Introduction:

Data quality makes it easier to ensure that observational studies are reliable.

Objective:

To describe assurance and quality control to maintain data reliability and validity in a cohort study.

Methodology:

We present the data management strategies implemented in a study that followed patients of chronic kidney disease who were in a renal protection program and compared them with those undergoing conventional treatment to observe its association with clinical outcomes. We assessed the changes in error frequency after implementing the plan along with the reproducibility of the strategies for entering records into the databases.

Results:

We documented a progressive decrease of data collection errors. The Kappa values among data collectors for the most important variables were: 0.960 for creatinine clearance <60 ml/min; 0.942 for renal ultrasound alteration; 0.871 for proteinuria >150 mg/dl; 0.730 for urinary sediment alteration and 0.956 for stage allocation upon admission. The intraclass correlation coefficient for the identification of systolic blood pressure was 0.996; for diastolic blood pressure, the coefficient was 0.993 and for serum creatinine levels at diagnosis, the value was 0.995.

Discussion:

Data quality begins with the recognition of the challenges and difficulties involved in responsible data collection, hence the contribution of standardized processes and personnel to carry them out in a suitable manner. Studies show that many improvement processes arise in the development of research without pre-established protocols.

Conclusion:

The reduction in error ratio and type during the data collection process are the result of the early identification of erroneously entered or missing data, the correction of the guidelines for completing forms as well as of the instruments for detecting errors and continuous training of the staff. The analysis showed good inter-rater reliability.