OBJECTIVES: A vital component of real-world research (RWR) is the accuracy of data capture as determined through source data verification. Accuracy rates are calculated to assess quality, and associated thresholds can direct timely corrective actions. We implemented a pilot chart abstraction process with a goal of addressing errors early in the chart abstraction process and improving accuracy.

METHODS: The pilot consisted of a 60-minute educational session on clinical and definitional aspects of assigned variables. During the 2-week pilot chart abstraction period, each abstractor was assigned 2 charts. Upon completion of the pilot period, the overall accuracy rate – the number of correctly abstracted variables divided by the total number of variables observed during chart abstraction data validation – was calculated. Reeducation was provided to abstractors that had individual variable accuracy rates below 80% and/or an overall accuracy rate below 90%. Once chart abstraction was complete, the remaining charts (considered full study charts) were reviewed, and the overall accuracy rate calculated. The Wilcoxon rank-sum test was used to compare average accuracy rates across the pilot and full study periods.

RESULTS: Fifteen retrospective observational chart abstraction studies within The US Oncology Network were included in the pilot chart abstraction exercise. The mean pilot accuracy rate for these fifteen studies was 94.5%. The mean full study accuracy rate was 97.1% and was a statistically significant increase over the mean pilot accuracy rate (p=0.046). Commonly seen variables that required reeducation included clinical characteristics and treatment history.

CONCLUSIONS: Implementing a pilot chart abstraction period and addressing potential errors increased the overall accuracy rate, ensuring higher quality data. Any opportunity to improve quality and the overall accuracy rate is beneficial to RWR. Enhancing training for data abstractors is a vital function to generate fit-for-purpose research data.