Ask CHPSO: Null Fields – Data Quality
In our previous newsletter, we discussed CHPSO’s new “No Nulls” initiative to improve data quality and how member organizations can help in the effort. This article will review the data pipeline involved in the process, touch on some of the challenges and successes we’ve observed in the effort so far, and suggest ways we can work together in a two-pronged approach to improve end-to-end data quality. The local event reporting system, which is a large component of the Patient Safety Evaluation System (PSES), is often considered to be the starting point of the patient safety data pipeline. A PSES is used to collect, analyze, and manage data for reporting to a Patient Safety Organization (PSO). However, this confidential Patient Safety Work Product (PSWP) data maintained in the PSES comes from several sources. Some of the data are sourced from a hospital’s Electronic Health Record (EHR) system, and some from patient safety risk analyses and event audits/reviews. Physically, the resulting data bundle usually ends up in an automated event recording system from a variety of vendor supported systems. Each member’s set up is unique and understanding the components and data flow housed within your specific hospital or health care system is critical in understanding where data might slip through the cracks and end up Null.
Automated event recording systems are built with an ability to export data into a spreadsheet. It is this spreadsheet data that facilities submit to our data intermediary (NextPlane). Based on an initial meeting with NextPlane and one or more members of the CHPSO team, this submitted data undergoes a transformative mapping from the format produced by the hospital’s automated event recording system into the standard Agency for HealthCare Research and Quality (AHRQ) format used by PSOs. The resulting output data is sent to CHPSO for ingestion into our database. Data that are not mapped present another opportunity for Nulls to creep into the picture.
Our work this past quarter has identified several interesting scenarios. In some circumstances, it was learned that key fields that specifically apply in incidents — reports where an event reached the patient — were never reaching the event recording system. In one situation the member event recording system did not have the fields to receive the data while in another the fields were there but the EHR-to-PSES programming needed to copy the data was not in place. Incident-related fields (age, date of birth, gender, harm, race, ethnicity) are of particular interest to CHPSO. In the case of race and ethnicity data – data were observed to be Null for every CHPSO member in incident reports over the past four years – our data intermediary (NextPlane) indicated that this data was generally not being seen in the spreadsheets produced by the member event recording systems. This suggests that some systemic challenges exist in the event recording systems or how they are being used.
Ensuring the data are mapped successfully is another ongoing challenge. In some situations, data were never mapped. This is understandable, as some of the mappings can be a bit tricky, such as mapping a discrete age field into the AHRQ age range categories. As the initial effort can sometimes be oriented toward “getting things going” instead of “getting everything captured”, some mappings may get skipped initially and not be revisited. Keeping things mapped properly was another common struggle. Something as simple as a spreadsheet column header change instituted by the hospital’s event recording system can cause the mapping to fail and cause more Nulls in data sent on to CHPSO.
Some of the examples above led to quick fixes. As a result, we are already seeing fewer Nulls in the CHPSO database. This translates into greater analytical power from the CHPSO data aggregate, and a corresponding increase in quality in our patient safety research and findings. Other scenarios will require more work. In both cases, our continuing thanks go to all the member organizations that have been working with us on this. Our recent work with you has proven to be enlightening.
Moving forward, we can all work together in a two-pronged approach. For our members, your prong would involve a review of your spreadsheet data before you upload it to NextPlane. This involves checking to see if the key fields you would expect to be in incident reports (event description, age, date of birth, gender, race, harm, and ethnicity) are in fact there. Then, observe any warnings or errors that come out of the data upload process. These two actions can help identify if you have a data problem, and one that may be within your power to fix.
On the CHPSO side, we will continue to review our data and look for patterns in the Nulls. When we see issues, we will reach out to members, sometimes with possible solutions in tow. As always, members are encouraged to initiate a data review with us. And in areas where a systemic issue is a possibility, we will engage the organizational part of data pipeline that we think is best equipped to work with us to address the issue.
If you would like to initiate a review of your organization’s data quality, please contact email@example.com and we will be happy to work with you.