Part 2’s 2019 draft made a record’s origin the thing to track
SAMHSA’s August 2019 proposal drew the line between a protected addiction-treatment record and an ordinary chart note by where the record came from. Health systems inherited a data lineage problem that outlasted every later rewrite of the rule.
Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.
Two-minute summaryThe dek and the first sentence of each section, quoted from the article
SAMHSA’s August 2019 proposal drew the line between a protected addiction-treatment record and an ordinary chart note by where the record came from. Health systems inherited a data lineage problem that outlasted every later rewrite of the rule.
- The week On August 26, 2019, the Substance Abuse and Mental Health Services Administration (SAMHSA), part of the Department of Health and Human Services, published a proposed rule in the Federal Register.
- Origin, not content Part 2 has always applied to specialized programs.
- Segregation without a definition The proposal’s safe path for recipients was segregation.
- Where origin gets lost A provenance rule is only as good as the provenance data.
- Flows that create new holders Much of the proposal was about letting Part 2 information go to more places, for good reasons.
- Building lineage that lasts The practical response to a provenance rule is data lineage: knowing, for any record or field, where it came from and what rules attach to it.
- With hindsight SAMHSA finalized the proposal, with some changes, in a final rule published on July 15, 2020 and effective August 14, 2020 (85 FR 42986).
The week
On August 26, 2019, the Substance Abuse and Mental Health Services Administration (SAMHSA), part of the Department of Health and Human Services, published a proposed rule in the Federal Register. It proposed changes to the Confidentiality of Substance Use Disorder Patient Records regulations, 42 CFR Part 2 (84 FR 44568). Comments were due by October 25, 2019. A second, shorter proposal appeared the same day (84 FR 44566). It would clarify that a court may authorize disclosure of a patient’s confidential communications to a Part 2 program when needed to investigate or prosecute an extremely serious crime, even one the patient was not alleged to have committed. HHS said the phrase limiting such orders to crimes allegedly committed by the patient had been added in error in 2017, and could hinder federal enforcement against rogue doctors and pill mills.
Part 2 implements a section of the Public Health Service Act, 42 U.S.C. 290dd-2. The proposal recounts that the regulations were first issued in 1975, and amended in 1987, 1995, 2017 and 2018. It describes their purpose by citing a 1983 proposed rule: patients receiving treatment in a Part 2 program should not be made more vulnerable to investigation or prosecution because of their association with a treatment program than they would be if they had not sought treatment.
The proposal tied its timing to the opioid crisis, citing data from the Centers for Disease Control and Prevention on rising overdose deaths. It said the crisis had increased pressure on the treatment system and highlighted the need to update the rules so that care could be coordinated, while keeping confidentiality protections for people in treatment.
Its central change concerned which records Part 2 covers. SAMHSA proposed to state in the regulation that a provider not subject to Part 2 does not, merely by recording information about a substance use disorder and its treatment, make that record subject to Part 2. That held provided the provider segregates any specific records received from a Part 2 program. A matching change to the definition of “records” said information a Part 2 program conveys orally to a non-Part 2 provider for treatment, with the patient’s consent, does not become a Part 2 record just because that provider writes it down.
SAMHSA said the existing text did not compel non-Part 2 providers’ own records to be covered, but that experience had shown the need for clearer regulatory language. A 2017 rule had extended Part 2’s restrictions to “lawful holders” who receive records from a Part 2 program, so that records stayed protected through a chain of redisclosures. SAMHSA said some confusion remained among providers about what information collected by non-Part 2 entities was covered, and that the 2017 change had not been meant to make their records subject to Part 2.
The proposal also included other changes. Patients could consent to disclosure to an entity without naming a specific individual there. Opioid treatment programs could report dispensing and prescribing data to state prescription drug monitoring programs, subject to patient consent. Non-opioid treatment providers could query central registries. Patient information could be disclosed to another Part 2 program or treatment provider during state or federally declared natural and major disasters. And 17 types of payment and health care operations activities would be listed expressly as permitted with consent. Other proposals concerned research, audits and evaluations, and undercover agents.
Part 2 is easy to read as a consent rule: who must agree, to what, and how often. The 2019 proposal changed consent rules too. But its most consequential idea was about data, not consent. It drew the line between a Part 2 record and an ordinary medical record by where the record came from, not by what it said.
Under the proposal, a primary care physician’s note about a patient’s opioid use disorder was an ordinary medical record, covered by HIPAA. A discharge summary that same physician received from a Part 2 treatment program was a Part 2 record, and remained one in the physician’s hands. The words could be nearly identical. The difference lay in provenance. Our argument is that the 2019 proposal turned record origin into a compliance attribute that health systems had to capture, keep and act on. That requirement has outlasted every later change to the rule.
The analysis has five parts:
- Origin, not content: how the proposal drew the boundary, and why it chose provenance.
- Segregation without a definition: what SAMHSA asked recipients to do, and what it declined to specify.
- Where origin gets lost: the ordinary workflows that strip provenance from a record.
- Flows that create new holders: how the proposal’s new disclosure paths multiplied the places a Part 2 record could go.
- Building lineage that lasts: what a health system can build, what it costs, and the case against bothering.
The first two describe the rule. The third and fourth describe how real data moves against it. The fifth asks what to do.
Origin, not content
Part 2 has always applied to specialized programs. The proposal recalls that a 1987 rule limited coverage to programs that hold themselves out as providing, and actually provide, diagnosis, treatment or referral for substance use disorders. General hospital emergency departments and medical or surgical wards were left out. The aim was to protect patients who sought specialized treatment without burdening general medical care.
The 2017 extension
In 2017, SAMHSA extended Part 2’s restrictions to recipients. Anyone who received a Part 2 record, from the program or from another lawful holder, had to respect the restrictions on redisclosure. The proposal explains the reason: to keep records from losing their protection as they moved downstream, so that patients would not be deterred from seeking treatment.
The extension worked for its purpose but blurred the boundary. If a primary care provider received a Part 2 record, discussed it with the patient and wrote a note, was the note now a Part 2 record? The proposal described significant confusion among non-Part 2 providers on this point, and fear of inadvertently violating Part 2. It said the result was the potential for impaired coordination of care, which could be life threatening for a patient with an opioid use disorder.
The proposal’s answer
SAMHSA’s answer was to anchor coverage to origin. Records created by a non-Part 2 provider in its own patient encounters would not be Part 2 records, unless records received from a Part 2 program were incorporated into them. Records received from a Part 2 program stayed Part 2 records in the recipient’s hands. The proposal noted that a non-Part 2 provider’s own record would still be subject to the HIPAA Privacy Rule.
This is a provenance rule. The same clinical fact, that a patient is in treatment for opioid use disorder, can sit in two records under two regimes. What decides the regime is the path the record took to get there.
Why provenance and not content
A content rule would have been simpler to state: any record mentioning a substance use disorder is protected. But it would have pulled every emergency department note, every primary care problem list and every pharmacy record that mentioned addiction into Part 2. That would have reversed the 1987 limit and multiplied the burden the proposal was trying to ease.
A provenance rule keeps the 1987 limit and protects what patients disclose to specialized programs. Its cost is that it asks every recipient to know where its records came from. That is a data management requirement, not a clinical one.
Provenance through a chain of holders
The 2017 concept of a “lawful holder” made provenance transitive. A record received from a Part 2 program stays protected when the recipient passes it on, and again when the next recipient does. The proposal described the effect as ensuring that records first created by a Part 2 program remain protected throughout a chain of redisclosures, even to a downstream recipient that is not itself a Part 2 program.
A chain is only as strong as its weakest link’s records management. The first recipient may be a health system with a capable EHR. The third may be a small practice, a benefits office or a researcher’s spreadsheet. Each link must know what it holds, and the only thing that tells it is the information that travels with the record.
The HIPAA layer underneath
It would be wrong to read the proposal as removing protection from non-Part 2 records. The proposal said a non-Part 2 provider’s own record remains subject to the HIPAA Privacy Rule. The choice is between two privacy regimes, with different rules on disclosure, consent and use in legal proceedings, not between protected and unprotected data.
That matters for design. A health system does not need to treat substance use information in its own notes as public. It needs to know which of two rule sets applies to each record, and its systems need to apply the right one.
Inside a combined health system
A health system that runs its own Part 2 program faces the provenance question inside its own walls. On our reading, the program’s records do not lose their status because they move to another department of the same organization. The current regulation applies Part 2’s restrictions to persons with direct administrative control over a Part 2 program, for information the program communicates to them.
So the boundary runs through the organization, not around it. A referral from the system’s addiction medicine unit to its own primary care clinic is a flow from a Part 2 source, and the receiving clinic needs the same labeling as if the record had come from outside.
Who this touches
The provenance rule matters most to organizations that sit between Part 2 programs and general care. A health system that runs a Part 2 program alongside general medical services holds both kinds of record. An ambulatory network that refers patients to community treatment programs receives Part 2 records. A health information exchange carries them. A health plan, as a third-party payer, receives them for payment.
Each of these organizations needs to know, for any record about substance use disorder, whether it was created internally or received from a Part 2 source. A standalone Part 2 program has a simpler problem, since nearly all its records are Part 2 records. The provenance question is hardest for the organizations in the middle.
Segregation without a definition
The proposal’s safe path for recipients was segregation. A non-Part 2 provider that segregated Part 2 records it received could keep its own records outside Part 2. SAMHSA described what segregation might mean and then deliberately stopped short of defining it.
What SAMHSA described
For paper records or an email attachment, the proposal said segregation could mean physically holding the records apart. For electronic records, it could mean logical segmentation within the electronic health record system. SAMHSA said segregation did not require a separate server, and that it did not intend the rule to lead to separate servers or health IT systems for Part 2 documents.
The proposal named one technical path. It said segmentation could be done electronically using an EHR platform compliant with the Data Segmentation for Privacy standard, known as DS4P. It said SAMHSA had led use-case discussions on implementing DS4P and contributed to a consent implementation guide built on the FHIR standard, called Consent2Share.
What SAMHSA declined to do
The proposal said plainly that it imposed no new requirement for data segmentation as a practice and set no new standards for EHR technology. SAMHSA considered defining “segmented” and “segmentation” and decided against it. It said a formal definition might have unforeseen technical consequences for EHR and health information exchange systems, and that providers with different capabilities should be able to choose the least burdensome approach.
That restraint had a consequence. Without a definition, each organization had to decide what segregation meant for its own systems and be able to defend that decision. The rule left the design to the organization, so the organization owned the risk that the design would not hold.
Paper that becomes electronic
The proposal called physical separation quite feasible for a received paper record or an email attachment. A health system may, however, scan incoming paper and faxes into the electronic record. At that moment a physically separate document becomes one more image in the chart.
The scanning step is therefore where paper segregation either turns into an electronic label or disappears. Scanning staff need a document type for received Part 2 records, a list of senders that qualify, and a check that the label was applied. Without those, a practice that segregated paper carefully loses the separation the day it goes digital.
Segregation is metadata
Whatever form it takes, electronic segregation is a matter of metadata. A record has to carry, or be linked to, a marker that says it came from a Part 2 source and is subject to redisclosure limits. Systems that store, display, exchange and analyze the record then have to read the marker and act on it.
That is a different problem from access control by department or role. A role-based model asks who may see a record. A provenance model asks what may be done with it once seen, and whether any copy or derivative inherits the restriction. The two call for different controls. The second requires the marker to survive every place the data goes.
Choosing a segregation design
With no definition to follow, organizations had several designs available. One is a dedicated document type or folder in the EHR for received Part 2 records, with restricted access and a banner on display. Another is an external document store linked from the chart, so the records never enter the main clinical data flows. A third is labeling at the data level, using security labels of the kind DS4P describes, so that each document or data element carries its own restriction.
Each has trade-offs. A dedicated folder is simple but relies on staff filing correctly and on every downstream system respecting the folder. An external store keeps records out of analytics but makes them harder for clinicians to use, which works against the coordination the proposal wanted. Data-level labels are the most precise but depend on every system in the chain understanding them.
What to ask of the EHR and exchange vendors
Because the proposal set no standard, the capability depended on vendors. A health system evaluating its position needed answers on a short list of points:
- Whether received documents can carry a source attribute and a restriction label that persists when copied.
- Whether interfaces can recognize and label documents from known Part 2 programs on arrival.
- Whether outbound exchange messages can include the label and the required redisclosure notice.
- Whether data extracted to warehouses and reporting tools keeps the label at the document or element level.
- Whether access to labeled records can be logged separately for audit.
The answers decide how much of the provenance burden falls on software and how much on people.
Admitting the gap
The proposal acknowledged that SAMHSA had recognized, in its 2017 rule, significant challenges in segmenting substance use disorder records within some EHR systems. It encouraged the development of DS4P standards and the adoption of EHRs that meet them.
Encouragement is not a requirement, and a standard that exists is not a feature that a given health system has switched on. A health system reading the proposal had to ask whether its own EHR, exchange interfaces and data warehouse could carry a provenance marker, and what it would do where they could not.
Where origin gets lost
A provenance rule is only as good as the provenance data. In ordinary clinical and analytical work, records are copied, summarized, merged and transformed. Each step is a chance for the marker to fall off.
Copy-forward and incorporation
The proposal’s boundary depended on whether a non-Part 2 provider “incorporated” a received Part 2 record into its own. Electronic records make incorporation easy. A clinician can pull text from an outside document into a note, copy a medication list or import a problem list with one action.
SAMHSA addressed the concern directly. It said the intent was to let a non-Part 2 provider discuss information with a patient and write her own records, not to encourage extensive transcription from a Part 2 record without a clinical purpose. For a health system, that distinction has to be built into workflow. Tools that import outside documents need to carry the provenance marker into whatever they create. Otherwise a protected record becomes an ordinary note one paste at a time.
The oral exception
The proposed change to the definition of “records” created a deliberate exception. Information a Part 2 program conveyed orally to a non-Part 2 provider for treatment, with patient consent, would not become a Part 2 record just because the provider wrote it down. Records otherwise transmitted, such as documents, would keep their status.
The exception follows the provenance logic but adds a branch. A faxed document from the program is a Part 2 record in the recipient’s hands, and a note that transcribes from it risks incorporating it. A note written from a phone call does not become a Part 2 record by being written. So the channel can matter, and a health system may need to show how information arrived as well as where from. That detail is easy to capture at the moment of the call and hard to reconstruct later.
Analytics and data warehouses
Clinical systems are only part of the picture. Health systems copy EHR data into warehouses and analytics platforms for quality measurement, population health and operations. Those copies may be built for analysis, with documents parsed into structured fields and joined to claims and other sources.
If the provenance marker does not travel into the warehouse, every downstream dataset that includes Part 2 data loses its restriction in practice. A risk model trained on such data, a dashboard that shows substance use treatment by clinic, or an extract sent to a vendor may all carry Part 2 information without anyone knowing. The proposal did not discuss data warehouses or analytics platforms by name. The provenance logic applies to them all the same.
Health information exchanges
Exchanges sit at the center of the provenance problem. They receive records from many sources and serve them to many recipients, sometimes by assembling a combined view of a patient from several organizations. The proposal said special instructions would apply to consents for disclosure to information exchanges.
An exchange that carries Part 2 records has to keep the source label attached as records are stored, combined and served. A combined medication list or problem list that merges entries from a Part 2 program and a primary care practice is a derivative record. Unless each entry keeps its origin, the combined view cannot be handled correctly by the recipient.
Summaries and derived text
Any process that produces new text from existing records raises the same question as transcription. A discharge summary, a referral letter, a care plan or a machine-generated summary may draw on a received Part 2 document. The proposal’s concern about extensive transcription without a clinical purpose applies to these derived documents as well.
Today, tools that summarize charts automatically make this sharper, because they can draw on every document in a record without a person choosing what to include. A summarizer that reads received Part 2 documents should either exclude them or mark its output as derived from them. The 2019 proposal did not address such tools. Its provenance logic tells an organization what to ask of them.
Copies in patients’ hands and beyond
Patients can obtain their own records and pass them on, for example through apps. The current regulation states that records obtained through patient access remain subject to the restriction on use in proceedings against the patient. A copy that a patient shares does not shed that protection, even though the patient may share it freely.
For a health system, the practical point is that exports to patients and their apps should carry the same notice and labeling as other disclosures, so that whoever receives the copy next can see what it is.
Communications outside the record
The proposal also addressed a narrower provenance problem. Part 2 defines records to include emails and texts. SAMHSA noted that, read together, the rules could suggest an employee’s personal phone that received a patient’s text might need to be sanitized when a program closed. It said that was not the intent, and that it did not wish to convey that such devices become part of the Part 2 record.
The guidance is a reminder that Part 2 information arrives through channels the record system does not see. The proposal said SAMHSA did not encourage patient communication through personal email and phones. It said an employee who receives such a message containing patient identifying information should forward it to an authorized channel of the program and delete it from the personal account.
Flows that create new holders
Much of the proposal was about letting Part 2 information go to more places, for good reasons. Each new path created a new recipient bound by Part 2’s redisclosure limits. Each therefore extended the provenance problem to a new organization.
Entity-level consent
The proposal would let patients consent to disclosure to an entity without naming an individual recipient. SAMHSA explained that patients seeking benefits or services, such as Social Security benefits or sober living programs, might not know the name of the person who would receive their information. It said special instructions would apply to consents for disclosure to information exchanges and research institutions.
From a data perspective, entity-level consent widens the set of organizations holding Part 2 records. Benefit agencies and social service providers are not health systems and may have no means to mark provenance at all. The redisclosure notice that accompanies a disclosure becomes the main carrier of the restriction. The proposal included amendments to that standard notice to clarify the limits.
Prescription monitoring and registries
The proposal would let opioid treatment programs disclose dispensing and prescribing data to state prescription drug monitoring programs, as required by state law and subject to patient consent. It would also let non-opioid treatment providers query central registries, which exist to prevent multiple enrollments.
Both are safety measures. Both create shared state systems that hold Part 2-derived data alongside data from ordinary pharmacies and prescribers. A clinician who queries such a system sees combined results. Whether the clinician’s organization can tell which entries came from a Part 2 source depends on how the state system presents them.
Payment, operations, audits and research
The proposal would list 17 types of payment and health care operations activities expressly as permitted with consent. SAMHSA said it believed these were already permitted but that stakeholders remained confused. It would clarify audit and evaluation disclosures to government agencies and payers, including their contractors. It would also widen research disclosures by HIPAA covered entities to researchers who are neither covered entities nor subject to the Common Rule.
Each clarification moves Part 2 data into administrative, financial and research systems. Claims platforms, audit contractors and research databases become lawful holders. The proposal allowed audit disclosures of identifiable data only where the audit or evaluation could not be carried out with de-identified information. That condition is itself a data capability. An organization that cannot readily de-identify Part 2 data may find itself disclosing identifiable data more often than the rule intends.
Audits on and off the premises
The proposal described how the existing audit rules distinguished between reviews done on the premises of a Part 2 program or lawful holder and reviews that involve downloading, copying or removing records. Auditors who review on site must agree in writing to the rule’s limits on use and disclosure.
The distinction is itself about where data goes. An audit that stays on premises leaves the records under the holder’s control. One that takes copies creates a new holder, often a contractor, which must keep the copies marked and protected. A health system that hosts audits should know which kind each one is, what leaves, and how the copies will be returned or destroyed.
Research outside HIPAA
The proposal noted that existing rules allowed research disclosures without consent only to HIPAA covered entities or business associates with proper authorization or waiver, or to institutions under the Common Rule. It proposed to allow disclosures by a HIPAA covered entity to researchers who are neither. It would also allow disclosures to workforce members for employer-sponsored research, and to recipients covered by FDA regulations on human subjects in clinical investigations.
Research datasets are copies by design, and they can outlive the projects that created them. Provenance labels need to travel into research extracts, and data use agreements need to say how Part 2-derived data will be held, reused and destroyed.
Court orders and legal process
The companion proposal published the same day concerned court orders for confidential communications, and the main proposal clarified how long undercover agents could be placed in a Part 2 program under court order. Both are reminders that Part 2 records are sought in legal proceedings, and that the rule’s strictest protections apply there.
Responding correctly to a subpoena or court order requires knowing which responsive records are Part 2 records. A health information management team that searches a chart for responsive documents needs the provenance label to tell it which documents need a Part 2-compliant order and which do not. Without the label, it must either over-withhold or risk disclosing records the rule protects.
Declared disasters
The proposal would allow disclosure to another Part 2 program or treatment provider during state or federally declared natural and major disasters, so patients could keep receiving care. Disaster disclosures may happen under pressure, through temporary arrangements.
Provenance is easiest to lose in exactly those conditions. A health system planning for disasters should include Part 2 data in its continuity plans: how records sent to another provider will be marked, and how the receiving provider will know what it holds.
Building lineage that lasts
The practical response to a provenance rule is data lineage: knowing, for any record or field, where it came from and what rules attach to it. This is good practice rather than something the 2019 proposal required, and SAMHSA was explicit that it set no technology mandate. In our judgment it is still the practical way to apply a provenance rule at scale.
Mark at the door
Provenance is cheapest to capture when data arrives. Every interface, document import and fax queue that brings in records from outside can tag them with their source. A registry of known Part 2 programs among the organization’s referral partners lets the intake process recognize Part 2 sources automatically.
Manual tagging is a fallback, not a strategy. Staff who scan a faxed record from a treatment program can mark it, but only if they know the sender is a Part 2 program and the scanning workflow offers the choice. The registry and the workflow have to be built first.
Carry it everywhere the data goes
The marker has to survive copies and transformations: into notes that incorporate received documents, into the warehouse, into extracts sent to vendors, and into exchange messages sent onward. Each system in the chain needs to read the marker and either honor the restriction or block the flow.
The data warehouse deserves particular attention because it is built for reuse. Part 2-sourced data there should carry its marker at the row or document level. Access, extraction and model-training processes should check it. A simple test is whether an analyst could produce a dataset containing Part 2 data without being told.
Record how information arrived
The oral exception means the channel matters. A note documenting a phone consultation with a Part 2 program should say so, in a structured way if possible. That lets the organization show, later, that information in a note came orally under the exception rather than from an incorporated document.
What it costs, and who pays
Lineage costs money and attention. Interface changes, EHR configuration, warehouse redesign and staff training all compete with other priorities. Smaller organizations, such as an ambulatory network or a community health center that refers to treatment programs, may lack the technical staff to build much of it.
For those organizations, the proportionate step is to reduce the number of places Part 2 records land. Received Part 2 documents can be routed to one controlled location in the record. Their incorporation into notes can be restricted. And they can be kept out of analytics extracts until a marker can travel with them.
Retention and disposal
Provenance matters at the end of a record’s life too. The proposal recalled that when a Part 2 program is discontinued, its electronic records must be sanitized within one year, so that patient identifying information cannot be retrieved. That obligation can be met only if the program’s records can be found, including copies in backups, exchange archives and warehouses.
A label applied at intake makes that search possible. Without one, a closing program, or a health system absorbing one in a merger, has to reconstruct where Part 2 data went before it can dispose of it.
Who owns the marker
A provenance marker crosses organizational lines inside a health system. Health information management handles incoming records. Clinical informatics configures the EHR. Integration teams build interfaces. Data teams run the warehouse. Privacy officers interpret the rule. If each owns a piece and no one owns the whole, the marker will fail at the boundaries between them.
We would name one owner for Part 2 provenance end to end, within privacy or data governance. That owner keeps the registry of Part 2 sources, the labeling rules, the list of systems that must honor the label and the results of testing.
Measures that show it works
A few measures tell a privacy officer whether lineage is holding. One is the share of documents received from known Part 2 sources that carry the label after intake. Another is the number of downstream systems holding Part 2-derived data and the share that enforce the label. A third is the result of periodic tests, such as tracing a sample of labeled documents through notes, exchange messages and the warehouse.
These measures are our suggestion, not something the rule requires. Their value is that they find gaps before an investigation, a subpoena or a patient complaint does.
The case against bothering
The strongest objection is that the rule kept changing. Congress amended the statute in 2020. HHS rewrote the rule in 2024 and, as the hindsight below shows, left a segregation requirement for only a small set of recipients. An organization that invested heavily in segmentation for the 2019 proposal might seem to have wasted the money.
The objection is right about segregation as a legal requirement. It is wrong about provenance. The 2024 rule and the regulation in force today keep certain protections attached to records because of where they came from. The restriction on using Part 2 records in proceedings against a patient is the clearest example. An organization cannot honor a restriction attached to a record’s origin without knowing the origin. The 2019 proposal made that visible, and later rules did not make it go away.
With hindsight
SAMHSA finalized the proposal, with some changes, in a final rule published on July 15, 2020 and effective August 14, 2020 (85 FR 42986). It adopted the applicability change. Records created by a non-Part 2 provider are not covered unless received Part 2 records are incorporated into them, and segregation of received records can be used to keep the provider’s own records outside Part 2. The final rule also adopted the oral-communication change to the definition of “records”, and the consent change allowing patients to name entities without naming individuals. SAMHSA said the combined effect was that oral and non-oral communications from a Part 2 program would be treated the same way: the receiving provider could create its own records outside Part 2, using segregation of any received record to keep the two apart.
By then the statute itself had changed. On March 27, 2020, Congress enacted the CARES Act, whose section 3221 amended 42 U.S.C. 290dd-2. HHS proposed rules to implement it in December 2022 (87 FR 74216) and issued a final rule in February 2024 (89 FR 12472). That rule took effect on April 16, 2024, and required compliance by February 16, 2026.
The 2024 rule, which HHS said would increase alignment with the HIPAA Privacy Rule, changed the segregation picture. It states that a Part 2 program, covered entity or business associate that receives records based on a single consent for all future treatment, payment and health care operations is not required to segregate or segment them. HHS said it removed language requiring segmentation or segregation. It added that, taking all the changes together, only a small set of recipients would still be subject to a segregation requirement.
Provenance did not go away. The same rule says that a record received under such a consent is still a Part 2 record. It remains subject to the prohibition on use and disclosure in proceedings against the patient. In responding to commenters who asked how records would be told apart, HHS wrote that this prohibition must be effectively operationalized. It also wrote that it was unclear how that could be done unless the recipient knows the records are subject to it. HHS said it might give further guidance on segmentation, tagging or tracking, but would not require specific technology.
The regulation as in effect on November 1, 2026, according to the eCFR, keeps both halves of the 2019 design. Section 2.12(d)(2)(ii) still provides that a treating provider not subject to Part 2 may record information about a substance use disorder and its treatment. It says the act of recording does not by itself make that provider’s record subject to Part 2. Section 2.12(d)(1) applies the restriction on use in criminal, civil, administrative and legislative proceedings to anyone who obtains a record from a Part 2 program, covered entity, business associate, intermediary or other lawful holder, whatever their status.
What held up is the provenance idea. The 2019 proposal’s boundary between a provider’s own notes and received Part 2 records survives in the regulation in 2026. What changed is the remedy. Segregation, which the proposal offered as the way to keep that boundary, is now required, in HHS’s words, of only a small set of recipients, while the obligation that depends on knowing a record’s origin remains. In our view, the lasting task the 2019 proposal set for health systems was to know where their records came from. It has turned out to be more durable than any single rule about how to store them.
Spotted an error? Tell us and we will check it and correct the article where needed.