Linguamatics provides the most comprehensive way of automatically extracting and surfacing patient level social risk factors through its SDOH NLP models. Pretrained to identify and characterize 14 different SDOH factors (figure 1), it is a fully customizable pipeline that enables healthcare organizations to identify information that is otherwise inaccessible to them at scale. The NLP models include categories mandated by CMS and the NCQA – enabling organizations to gain better targeted insights into the SDOH that impact quality and performance programs.
Figure 1: Automated NLP workflow for SDOH screening
The Linguamatics SDOH module can be interacted with in a number of different ways, enabling its impact to be realized across different teams and user types in healthcare. As part of the Linguamatics NLP platform, the SDOH models are available out of the box to license holders. This includes those with on premise, in cloud or hybrid solutions. There is also the ability to use an IQVIA hosted cloud API via the API marketplace.
Putting SDOH into action with Natural Language Processing (NLP)
The US healthcare industry spent over $2.5 billion on SDoH programs between 2017 and 2019, primarily focused on housing interventions, employment, education, food security, social and community context and transportation.
While there has been a significant increase in the importance that SDoH play in healthcare over the past decade, with increased investment as mentioned above – it is only recently that the screening for and management of social risk factors has become tied to quality and performance metrics. With these recent changes – more organizations are looking towards Artificial Intelligence to support in their SDoH screening efforts in order to ensure they are delivering the quality of care required.
Figure 2: Example Social worker EMR patient queue, prioritized by SDOH need
In the United States and most western countries, medical care (in a hospital or clinic) accounts for only 20% of health outcomes. By contrast, 40% is attributed to socio-economic factors, 30% to health behaviors (e.g. obesity, tobacco use or alcohol abuse) and 10% to the physical environment.
Collectively, these factors represent the Social Determinants of Health (SDoH) defined by the World Health Organization (WHO) as “conditions in which people are born, grow, live, work and age”.
This definition recognizes that social determinants of health such as ambulatory issues, food insecurities, transportation issues, living conditions and cultural beliefs and habits due to race and/or ethnicity have a profound influence on patients’ health and lives.
The patient impact of social risk factors
There are numerous determinants of health that can be considered “social”, for example, average life expectancy is reduced by 15-20 years for people living in low-income communities due to the increased risk of chronic conditions such as Diabetes and heart disease. Poverty can limit access to healthy food, limit the ability to afford safe neighborhoods (where they may be more likely to be victims of crime and/or violence, restrict healthy living conditions and promote transportation issues.
Most recently, social isolation has been linked with several adverse outcomes, such as dementia, stroke and heart disease.
Supporting Quality Reporting with NLP
Social determinants of health are typically divided into several categories, with definitions from organizations such as the World Health Organization (WHO) and the Centers for Medicare & Medicaid Services (CMS). This is a rapidly evolving landscape, and our SDOH package is regularly updated to ensure outputs are aligned with population health and reporting needs.
Example output aligned to Social Need Screening:
In the above example, the NLP has identified that the patients is receiving SNAP benefits, and therefore has food insecurity. This is presented alongside other SDOH risk factor that the patient has. The end user (social worker, ED Nurse etc) can use the interface to “accept” that this patients has this risk factor, edit or reject this NLP finding. This is augmented intelligence in practice.
Example output aligned to Social Drivers of Health screening:
In this example, the patient was discharged to a shelter – showing they have housing insecurity. This identifies the patient as part of the numerator social risk, from data that has no corresponding structured data (such as Z codes) in the patient’s medical record.
The Linguamatics NLP models cover the following 14 domains, and are completely customizable by our clients should they wish to edit or extend the out of the box pretrained models:
Victim of Crime and Violence
Weight Range Category
Victim of Crime and Violence
Weight Range Category
Other ways NLP adds value in Health Equity
Artificial Intelligence (AI) platforms like Linguamatics Natural Language Processing (Linguamatics NLP) allow HCOs to unlock social determinants of health, giving them a far more complete picture of each patient’s circumstances.
The same sophisticated tools can also enable payers to analyze member-supplied data, including sources such as online chats between patient and nurse. NLP can even be used to review social media posts and provide relevant insights about exercise routines, diet and behavior.
Lack of insight into social determinants of health such as ambulatory status, food insecurity and social isolation can have a major impact on making the right decision for a member, and on their satisfaction with the health plan.
Beyond basic sentiment analysis, a sophisticated NLP platform is needed to extract key concepts and relationships, for example: to identify social isolation issues, transport problems and cultural factors, which can be used to improve understanding and customer satisfaction.
For both providers and payers, the volume of available data is increasing exponentially, and so is the need to analyze unstructured patient data in real-time. By deploying sophisticated, predictive clinical models driven (in part) by social determinants, providers can identify which patients are at higher risk and act accordingly.
Applying natural language processing enables the capture of information from unstructured patient data in a timely manner and facilitates its use for analytical purposes. Unlike earlier systems, tools such as Linguamatics NLP enable open and flexible development of queries and are not as reliant on expensive data sets manually annotated by clinicians.
Predictive Analytics in Healthcare using Social Determinants
Social determinants of health data can feed analytics tools, like machine learning algorithms, predictive analytics and risk stratification models, to forecast future health outcomes. Using these models, providers and payers can discover if patients are at risk of hospital re-admission or failing to take their medication and take appropriate action to mitigate that risk.
Healthcare organizations able to harness the power of predictive analytics can estimate the likelihood of future outcomes based on patterns in the historical data. This data can also be used to identify resource gaps so new care programs and interventions can be developed.
Coupled with predictive analytics, healthcare companies that can track interventions and their correlations to better health outcomes, can achieve greater savings and better outcomes - especially when an intervention makes a positive impact on a patient’s social needs.
By automating the analysis of social determinants of health and delivering results at the point of care, predictive analytics solutions can help care teams identify patients that need special attention and give them a head start on the types of issues to expect.
In some cases, these tools can even tie into care team workflows and automatically generate preliminary plans of care that specifically address non-medical risk factors.
Predictive analytics using social determinants of health therefore hold enormous promise for reducing costs and improving outcomes, but success depends on the quality of the data employed. For example, the more geographically precise and patient-specific the data, the more accurate the predictive model.
However, for both providers and payers, predictive analytics is often complicated by the heterogeneous nature of patient-related data. The ability to automatically extract precise data from unstructured text is invaluable for organizations participating in value-based payment models. By leveraging NLP, providers can look at both the structured and unstructured data for a complete picture of each patient's lifestyle.
Easily extract social determinants of health with multimission NLP