A. Types of Data

First, it is important to understand the types of relevant data. Their sources tend to fall into two categories: health and law enforcement. There are several sources of health data, including PDMPs, state coroners/medical examiners, death certificates, Medicaid and other claims, insurance providers, and state emergency services.^{Reference Murphy50} Within each of these sources lie potentially useful data points. Emergency services data, for example, could include the number of overdoses, treatment admissions, emergency department admissions, and the number of 911 and poison control calls,⁵¹ while death certificates may include information about the cause of death, the drugs involved, and demographic information, such as age, sex, marital status, race/ethnicity, and education level.⁵²

Meanwhile, law enforcement data may also be a source of data. Certain aspects of the crisis intersect with illegal activity, such as prosecutions in relation to overdose deaths.^{Reference Nedelman53} Additionally, police officers are often first responders to overdoses, and may be responsible for deploying naloxone.^{Reference Leger54} Data from law enforcement sources may include state forensic labs, drug arrests, drug seizures, crime or incident reports, and urinalysis results.⁵⁵ Past programs have sought to collect and use data from law enforcement sources in surveillance systems of drug use and trends. The now-defunct Arrestee Drug Abuse Monitoring Program, for example, interviewed volunteer arrestees who provided urine for testing and information from these interviews was used to “help monitor trends in drug use.”⁵⁶ Another example is the DEA's System to Retrieve Information from Drug Evidence, an former online database of drug seizure data.⁵⁷ Revitalization of these systems may yield insights for law enforcement in investigating and prosecuting, for example, suspected drug traffickers.

There are other sources of health and law enforcement data that may be useful. Successful data-driven efforts will therefore need to identify these sources, identify who collects the data, and determine ways to gain access before other issues — such as sharing and analysis — can be considered.

B. Data Reporting and Collection

Once potentially useful sources of data are identified, data sharing initiatives will need to ensure that the data is reported and collected properly. One major issue with existing data is the lack of standardization in how it is reported and stored.^{Reference Schneeweiss58} Standardization is necessary in “(1) defining what to collect, (2) deciding how to represent what is collected (by designating data types or terminologies), and (3) determining how to encode data for transmission.”⁵⁹ The lack of data standardization poses a significant hurdle to data sharing initiatives, as it makes it difficult to match datasets and derive insights through analysis. Ex-post standardization is an extra, unnecessary step that can be avoided by imposing common data standards.

Another issue is the form in which data is collected. Clinical data is collected in a variety of forms. In recent years, there has been a movement toward digitization through electronic health records (EHRs). EHRs can “contain a patient's medical history, diagnoses, medications, treatment plans, immunization dates, allergies, radiology images, and laboratory and test results.”⁶⁰ Because EHRs are in a digital format, they can be shared more easily with other health care providers and agencies,⁶¹ and therefore may more easily facilitate the sharing of data.

But EHR systems have not yet been adopted by all healthcare systems. The 2009 Health Information Technology for Economic and Clinical Health Act authorized the HHS Secretary to “invest in the infrastructure necessary to allow for and promote the electronic exchange and use of health information.”⁶² But as of 2016, only a little over sixty percent of all office-based physicians, twenty percent of nurse practitioners, and two percent of physician assistants have adopted EHR systems.⁶³ Adoption of digitization is ongoing, and until it has fully penetrated the market, some data will inevitably fall through the cracks. Meanwhile, non-digitized data can hamper data sharing and matching because datasets in non-digitized formats are unable to “talk” to each other or to datasets in digitized formats stored electronically.

Further, much of the data currently available is in data “silos.”⁶⁴ When data is “siloed,” that is, when it is kept or stored in different repositories that are unable to “talk” to each other, it is difficult to “match” data-sets and identify trends across data sources.^{Reference Cherico-Hsii65} As the Department of Commerce's Internet Policy Task Force noted, “[r]ather than coming up with an overall picture and then breaking it up into smaller pieces that mesh together, Congress has been sporadically creating individual pieces of ad hoc legislation.⁶⁶ This system of “sectoral data protection” causes “severe regulatory data gaps,” “encourag[es] different levels of protection,” and inhibits interagency cooperation.^{Reference Terry67} Funding programs may also be siloed, further complicating the picture.^{Reference Gasner68}

These gaps in the data are problematic because they can hamper the capability of relevant actors to act. Consider, for example, the numbers of children placed in foster care, which has reportedly exploded in recent years.^{Reference Lachman69} Some connect this increase to the opioid crisis, but it is difficult to make that connection directly when state child welfare databases do not collect information on what type of drug was a contributing factor to the child's removal.^{Reference Duncan70} According to one analyst, it is “hard to have enough substance abuse services for these families if we don't have an accurate idea of how many families are actually being affected by opioid abuse.”⁷¹ Thus, the lack of data in this context may lead to poorer outcomes or policies for children and families at risk. Or consider, for example, a possible data gap between law enforcement and hospitals with regards to fentanyl. Law enforcement may have access to data on the potency of heroin seized in drug arrests. That same data may be relevant to emergency rooms in treating potential overdose cases,⁷² but it is worthless to them unless that data is communicated between them. These problems are compounded as new dimensions to the crisis manifest, such as the proliferation of fentanyl.

For data-driven initiatives to work, uniform data standards needed to be implemented and enforced. And while an overhaul of the existing American data regulation system is beyond the scope of this article, coordination between existing systems will be necessary for data-driven initiatives to succeed.

C. Data Sharing

The issue of “siloing” of data is also related to existing problems with data sharing. First, as noted, potentially relevant data is held in different databases that are not connected, preventing the data from being analyzed together. Second, agencies that may be in the best position to respond only have one piece of the puzzle — the data they themselves hold — and are therefore unable to approach the crisis from a wide enough angle. The fact that data is held in both public health and law enforcement sources creates difficulties when agencies seek to share information across such categories.

The CDC defines data sharing as “[g]ranting certain individuals or organizations access to data that contain personally identifiable information …”⁷³ To be sure, not all data related to the opioid crisis necessarily contains personally identifiable information, and indeed, many examples of data sharing initiatives do not use personally identifiable information at all. But not only are there issues with data sharing within each category, e.g., between hospitals and public health agencies, there are additional issues with data sharing across categories, e.g., between public health agencies and police departments. Thus, agencies must be able to either share data with each other or contribute to a central data repository to which each agency has access.

Again, the call for coordination between public health agencies and law enforcement is not new—particularly in the pandemic or bioterrorism context.⁷⁴ Here, however, the situation is not wholly health-related, differentiating it from the opioid crisis context, which calls on data from both health and law enforcement sources, and thus potentially requires the involvement of a larger number of actors. Further, public health or syndromic surveillance systems developed for bioterrorist threats often concern cooperation for the purposes of prosecution. While criminal justice solutions, such as arrests and prosecutions of drug traffickers and prescribers, are being used by some areas of law enforcement,⁷⁵ this article is concerned with data sharing for non-prosecutorial reasons. Thus, other considerations must be made.

Lack of coordination presents additional difficulties for data sharing.⁷⁶ Researchers working within New York City public health agencies have noted that “state and local laws that authorize the collection of data may actually impede the ability of health departments to share, analyze, and make use of data across separate disease programs.”⁷⁷ The researchers explained the genesis of some of the major barriers to data sharing:

This situation described here applies not only to New York City, but is replicated across the country. Thus, city, state, and federal health codes will need to be examined to determine how they interact with each other and how they pose barriers to data sharing.

The phenomenon of information blocking also hampers data sharing. Information blocking “occur[s] when persons or entities knowingly and unreasonably interfere with the exchange or use of electronic health information.”⁷⁹ A 2015 report found a “substantial body of complaints and other anecdotal evidence” of information blocking.⁸⁰ Complaints directed at healthcare providers also allege that “privacy and security laws are cited in circumstances in which they do not in fact impose restrictions.”⁸¹ Information blocking therefore demonstrates the existence of social barriers preventing data sharing. It also indicates a need for educational efforts to inform data holders of what legal obligations they have. Establishing clear, uniform standards and expectations may assist in preventing barriers like information blocking.

Such barriers may also be overcome with the designation of a “lead” agency that is ultimately responsible for collecting data from other agencies. While the “lead” agency would take on additional costs as the central repository, having a designated central source a) may make data sharing easier and more streamlined, b) may help to solve collective action problems, and c) could provide useful models as it is replicated across agencies.

For example, Connecticut publishes a dataset — downloadable by the public and available in a visualization format through graphs and maps ⁸²— of each accidental death associated with an overdose and contains information on the victim's demographics, cause of death, substance detected, and the manner of death.⁸³ The data is provided from investigations by the Office of the Chief Medical Examiner.⁸⁴ The data, however, is published not by the Office, but rather through Connecticut's Open Data Portal, which was launched by an executive order in 2014.⁸⁵ That executive order also established a state Chief Data Officer.⁸⁶ In this example, the Chief Data Officer and the Open Data Portal serve as the lead agency, while agencies like the Chief Medical Examiner are cooperating agencies. Such a system may help overcome barriers to data sharing, although this method does depend on the cooperation of state agencies.

Of course, the designation of a “lead” agency is only one potential solution. Further, the “lead” agency model serves mainly to address logistical issues in developing data-driven initiatives. There are further issues, such as legal barriers to data sharing that must be addressed. “Lead” agencies could take the lead on identifying and dealing with these barriers, although removing them will likely require coordination with state legislatures as well.

D. Benefits from Improved Data Collection and Sharing

Some have already noted the potential benefits of improved data collection and sharing.^{Reference Olson and Downey87} Consider, for example, the use of emergency medical services (EMS) data. One article notes that EMS naloxone administration data was used to observe trends on a population-wide basis, “such as increased use [of opioids] in the summer, weekends, and late afternoon.”^{Reference Garza and Dyer88} Another study acknowledges that “EMS records may include objective indicators of opioid involvement; information about circumstances of overdoses, including time; medical care provided; and characteristics of persons who overdose …”^{Reference Knowlton89} Indeed, EMS data has been used as proxies for suspected overdoses by researchers.^{Reference Merchant90}

Cincinnati's Heroin Tracker⁹¹ is instructive in combining publicly available EMS data and visualization tools. It maps overdose incidents neighborhood-by-neighborhood, as well as month-by-month and day-by-day. The tracker is updated daily and allows users to search for incidents down to the individual street.⁹² It has helped Cincinnati see “trends and ‘geographic hotspots,’ helping public safety identify key areas for strategically deploying personnel and medical resources.”⁹³ The Heroin Tracker has reportedly enabled Cincinnati to “strategically deploy roving medic units to [high usage locations] …” with the intention of directing increased resources to specific areas in order to be better positioned to respond to overdoses.^{Reference Burry94} For example, say the Heroin Tracker identifies a public park as a “hotspot,” particularly on summer weekends. Officials could use this insight to deploy more patrols or EMS teams to that area during those times, or to increase access to Naloxone in that area. If the Heroin Tracker achieves its intended aim, Cincinnati should see less overdoses in the identified areas over time. According to Cincinnati's Chief Data Officer, Brandon Crowley, this is already happening: “Every day, dozens of public-safety and health-department hospitals … check the Heroin Tracker. It helps them plan more targeted patrols in neighborhoods with the most serious overdoses hotspots and anticipate overdose spikes based on previous trends.”^{Reference Crowley95} And because both the Heroin Tracker and the underlying datasets are publicly available, Cincinnati-area nonprofits and charities can also use the data.⁹⁶

As another example, improved data collection and sharing can help identify “outbreaks” of substances laced with fentanyl. In June 2016, the Yale-New Haven Hospital in Connecticut faced twelve overdoses in less than eight hours after a white powder sold as cocaine — and later identified to have been laced with fentanyl — hit the streets.^{Reference Tomassoni97} A coordinated effort between the Office of Emergency Management, Department of Public Health, DEA, local police, Connecticut Poison Control, and Mayor's Office led to immediate action as information came in about the outbreak in real time. Such action included advice to EMS crews to increase the dose of Naloxone for suspected overdoses and health alerts warning the public that “a high potency opioid marketed as cocaine [was] causing deaths in the region.”⁹⁸ This coordination may have helped “to confine the outbreak, save patient lives where possible, alert the public, and gather additional information.”⁹⁹

Efforts like those in Cincinnati and New Haven are certainly promising, although it is unclear whether such a project could be implemented on a larger scale. The Heroin Tracker is an example of how relevant actors' interests must align, along with institutional support from the government, for new initiatives to develop and thrive. Nevertheless, it does demonstrate a new way of using data visualization tools. Other cities have followed suit with either similar maps, such as Seattle's overdose “heat map,”¹⁰⁰ or by publicly publishing datasets, like the Northampton Massachusetts Police Department.¹⁰¹ These efforts indicate the potential for the implementation of local and publicly available overdose trackers and datasets.

But Cincinnati's Heroin Tracker relies on only one source of data: emergency response data. Thus, while it may be instructive from a data collection standpoint, it may not necessarily be useful from a data sharing standpoint. Moreover, EMS data and records are not the only useful information available.

Maryland's efforts illustrate this point. In June 2014, Governor Martin O'Malley issued an executive order establishing an Overdose Prevention Council, consisting of the Department of Health and Mental Hygiene, the Department of State Police, the Department of Public Safety and Correctional Services, the Department of Juvenile Services, and the Institute for Emergency Medical Services Systems.¹⁰² As part of the Council, the agencies are to “seek opportunities to share data with one another….”¹⁰³ The executive order, however, provided no guidance on how to share data, for what purpose, or how to deal with state data sharing regulations.

The Council sought to overcome these difficulties through a six-step method: “identify available data, set priorities, identify a lead agency, address legal concerns, draft memoranda of understanding (MOUs), and share data.”¹⁰⁴ Using this method, the Council prioritized identified sets of data for “matching,” such as matching overdose records with “prior hospitalizations and emergency department visits for overdose …”¹⁰⁵ The Council also determined that data privacy and confidentiality laws would “not prohibit data sharing if the data were exchanged in the limited manner as proposed by the Council, and with appropriate MOUs in place.”¹⁰⁶ For example, the Council concluded that overdose registry data that originated from autopsy records was not in violation of privacy laws.¹⁰⁷ The Council also worked with attorneys to draft MOUs that outlined responsibilities for data use and data security and confidentiality between the agencies.¹⁰⁸ The resulting matches were received by the Department of Health and Mental Hygiene, who found that analyses using the data resulted in “important insights and policy initiatives.”¹⁰⁹ Maryland's efforts demonstrate a potential success in matching and sharing datasets, and its use of MOUs to address privacy and security concerns may prove instructive.

The Cincinnati, New Haven, and Maryland examples also gesture towards other, more theoretical ways in which improved data sharing and collection could be used in public health surveillance systems. For instance, if matching datasets from disparate agencies can help identify particularly vulnerable populations in a city — whether demographically, by location, etc. — it may help policymakers or researchers design programs to reach those populations, such as clinics, treatment services, information campaigns, or increased deployment of naloxone. Maryland has already implemented this to some extent. Insights derived from Maryland's efforts to match datasets found that individuals recently released from correctional facilities were at a higher risk of overdose immediately following release compared to periods of time three months to a year later.¹¹⁰ This insight had an immediate impact on policy, as the state corrections agency “became a leading partner in distributing information on overdose prevention and treatment programs to inmates at the time of release,” and is currently exploring the possibility of providing vulnerable individuals with naloxone upon release.¹¹¹

Additionally, public health and syndromic surveil-lance systems may also help predict patterns or trends. Some researchers have already pointed to this possibility by looking at past datasets and known fentanyl outbreaks to determine whether a real-time analysis of data leading up to the outbreak would have helped predict it. One study, for instance, used a five-year dataset of calls made to the Illinois Poison Center and conducted a prospective analysis to “determine when [a 2006 Illinois fentanyl] outbreak would have been identified if the data were being analyzed in real time.”^{Reference Friedman112} The study found that the outbreak would have been identified one month before the first reports were made to the CDC.¹¹³ The ability to predict or track particularly potent batches of heroin could help both law enforcement and hospitals contain and respond to overdoses.

Improved data collection and sharing may also help to identify who may be at risk for developing substance use disorders in order to target them for treatment, such as medically assisted treatment or counseling. Such screening tools have been tested in the past. For example, one study used certain behaviors and risk factors, such as requests for early refills, the type of pain involved, or resistance to therapy changes or alternative therapy, as part of a screening tool to identify which patients were at a higher risk of abusing opioids.^{Reference Webster and Webster114} Another study applied identified risk factors, such as medical, family, and social history, to health insurance claims data, enrollment data, and PDMP data as part of a risk analysis to identify patients that may be at risk of developing a substance use disorder.^{Reference Ciesielski115} A third found, in looking at the electronic health records of patients admitted to hospitals for an overdose, that healthcare costs for a patient “often spike just before an overdose.”^{Reference Williams116} Such tools, if implemented more widely, and using data from different sources, such as a patient's medical history, the state PDMP, and insurance claims data, could help physicians identify which of their patients may be at risk of developing a substance use disorder and to adjust monitoring systems or treatment plans accordingly.

It is also possible that improved data collection and sharing could help individuals with substance use disorders already undergoing treatment. Better aggregation and analysis of data on a population-wide scale, or within specific populations or communities, could help researchers identify which types of treatment help particular individuals. For example, integrating both healthcare data and law enforcement data for incarcerated individuals with substance use disorders could help identify risks for that specific population, determine whether their unique needs are being met, evaluate whether existing treatment programs are adequate, and reveal insights into what treatment may work best for them. Similarly, data analysis could be used to make treatment policy decisions on a more general level. For example, the Healthcare Fraud Prevention Partnership proposed using their cross-payer data to “develop and implement data monitoring algorithms to identify geographic locations or payer networks with comparative shortages of MAT providers and/or MAT utilization.”¹¹⁷ These kinds of efforts could help officials ensure that treatment options are reaching the neediest populations.

Other benefits may accrue from improved data sharing, many of them practical in nature. For instance, the centralization of data may be useful, if only to “minimize [requests for information] among numerous data sources.”¹¹⁸ Data sharing may also help address issues with underreporting. It may also have direct impacts on the opioid crisis by identifying excessive opioid prescribing and doctor shopping for more targeted investigations by law enforcement.¹¹⁹

Finally, simply having more data collected and shared about the opioid crisis may help forge the political will necessary to act. A lack of political will may very well be a barrier to implementing important policies. Understanding the human and economic impact of the opioid crisis may either spur political actors to act on their own initiative or give the public the ammunition necessary to put pressure on public officials themselves.

Of course, it will be difficult to measure the impact of sharing data and policy changes immediately, and it is difficult to assess the effectiveness of currently-implemented data sharing initiatives. Some metrics are easy to measure. For instance, if Cincinnati's Heroin Tracker is effective, we should see less overdoses. But other success criteria are less clear. At this stage, many of these benefits are merely hypothetical. It is important to note that data sharing initiatives are not cheap and bring not only monetary costs in their implementation, but also in the redirection of resources. Further, not all data sharing initiatives are created equal. Some may cost far more than they benefit. Thus, a cost-benefit analysis is a key concern and must be a critical piece of any campaign for improved data sharing and collection.

A. Current Federal Regulations

As discussed, there are no comprehensive federal regulations governing the collection of data. Instead, there is a sectoral framework, coupled with a system of rules and guidelines issued by different governmental agencies. This complex web of legislation and regulation contributes to difficulties in determining what kind of information can be collected and shared.

The major piece of federal legislation at issue is the Privacy Rule of the Health Insurance Portability and Accountability Act (HIPAA). HIPAA applies to protected health information.¹²⁰ Such information may not be disclosed by a covered entity¹²¹ without valid authorization.¹²² HIPAA does permit disclosure by a covered entity to “[a] public health authority that is authorized by law to collect or receive such information for the purpose of preventing or controlling disease, injury, or disability.”¹²³ Another provision of HIPAA also permits a covered entity to disclose protected health information if it “in good faith, believes the use or disclosure: [i]s necessary to prevent or lessen a serious and imminent threat to the health or safety of a person or the public; and [i]s to a person or person reasonably able to prevent or lessen the threat.”¹²⁴

One regulation for the records of patients with substance use disorders, 42 C.F.R. Part 2, “prohibit[s] the disclosure and use of patient records unless certain circumstances exist.”¹²⁵ Patients can sign consent forms allowing disclosure, but such consent can also be revoked.¹²⁶ The regulation contains exceptions for notification to medical personnel in an emergency or to law enforcement if there is an immediate threat to the health or safety of an individual, but these exceptions are limited.¹²⁷ While some have praised 42 C.F.R. Part 2 for its protection of individuals with substance use disorders before HIPAA existed, it may now impede the delivery of quality medical care to those same individuals.^{Reference Butler128}

Thus, data protections concerning the sharing of identifying protected health information, on the federal level may prevent extensive use of such information in public health or syndromic surveillance systems. However, it may be possible to use de-identified information. HIPAA's Privacy Rule “permits a covered entity or its business associate to create information that is not individual identifiable by following the deidentification standard and implementation specifications in § 164.514(a)-(b).”¹²⁹ Covered health entities have two options to create de-identified data: “1) a formal determination by a qualified expert; or 2) the removal of specified individual identifiers as well as absence of actual knowledge by the covered entity that the remaining information could be used alone or in combination with other information to identify the individual.”¹³⁰ De-identification could therefore both address privacy concerns and provide a means to share protected health information.

But regulations concerning protected health information are only one side of the story. Many of the public health or syndromic surveillance systems currently used rely on emergency response data. Public health data can also be matched with emergency response data. Many of the systems discussed do not rely on protected health information, although arguments could certainly be made that access to such information would provide more targeted responses. Further, sharing protected health information may be more productive in combating the opioid crisis long-term, rather than addressing short-term objectives and goals, such as where to direct emergency resources for overdoses. Still, while existing regulations initially appear to be a barrier, multiple sources of data exist that can be harnessed effectively.

B. Current State Regulations

State regulations regarding health information differ both in their attempts to conform to HIPAA and their own regulatory schemes.^{Reference Habte131} On a state level, public health agencies are sometimes exempt from patient consent and authorization for data “used in the course of fulfilling public health functions.”¹³² Further, state governors may have emergency powers to issue “orders that mandate reporting of health information needed to respond to a public health threat.”¹³³ Certain states also permit health care providers to send clinical information to public health agencies for “studies that address a wide range of public health concerns …”¹³⁴ When it comes to emergency response data, many departments are permitted to report information to state public health agencies, while some states actually require hospital emergency departments to report.¹³⁵ Ultimately, many of the issues discussed in the previous section replicate themselves on the state level; albeit in a more complicated fashion given the multitude of actors involved.

PDMPs illustrate differences between states in how data-based programs operate. For instance, states differ in how long PDMP data can be retained.¹³⁶ Arkansas, for instance, requires that data be maintained for three years,¹³⁷ while Pennsylvania allows for the removal of patient identifying information that is more than seven years old, along with complicated rules for extensions in the event of requests made by law enforcement or licensing boards.¹³⁸ Other states do not appear to have statutory data retention timelines.¹³⁹

Similarly, states differ when it comes to authorized users and the conditions of access.¹⁴⁰ Colorado requires users to sign a written agreement that “ensures compliance” with the statute.¹⁴¹ Oklahoma, on the other hand, describes the disclosure of certain statistical information that is not prevented.¹⁴² Other states have different standards for disclosure to authorized users.¹⁴³

Finally, states differ in how their statutes approach interstate sharing of PDMP data.¹⁴⁴ Connecticut, for example, permits it for the purposes of licensing or criminal investigations.¹⁴⁵ States differ between what agencies or officials in other states may receive information from PDMPs, with certain caveats, such as data use agreements.¹⁴⁶ These issues with PDMPs illuminate many of the problems that arise when states have their own data sharing regulations.

Although a full survey of both federal and state data sharing laws is beyond the scope of this article, these differences exist and must be addressed if state public health and syndromic surveillance systems are to coordinate. Data-sharing initiatives on both a state and federal level will have to carefully consider the existing legislative and regulatory environment to ensure that their programs are effective while still complying with the law.

C. Legislative Proposals

This article has thus far discussed existing barriers to data sharing and collection and highlighted a few key examples of current data sharing initiatives used in combating the opioid crisis. It also discussed potential benefits that may result from existing initiatives, as well as other ways in which data could be used to yield important insights. The article now turns to existing calls for reform and makes a few recommendations for the relevant actors to keep in mind. Finally, the article considers objections to improved data collection and sharing — such as privacy interests, data security, and the potential cost of data sharing initiatives.

Some states and policy groups have already recognized that there are existing issues with data collection and sharing, and several have called for changes in the data sharing regulations as a method of combating the opioid crisis. The National Governor's Association recommended the establishment of a “collaborative information sharing environment that breaks down silos across state agencies,”¹⁴⁷ while the Safe States Alliance called for states to “establish and enhance capacity to conduct opioid drug overdose surveillance and prescription opioid monitoring” and for local public health agencies to “[a]ccess and use multiple sources of data and tools to monitor and describe the opioid overdose epidemic.”¹⁴⁸ Meanwhile, Washington passed a bill in 2017 in which the legislature found that “public health data is critical to the [D]epartment [of Health]'s ability to respond to emerging health threats.”¹⁴⁹

Further, some states have already made efforts to improve data collection and sharing. In April 2017, West Virginia passed a bill creating an Office of Drug Control Policy.¹⁵⁰ It requires the Office to “[m]onitor, coordinate and oversee the collection of data and issues related to drug [] access [and] substance use disorder policies,” to encourage coordination among agencies, organizations, and service providers, and to “[d]evelop and implement a program … to collect data on fatal and nonfatal drug overdoses” from different agencies.¹⁵¹ The Office is also required to establish and maintain a “central repository”¹⁵² that contains a wide range of information that pharmacies, health care providers, medical examiners, emergency response providers, and law enforcement agencies are required to report, including “emergency medical or law-enforcement response[s] to a suspected or reported overdose … [m]edical treatment for an overdose; [t]he dispensation or provision of an opioid antagonist; and [d] eath attributed to overdose or ‘drug poisoning.’”¹⁵³

Such systems also have the potential to be useful between states and in conjunction with federal agencies. Already, policymakers have called for greater interstate data sharing efforts. The President's Commission on Combating Drug Addiction and the Opioid Crisis called on President Trump to “[p]rovide federal funding and technical support to states to enhance interstate data sharing among state-based prescription drug monitoring programs” and to “[e]nsure federal health care systems … participate in state-based data sharing.”¹⁵⁴ Additionally, some states have attempted to share data across state borders, particularly with data collected by PDMPs. Federal policymakers and legislators should devise plans to encourage the implementation of such systems and to facilitate sharing between them, while keeping in mind important concerns regarding privacy and security.

There may be benefits to focusing new data sharing efforts on the federal — rather than state — level. Federal agencies may have more funding or be better situated to respond to nationwide issues, as opposed to the local responses of cities and states. Further, aggregating data nationwide will allow for broader analyses of the opioid crisis. Involving the federal government can also provide incentives for states to cooperate in interstate data sharing. Implementing public health and syndromic surveillance systems on a nationwide level may also help in creating uniform data standards, protecting privacy interests, and securing data from breaches. Finally, federal agencies may find it easier to coordinate data sharing between states, rather than relying on individual states to broker data sharing agreements. Thus, federal agencies and policymakers are a key component of using data to fight the opioid crisis.

What models of public health or syndromic surveil-lance should be used? The West Virginia example, as well as the Maryland example discussed earlier, may prove instructive for future data sharing initiatives on both the state and federal level. Both examples involve the designation of a “lead agency” that acts as a central repository for the collection of data relating to the opioid crisis. The Maryland Council's six step method could be implemented across state and federal governments to assist agencies in deriving meaningful analysis through the “matching” of different datasets. And along with the identification of a lead agency for a central repository, state and federal governments must implement data standardization procedures for collection among their various agencies to ensure that any data collected can be used effectively. They must refine and implement de-identification procedures for any patient protected health information. HIPAA's Privacy Rule, as well as guidance from government agencies, can be helpful here. However, as noted earlier, there are a few potential obstacles to sharing data between states and the federal government. Moreover, implementing public health and syndromic surveillance systems on a nationwide level will require the alignment of political interests and actors, which may prove difficult in the current political climate.

Still, this article ultimately argues that there are important potential benefits to improved and increased data collection and sharing as well as the development of new data sharing initiatives as tools to combat the opioid crisis. State and federal officials should carefully examine their existing datasets and sources, devise new standards for data collection and the format it is stored in, explore ways of creating central data repositories, and identify existing barriers to sharing data across and between local, state, and federal agencies. The sheer human and economic cost of the opioid crisis necessitates assurance that the correct data is being collected so that we can truly understand the scope of the opioid crisis and be better positioned to respond. The individual efforts by cities and states discussed throughout this article are positive indicators of the potential role data sharing initiatives could play in combating the opioid crisis and should signify to policymakers the importance of ensuring they can work effectively.

D. Potential Objections

This section briefly addresses objections to expanding data collection, sharing, and analysis. Although devising thorough solutions to these problems are beyond the scope of this article, considerations and awareness of these problems are important for policymakers seeking to improve data collection and sharing and to implement data sharing initiatives.