EHR audit log product research is not a search for the most feature-rich dashboard. It is an exercise in matching a product's recorded data elements, tamper-resistance controls, and query behavior to the specific questions an organization must answer. The HL7 May 2026 normative ballot for audit and disclosure logs, which carries forward the former ASTM E2147-18 specification, defines the core content that should be captured for every access to patient information [1]. The specification states that audit logs shall track data entry and modifications, simple access and viewing of the patient record, and whether any modifications are made during that access [1]. ONC certification criteria then translate those content expectations into testable requirements for certified health IT [3]. A systematic review of 85 published studies shows that researchers use these logs for far more than privacy monitoring, including workflow measurement and duty-hour estimation, but the review also warns that validity testing is rarely reported [2]. Product evaluation therefore requires separating what a standard mandates from what a vendor claims its analytics can prove.

The Content Baseline: What a Log Must Record

The HL7 specification states that an audit log should record all activities impacting a medical record, including creating a new record, entering data, changing or deleting an existing record, and additional user access data such as identification, location, and date and time [1]. The specification also states that audit logs shall track simple access and viewing of the patient record, and whether any modifications are made during that access [1]. The specification further lists the information expected for each access of patient medical information: user identification, date and time documenting both start and exit times, total duration of access, specific terminal, hardware, or location, type of action such as copy, print, addition, modification, or deletion, and the specific patient data accessed [1]. The specification also states that all accesses shall be recorded, and the entire access record shall be provided when an access record is requested [1]. These elements form the minimum vocabulary for any product comparison.

ONC certification narrows this baseline for certified health IT. The auditing criterion requires that health IT, by default, record actions related to electronic health information in accordance with sections 7.1.1, 7.1.2, and 7.1.6 through 7.1.9 of ASTM E2147-18, and record date and time according to any Network Time Protocol standard [3]. The ONC guidance states that only those sections specified from section 7 of ASTM E2147-18 are the minimum required for certification [3]. The audit report criterion references the same ASTM sections for audit log content and requires synchronized clocks using any NTP standard [6]. The criterion also requires that the audit log record changes to user privileges when health IT is in use [6]. A product that lacks these recorded fields, or that cannot demonstrate NTP synchronization, fails the baseline before any analytics question is asked.

Tamper Resistance and the Limits of Certification

The certification criterion states that actions recorded related to electronic health information must not be capable of being changed, overwritten, or deleted by the technology [3]. The criterion also requires that the technology be able to detect whether the audit log has been altered [3]. The companion guidance clarifies that ONC encourages the use of hashing algorithms consistent with standards adopted for hashing electronic health information, including a hashing algorithm with security strength equal to or greater than SHA-2 as specified by NIST in FIPS Publication 180-4 [3]. This is a technical decision point that product researchers should verify directly rather than assume from a certification badge.

Certification also has explicit scope boundaries. The ONC guidance states that copy and paste functions of Microsoft Windows originate outside the health IT environment and are outside the scope of certification, while copy actions originating from within the health IT interface, such as exporting or downloading a copy of electronic health information, are required to be tracked [3]. Demonstration of the ability to use NIST time servers is required for certification, but vendors are not required to use NIST servers post-certification [3]. These boundaries mean that two certified products can still differ materially in what their logs capture during real clinical use.

A Suggested Evaluation Scenario

As a suggestion, not a documented test, a product researcher could construct a hypothetical scenario: a privacy officer receives a request to identify every access to a specific patient's record over a 90-day period, including user, location, action type, and duration. The researcher would then ask each vendor to demonstrate, in a test environment, how the query is constructed, how long it takes against a realistic record volume, whether start and exit times are both returned, and whether the result set can be exported without altering the underlying log. The VAS documentation's emphasis on search accuracy, filtering behavior, and rows-per-page controls provides a concrete checklist for observing that demonstration [5]. The VAS user guide also documents a default of 10 rows per page, with the option to select a different number of rows [5]. The researcher would also ask how the product proves that the exported report has not been altered, referencing the hashing requirement in the ONC guidance [3]. This scenario tests content completeness, query performance, and tamper evidence in one pass, without requiring any inference about clinical outcomes.

Workflow Measurement: Promise and Reporting Gaps

A systematic review of 85 articles found that most studies used audit logs to study EHR use directly, including how often providers accessed individual pieces of information, patterns of EHR use across features, and total duration of EHR use [2]. The review also found that studies derived measures including counts of actions captured by audit logs, counts of higher-level activities imputed by researchers, activity durations, activity sequences, activity clusters, and networks of EHR users [2]. More recently, studies began to use audit logs to examine clinical workflows extending beyond the EHR, using audit log timestamps to mark clinical event boundaries [2]. A few articles calculated resident duty hours using EHR login and logout timestamps, assuming these occur near shift boundaries [2]. These applications show the analytical range that a well-structured log can support.

The same review, however, found that relatively few articles report testing the validity and sensitivity of audit log measures [2]. The review also noted that the process of turning raw audit logs into insights is complex, requires professional judgement, and varies from study to study when it is even reported [2]. The authors concluded that this lack of rigor and reporting prevents synthesis and comparison across studies, and that continued growth of the field will require greater methodological transparency and validated standards to support replication and cross-study knowledge discovery [2]. For product researchers, this means that a vendor's claim that its log supports workflow analytics should be treated as a hypothesis to test, not an established capability. The underlying data may be present, but the derived measures may not have been validated in the customer's clinical setting.

A Concrete Reference: The VistA Audit Solution

The Department of Veterans Affairs' VistA Audit Solution (VAS) provides a documented example of a production audit log interface. According to its user guide, VAS is a real-time web-based interface designed as a nationwide HIPAA-compliant audit tracking solution with the ability to track and report on access logs for patients' personally identifiable information and protected health information across all VistA instances [8]. The VAS database contains billions of audit records with millions more arriving every hour, sourced from approximately 130 individual VistA systems [5]. The VAS user guide also states that the system is only accessible within the VA intranet to authorized users such as Privacy Officers [5]. This scale illustrates the query performance demands that any enterprise audit log product must meet.

The VAS documentation also shows how product capabilities evolve through versioned fixes. The November 2026 version 2.5.5 release notes describe targeted updates to improve search accuracy, filtering behavior, and system stability, including a remediation for a patient multiple MVI selection issue [5]. The release notes also describe an earlier version 2.5.4 update that standardized sort indicators and updated SQL to improve system latency [5]. An earlier April 2026 version 2.5.3 updated incoming VistA records with no or incorrectly formatted date of birth to instead use a default date of 1900-01-01 [8]. The same release notes also document a version 2.5.1 fix for issues with VAS locking for VAS Privacy Officer names with unexpected characters [8]. These documented changes are useful evaluation prompts: a product researcher can ask any vendor how it handles malformed source records, identity matching during patient searches, and latency under billions of records.

Decision Points for Product Comparison

A comparison should be organized around decision points that the documentation can actually support. The first is recorded content: does the product capture the ASTM E2147 sections 7.1.1, 7.1.2, and 7.1.6 through 7.1.9 fields, including start and exit times and total duration [1][3]? The second is tamper evidence: can the product demonstrate that recorded actions cannot be changed, overwritten, or deleted, and does it use a SHA-2 or stronger hashing algorithm [3]? The third is query behavior: how does the interface perform when filtering, sorting, and viewing records at the scale of billions of entries, as VAS documents [5]?

The fourth decision point is analytical transparency. The systematic review's finding that validity testing is rarely reported means that a buyer should ask for evidence that any workflow measure derived from the log has been tested for accuracy in a comparable setting [2]. A separate commentary on audit log research recommends creating a publicly available repository to share existing audit log measures, allowing researchers and vendors to be more transparent and iterate collaboratively [7]. The commentary also recommends that measures available from such a repository specify sources of audit log data needed, secondary data requirements, query for retrieving data, data processing pipelines, and publicly available algorithms or code for deriving the measures [7]. A vendor that participates in such sharing, or that can point to validated measures, offers a different evidentiary position than one that only provides raw log export. The commentary also notes that a repository would identify where data classes and elements are sufficiently mature to be proposed for incorporation into USCDI [7].