The Healthcare Effectiveness Data and Information Set (HEDIS) is a standardized set of performance measures developed and maintained by the National Committee for Quality Assurance (NCQA). It is widely used by health plans and other organizations across the United States to evaluate and compare the quality of care and services they provide. By tracking key areas such as preventive care, chronic disease management, and member satisfaction, HEDIS helps ensure accountability, improve health outcomes, and inform consumers about plan performance.
The Social Need Screening and Intervention (SNS-E) measure first appeared in HEDIS for Measurement Year (MY) 2023 as a new, first-year measure developed by NCQA to standardize how health plans capture and respond to members’ unmet social needs. This focus grew out of a larger industry shift toward addressing social determinants of health (SDOH) as crucial factors in overall health outcomes, alongside NCQA’s emphasis on health equity.
In creating SNS-E, NCQA worked to align the measure with the Gravity Project, a multi-stakeholder initiative dedicated to developing standardized terminology and data exchange for SDOH documentation. To ensure consistency and reliability, NCQA incorporated recognized, evidence-based screening instruments (identified by specific LOINC codes) and tied the screening results directly to corresponding intervention codes (in SNOMED CT or CPT). By requiring plans to document both the presence of unmet needs and any subsequent intervention, NCQA hoped to promote more comprehensive data collection, better care coordination, and ultimately improved outcomes for members facing food insecurity, housing instability, or transportation insecurity.
Because health plans were finding it difficult to extract LOINC data from EHRs, the original SNS-E specification also allowed the use of administrative or case management data—part of NCQA’s ongoing effort to expand data sources and reduce reporting burdens. These early efforts laid the groundwork for the upcoming proposed revisions, such as adding certain Z codes (to capture social needs via diagnostic coding) and including newly introduced G codes (to document social needs assessments and interventions), all intended to further improve the reach and accuracy of social need screening and intervention.
Currently, the SNS-E measure requires health plans to document that members were screened (via approved, LOINC-coded tools) for unmet food, housing, and transportation needs at least once during the measurement year, and, if screening is positive, that a corresponding intervention (documented via specific SNOMED CT or CPT codes) was provided within one month. Plans must report six rates: screening and intervention for each of the three social needs (food, housing, transportation). The measure excludes members in hospice, those residing in long-term care, or enrolled in Institutional SNPs, and is stratified by age. Currently, a screening “counts” if it uses an approved instrument with an associated LOINC code (e.g., Hunger Vital Sign™, PRAPARE). A positive screen then places members in the denominator for the intervention measure .
Under NCQA’s proposed changes, several updates would expand how needs are identified and how interventions are counted:
• Adding HCPCS G0136 to the screening indicators
NCQA proposes that members can be counted in the screening numerator if a provider performs a standardized social needs assessment (billed under G0136), instead of relying exclusively on LOINC-coded screening tools.
• Adding Z codes to the intervention denominators
Currently, a member’s housing, food, or transportation need must be documented via a positive screen. NCQA would allow members with relevant ICD-10-CM Z codes (e.g., Z59.41 for food insecurity, Z59.82 for transportation insecurity) to be included in the intervention denominator, meaning they are identified as having a social need based on either a screening or a Z code.
• Adding new G codes to the intervention numerators
Three new G codes (G0019, G0023, G0140), which capture services by ancillary staff like community health workers, would now count toward meeting the intervention numerator for food, housing, and transportation needs.
• Removing “assessments” from the list of allowable interventions
The measure previously allowed certain “assessment” codes to fulfill the intervention requirement; these codes (e.g., 96161) will be removed. As proposed, assessments will instead align with screening (via G0136) rather than counting as an intervention.
Overall, these changes aim to capture more complete data using administrative codes, align with new CMS policies on SDOH billing (notably Z and G codes), and simplify the measure’s framework by moving all assessments into the screening phase and preserving the intervention phase for actual services—such as referrals, assistance, provision, and counseling—delivered in response to a positive social need.
Despite the promise of these administrative coding changes, many health plans still struggle extracting data from electronic health records (EHRs) and case management systems. This is where artificial intelligence tools such as data integration, natural language processing (NLP), and predictive analytics can be brought to bear. The following describes how AI today can support collecting data to support the SNS-E measure and what might be expected in the near future.
Current State‐of‐the‐Art AI Applications Supporting SNS‑E
Advanced Data Extraction and Integration
An immediate challenge in meeting SNS‑E requirements is the extraction and standardization of social need data from heterogeneous sources. AI-powered NLP systems now routinely parse unstructured clinical notes, patient questionnaires, and administrative records. Transformer-based models such as BERT (Bidirectional Encoder Representations from Transformers) and its healthcare adaptations (e.g. ClinicalBERT) are able to identify key phrases and context that indicate a patient’s social need status. By training such models on annotated EHR datasets, AI can automatically recognize when a screening has been performed using the approved LOINC codes or when a provider’s narrative contains references that align with the HCPCS G0136 assessment code.
AI can also fuse data from multiple systems—EHRs, case management tools, and administrative claims databases—by aligning disparate data with a common ontology. Using data fusion techniques, machine learning systems are capable of identifying redundancies or gaps in SDOH documentation. This is particularly relevant for SNS‑E where the measure’s numerator must accurately capture screening events performed by ancillary staff or automated methods in addition to provider assessments. Automated data integration not only reduces manual coding errors but helps streamline reporting and auditing processes.
Predictive Analytics and Risk Stratification
Machine learning models are already being used to predict adverse health outcomes based on SDOH. By leveraging historical data, gradient boosting machines and deep neural networks can identify patterns that suggest a higher risk of food insecurity, housing instability, or transportation barriers. These predictive models can flag patients who are likely to benefit from additional screening and targeted interventions. An AI system might, for example, analyze trends in a patient’s clinical history, geographic data, and socioeconomic status to predict whether they are at risk for housing inadequacy—thus prompting earlier or more frequent screenings.
These models can be embedded within clinical decision support systems (CDSS) that alert care coordinators or community health workers when a patient’s risk profile suggests an unmet need. By automating the risk stratification process, health plans can allocate resources more effectively, ensuring that interventions are delivered to those most in need.
Geospatial Information Systems (GIS) play a key role in these predictive models. By examining the neighborhood or region where a patient resides—along with local metrics such as area‑level poverty rates, food desert designations, crime statistics, and availability of public transportation—AI systems can further refine risk predictions. For instance, a patient living in a region identified as a “food desert” may be flagged as having a higher likelihood of food insecurity, especially if they also belong to a demographic group with known transportation or economic barriers. Integrating geospatial analytics into risk stratification models not only captures important environmental factors but also ensures that health plans can proactively target high‑risk areas with more intensive outreach and resource allocation.
Embedded within clinical workflows, these AI‑based risk stratification tools can generate alerts prompting earlier or more frequent screening for social needs. In this way, clinicians and care coordinators can focus on members most in need of intervention, improving the measure’s overall performance and ensuring resources are used efficiently.
Real-Time Monitoring and Adaptive Learning
Another breakthrough of modern AI is its ability to operate in real time. In the context of SNS‑E, real-time analytics can monitor screening rates and intervention delivery on a continuous basis. AI systems integrated into health IT infrastructure can flag delays in intervention—such as when a patient who screened positive for food insecurity has not received a corresponding intervention within the 30-day window. Alerts generated by these systems can trigger immediate follow-up actions, thereby improving the timeliness of care.
Furthermore, adaptive learning algorithms continuously refine their predictions based on new data. As more information is collected via the SNS‑E measure, these systems improve their ability to predict which interventions are most effective for which patient populations. Over time, this iterative learning process will likely lead to more personalized and impactful strategies for addressing social needs.
Enhanced Natural Language Understanding for Code Mapping
A key challenge in SNS‑E is ensuring that only screenings documented with the prescribed LOINC codes are counted. AI can help standardize and validate these data points by mapping free-text entries and other non-standardized inputs to the correct coding schemes. For instance, by applying named entity recognition (NER) techniques, an AI model can accurately extract mentions of social need screening from clinical notes, even when providers use slightly different terminologies. This enhances data accuracy and minimizes the risk of underreporting due to variations in documentation practices.
Provider Burden Reduction and Workflow Optimization
The SNS‑E measure introduces additional coding requirements—such as the use of HCPCS G0136 for provider assessments—that can be burdensome for physicians. AI-driven automation can help reduce this burden. Voice recognition systems combined with NLP can transcribe and code verbal assessments during patient encounters. Similarly, automated documentation tools can populate required fields in the EHR based on the conversation between providers and patients, ensuring that assessments are captured accurately without adding significant time to the clinical encounter.
Future Directions: AI’s Evolving Role in SNS‑E
As AI research and technology continue to advance, several emerging trends promise to further enhance support for the SNS‑E measure:
Multimodal Data Analysis
Future AI systems will increasingly integrate multimodal data sources. Beyond text and administrative codes, AI will be able to incorporate imaging, sensor data, and even geospatial information. For example, wearable devices or mobile apps might collect data on a patient’s physical activity or location, providing additional context that can inform social need assessments. When combined with EHR data and administrative records, these multimodal inputs can help create a more holistic picture of a patient’s social environment. In the future, such comprehensive data integration may enable AI to identify subtle indicators of SDOH that are currently overlooked.
Improved Interoperability Through Standardized Data Models
One of the barriers to effective SDOH data capture has been the lack of interoperability between different health IT systems. In the coming years, advances in AI and blockchain technologies may foster greater data standardization and secure data sharing. This, coupled with initiatives like the Gravity Project and HL7’s efforts to standardize SDOH data elements, will allow AI systems to access and analyze data from multiple sources more seamlessly. Enhanced interoperability will lead to more robust AI models, as they can be trained on larger, more diverse datasets—thereby improving their predictive accuracy and clinical utility.
Explainable AI and Trustworthy Decision Support
One of the challenges in deploying AI in healthcare is the need for transparency and explainability. Future AI systems will increasingly incorporate explainable AI (XAI) frameworks that make it clear how decisions are reached. For SNS‑E, this means that when a model flags a patient as high-risk for housing instability or food insecurity, the system can also provide an explanation in terms of the specific data points that led to that conclusion. This transparency will not only build trust among clinicians and patients but will also facilitate more targeted interventions, as providers can better understand the underlying factors driving a patient’s risk.
Personalized Intervention Strategies via Reinforcement Learning
As AI matures, reinforcement learning (RL) models could be applied to optimize intervention strategies over time. RL algorithms learn by interacting with an environment and can adjust their actions based on feedback. In the context of SNS‑E, RL could be used to test different intervention strategies—such as varying the type, timing, or intensity of interventions—and learn which combinations yield the best outcomes for different patient subgroups. Over time, such systems will be able to recommend highly personalized intervention plans that are tailored not only to a patient’s clinical profile but also to their unique social circumstances.
Edge AI and Real-Time Remote Monitoring
Advances in edge computing mean that AI can increasingly be deployed on local devices rather than relying on centralized data centers. For patients in rural or underserved areas—populations that often face significant social challenges—edge AI could enable real-time remote monitoring and intervention. Smartphones and IoT devices equipped with AI capabilities might continuously track indicators of social need (such as missed appointments, changes in medication adherence, or fluctuations in daily activity) and alert care teams instantly. This decentralized approach could significantly reduce the latency between screening and intervention, ensuring that social needs are addressed in a timely manner.
Integration of Conversational AI and Virtual Care
The next few years will likely see a broader integration of conversational AI and virtual care platforms. Chatbots and virtual assistants are already being deployed in patient-facing applications to guide users through self-assessments and screenings. In the future, these systems will become more sophisticated, capable of handling natural language queries and providing personalized recommendations based on real-time interactions. Such conversational agents could conduct preliminary social need screenings, document responses using the appropriate coding systems, and even schedule follow-up interventions. By automating these tasks, conversational AI can alleviate the administrative burden on clinical staff while ensuring that patients receive consistent and empathetic care.
Ethical AI and Bias Mitigation
As AI becomes more integral to SDOH assessments, addressing issues of fairness and bias will be critical. Future research will focus on developing algorithms that are not only accurate but also equitable. By incorporating fairness constraints into AI models, researchers can ensure that the predictive algorithms used in SNS‑E do not inadvertently reinforce existing disparities. Techniques such as bias auditing, fairness-aware learning, and adversarial de-biasing are already in development and will be refined in the coming years. Ensuring that AI-driven interventions are both effective and just will be a key factor in improving overall health equity.
Enhanced Data Privacy and Security with AI
The sensitive nature of SDOH data calls for robust privacy and security measures. AI techniques, including differential privacy and federated learning, will become increasingly important in protecting patient data while still allowing for high-quality model training. Federated learning, in particular, enables AI models to be trained on decentralized data (for example, across multiple health systems) without compromising individual privacy. As these methods mature, health plans will be able to leverage richer datasets to train their SNS‑E models, further enhancing predictive accuracy and intervention effectiveness while ensuring compliance with data privacy regulations.
Implications for Health Equity and Patient Outcomes
The integration of state-of-the-art AI in SNS‑E has far‐reaching implications for health equity. By automating and enhancing the screening process for social needs, AI can help ensure that vulnerable populations are identified more accurately and receive timely interventions. For example, the ability to seamlessly extract and standardize LOINC, HCPCS, and ICD‑10 data means that patients who might otherwise fall through the cracks are more likely to be connected to resources. Furthermore, AI’s predictive capabilities can allow health plans to allocate resources more effectively, focusing on those most at risk and tailoring interventions to meet individual needs.
Moreover, as AI systems evolve, they will enable a more dynamic and responsive approach to managing social needs. With real-time monitoring and adaptive learning, interventions can be modified on the fly, ensuring that patients receive the right support at the right time. This level of personalization, combined with enhanced data integration and predictive analytics, can lead to improved patient outcomes, reduced hospitalizations, and lower overall health care costs.
Conclusion
In summary, current AI technologies already offer a powerful set of tools to support the SNS‑E measure. From advanced NLP systems that extract standardized screening codes to machine learning models that predict risk and trigger timely interventions, AI is transforming how social determinants of health are captured and addressed. Looking ahead, the evolution of multimodal data analysis, explainable AI, reinforcement learning, edge computing, and conversational AI will only deepen this impact. As these technologies mature, they promise not only to reduce provider burden and improve reporting accuracy but also to drive meaningful improvements in health equity and patient outcomes.
Health plans and providers who embrace these state‐of‐the‐art AI techniques today will be well positioned to lead the way in SDOH documentation and intervention. In the next few years, as AI models become more sophisticated, more interoperable, and more ethically robust, the potential to transform care for populations with unmet social needs will grow exponentially. This is not just about better compliance with SNS‑E measure requirements—it is about fundamentally rethinking how we capture, interpret, and act on the social factors that have a profound impact on health. Ultimately, these advancements in AI will play a critical role in ensuring that every patient, regardless of their social circumstances, has access to the care and resources they need to thrive.
By leveraging both current AI capabilities and anticipating the transformative breakthroughs of the near future, stakeholders in healthcare can work together to drive the next generation of SDOH interventions, fostering a more equitable and effective healthcare system for all.
Ready to Transform?
Ready to leverage AI to improve your SNS-E compliance and optimize social need interventions?
Contact us today to explore custom AI solutions for health plans and stay ahead of regulatory changes.
The latest
