D: The Most-Promising Data Sources

Unemployment Insurance Claims

Overview and Relevance for Behavioral Health Monitoring

Unemployment is a widely recognized social determinant of health. Systematic reviews find consistent evidence linking unemployment with all aspects of BH (Milner, Page, and LaMontagne, 2014). Job loss has both psychological and material consequences that increase the likelihood of poor health and declining BH. In addition, poor and declining BH are recognized to increase the likelihood of job loss (e.g., due to substance use in the workplace). Understanding the patterns of unemployment across a population, particularly after a PHE, can provide information about when and where there is a need to strengthen BH system capacity and take action to support BH resilience. Along with housing and property losses, employment is one of the first and most directly related social determinants of health that is affected by PHEs.

The U.S. Department of Labor (DOL) and Bureau of Labor Statistics (BLS) collect and disseminate data on employment, unemployment, and participation in the federal UI program. The timeliest data on emerging labor market conditions are weekly counts of initial claims for benefits from the UI program (BLS, 2014; DOL, 2022a). The process for filing claims is as follows:

• An initial UI claim is created when an individual separates from an employer and applies with the state to have their separation evaluated for eligibility for UI benefits. Initial claims capture an individual’s earliest interaction with the UI program; for this reason, this information is used by economists and federal agencies as a leading indicator of emerging labor market conditions.
• After filing the initial UI claim and experiencing a week of unemployment, the individual then files a continued claim. The DOL interprets continued claims information as another good indicator of labor market conditions that can help provide confirmatory evidence of U.S. economic trends.

In addition to the timeliness of UI data on unemployment, the availability of disaggregated data can be used to understand geographic patterning (by state and county), as well as sociodemographic differences (e.g., by gender, race, ethnicity, age, or education) in the geographic patterning.

Accessing the Data

The DOL and state labor and workforce agencies disseminate administrative data on UI claims. Table 1 summarizes the various sources of UI data and how to access them. The DOL publishes initial UI claims data reported by states as part of the legislatively mandated oversight process. Data are aggregated to the state level each week and published online (DOL, 2022a) (see Rows 1 and 2 in Table 1). In addition to these weekly state-level data, the DOL publishes monthly state-level data for selected sociodemographic groups (DOL, 2022b) (see Row 3 in Table 1).

More–geographically granular UI claims data are publicly available from many state labor and workforce agencies (Row 4 in Table 1). Some states access these data using the BLS Program to Measure Insured Unemployed Statistics (PROMIS), which is a database and query tool. The BLS uses PROMIS to collect initial UI claims inputs used in conjunction with other BLS and Census Bureau data to generate the monthly Local Area Unemployment Statistics (LAUS).

The LAUS is the BLS program for estimating unemployment for local areas (e.g., counties and metro areas). This data source may also be useful for understanding the impact of a PHE, but data are released on a monthly basis and with a greater production lag than the UI data. Additionally, the LAUS estimates come from a complex data integration in which the UI data are combined with additional data from the BLS and Census Bureau. Because these additional data sources have less granularity (in spatial units, time units, or both) and longer time lags, LAUS estimates also have less granularity and longer production lags than the UI data. In addition, the statistical methods used to produce the LAUS may smooth out changes in unemployment patterns that may be associated with a PHE.

State and federal sources of UI data differ in the data they make publicly available, with states providing the most-granular initial UI claims data (e.g., county-by-week strata or, in some cases, county-by-week-by-sociodemographic strata). As of 2022, the DOL provides the following additional data on outcomes other than initial UI claims: a count of continued claims, a count of covered employment used by the DOL as the size of the population at increased risk for initial and continued UI claims, and the ratio of the continued claims to the covered employment counts (which the DOL defines as the insured unemployment rate). At the time of this writing, no state dashboard provides a similar count of the covered employment from which the provided initial UI claims could be calculated as a ratio to provide an estimate of the incidence of insured unemployment.

Strengths and Limitations of These Data

Learn more about how this data source compares with others. Access a list of real-world examples of the use of this data source for BH surveillance in the PHE context. Access selected publications that have used this data source for monitoring and surveillance.

2-1-1 Calls

Overview and Relevance for Behavioral Health Monitoring

When an individual has a need for information or referral to social services and other assistance, such as housing support, help with food insecurity, or a mental health crisis, information on COVID-19, or help with child care (just to name a few), they can call 2-1-1, 24 hours a day, seven days a week (“About 211,” 2021). The caller is routed by the local telephone company to one of 200 agencies across the United States that are staffed by trained community specialists and operated by United Way (70 percent of agencies), local crisis centers, Goodwill, and others. Callers are provided with information and connected with services in their area based on self-reported zip code. 2-1-1 call center operators collect data from each caller, including the date and time of the call, limited demographic information (age, gender, race or ethnicity), the major need (i.e., the primary reason for the call), and other needs the caller has. Most 2-1-1 call centers use a standard taxonomy to categorize the major need for each call (e.g., “mental health & addictions”), but others use their own classification systems to categorize needs, which can make it more difficult to compare call data across multiple centers.

Relevant to BH surveillance, 2-1-1 call centers across the country reported fielding more than 1 million requests in 2020 related to mental health and substance use, including 900,000 related to suicide or mental health crisis. Even if a caller is not expressing an explicit need for mental health support, recent research suggests that unmet needs related to social determinants of mental and physical health can be a source of significant stress (Carod-Artal, 2017; Allen et al., 2014), and our preliminary analyses suggest that total call volume and call volume for BH needs are moderately correlated. With better documentation of call reason, misclassification would be less of an issue, and correlation would likely be higher. Therefore, depending on (1) the sample size of 2-1-1 calls in your jurisdiction, (2) the completeness and accuracy of call reason in your 2-1-1 call data, and (3) your goals for using 2-1-1 calls as a proxy for BH impacts of PHEs, you may choose to either monitor trends in 2-1-1 calls for any reason (total calls), indicating a variety of stressors in your community, or to narrow your focus and monitor only calls with a documented need for mental health and addiction assistance.

Accessing the Data

To access data that have sufficient granularity (i.e., temporal, spatial, demographic characteristics) for the types of analyses that you will likely want to perform, you will probably need to establish a data-sharing arrangement with your local 2-1-1 call center.

Depending on your location, you may be able to obtain state-level 2-1-1 data or data from several call centers through a single entity. For instance, 2-1-1 call data for Texas, encompassing 25 call centers, are available by submitting a public records request to Texas Health and Human Services (undated). Requests like these may involve a data processing fee.

A website called 2-1-1 Counts (Health Communication Research Laboratory at Washington University in St. Louis, undated) is a publicly available source of aggregated 2-1-1 calls from 37 states (as of November 2021), housed at Washington University in St. Louis. Again, depending on your location and whether 2-1-1 agencies contribute their data to this aggregator, you may be able to use this website’s interactive dashboard to drill down to calls with a mental health need at the county level, with near real-time results available (Summary 1).

In this example, the 2-1-1 counts website shows counts of 2-1-1 calls in Cumberland County, Maine, for November 5-11, 2021, by documented need and the breakdown of calls within the “mental health & addictions” category.

Source: Reproduced from Health Communication Research Laboratory at Washington University in St. Louis, undated.
Note: This screenshot shows counts of 2-1-1 calls in Cumberland County, Maine, for November 5–11, 2021, by documented need and the breakdown of calls within the “mental health & addictions” category.

2-1-1 Counts also allows you to view overall trends in counts of calls by type. Summary 2 shows calls for assistance with “mental health & addictions” for Cumberland County, Maine, in 2020 (orange) and 2021 (green). Although it can be helpful to see the big picture, you may prefer analyzing the raw data from your local 2-1-1 call center to fully interpret these figures (e.g., put these trends in the context of total calls to 2-1-1 for your location of interest).

This example from the 2-1-1 counts website shows the number of “mental health & addictions” requests in Cumberland County, Maine, from November 2019 to October 2021, with a spike in February 2020.

Source: Reproduced from Health Communication Research Laboratory at Washington University in St. Louis, undated.
Note: This screenshot shows counts for “mental health & addictions” requests in Cumberland County, Maine, from November 2020 to October 2021. The number of calls is shown on the y axis; the original figure does not have a y axis label.

Strengths and Limitations of These Data

Learn more about how this data source compares with others. Access a list of real-world examples of the use of this data source for BH surveillance in the PHE context. Access selected publications that have used this data source for monitoring and surveillance.

Poison Control Center Calls

Overview and Relevance for Behavioral Health Monitoring

If you are concerned about an exposure to a potentially harmful substance, calling 1-800-222-1222 connects you to one of 55 the PCCs across the country that is closest to you. The center provides information and help triaging and treating the exposure. Callers are asked for details about the exposure, such as the substance, if known; when the exposure happened; and the age and gender of the person who was exposed to the potentially harmful substance. Callers are also asked whether the exposure (e.g., ingestion, inhalation) was unintentional or intentional; if intentional, they are asked whether the exposure should be classified as a suspected suicide attempt.

PCCs submit these data on each call they receive (i.e., case records) to the National Poison Data System (NPDS), a data warehouse managed by the AAPCC (undated-b), a nonprofit organization that represents all 55 PCCs in the United States. As of late 2021, the NPDS contained nearly 75 million case records and is updated every eight minutes.

Calls to PCCs may be classified as “intentional,” including subcategories for “suspected suicide,” “misuse,” “abuse,” and “unknown.” In the context of PHEs, intentional exposures may serve as a proxy for the population’s BH status.

Accessing the Data

The NPDS has a publicly available interactive dashboard (AAPCC, undated-b) that allows you to view aggregate data, including timely data on emerging hazards (AAPCC, undated-d). (For an example, see Summary 3.) However, this resource lacks some of the necessary granularity (e.g., intentional vs. unintentional exposure) to be used specifically for BH surveillance.

Hand sanitizer exposure cases in children 12 years and younger in 2021 ranged from 1,696 to 2,195 and were highest in the first half of the year, declined in June and July, and began to increase again in August.

Source: AAPCC, undated-a.

Instead, you have a few options for accessing PCC data, depending on how granular and how timely you need the data to be, as well as whether you have funding available to purchase these data and have a relationship with your local PCC.

• PCCs have access to the NPDS data portal, which provides them with detailed and timely data on PCC calls in their area. To access data that have sufficient granularity for the types of analyses you will likely want to perform (whether the exposure was intentional or unintentional, the age of the individual), you will likely need to establish an arrangement with your local PCC to obtain access to the NPDS data portal or develop a data-sharing agreement with the PCC. Some local PH agencies and PCCs have these relationships already in place, but many do not. (Furthermore, some PCCs are considered part of the local health department, based on state statutes and their funding streams).
• Another option might be to purchase NPDS data for your local jurisdiction through the AAPCC (undated-b). This type of arrangement can take several weeks to establish. To be useful for ongoing monitoring and surveillance, this approach would require timely, ongoing data transfers from the NPDS. More information on the NPDS can be found on its website (AAPCC, undated-d).

Strengths and Limitations of These Data

Learn more about how this data source compares with others. Access a list of real-world examples of the use of this data source for BH surveillance in the PHE context. Access selected publications that have used this data source for monitoring and surveillance.

Prescription Medication Fills

Overview and Relevance for Behavioral Health Monitoring

In 2019, approximately 16 percent of U.S. adults had taken a psychotropic medication, including antidepressants, anti-anxiety medications, mood stabilizers, antipsychotics, and stimulants (Terlizzi and Zablotsky, 2020; Moore and Mattison, 2017). To examine the BH in your jurisdiction in the PH emergency context, you could consider monitoring pharmacy data: specifically, fills of certain common psychotropic medications, such as antidepressants or anti-anxiety medications. A change in the rate of SSRI fills could signal a change in the incidence or intensity of depressive symptoms or clinical depression. Importantly, such a change also might indicate changes in access to mental health care (e.g., expansion of telehealth).

There are several potential sources of pharmacy data, including sales data, prescription fill data, and longitudinal prescription claims data. Sales and prescription fill data tend to report summaries of prescription drug dispenses at a given spatio-temporal level. Claims data are more granular, as they track individual dispensing events per patient. There are strengths and limitations to each of these data sources.

One potential data source you could use to monitor this indicator is the IQVIA Xponent dataset (IQVIA, undated-a). IQVIA is a data science company that curates datasets on pharmacy fills. One of its products, Xponent, is a dataset that estimates the monthly volume of prescription drugs dispensed in a given municipality, with spatial granularity ranging from zip code to state level.¹ IQVIA produces these estimates by obtaining prescription fill information from a broad sample of retail settings (93 percent of all retail-based transactions), long-term care (LTC) facilities (74 percent of all LTC transactions), and mail order services (77 percent of all mail order transactions). IQVIA then uses proprietary methods to project the population-level prescription volume dispensed via retail and LTC facilities in a given municipality. Then, IQVIA sums the projected retail volume, the projected LTC volume, and the raw mail order volume for the following data elements for a given product:²

• number of new prescriptions and total prescriptions (i.e., new prescriptions plus refills)
• new quantity dispensed and total quantity dispensed (i.e., new quantity plus refills)
• consumer and pharmacy spending on new and total prescriptions.³

The IQVIA Xponent dataset also includes prescriber identification number, patient age and gender, and payer source.⁴

Accessing the Data

On the IQVIA website, you can fill out the “Contact Us” form (IQVIA, undated-b). You will be asked to describe the specifications of the data you are interested in (e.g., the period, the geography, the medication class(es), and how often you need new data transferred).

A representative from IQVIA will then contact you to get additional information on your data request and to schedule a meeting. At this meeting, you will further define the scope of the data that you will need, how often you will need the data transferred to you for analyses, financial resources available, and other details.

The IQVIA team will work with you to determine an appropriate scope for a data purchase agreement depending on several factors, including funding available for purchasing the data:

• IQVIA can provide you with data aggregated to different time frames (e.g., week, month, quarter) and different geographies (e.g., zip codes, metropolitan statistical area, state).
• IQVIA can also drill down to a specific active ingredient of interest (e.g., sertraline, alprazolam, bupropion) or aggregate the medications to higher levels, such as the drug class level (e.g., SSRI) or therapeutic class level (e.g., antidepressant, anti-anxiety medication).
• After iterating on a data specifications template in collaboration with IQVIA and finalizing the data purchase agreement, you will begin receiving the data, either as a CSV (comma-separated values) file or Microsoft Excel workbook.

Strengths and Limitations of These Data

Learn more about how this data source compares with others. Access a list of real-world examples of the use of this data source for BH surveillance in the PHE context. Access selected publications that have used this data source for monitoring and surveillance.

Over-the-Counter Sleep Aid Sales

Overview and Relevance for Behavioral Health Monitoring

When a grocery store, pharmacy, or convenience store scans a barcode on a product, data about that product, known as scanner data, are anonymously collected. Collecting scanner data from a large and representative sample of retailers gives us a good estimate of total sales volume. This section will describe why these data are useful, how to get them, and how to use them.

Sales data for OTC medications show promise for BH surveillance. For instance, sales of anti-anxiety medications or sleep aids may be indicative of underlying BH issues because of their use as coping mechanisms for anxiety or insomnia. These sales data may therefore enable both the detection of the BH impacts of PHEs at the population level and corresponding PH interventions before more-severe downstream impacts occur.

There is a long history of using OTC medication data for PH, particularly in syndromic surveillance of infectious disease outbreaks. They are often used in combination with other signals, such as ED and PCC data. OTC medication sales have been used in a wide variety of infectious diseases, including influenza (Liu et al., 2013), H1N1 (Holtry, Hung, and Lewis, 2010), and gastrointestinal illnesses (Hines et al., 2017). OTC medication sales data are a timely signal; people will often purchase medication when they start to feel unwell, so we see spikes in sales data before hospitalizations. One 2003 study (Hogan et al., 2003) found that electrolyte sales could be used to predict hospital admissions for diarrheal pediatric diseases. There was a correlation of 90 percent between these signals, but the electrolyte sales preceded hospital admissions by 1.7 weeks. Follow-up research validated this finding by talking to caregivers who had taken their children to the ED for diarrheal disease: 41 percent had purchased electrolyte medications OTC before the ED visit (Johnson, Wagner, and Saladino, 2005). Use of OTC medication sales data could allow hospitals to prepare or PH agencies to engage in preventative action more quickly. OTC data have also been used to look at other conditions and coping strategies. One study demonstrated that access to recreational cannabis in Colorado was associated with a decrease in sleep aid sales (Doremus, Stith, and Vigil, 2019).

There are several organizations that collect, clean, and distribute scanner data. Nielsen IQ and IRI Marketing both maintain large datasets of retail scanner data from across most of the United States. These datasets are commercial and can be costly to access, especially on an ongoing basis. Licensing agreements with retailers prevent data being shared if they could be exploited for market research, so products are aggregated into categories and stores into geographies. Nielsen also maintains an academic dataset that can be accessed at a lower cost, but it is available only to academic institutions for research and is updated annually, so it is not suitable for surveillance.

Some jurisdictions, such as New York City, collect their own OTC medication data. The Department of Health and Mental Hygiene maintains a dataset on anti-diarrheal, fever, and allergy medications sold through the largest retail chains in the city. This information is used for syndromic surveillance of infectious disease outbreaks. Similarly, some states complement the ED data in their Electronic Surveillance System for the Early Notification of Community-Based Epidemics (ESSENCE) with OTC medication data, also for infectious disease outbreak monitoring. These resources are location-specific and may not be suitable for your PH department.

Accessing the Data

An NRDM subscription is required to access the OTC medication data. PH agencies may obtain a subscription by contacting the NRDM program manager (Real-time Outbreak and Disease Surveillance, undated-a). Subscription fees are based on the number of stores in a health department’s jurisdiction and, as of 2021, were approximately $42 per store per year. (Costs are shared across all subscribers, so fees could decrease if more PH agencies were to subscribe.) After a departmental data use agreement is signed, a PH supervisor assigns users to the system who must then sign an individual data use agreement. Once a subscription is active, the NRDM platform can be accessed at any time and is automatically updated daily with new data.

As of 2021, the NRDM had licensing agreements with ten chains of retail pharmacy, grocery, and mass merchandise stores. These retailers send data on daily product sales in each store to the NRDM, which then aggregates these sales into zip codes and product categories. Most categories are focused on infectious disease outbreaks. This means that, out of the 20 product categories (e.g., anti-diarrheal medications, cold-relief products, personal protective equipment), only one category—sleep aids (i.e., medications to help with insomnia)—could be used as a proxy for BH concerns. Sleep difficulties can be a manifestation of anxiety, depression, or simply stress after a PHE. However, new categories, such as medications or supplements to cope with anxiety, could be added in the future.

Some medications that people take to help them sleep, such as diphenhydramine (Benadryl), would show up in the “cough and cold” category in the NRDM, while the sleep aid category excludes these so-called dual-use products. A subscription with the NRDM would include all categories of medications, which could be used for other PH monitoring programs.

To protect against the use of these data for market research (e.g., trying to figure out which products are selling the most), the NRDM reports daily sales volumes for each category (e.g., cough syrups) at the zip code level. These data can be aggregated to larger geographies, such as counties or states, or to longer periods within the the NRDM’s platform. The lag between the sales and the data being available on the NRDM platform is short, typically less than a day. There are no privacy or Health Insurance Portability and Accountability Act (HIPAA) concerns with the NRDM data because customer details are not collected and licensing agreements between the NRDM and retail chains ensure that product categories and geographies are adequately aggregated before being made available.

Viewing the Data

The NRDM has a well-developed platform and graphical user interface with two main views. The first view, known as EpiPlot, provides views of OTC sales for specific categories, regions (down to the zip code level), and periods. Data can be shown with or without normalization and downloaded in multiple formats, including for Excel or Statistical Analysis Software (SAS). An application programming interface is also available to programmatically obtain data. The time series plot in Summary 4 shows daily sales of thermometers for 536 stores in the state of Washington since 2003 as viewed in EpiPlot. The NRDM tracks both promoted sales (blue line showing sales associated with a coupon or sales) and unpromoted ones (green line). You may overlay an alert threshold determined by a time series analysis (red line). For instance, in Summary 4, the threshold is defined as the moving average of the three standard deviations above the mean. When the blue or green lines cross the threshold, this represents an alert. Sales driven by seasonal influenza and the start of the COVID-19 pandemic are apparent.

The National Retail Data Monitor’s EpiPlot view allows users to plot data on OTC sales according to certain temporal and spatial specifications. This example uses the number of thermometer sales from 536 stores in Washington State from 2003 to 2021, including unpromoted sales and all sales. From 2002 to 2010 sales gradually increased, until a large spike in 2020 that corresponds with the start of the COVID-19 pandemic.

Source: NRDM, undated.

Note: This view displays the number of thermometer sales from 2003 to 2021.

The second view, MapPlot, gives a geographic view of OTC sales. MapPlot shows a map of your jurisdiction with OTC sales by subarea, allowing PH agencies to see which zip codes in each county have the highest or lowest OTC sales. Using the MapPlot view, Summary 5 shows the number of standard deviations above the expected daily thermometer sales in jurisdictions in Washington state near the start of the COVID-19 pandemic (March 1, 2020).

A state or local health department could use this MapPlot view to look at state sales, county sales, and (by zooming in on the map) zip code sales. A drill-down function can be used to produce more-detailed plots for any map region.

Daily thermometer sales at the start of the COVID-19 pandemic (March 1, 2020) in Washington State varied widely by zip code anywhere from 0.5 standard deviations above the expected daily sales to greater than four standard deviations above the expected daily sales.

Source: NRDM, undated.

Note: This map displays findings from a wavelet model showing the number of standard deviations above the expected daily sales of thermometers for March 1, 2020.

Strengths and Limitations of These Data

Learn more about how this data source compares with others. Access a list of real-world examples of the use of this data source for BH surveillance in the PHE context. Access selected publications that have used this data source for monitoring and surveillance.

Suicide and Crisis Hotline Calls (e.g., 9-8-8)

Overview and Relevance for Behavioral Health Monitoring

In 2020, Congress passed the National Suicide Hotline Designation Act (Public Law 116-172) establishing a new three-digit dialing code (9-8-8) to serve as the mental health emergency equivalent to 9-1-1. On July 16, 2022, the 9-8-8 dialing code went live and began routing callers to the 988 Suicide and Crisis Lifeline, formerly the National Suicide Prevention Lifeline. This network, consisting of more than 200 crisis centers operating 24 hours a day, seven days a week, provides such services as emotional support, suicide risk assessment, treatment referrals, and transfers to emergency services (SAMHSA, undated-a). Some jurisdictions also have the capacity to deploy mobile BH crisis response teams (Mishara et al., 2007a; Mishara et al., 2007b). People in crisis can call, text, or chat 9-8-8 to be connected to a network of trained counselors. If the caller’s distress level is thought to be life-threatening, the caller is connected to 9-1-1.

The aim of the 2020 legislation was to create a number that was easier to remember than the previous number for the National Suicide Prevention Lifeline (1-800-273-8255). Legislators also hoped that the new number’s similarity to the well-known 9-1-1 dialing code would contribute to reducing mental health–related stigma (Cantor et al., 2022).

SAMHSA provides funding support for the 988 Suicide and Crisis Lifeline. With the transition to the new dialing code, SAMHSA and Vibrant Emotional Health (the administrator of the network) support 988’s implementation at the state, tribal, local, and territorial levels (SAMHSA, 2021). As of late 2022, the 9-8-8 dialing code had launched only recently. As the code becomes more fully implemented and as public awareness of this resource grows, data at the sub-state level on calls, texts, and chats to crisis call centers may serve as a useful data source for population-level BH impacts of PHEs in certain jurisdictions. While we do not highlight crisis text lines in this toolkit, click the button to explore relevant studies that have used this data source.

Access selected publications that have used crisis text line data for monitoring and surveillance.

Emergency Department Visits and Other Patient Encounters for Public Health Insights: The National Syndromic Surveillance Program

Overview and Relevance for Behavioral Health Monitoring

Syndromic surveillance is a PH monitoring activity that provides near real-time insights to relevant leadership and decisionmakers. It involves the use of anonymized patient encounter data (e.g., ED visits) to monitor a specific set of symptoms (i.e., chief complaints) and diagnoses—together, syndromes—that can serve as an early warning system for understanding and addressing PH issues (CDC, 2022b). Common uses of syndromic surveillance are to understand disease prevalence, incidence, and spread (e.g., influenza outbreaks) and to help inform PH policy, practice, and program design and implementation.

NSSP, established in 2003, is a collaboration spanning federal, state, and local government; academia; and the private sector. It focuses on leveraging patient encounter data for PH insights (CDC, 2022b). NSSP “promotes and advances development of the cloud-based BioSense platform, a secure integrated electronic health information system that hosts standardized analytic tools and facilitates collaborative processes” (NSSP, 2017; see Summary 6). NSSP’s BioSense Platform (NSSP, undated-a) hosts ESSENCE, which allows PH officials to “analyze events of PH interest, monitor healthcare data for events that could affect PH, and share data and analyses” (NSSP, 2017). ESSENCE “provides insight into the data sent to the BioSense Platform” (NSSP, 2017).

Steps taken by the National Syndromic Surveillance Program at CDC are as follows: People seek treatment in a medical facility; the facility sends data to state and local health departments or data aggregators, which contribute data to the NSSP Biosense Platform. CDC provides analytic tools, services, system infrastructure, funding, technical assistance and training, and data analysis support, and the NSSP Community of Practice shares data and knowledge, builds skills, and collaborates to develop methods and respond to emergencies.

Source: NSSP, 2021a.

As shown in Summary 6, when a person visits a participating health care facility, de-identified data from the patient encounter are sent to PH agencies or Health Information Exchanges, then uploaded to the NSSP BioSense Platform. The patient encounter data include the chief complaint (as free text), ICD-9 or -10 diagnosis codes, Systematized Nomenclature of Medicine (SNOMED) codes, patient characteristics (e.g., demographics, insurance type) and location, and other variables. Combinations of keywords based on patient-reported symptoms, provider assessments, and diagnosis codes can be used to identify and monitor ED and other visits associated with various health conditions. The patient encounter data can be available in the BioSense Platform as quickly as within 24 hours of the encounter. As of late 2022, “more than 6,000 health care facilities” in 49 states and the District of Columbia contribute ED data daily and “71% of the nation’s EDs contribute data to NSSP” (CDC, 2022b). Currently, ED data are the primary source in ESSENCE, but some PH agencies participating in NSSP have reported augmenting their analyses of patient encounter data with other data sources, such as EMS data and OTC medication sales data, that might support the identification of BH needs and crises. In addition, the NSSP Community of Practice (NSSP, undated-b) is exploring the feasibility of integrating such data sources into the BioSense Platform.

NSSP has established syndrome definitions related to BH to help state, tribal, local, and territorial PH agencies understand BH needs in their jurisdictions. The NSSP Community of Practice is also exploring how different mental health conditions might be grouped according to surveillance definitions to better support BH surveillance activities; this might be particularly useful for differentiating new-onset conditions from exacerbations of chronic BH conditions, or simply the presence of a BH condition that is currently well managed but appears in the patient’s list of diagnoses when they visit the ED. These definitions will be posted to the NSSP Community of Practice “Knowledge Repository” (NSSP Community of Practice, undated) when available.

BH surveillance activities typically include tracking the prevalence and incidence of particular visit types, as determined through diagnosis codes. The ESSENCE system enables analyses that look across space, time, and demographics, allowing for the identification of which localities at the state, tribal, local, and territorial levels are experiencing spikes or anomalies and associated population attributes or driving factors.

Throughout the COVID-19 pandemic, PH agencies and federal PH officials have leveraged these data to understand mental health service use across the life span. A Morbidity and Mortality Weekly Report from February 2022, for instance, analyzed pediatric mental health needs through ED visits before and during the pandemic; the researchers found that weekly ED visits among adolescent girls increased for eating and tic disorders, potentially demonstrating increased need (Radhakrishnan et al., 2022). Another study using NSSP data found that ED visits for suspected suicide attempts around February to March 2021 were 51 percent higher among girls ages 12–17 years than during the same period in 2019 (Yard et al., 2021). Similarly, a 2022 article published in JAMA Psychiatry assessed the relationship between COVID-19 surges and mental health–related ED visits among adults, finding that surges may have had impacts on mental health, causing an increase in related ED visits (Anderson et al., 2022). This research has implications for PH messaging and intervention targeting. Some local PH agencies are beginning to use NSSP’s definitions for BH and relevant patient encounter data in collaboration with their local EMS agencies to direct resources to specific communities demonstrating increased need for BH support.

Accessing the Data

NSSP is funded through CDC; its data, through the BioSense Platform, are available to all PH practitioners with demonstrated PH agency affiliation. The BioSense Platform and local instances of ESSENCE include statistical tools to enable timely monitoring, data analyses, and data visualizations, as well as the functionality to generate syndrome-specific reports that can be shared at varying frequencies (e.g., daily, weekly).

For a quick-start guide to ESSENCE, see Summary 7 (NSSP, 2017).

Cover page of the BioSense Platform Quick Start Guide to Using ESSENCE (Electronic Surveillance System for the Early Notification of Community-Based Epidemics) from December 2017.

Source: Reproduced from National Syndromic Surveillance Program, 2017.

The NSSP Community of Practice also offers technical resources on its website, including the following:

• the NSSP Community of Practice “Knowledge Repository” (NSSP Community of Practice, undated), which includes Community of Practice call recordings, resources on syndromes, data analytics, and data-sharing
• an overview of ESSENCE, its components, and its PH applications (Burkom et al., 2021)
• the NSSP Technical Resource Center (CDC, 2022c), with resources to practice syndromic surveillance, get onboarded into the BioSense Platform, start using BioSense Platform tools, and more
• a monthly newsletter
• a Slack workspace for the Community of Practice.

Although distinct from the NSSP, select localities (such as New York City) have created platforms that extract data from local instances of ESSENCE and integrate them with other data sources.

Strengths and Limitations of These Data

Learn more about how this data source compares with others. Access a list of real-world examples of the use of this data source for BH surveillance in the PHE context. Access selected publications that have used this data source for monitoring and surveillance.

Emergency Medical Services Activations

Overview and Relevance for Behavioral Health Monitoring

When someone calls 9-1-1 for emergency assistance, they are connected to an emergency dispatch office that can send emergency responders, such as police, firefighters, and EMS, to a caller’s location. In addition to collecting a caller’s phone number and location, dispatchers collect patient care information for calls that require EMS and then transmit the information to EMS responders. This patient care information becomes part of an EMS database, which is then updated by EMS responders after they respond to the 9-1-1 call. In addition to 9-1-1, EMS data also contain information about transports that require an EMS team.

EMS data include detailed information for each activation (i.e., each time EMS responds) about the scene (e.g., number of patients at a scene, location, time of EMS arrival and departure), situation (e.g., chief complaint, initial patient acuity), patient vitals (e.g., pulse oximetry, heart rate), medications administered and procedures performed (e.g., naloxone to try to prevent an overdose), and where the patient was transported (e.g., substance use disorder treatment center, hospital).

Specific indicators of BH needs or crises that are tracked through EMS data include whether patients displayed symptoms of alcohol or drug use, signs of an overdose or intentional ingestion of a harmful substance, or symptoms suggesting that psychiatric care may be needed (e.g., suicide attempt, delusional behavior). Geographic identifiers (addresses and zip codes) are collected for the scene of the incident and for a patient’s residence.

EMS activation data demonstrated their sensitivity to change after COVID-19 (Slavova et al., 2020; Khoury et al., 2021; Glober et al., 2020; Friedman et al., 2021), with changes detected at national, state, and county levels. Thus, EMS data may be useful for examining patterns and changes over time in EMS responses related to BH (e.g., overdose, suicide attempt) after other PHEs (e.g., hurricanes). However, because EMS is typically dispatched only in an emergency, the data likely represent “crisis-level” BH problems rather than more-upstream or less severe concerns.

Most states collect a standardized set of data elements from EMS agencies and submit them to NEMSIS (undated-a), which collects, stores, and shares state and national EMS data. For example, NEMSIS provided a briefing (Mann, 2023) that showed preliminary data on how BH-related activations have increased since COVID-19 (Summary 8). More-specific information about these and other COVID-19 analyses conducted by NEMSIS can be found online at NEMSIS (2022).

Data from the National Emergency Medical Services Information System show changes in the rate of BH-related EMS activations from 2018 to 2022, with the highest percentage tied to the declaration of the COVID-19 pandemic, higher than expected rates in EMS activations in week 10 after the onset of the pandemic, and beginning to drop off in week 14.

Source: Mann, 2023.

Note: This figure displays changes in the rate of BH-related EMS activations from 2018 to 2022, with the declaration of the COVID-19 pandemic shown as a dark red vertical line and the higher than expected rates in EMS activations in week 10 after the onset of the pandemic and beginning to drop off in week 14 shown in the blue shading.

Accessing the Data

There are three options for obtaining EMS activation data.

Three Options for Obtaining EMS Activation Data

Option 1: NEMSIS	NEMSIS maintains a uniform national database of EMS data from across the United States. The database is built from data shared by states and local agencies, all of which use a common computer language to move the data from their software systems to NEMSIS. NEMSIS then performs a series of quality checks and standardizes the data through a set of automations. The NEMSIS data standards and data dictionary can be found in the NEMSIS Data Dictionary (NEMSIS, 2021). In addition to providing a core set of data elements, states can opt to share additional data elements with NEMSIS according to their unique needs and interests. NEMSIS then makes these standardized state-level databases available for download directly from the NEMSIS website state map (NEMSIS, undated-b; click on your state). In addition to state-level databases, NEMSIS creates a series of state-level reports, including Tableau dashboards that visualize EMS data. Visualizations of trends over time and by sociodemographics are available through these dashboards. (See Summary 9 for an example from the national NEMSIS opioid dashboard.) In November 2021, NEMSIS was considering adding a dashboard focused on mental health and BH activations; however, plans have not been finalized. The rate of update for the datasets may vary by state, and these data are not typically provided at the county or subcounty level.
Option 2: State EMS agencies	To help facilitate data collection, NEMSIS has identified a key point of contact in each state for EMS data available in the NEMSIS state map (NEMSIS, undated-b), and data dictionaries for state EMS data are available online in the NEMSIS Data Dictionary (NEMSIS, 2021). Local PH agencies may also be able to request state-level EMS data directly from these key points of contact.
Option 3: Local EMS agencies	Local PH agencies could also obtain the data directly from local EMS agencies. This option will require data use agreements with each agency and will require the PH agency to acquire, standardize, clean, and compile the data on an ongoing basis. In some cases, data obtained directly from local EMS agencies may be timelier than data from NEMSIS. Different states, counties, and EMS agencies have different policies and processes affecting how quickly a record is closed after an EMS activation. For example, some agencies close the case and send the data immediately; others may take days or even weeks before the data are sent to the state EMS data manager, who then may wait even longer to gather data from all counties before submitting those data to NEMSIS. Therefore, in some communities, it may be as efficient or more so to receive data directly from the local EMS agency rather than waiting for the data to be received, checked, cleaned, and shared by the state EMS point of contact or by NEMSIS. However, larger geographic areas may have many EMS agencies, making it inefficient to collect and aggregate those data (e.g., Pennsylvania has 1,319 local EMS agencies).

According to our conversations with EMS data managers and users, the data are free for PH agencies. Standard statistical software (e.g., R, SAS, or STATA) or even Microsoft Excel can be used to analyze the data, depending on the complexity of your analyses.

The National Emergency Medical Services Information System Opioid Dashboard shows that the rate of Naloxone administration during EMS activations increased from 2014 to 2016, with projections of further increases through 2017. The rate was approximately 0.45% in January 2014, increased to nearly 0.80% by the end of 2016, with projections approaching 0.90% by the end of 2017.

Note: This figure shows the rate of naloxone administration during EMS activations from 2014 to 2016, with projections through 2017.

Source: Reproduced from NEMSIS, 2023.

Strengths and Limitations of These Data

Learn more about how this data source compares with others. Access a list of real-world examples of the use of this data source for BH surveillance in the PHE context. Access selected publications that have used this data source for monitoring and surveillance.