ESSENCE Data Quality Filters: Percent Values for Discharge Diagnosis Informative

NSSP has been consistently working with state partners to improve completeness of data elements received. As a result, Discharge Diagnosis (DD)—a key syndromic element—has steadily improved in completeness over the years. However, this raises concerns about the impact of changing completeness on the observed syndromic trends. For example, if a facility has NOT been sending DD code values (or has sent values with low completeness) and that facility suddenly starts sending these values during a study period, the trends can be falsely influenced.

This led the NSSP team to create various ESSENCE data quality (DQ) filters to reduce the impact of changes in data quality across syndromic trends. Discharge Diagnosis Informative (DDI) is one such filter. By selecting DDI = Yes, the user’s choice of subsets will only include data with complete and informative (or, valid) diagnosis codes. To identify which values were invalid, the NSSP team sought feedback from the user community. They defined invalid codes as those where, although the data element is populated, the value received is not meaningful to syndromic surveillance.*

Using the DDI filters improves upon the existing DQ filters. The DDI values are percentage cutoffs for including facilities in the trends observed. Variations of the DDI filter are available based on the time period used to create the filter; the time period chosen for the filter should be based on the period of study.

For example, let’s say you want to look at DDI greater than 50% during the past year for Influenza-like Illness (ILI) visits. By selecting this filter, you would create a trend of ILI visits limited to facilities that sent DDI values greater than 50% of their total visits for ILI visits in the past year.

There is no hard rule by which to choose a cutoff for DDI percent value. Ideally, you want to graph multiple values and let the data speak for itself. If varying DDI percent values are overlaid perfectly, or are close to each other, you are exercising the least control (smaller percent value) to include more facilities (and therefore more visits). However, if you observe obvious differences in trends, you may want to select a percent value that is more restrictive. Any DDI percent value can always be paired with limiting the standard deviation to less than 20%, given the standard deviation cutoff does not vary.

You can find the DQ filters in the Query Wizard (shown below). Several different time period cutoffs are provided (6 months, 1 year, 2 years, all time, etc.):

Hopefully, this explanation will help you use this additional ESSENCE DQ filter. As usual, please send your questions to nssp@cdc.gov.

* Non-Informative Discharge Diagnosis values include these terms: ‘””’ , ‘null’ , ‘unknown’ , ‘unk’ , ‘n/a’ , ‘na’ , ‘Chief complaint not present’ , ‘ed visit’ , ‘ed’ , ‘er’ , ‘Advice only’ , ‘Other’ , ‘xxx’ , ‘Evaluation’ , ‘Follow up’ , ‘medical’ , ‘illness’ , ‘General’ , ‘General symptom’ , ‘EMS’ , ‘AMR’ , ‘Medic’ , ‘Ambulance’ , ‘EMS/Arrived by’ , ‘EMS/Ambulance’ , ‘;’ , ‘;;’ , ‘Triage’ , ‘Triage peds’ , ‘Triage-’ , ‘Triage peds-’ , ‘See chief complaint quote’ , ‘See CC quote’ , ‘Sick’ , ‘Injury’ , ‘ill’ , ‘Eval’ , ‘squad’ , ‘Referral’ , ‘Code 1’ , ‘Code 3’