By: Jim Hoover, Principal Investigator
Accurate healthcare data is vital—from tracking disease trends to prescribing the right treatments. Even minor data errors can ripple into major patient safety issues. Avant Health Sciences recently conducted an error detection project to uncover wrong patient associations using a statistical technique called delta analysis, focusing initially on height and weight measurements. The study revealed both insightful and concerning findings.
In 2024, we used a delta analysis approach to identify discrepancies in patient height and weight measurements for a country health department with a population of just under half a million residents. Our analysis revealed a significant 3.04% error rate in height and weight measurements that fell outside a 1.5 standard deviation threshold above or below the mean value for each measurement.
A Concerning 3%
While 3% may sound minor, in a dataset of this scale, it represents thousands of potential errors. These discrepancies can distort BMI calculations, lead to incorrect medication dosages, and compromise public health analytics.
What is Delta Analysis?
Delta analysis involves comparing individual data points against a statistical benchmark. In our case, we examined height and weight measurements, flagging any values that deviated by more than 1.5 standard deviations from the mean. This approach allows us to identify outliers that might indicate data entry mistakes, faulty readings, or even that the data belongs to another patient. For example, a patient’s height is recorded as 5 feet 3 inches during an encounter, which would trigger a flag when the patient’s height is typically captured as 6 feet 2 inches.
Implications Beyond Height and Weight
The discovery of a 3% error rate in height and weight measurements raises serious concerns about the accuracy of other vital signs collected. If a measurable error rate exists in these relatively straightforward measurements, we must consider the potential for similar or even higher error rates in more complex data points, such as blood pressure, temperature, pulse oximeter readings, or extensive lab work measuring cholesterol levels, glucose readings, and other biomarkers.
Implications on Patient Identification
If a patient’s recorded weight significantly differs from prior measurements, it might indicate the data belongs to another patient. The more vital sign measurements outside of an error tolerance band, the more likely clinical data has been associated with the wrong patient. Overlay data, or wrong patient errors, occur frequently, and we must do all we can to identify and correct these errors as soon as they occur.
Strengthening Clinical Data Accuracy
Empowering Patients
Patient engagement is an extremely valuable tool for improving data quality. Patients will notice when height and weight measurements are incorrect and should be encouraged to report data discrepancies. Astute patients may also notice when glucose, cholesterol, or other biomarkers are outside the normal range.
Automating Checks and Thresholds
More deterministically, clinical measurement data should be monitored and automatically validated when input into a patient encounter record. Specific thresholds for each validated vital sign or blood panel value can be set. Investigations and corrections must occur when values fall outside the expected range. Care must be taken to avoid false alarms, such as falsely reporting data errors for growing children, blood panel values expected to change due to medication addressing certain illnesses, etc.
Auditing Regularly
The height and weight error rates found in this effort indicate further studies are needed to fully scope the problem of incorrect clinical data in patient records. Measurement errors, transcription problems, or wrong patient errors are all problematic. A delta analysis approach can validate that data is within each patient’s expected range of reasonableness. When the values are outside of the expected range, these should be flagged as potential medical data errors and reviewed. Periodic and comprehensive data audits should be considered to identify and rectify historical patient record errors.
Ensuring accurate clinical data is not optional; safe, effective care is essential. As delta analysis shows, even “routine” measurements can reveal hidden risks. Now is the time to prioritize data validation and regular audits across all healthcare systems.