Electronic Health Data Sources: A Comparison

The ongoing disruptions caused by COVID-19 – including difficulties in obtaining paramedical exams and traditional APS reports- has accelerated the already active trend towards insurers exploring the use of electronic health data. Many companies are far down the path of researching these tools or have implemented them in production.

For those newer to digital health data, the following is a primer on the different types of data available and comparisons of their benefits and challenges.

These sources include:

• RX Fill Reports
• Medical Claims data
• Historical Lab Reports
• Electronic Health Records

In evaluating how these tools can be incorporated into an underwriting program, there are many factors to consider including cost, type of data available, coverage and completeness of the data.

Deciding which tools to use can be complicated – most are not a one to one replacement for existing requirements, but rather a new source of data that provides insights into the risk factors for a proposed insured from a view point that may differ from those in traditional requirements.

However, the benefits can be game changing. Many carriers look at new data sources through a traditional lens and fell into one of two common traps.

Trap 1.

All that matters is net expense. What is the cost of these new data sources and, do the savings achieved make up for the assumed negative mortality impacts that will result in foregoing traditional requirements?

 A true cost/benefit study would also include the indirect impacts of going digital such as significantreductions in producer and client experience, underwriter review time, increased placement rates and new sales opportunities. These are metrics that are not easily quantified, but ignoring them in the will produce flawed results. Assuming that mortality can only get worse is also not a given –  many carriers using digital health data are discovering underpriced mortality in their existing fully underwritten cases and the price points of digital data can enable use on many more cases than traditional tools. Additionally, the long-term benefits for having a complete digital repository of the structured health data for study and use in future pricing and valuation must also be considered.

Trap 2.

A new source of data must meet an arbitrarily high “hit rate” before it can be useful in the process.

While a source must clearly meet some minimum coverage thresholds to justify the operational burdens of onboarding, hard and fast rules can be counterproductive, particularly as most of these sources are free of charge unless data is available. Consideration for quality, cost, speed and trends in coverage are also important.

While many companies historically take the “conservative” approach when new technologies or processes emerge – “fast follower”, “wait and see” or conduct some small-scale pilots, the landscape in the industry is changing.  Leading companies who are embracing digital data and moving aggressively to simplify the underwriting process have been at work on these tools for more than a year and producers and consumers are seeking them out.  The risk of a carrier being left behind is real.

This risk can be exacerbated because these tools are different and implementing them in existing processes can be complex. It will take time to develop the underwriting rules for their use, potentially requiring a carrier to  use them in a research and development environment for a period of time Underwriters will need time and training on the use of digital data and how to incorporate it in their process.

Traditional mortality studies can present challenges. Back-ordering much of this data is not feasible so waiting for credible mortality results to emerge is problematic. Some carriers are choosing to aggressively order the most promising new requirements in production, even as additions to existing requirements, to build up sufficient sample sizes to enable effective data studies on the back end.

An additional insight that may be gleaned from using multiple tools in an R&D environment – not only do these tools provide different data but they source data in different ways. Having sufficient samples of multiple data sources on individual cases may surface the protective value provided in obtaining information from different combinations of sources. This data can then be used to develop an optimized ordering strategy to meet the goals of mortality at or below pricing with minimal expense and superior client experience.

None of these electronic health data sources were initially developed for insurance underwriting or non-health claims adjudication. They each tap into the existing communication networks used by healthcare providers and systems. Feedback from early adopters will enable significant improvement in the look, feel and usability of the data received from these sources.

Sources of Digital Health Data

RX Fill Reports

While well established, they will continue to play a key role in underwriting and can be used in conjunction with newer forms of digital data.

Rx records provide a one-dimensional look at a patient through the lens of their prescription fill history and enable inference of medical conditions. While there is not always a 1:1 correlation from a drug to a medical condition due to issues like off-label uses, the algorithmic tools available to accurately predict underlying conditions have gotten sophisticated over the years and continue to improve over time.

Strengths

• Coverage – while not perfect (no source ever is) this market has matured to the point where data is often available on over 90% of applicants – notable exceptions include drugs administered in some hospital settings (including some chemotherapy treatments) and items paid for in cash.
• Familiarity – Carriers have significant experience with Rx data – rules are in place and underwriters are comfortable with its use.
• Delivery Time – available in near real-time.

Challenges

• Limited information – only provides information on risk factors that are treated with a prescription – absent data like medical diagnosis, lab results, height, weight, smoking status, etc. it is a partial solution to establish fully underwritten mortality.
• Information Time Delays – this data is available through the health insurance adjudication process and there are delays between a prescription being written and data being available.

Medical Claims Data

Available more recently, claims reports capture the billing codes (ICD-9, ICD-10, HCPCS, or SNOMED) documented by healthcare providers for the purpose of billing and reimbursement to health insurance providers. They provide a valuable tool to uncover material medical impairments not disclosed on an application or discovered via other sources. Coverage is growing over time and hit rates of 35-50% are available with certain data suppliers.

Data will be available if the medical procedures and exams were submitted for health insurance reimbursement through a 3^rd party who participates in the life insurance use case.

Strengths

• Availability- Can be received from vendor in real time, often needing only an attestation of having a valid HIPAA authorization.
• Coverage– hit rates are moderate but expected to continue growing.
• Triage- May enable discovery of a “big-miss” mortality assessment due to undisclosed information

Challenges

• Ingestion- carrier will need to find a way to filter and consume the data. It is more likely to raise a red flag for the underwriter than it is to provide the complete information needed to decide on a case.
• False Alarms- Processes need to be developed to handle “differential diagnosis” items – health insurance claims billed for procedures with a negative result. Ex. a patient presents at an ER for a severe headache and the hospital conducts a brain tumor scan. There will be a brain tumor scan code billed even though the test was negative.
• Delays- Like Rx, this information is not available until the end of the medical claims adjudication process and may be weeks or more following treatment.
• Gaps- if an applicant has changed health insurance carriers, one of which is not contributing data to the source, there may be extended periods of time for which data is unavailable.

Historical Lab Data

Copies of a patient’s historical lab work conducted as part of their medical treatment. Data will be available if lab work was conducted by a laboratory participating in the insurance use case (The two biggest labs in the US are participating).

Strengths

• Insight- Can provide a significant amount of the information traditionally obtained from an insurance lab without the need for a paramedical visit.
• Value- Lab data can provide critical information particularly to validate preferred underwriting on accelerated cases.

Challenges

• One-dimensional View- no insights on risk factors that do not evidence themselves in lab results.
• Not an Insurance Lab Often missing some typical insurance specific lab values not common in patient treatment including cotinine (smoking status) or HIV status.
• Interpretation- Reference ranges are not always included in the raw data. For an underwriter to interpret normality, ie. whether the results are high, low, or abnormal can present a challenge

Electronic Health Records

Data extracted from the systems that healthcare providers use to store and share patient health information. Accessed by utilizing a patient portal tool or through contractual arrangements with healthcare organizations to provide information pursuant to a valid HIPAA authorization.

Data will be available if the patient’s physician or hospital participates in the life insurance use case through their EMR vendor, Health Information Exchange or if they make patient portal data available with proper patient credentials.

Note – the following strengths and weaknesses pertain to EHR data in general. The next section provides a deeper dive into the individual characteristics of the different types of EHR data available.

Strengths

• Multi-dimensional- a more holistic look at a patient’s medical history including diagnosis codes, medical procedures, lab results and prescribed medications. Often include height and weight, blood pressure and smoking status. These reports may include physician notes, providing substantial insight into the risk factors for complex underwriting situations.
• Completeness– Potential to have everything needed in one data source to enable a final underwriting decision.
• Value- Low cost, frictionless process and near real-time delivery could enable significantly higher overall usage than traditional APS and paramedical exams leading to improved overall carrier mortality experience than existing underwriting processes.
• Regulation– recent government mandates are reinforcing that healthcare organizations participate in interoperability initiatives and that patients have a right to access this data electronically.

Challenges

• Hit Rates – networks are growing but coverage today is the lowest of the digital sources discussed here.
• Rules Will Be Complex. Given the vast amount of structured data delivered in various formats (diagnosis codes, lab codes, medication codes, procedure codes, etc.) as well as unstructured data (clinical notes), development of carrier specific rules are more complex
• Variability– in data look and feel. This is driven by two issues:
  • Data comes from a multitude of data sources. Add-on services are available to standardize, normalize, de-duplicate and present with one unified look.
  • There are significantly more potential data points than other sources and patients’ conditions and type of medical care increase variability.

Quick Comparison of Sources

	RX Fill	Historical Lab	Medical Claims	EHR
Data Updated in Real-time		X		X
Rx	X			X
Lab Results		X	X	X
Diagnosis Codes			X	X
Medical Encounters				X
Height/Weight (BMI)				X
Smoking Status				X
Clinical Notes				X

EHR Data Primer

All EMR data available for the insurance use case is extracted from the Electronic Medical Record (EMR) software in use at the physician’s office or site of care. While there are well over 100 separate EMR platforms in use, several large players like Epic, Cerner and Allscripts have significant market share.

There are several national networks with a mission to enable broader healthcare interoperability such as CommonWell or Carequality. Some vendors tout their affiliations here as a strength but though they may play some future role their current contribution of data for the insurance use case is minimal.

The most common methods of accessing EMR data for the insurance use case include:

• Utilizing data available through a Health Information Exchange
• Through Insurance specific interfaces offered by an EMR vendor
• Via a portal offered for patient use.

Despite their differences, these methods share several things in common.

• Data is extracted from the EMR system in use at the site of care.
• Data is delivered utilizing some variation of a standard healthcare interoperability format (CCDA) known as a continuity of care document.
• The CCD is stratified into distinct sections such as a Problem List (medical diagnosis like ICD 9, 10 or SNOMED codes) Lab results, Rx history, Medical procedures, Vital signs, Family History etc.
• The data is a combination of structured data (diagnosis codes, lab results, vital signs) and unstructured data (physician notes).
• There may be significant data variability and instances of misclassified data do exist.
• Records are available in a machine-readable format such as xml and can be rendered into a human readable document for manual review.

Despite these similarities there are significant differences between them in the way they access and deliver data.

A look at each in more detail:

Health Information Exchanges

Often non-profit organizations, HIE’s play a critical role in healthcare interoperability usually operating on a regional basis. They act as a hub, connecting physicians, hospitals, labs and other healthcare providers to enabling sharing of health data for use in patient treatment. 

Physicians who participate in the exchange for patient treatment usually contribute data for the insurance use case. This enables very high participation rates in the areas where the HIE operates. Coverage rates of 70% or higher are not uncommon even for HIE’s that cover an entire state.

Breadth of data providers is another key advantage of HIE’s. They provide data from any care providers in the exchange with patient data. In our experience at Clareto, this results in records from over four different physicians per successful search.

This ability to provide data from multiple, unaffiliated providers for a single fee make the HIE a very economical way to gather large amounts of data on an individual.  While success rates for replacing an APS are relatively good, HIE’s are also an excellent tool in accelerated underwriting or triage use cases given the breadth of data they provide.

HIE’s synthesize data from multiple EMR systems and typically provide summary level information. While usually high in quality, it may lack the intricate detail necessary for a final decision in particularly complex cases.  

Data delivery times for HIE’s vary from under a minute to two days but there is a distinct trend towards instant release of data.

EMR Vendor Insurance Products

Several major EMR vendors including Epic, Allscripts, Practice Fusion and NextGen currently offer a proprietary product to enable healthcare providers to make records available for the insurance use case. Other EMR vendors are evaluating participation in the use case.

EMR vendor records generally provide the most complete electronic medical data – information is available at a summary level, but details for each physician encounter are often included as well. Encounter level details may be delivered via separate documents (instances of over 100 separate documents occur) but there are products available to consolidate these into a single document to simplify the review process.

Medical information will be available if the physician who has records utilizes a participating EMR vendor AND the healthcare provider opts to participate in the insurance use case. The number of physicians participating continues to increase.

Release of records require evidence of patient consent, but most physicians accept an insurers standard HIPAA authorization. An increasing number of physicians enable auto-release of records, but many conduct a manual review. Record return service levels span from minutes to several days.

If a patient has received care from multiple physicians who are in unaffiliated practices or who use different EMR systems, separate fees are charged for each record as with a traditional APS.

Patient Portal

Portal solutions access medical records using the client facing internet tool physicians use to share medical data directly with their patients. They require a unique set of patient credentials (username and password).

Introduced more recently, this is a more fundamental paradigm shift than the switch from paper to digital. Here, the client must actively participate in the data exchange versus other sources which operate “behind the scenes” allowing an insurer to access information directly from a data source using the permission granted in the HIPAA authorization.

The coverage potential here is very high given that virtually all physicians offer a portal.

However, the need for direct action by the client provides unique considerations in implementing portal solutions.

Actual coverage rates will be affected by a client’s being willing and able to share their credential(s) and by an insurers ability to successfully interact with the client and gather this  information.

Consideration will need to be made here for potential anti-selection – this model relies heavily on the client properly disclosing all physicians who may have access to pertinent medical data.

The Federal Government is acting to improve access and data availability via patient portal. New Office of the National Coordinator (ONC) regulations scheduled for 2022  mandate that all of a patient’s data be made available utilizing Fast Healthcare Interoperability Resources (FHIR) – a standard data format and elements as well as an application programming interface (API) for exchanging electronic health records. Enforcement mechanisms are not finalized, and these rules could be delayed again due to Covid but many major EMR vendors and healthcare providers are already in compliance.

Conclusion

Healthcare interoperability and data exchange are complex issues, but they lie at the heart of what we do at Clareto. We own and operate a Health Information Exchange in central Virginia that has been exchanging patient data for over 20 years. We process over 5 million electronic health data requests per month for patient Treatment use cases are connected to major hospital systems across the country as well as federal agencies including the Veterans Administration, Social Security Administration and the Department of Defense.

We are bringing the knowledge, expertise and systems capabilities developed working on patient treatment use cases to life, disability and long-term care businesses for use in their underwriting and claims adjudication processes.