EMR data quality – Data quality actions

Updated as of January 22, 2016

Estimate impact of data quality

The data are almost certainly not a definitive estimate of your team’s actual performance. However, they might be “good enough” to help you decide if your team needs to improve or not. To determine if the data are “good enough” for that, estimate how likely it is that one or more of the issues outlined in the interpretive notes are a problem with your team. Then, run the “imperfect data impact calculator” to see if the issue(s) could lead to a different decision related to the need for improvement. To do this, work with your clinical leaders and staff to establish an approximate impact of data quality – i.e. is the data quality issue causing your performance to look like TWICE or HALF or 10% (or other number) less or more than it actually is? Plug that number into the “imperfect data impact calculator”. It will show you whether the data quality issue(s) you think you have would change your initial decision regarding the need to improve. Click here to access the Imperfect Data Impact Calculator. You may find it hard to generate consensus about the impact of data quality issues on the level of performance shown in the D2D 3.0 report. In that case, consider the following options:

• Track the next 10 (or 20 or other small number) encounters to get a better estimate of the extent of the data quality issue. Perhaps the rate among these patients will shift your team’s overall rate to be TWICE or HALF or 10% (or some other number) of the rate in the report. Plug that number into the “imperfect data impact calculator” and proceed accordingly.
• Estimate how many of your patients eligible for cervical or colorectal cancer screening have tests or labs recorded properly in the EMR. Perhaps the rate among these patients will shift your team’s overall rate to be TWICE or HALF or 10% (or some other number) of the rate in the report. Plug that number into the “imperfect data impact calculator” and proceed accordingly.
• Estimate how many of your patients have a smoking status coded in a consistent manner in your EMR. Perhaps the rate among these patients will shift your team’s overall rate to be TWICE or HALF or 10% (or some other number) of the rate in the report. Plug that number into the “imperfect data impact calculator” and proceed accordingly.
• If none of the above is helpful, consider instead experimenting with possible “error” rates to see how much error (i.e. TWICE or HALF or 10% of some other number) would be needed to change the decision made on the basis of the performance of the indicator in D2D 3.0. If, in the opinion of the team, such an amount of error is reasonable, then it may be worth considering efforts to improve data quality. Alternatively, if that amount of error is considered to be unlikely, then the data are likely good enough to support the initial decision regarding the need to improve, based on the performance shown in D2D 3.0.

If the “imperfect data impact calculator” points to the same decision (e.g. a need to improve or NOT) even after data quality issues are considered, the data are likely “good enough” to base your decision on regarding the need to improve. The next step is to consider strategies to improve, assuming the area of care measured by the indicator is a priority for your team. If your data are not “good enough”, you may then consider taking action to improve your data quality, before or at the same time as you try to improve processes of care.

Increase quality of the data

…if the “imperfect data impact calculator” shows that the issues in your data may point you to a different decision than suggested in D2D 3.0.

• The goal of this indicator is to inform and motivate action to improve data quality in the EMR and also serve as a measure to monitor progress for actions to improve data quality such as those described for all the other D2D indicators.
• The actions to improve data quality do not need to be limited to cancer screening data or smoking status data, although these might be of immediate interest. Accurately identifying deceased patients could be a focus, or accurately recording lab results in appropriate fields and in appropriate language, or coding diagnoses consistently are other areas
• Patients served data: confirm which patients are alive and active (by whatever definition you use in your EMR/team) as rates of indicators based on incorrect denominators (i.e. all patients ever seen vs just the patients who are alive and who are active) will be incorrect.
• Provide feedback to clinicians:
  • Create lists of patients with particular characteristics identified via standardized EMR queries for their review/confirmation. These have been developed by QIDS Specialists and the EMR CoPs (e.g. childhood immunization, COPD, diabetes – CHF and depression are coming soon)
  • See how six teams optimized EMR queries to gather precise, meaningful data
  • Participate with external partners like CPCSSN, EMRALD, and eCHN to use your EMR data in more meaningful ways
• Consider hiring a student to help you clean up your data (see suggestions in this handbook for cleaning up your roster and smoking/alcohol status)
• Participate in the EMR communities of practice and join your peers in developing new tools and processes for standardizing access to EMR data. Contact improve@afhto.ca to get connected.
• Tap into external resources to support clinical process changes using PDSAs from HQO or others (also check with your QIDSS).