In Overdoses’ Wake, City Dives Into Data

The next time New Haven sees a spike in overdoses, New Haven might have time to plan, thanks to a new data-sharing initiative.

That plan might be short-term, with quicker information about unfolding cases, thanks to newly broadened access to a state database allowing local public health officials and emergency responders to track opioid-related overdoses in real time.

And the plan might be longer-term, if a New Haven tech start-up succeeds with efforts to track statewide trends to predict the next big incident like August’s mass K2 overdose episode on the Green or one prompted by the growing misuse of opioids.

State and city officials and a representative from a local health-tech company described those efforts to combat the opioid epidemic in interviews with the Independent this week.

They singled out the wealth of anonymized information that the state collects on overdoses on a minute-by-minute basis as presenting a unique opportunity for towns and cities to identify vulnerable populations and to figure out how best to allocate resources when responding to drug-related public health crises, such as the 100-plus synthetic-cannabinoid poisonings that took place on the New Haven Green in August.

“That means that the fire chief can change his deployment schedules,” New Haven Special Assistant to the Mayor Michael Harris said about what would be possible if city officials could predict when and where the next overdose outbreak might take place. ​“He can make sure he has Narcan out on the streets. He can coordinate with AMR.”

Depending on the level of information available, city officials may also be able to better understand in real time which kinds of addiction treatments are the most effective at preventing repeat overdoses by people who would otherwise cycle in and out of the hospital.

When a bad batch of K2 (synthetic marijuana) hit New Haven in August, the city saw around 120 overdose-related medical transports over the course of a three-day period. The total number of victims making those trips was only 47, since so many people were poisoned multiple times after smoking from the same batch of K2 after being released from the hospital. 

Sharing Overdose Data In Real Time

During the last week of June, the state Department of Public Health (DPH) provided New Haven’s public health department with access to the trove of overdose-related information that the state collects in its EpiCenter Syndromic Surveillance System.

That’s the centralized database that the state uses to gather pre-diagnostic patient data about emergency room visits from hospitals and urgent care centers in an attempt to better understand the latest trends in diseases and other public health concerns.

The database is updated every 15 to 30 minutes based on intake information uploaded by hospital staff. It records anonymized patient information like age, sex, race, home zip code, and reason for emergency room visit, but not name, address, or phone number.

Brian Weeks, the city health department’s epidemiologist, told the Independent that New Haven was one of the first Connecticut cities to receive access to the state database as part of a pilot program conducted at the beginning of the summer. He and state DPH Director of Communications Maura Downes said that access to the state database was rolled out to all other local public health department in the state later in the summer.

The state has used the syndromic surveillance system since 2004 to monitor emergency room visit data for everything from seasonal and pandemic influenza outbreaks to the spread of gastrointestinal illnesses to weather-related emergencies.

In January 2018, the state added syndromes related to ​“drug and alcohol, including drug/opioid/heroin overdoses” to its list of reportable emergency health conditions. So, from the beginning of this year, the state’s been collecting anonymized demographic patient information for people who land in the hospital after overdosing.

“One of the benefits of the EpiCenter syndromic surveillance system,” Downes told the Independent Wednesday, ​“is that it allows local health department and district staff to access line-level de-identified reports for residents of their jurisdiction as well as aggregate statewide and regional data.”

She said that examining region-specific trends over time will allow state and local officials to establish an expected overdose threshold for each area of the state.

“If a spike, or greater than expected number, of overdoses occurs in a region over a short period of time,” she said, ​“the local health departments will be alerted via a public health alert.”

In addition to the real-time information available in the state’s syndromic surveillance database, Downes said that local public health departments now also receive a series of quarterly reports related to overdoses.

Those reports include statewide and municipal-level death data on unintentional drug overdoses from the Connecticut State Unintentional Drug Overdose Reporting System (SUDORS).

Local health departments also receive statewide prescription opioid data from the Connecticut Prescription Monitoring and Reporting System, as well as statewide and municipality-level data on Neonatal Abstinence Syndrome (NAS) hospital discharges.

The Benefits, And Limitations, Of The Syndromic Surveillance System

Weeks said that city access to the syndromic surveillance system’s overdose data gives the local department a wealth of demographic information that it can use to inform outreach events, health education, interventions, and policy development related to addiction and substance abuse.

But he cautioned the state’s emergency room visit database does have its limitations.

First, he said, the system allows local officials to see patient data only for New Haven residents, and not for patients who live outside of the city but who overdose and are treated in New Haven.

Even for a New Haven patient, he said, the only identifying address information the local health department can see is the patient’s zip code. In a city as small as New Haven, that makes it difficult to identify whether or not an overdose epidemic is concentrated in a particular neighborhood or on a particular block.

Second, he said, the system is only as useful as the level of detail entered by hospital intake staff.

“With the K2 incident,” he told the Independent by email, ​“some intake staff only put ​‘overdose’ without specifying the substance or location of incident. This reduces the accuracy and ability to capture all relevant cases of the K2 incident on the New Haven Green.”

Since the syndromic surveillance system parses the International Classification of Diseases (ICD) codes and other text and keywords associated with a patient’s intake, incomplete information can skew the local health department’s ability to identify what exactly the patient has overdosed from.

But on the flip side, he said, he now has access to emergency room overdose information within half an hour of it being entered into the database. Before this summer, he had to wait for a biannual or annual emergency room visit report from the state public health department.

More importantly, he said, the syndromic surveillance system provides information on non-fatal overdoses in addition to fatal overdoses, so the data that he can review is not just limited to patients who have died.

He said that the city’s health department, police department, and fire/EMS department are working with Alfredo Herrera, the city’s geographic information systems (GIS) analyst, to compile each of their respective datasets to make sense of New Haven’s substance use situation as a whole.

“This comprehensive and highly specific dataset will be cutting-edge,” he told the Independent, ​“providing the opportunity for near real-time alerts, but also a holistic picture of the situation (specific to the location of the incident) and improved opportunities to infer and to address the opioid epidemic and substance use in the city in general. This also provides the opportunity to assess the success/failures of interventions and policies.”

While Herrera develops that integrated dataset, Fire Chief John Alston said his department’s emergency responders are relying on their close relationship with American Medical Response (AMR), the private ambulance company that the city contracts with to handle hospital transports.

For the past two months, the fire department has had access to a stripped-down version of AMR’s First Watch application, which provides real-time updates for overdose-related transports without sharing revealing those patients’ personal information.

“If I get more than 10 overdoses in a day,” he said, ​“it sets off a trigger for AMR and for us.”

He said that the city is currently working with Yale-New Haven Hospital and other local treatment centers to get those medical providers to share more of their overdose patient data. That way, he said, the city can track in real time not just when and where overdoses are taking place, but also which different types of treatment patients are receiving.

Predicting The Next Outbreak

Mayoral special assistant Harris said that the city is interested not just in getting access to real-time overdose information for New Haven patients. He said the city, in collaboration with a local health-tech start up, would also like to see overdose information for surrounding cities and towns, so it can predict where and when the next overdose outbreak might take place.

That tech company is called Telesphora. It is based out of the new Elm Street health-tech business incubator, Health Haven Hub.

David Pearlstone, the executive director of Health Haven Hub, said the start up grew out of an opioid crisis code-a-thon hosted by the federal Department of Health and Human Services (DHHS) in Washington D.C. in 2017.

Yale University medical students behind the design lab Origami Innovations partnered with a machine learning specialist and a pain management specialist to create a machine-learning model that analyzed Connecticut opioid fatality data from 2012 to 2016. They used chunks of that information to see if they could predict when and where overdose outbreaks would take place within the historical timeframe, and found that their algorithm was right 90 percent of the time.

“If you look in the past,” Pearlstone said, ​“you can predict where and when” overdoses will happen next.

Pearlstone said the group won the top prize the DHHS code-a-thon, and then formed Telesphora in New Haven.

He said his company is now interested in using regional, real-time overdose data as a way to predict where and when Connecticut’s next overdose outbreak might take place. He has met with state and city public health officials in the wake of August’s K2-related overdoses, and is working with DPH now to come to see if it can get access to overdose-related patient data for more than just New Haven.

“We’re all in agreement about the principle,” Pearlstone said. ​“We have lawyers setting up the data sharing agreement. We’re hoping to get access to real-time data ASAP so that we can start processing the data through the algorithm.”