appropriateness, or review of the population assumptions in the Analysis Plan. In the example discussed, drawn from real life, it emerged that a chronic hypertension clinic had inadvertently been allowed into the study - and Outlier Tracking facilitated early corrective steps that stopped this high-enrolling site from skewing the dataset.

Key Features

Full filtering:
- Site, patient, form, visit date range
Sortable details, including:
- Site, patient, form, visit date
- Field name and value entered
- Edit check rule triggered
- Person who entered the data
  - With direct email query link
- Drill down links for in-depth form/patient review

Key Benefits

Delivers real-time, fully automated signal detection
Enables granular analyses without third party tools or programming
Increases the probability of a successful study