Quality Control – Sampling and Data Recording
Often public health research includes the investigation of the presentation, geographic distribution, management, behavior, opinions and feelings about a particular disease, illness or malady using a survey approach. If your research is disease or illness related, make sure that you have an agreed upon and precise case definition before you begin data collection and that this definition is written down and reviewed by your supervisor. All field assistants should be trained on the proper case definition and closely supervised at the beginning of data collection and in the first few days of data collection random re-inspections of data collected by field assistants is warranted.
If you are not accompanying field assistants to every interview, focus group discussion or surveyed household, then you must design a plan to sample the activities performed in the field to ensure that they are performed according to sample selection, schedule and at the appropriate quality. To do this establish a visitation schedule that is random, meaning don’t always visit the same field assistant on the same day each week. During these visits, observe the delivery of the data collection instrument and speak to the interviewed participant afterwards to verify questions and answers. Discuss any feedback with field assistants as necessary.
If your thesis research involves laboratory tests, design a method to ensure the accuracy of the lab readings. Throughout the course of the project, randomly select 15% – 20% of all samples collected for re-examination. You should also visit the lab often and observe laboratory practices and hold feedback sessions with lab assistants as necessary.
Plans for ensuring the quality of collected and recorded data and laboratory data should be written before data collection commences and shared with your supervisor.
Quality Control – Recorded Data
When collecting data using printed forms human error is a potential risk. Therefore, it is necessary to design a quality control process prior to beginning data collection. In general, your quality control process should review person-recorded data for three things:
- Readability – Ensure the handwriting is legible for all data collected.
- Odd Indicators – Focus on issues like birth dates, gender, and age as these are variables that often get recorded incorrectly.
- Data Reliability – Review the answers to survey questions to make sure that they make logical sense in light of the gender, age, ethnicity and location of the respondent. For example, if the age of the respondent is 18 and the form indicates that they have a 20 year old child, this would be illogical. A quality review of recorded data through the lens of reasonability will often highlight questionable data. When in doubt about the validity of a response, return to the surveyed participant for clarification or remove the data entirely.
Design a strict procedure for the random quality control of non-lab data recorded forms before commencing your filed work and review this procedure with your supervisor. You can begin by randomly reviewing 25% of the forms. If you encounter a high number of errors, then you should review all the forms for reliability.
Quality Assurance – Data Transfer
It is also possible to encounter errors when uploading survey data to a database. Data may not load correctly, so it is necessary to check that all data has been correctly transferred. For short term projects, it is recommended that you check 100% of the uploaded data by comparing entries in the database (PC) with the data recorded on the form to ensure they match. You may also automate some checks, including:
1. Duplicate records
- Duplicate household IDs (if not permitted)
- Duplicate participant IDs
2. Errors in free text data entry
- Check for errors in Community Codes
- Check for errors in interview date format
- Check for errors in field worker codes
- Check for errors in household ID
- There should be no IDs that are not on the list of randomly selected participants
- Check that all fields in which “Other” has been chosen also has a specification in the text box
- Check for the consistent spelling of words
3. Errors Uploading/Transferring Data
- Check each question for blank fields
- Check that the variable type is correct (numeric vs. alpha numeric)
- Check that single answer questions only have one answer
The exact nature of your quality assurance plan will vary depending on the study type and the data collection protocols, but be sure to define your plan before commencing field work and review the plan with your supervisor.
Learn about the project execution phase in the next lesson.