BILL ANALYSIS Senate Research Center C.S.S.B. 1822 89R21001 SCL-D By: Johnson Health & Human Services 3/26/2025 Committee Report (Substituted) AUTHOR'S / SPONSOR'S STATEMENT OF INTENT Artificial intelligence-based algorithms are increasingly used in healthcare to streamline time consuming processes, such as utilization reviews, in which health coverage policy issuers or utilization review agents evaluate the necessity of treatment and expenditure of resources relating to medical services provided. Under current statute, algorithms cannot be used to deny services unless explicitly reviewed by a physician, but there are limited transparency requirements in place to enforce this restriction. Meanwhile, there have been reports of artificial intelligence-based algorithms wrongly denying claims based on biases and public concern about data disclosure. S.B. 1822 would address the lack of transparency surrounding the use of artificial intelligence (AI) in utilization review to protect patients against the improper use of algorithms to deny services. Under the current law, there is no guarantee that AI algorithms are being used in a way that prevents bias based on gender, race, age, and other protected classes. Lawsuits against insurance companies like Humana and United Healthcare detail accounts of AI algorithms wrongfully denying claims for vulnerable populations, including the termination of policies. Requiring that issuers provide notice to individuals and provide public, easily accessible disclosure of their use of artificial intelligence would improve public oversight to safeguard against the improper use of these algorithms. Meanwhile, requiring that the associated algorithms are trained in a way that would minimize these biases, utilize scientifically based clinical guidelines, and provide proof of compliance is necessary to better gauge the scope of this practice and better protect Texas residents. S.B. 1822 would require issuers of health policies to publicly disclose the usage of artificial intelligence-based algorithms in utilization reviews in accessible place on their website and provide notice to insured or enrolled individuals. S.B. 1822 would also require issuers and utilization review agents to attest that algorithms used for utilization reviews and associated training of data sets minimize the risk of bias, comply with clinical guidelines, and submit this attestation and applicable algorithms to TDI each year. Increasing the transparency of these processes, better protecting data, and curtailing the risk of bias for those in need of medical care is vital to maintaining the integrity of this highly sensitive issue area and providing accurate, quality care. Key Provisions: Requires an issuer of a health insurance policy, including health maintenance organization that issue health benefit plans, to publicly publish a disclosure regarding whether artificial intelligence-based algorithms are used to conduct utilization reviews Requires that the disclosure be provided on the issuer's website and provided in writing to insured or enrolled individuals. Requires that issuers and utilization review agents ensure that artificial intelligence-based algorithms used for utilization reviews and the training of data sets minimize the risk of bias and comply with evidence-based clinical guidelines. Requires that issuers and utilization review agents submit an attestation of compliance along with the artificial intelligence-based algorithms used in conducting utilization reviews and training data sets to TDI no later than December 31st of each year. Committee Substitute The committee substitute no longer requires the submission of the AI-based algorithm used for utilization reviews unless the commissioner determines additional documentation is needed based on belief that there is noncompliance. Requires issuers and utilization review agents to submit a compliance statement disclosing certain characteristics, including: o The use, scope, and function of AI in utilization reviews. o The logic or decision tree of an AI-based algorithm used for utilization reviews. o A description of each training data set used by AI-based algorithms for utilization reviews. o Compliance of training requirements for algorithms that minimize the risk of bias and comply with evidence-based clinical guidelines. o A description of the issuers or agent's oversight and validation process of the performance and compliance for the AI-based algorithm. Allows the commissioner to require additional documentation to evaluate compliance if there is reason to believe noncompliance C.S.S.B. 1822 amends current law relating to the use of artificial intelligence-based algorithms in utilization review conducted for certain health benefit plans. RULEMAKING AUTHORITY This bill does not expressly grant any additional rulemaking authority to a state officer, institution, or agency. SECTION BY SECTION ANALYSIS SECTION 1. Amends Subchapter D, Chapter 4201, Insurance Code, by adding Section 4201.156, as follows: Sec. 4201.156. ARTIFICIAL INTELLIGENCE-BASED ALGORITHMS. (a) Requires an issuer of a health insurance policy or a health maintenance organization that is the issuer of a health benefit plan to publish on a publicly accessible part of the issuer's Internet website and provide in writing to each insured or enrollee, and to any physician or health care provider contracting with the issuer or providing services to an insured or enrollee, a written disclosure regarding whether the issuer or the issuer's utilization review agent uses artificial intelligence-based algorithms in conducting utilization review. (b) Requires an issuer described by Subsection (a) or a utilization review agent for an issuer described by Subsection (a) to ensure that any artificial intelligence-based algorithm used by the issuer or agent in conducting utilization review and the training data sets the algorithm uses have minimized the risk of bias based on an individual's race, color, religion, ancestry, age, sex, gender, national origin, or disability and comply with evidence-based clinical guidelines. (c) Requires an issuer described by Subsection (a) or a utilization review agent for an issuer described by Subsection (a), not later than December 31 of each year, to submit to the Texas Department of Insurance (TDI) for review, in the form and manner prescribed by the commissioner of insurance (commissioner), any artificial intelligence compliance statement. Requires that the statement include certain information. (d) Authorizes the commissioner, if the commissioner has reason to believe that an issuer described by Subsection (a) or utilization review agent for an issuer described by Subsection (a) is not in compliance with this section, to require additional documentation or conduct an inspection of any artificial intelligence-based algorithm used by the issuer or agent in conducting utilization review and each training data set the artificial intelligence-based algorithm uses. (e) Provides that information submitted to TDI under this section is confidential and not subject to disclosure under Chapter 552 (Public Information), Government Code. SECTION 2. Effective date: September 1, 2025.
BILL ANALYSIS
Senate Research Center C.S.S.B. 1822
89R21001 SCL-D By: Johnson
Health & Human Services
3/26/2025
Committee Report (Substituted)
Senate Research Center
C.S.S.B. 1822
89R21001 SCL-D
By: Johnson
Health & Human Services
3/26/2025
Committee Report (Substituted)
AUTHOR'S / SPONSOR'S STATEMENT OF INTENT
Artificial intelligence-based algorithms are increasingly used in healthcare to streamline time consuming processes, such as utilization reviews, in which health coverage policy issuers or utilization review agents evaluate the necessity of treatment and expenditure of resources relating to medical services provided. Under current statute, algorithms cannot be used to deny services unless explicitly reviewed by a physician, but there are limited transparency requirements in place to enforce this restriction. Meanwhile, there have been reports of artificial intelligence-based algorithms wrongly denying claims based on biases and public concern about data disclosure. S.B. 1822 would address the lack of transparency surrounding the use of artificial intelligence (AI) in utilization review to protect patients against the improper use of algorithms to deny services.
Under the current law, there is no guarantee that AI algorithms are being used in a way that prevents bias based on gender, race, age, and other protected classes. Lawsuits against insurance companies like Humana and United Healthcare detail accounts of AI algorithms wrongfully denying claims for vulnerable populations, including the termination of policies. Requiring that issuers provide notice to individuals and provide public, easily accessible disclosure of their use of artificial intelligence would improve public oversight to safeguard against the improper use of these algorithms. Meanwhile, requiring that the associated algorithms are trained in a way that would minimize these biases, utilize scientifically based clinical guidelines, and provide proof of compliance is necessary to better gauge the scope of this practice and better protect Texas residents.
S.B. 1822 would require issuers of health policies to publicly disclose the usage of artificial intelligence-based algorithms in utilization reviews in accessible place on their website and provide notice to insured or enrolled individuals. S.B. 1822 would also require issuers and utilization review agents to attest that algorithms used for utilization reviews and associated training of data sets minimize the risk of bias, comply with clinical guidelines, and submit this attestation and applicable algorithms to TDI each year. Increasing the transparency of these processes, better protecting data, and curtailing the risk of bias for those in need of medical care is vital to maintaining the integrity of this highly sensitive issue area and providing accurate, quality care.
Key Provisions:
Requires an issuer of a health insurance policy, including health maintenance organization that issue health benefit plans, to publicly publish a disclosure regarding whether artificial intelligence-based algorithms are used to conduct utilization reviews
Requires that the disclosure be provided on the issuer's website and provided in writing to insured or enrolled individuals.
Requires that issuers and utilization review agents ensure that artificial intelligence-based algorithms used for utilization reviews and the training of data sets minimize the risk of bias and comply with evidence-based clinical guidelines.
Requires that issuers and utilization review agents submit an attestation of compliance along with the artificial intelligence-based algorithms used in conducting utilization reviews and training data sets to TDI no later than December 31st of each year.
Committee Substitute
The committee substitute no longer requires the submission of the AI-based algorithm used for utilization reviews unless the commissioner determines additional documentation is needed based on belief that there is noncompliance.
Requires issuers and utilization review agents to submit a compliance statement disclosing certain characteristics, including:
o The use, scope, and function of AI in utilization reviews.
o The logic or decision tree of an AI-based algorithm used for utilization reviews.
o A description of each training data set used by AI-based algorithms for utilization reviews.
o Compliance of training requirements for algorithms that minimize the risk of bias and comply with evidence-based clinical guidelines.
o A description of the issuers or agent's oversight and validation process of the performance and compliance for the AI-based algorithm.
Allows the commissioner to require additional documentation to evaluate compliance if there is reason to believe noncompliance
C.S.S.B. 1822 amends current law relating to the use of artificial intelligence-based algorithms in utilization review conducted for certain health benefit plans.
RULEMAKING AUTHORITY
This bill does not expressly grant any additional rulemaking authority to a state officer, institution, or agency.
SECTION BY SECTION ANALYSIS
SECTION 1. Amends Subchapter D, Chapter 4201, Insurance Code, by adding Section 4201.156, as follows:
Sec. 4201.156. ARTIFICIAL INTELLIGENCE-BASED ALGORITHMS. (a) Requires an issuer of a health insurance policy or a health maintenance organization that is the issuer of a health benefit plan to publish on a publicly accessible part of the issuer's Internet website and provide in writing to each insured or enrollee, and to any physician or health care provider contracting with the issuer or providing services to an insured or enrollee, a written disclosure regarding whether the issuer or the issuer's utilization review agent uses artificial intelligence-based algorithms in conducting utilization review.
(b) Requires an issuer described by Subsection (a) or a utilization review agent for an issuer described by Subsection (a) to ensure that any artificial intelligence-based algorithm used by the issuer or agent in conducting utilization review and the training data sets the algorithm uses have minimized the risk of bias based on an individual's race, color, religion, ancestry, age, sex, gender, national origin, or disability and comply with evidence-based clinical guidelines.
(c) Requires an issuer described by Subsection (a) or a utilization review agent for an issuer described by Subsection (a), not later than December 31 of each year, to submit to the Texas Department of Insurance (TDI) for review, in the form and manner prescribed by the commissioner of insurance (commissioner), any artificial intelligence compliance statement. Requires that the statement include certain information.
(d) Authorizes the commissioner, if the commissioner has reason to believe that an issuer described by Subsection (a) or utilization review agent for an issuer described by Subsection (a) is not in compliance with this section, to require additional documentation or conduct an inspection of any artificial intelligence-based algorithm used by the issuer or agent in conducting utilization review and each training data set the artificial intelligence-based algorithm uses.
(e) Provides that information submitted to TDI under this section is confidential and not subject to disclosure under Chapter 552 (Public Information), Government Code.
SECTION 2. Effective date: September 1, 2025.