With another flu season come and gone, the summer season is a prime opportunity to make improvements and prepare for the next year’s response. As such, NACCHO staff took time to speak with Applied Public Health Informatics Fellow Robyn Matthews, MPH, to learn about her project to improve public health response to influenza by leveraging the County of San Diego Public Health Services’ (PHS) syndromic surveillance system.
Tell me about your project at the County of San Diego Health & Human Services Agency.
Like many other jurisdictions doing syndromic surveillance, San Diego County utilizes raw data text classifiers to detect many types of communicable diseases and outbreaks. This includes potential influenza outbreaks in the county based on chief complaints that are similar to influenza-like illness (ILI). Using data from January 1st, 2016 to April 30th, 2017, San Diego linked lab-confirmed influenza reports with the original chief complaints found in the syndromic surveillance system to more accurately classify the chief complaints experienced by those with confirmed influenza. More specifically, since many of the large hospitals in San Diego County are now participating in electronic lab reporting (ELR) and syndromic surveillance to meet Meaningful Use requirements, we were able to link the ELR reports with the original Health Level-7 (HL7) Admit, Discharge, Transfer (ADT) encounter messages containing chief complaints via medical record numbers (MRNs), if the MRNs were present in both ELR and the HL7 messages. By undergoing this work, we were able to refine our syndromic surveillance text classifiers based on information gained from the activity with the hopes of detecting potential cases or outbreaks and initiating public health response sooner.
What surprised you the most when you began to classify the chief complaints experienced by patients with lab-confirmed influenza?
The text classifiers specific for ILI didn’t pick up on as many cases of ILI as we anticipated – in fact, while about 88% of the patients presented with at least one symptom related to influenza (e.g., fever, cough), only about a third fit the CDC case definition of ILI. An analysis of the ILI chief complaints revealed the majority of patients had chief complaints specifically mentioning “ILI” or some variation of the word, while others had a combination of symptoms consistent with ILI with the majority having fever and cough.
What challenges or barriers did you face during the project?
While it sounded simple to merge these two data systems, this project required hours of data format restructuring and data processing in order to link the ELR reports with the original HL7 messages and create one analysis-ready visit containing all corresponding messages per patient, including all registration, admission, and discharge data. Sometimes the MRNs from the ELR contained leading or trailing numbers or characters that had to be omitted before merging with the chief complaint data. Also, for some cases, we had to account for preceding and subsequent hospital visits to the primary visit for ILI.
How would a faster public health response impact those most at-risk for becoming infected with influenza in San Diego County?
Syndromic surveillance is powerful because it utilizes near real-time data to help us see what’s going on with the health of populations, as opposed to sources like lab reports that may not be available until after a diagnosis is made, which is sometimes a week or two later. This project allowed us to assess the usefulness of chief complaints for quicker public health response related to influenza. We encourage providers to perform influenza testing on patients presenting with ILI and subsequently report positive results to public health. With syndromic surveillance, we can find and track the percentage of influenza- and ILI-related emergency department encounters over time. When we see increases in these encounters, we can inform providers of the trends and use the information as a basis for advocating vaccinations, which can be particularly important in long-term care facilities where flu outbreaks often occur. The project enabled us to know for certain what patients with lab-confirmed influenza were complaining of, and consequently, we were able to add new syndrome terms and variations into our text classifier syntax to more thoroughly pick up influenza-related chief complaints. A more accurate picture of influenza-related chief complaints may help to inform both providers and the public about the potential for an influenza outbreak and provide a basis for vaccinations in the county where thousands are burdened by the disease each flu season.
What was the County of San Diego Health & Human Services Agency’s role in the project and your role as an Applied Public Health Informatics Fellow?
The County of San Diego PHS Division served as host to my Fellow tenure. The Data Analysis and Surveillance Unit (DASU) team of the Epidemiology and Immunizations Services Branch (EISB) of the County of San Diego PHS was involved in developing and carrying out the activity. As a fellow, I work alongside the DASU and closely with the ELR specialist. The project came up as an idea during a surveillance staff meeting, and I was given the task of leading the project and working with the syndromic surveillance subject matter expect to develop the methodology, conduct the data merge, analyze the findings, and present at the 2017 CSTE Annual Conference.
What improvements in the syndromic surveillance system may be coming next?
We are in the process of applying the matching methodology of this project to a current Hepatitis A outbreak burdening our homeless population in the community. We hope results from the project will help us better understand not only the chief complaints associated with the outbreak, but also ways in which we can apply the findings to develop and refine our syndromic surveillance classifiers for diseases such as Hepatitis. Additionally, we’re interested in looking at the hospital visits leading up to the Hepatitis A lab specimen collection dates among those with confirmed Hepatitis A and assessing the severity of symptoms expressed.
What advice would you share with other local health departments seeking to improve their syndromic surveillance system?
I think we have been successful in San Diego for a few reasons. We have a mature understanding of and experience conducting syndromic surveillance. Also, we have a close partnership with San Diego Health Connect, who provides our local health information exchange (HIE) where most of our syndromic surveillance and ELR data are channeled through. So, identifying and maintaining potential partnerships with organizations involved with the data you want are crucial. Prioritizing your data sources is also important. For us, hospital emergency department (ED) encounter data sent as a result of Meaningful Use requirements is our top priority. As such, we make sure we consistently validate test messages from onboarding hospitals and clinics, especially since the testing phase can take several months to accomplish as it is. I also feel our DASU team of epidemiologists are equipped to successfully carry out informatics-related tasks necessary for the development and maintenance of our syndromic surveillance system, ELR, our communicable disease registry, and eventually electronic case reporting (eCR). See if you have staff capable of carrying out activities such as these before resorting to hiring an outside contractor. It wouldn’t be a bad idea to emphasize a need for informatics skills when hiring new staff, either.
Additionally, I recommend participating in workgroups such as those coordinated through the International Society for Disease Surveillance (ISDS). In our ISDS Region 9 Data Sharing Workgroup, San Diego County has worked together with other jurisdictions in the region to review an opioid-related syndromic surveillance text classifier provided by the CDC, provide recommendations for adapting it to our region, and define ways to display and report the data. I think there’s a lot that can be learned by participating in these workgroups because we are able to brainstorm together and discuss each jurisdiction’s use of their data, successes, and challenges faced.
Click here to access Ms. Matthews’ paper and presentation slides from the 2017 CSTE Annual Conference and read how else NACCHO is supporting local health departments advance the use of informatics at their agencies.