RGA has been studying the evolution of Digital Health Data (DHD) for the past decade and has developed tools that can aid our clients in optimizing their use of electronic health records (EHRs) in underwriting.
To many, the EHR and the Attending Physician Statement (APS) have become interchangeable terms referring to patient medical records. In this article, for clarity, an APS refers to the handwritten or typed notes that contain office visit summaries and medical histories as well as the imaging and test and procedure results that make up a patient’s medical file. An EHR denotes the digitized version of these records. Much of the information in an APS may be contained in an EHR, but in digitizing, some of the detail and clarity may be lost due to constraints in how the information can be recorded.
Reviewing this data has given RGA underwriters some valuable insights into how best to reconcile EHRs with traditional medical records. While we are excited about the potential utility of DHD and believe it has become an important underwriter resource, we do not yet think an EHR should routinely be substituted for an APS. Many in the industry are currently working to understand and define use cases around DHD, and it may be helpful to share some lessons learned.
DHD and Underwriting: Still a Lot to Learn
Today’s electronic health records (EHRs) are the result of the ongoing drive to digitize the information contained in physician patient records. They aim to automate and streamline provider workflow and are increasingly used by medical professionals and service providers to maintain patient histories and records. However, the EHR and its structured digital data is still not a perfect substitute for the unstructured data found in many attending physician statements (APSs).
There are several reasons for this: first, an EHR might not always provide a full and comprehensive view of the history, longevity, and severity of each condition in a record. For instance, the cancer diagnostic codes in an EHR may not always include tumor stage or grade, and the cardiac diagnostic codes may not provide the detail available in an echocardiogram or cardiac catheterization report.
Interoperability also continues to be a challenge. Sometimes the lack of coding detail depends on which medical entity is providing the EHR. A primary care physician, for example, might not have the diagnostic detail found in a specialist’s records if good interoperability does not exist. This can be especially true if the specialist is outside of the primary physician’s practice or network.
Surprisingly, information such as “status” of a specific condition or effective time (when that condition first emerged) are both optional EHR fields. In addition, digital information-gathering and recordkeeping are not yet universal practices in the medical profession: EHRs are only available from providers or practices which have chosen to implement these systems.
The digital medical history in an EHR may be limited, as unstructured data can be difficult to standardize and interpret on an automated basis. Underwriters often need the unstructured data found in an APS for diagnoses such as alcohol and drug abuse or depression.
Finally, diagnostic codes in EHRs may be based on a differential and not a final diagnosis, especially if a test or procedure was ordered to confirm or rule out a diagnosis. The actual diagnosis is therefore not always clear or needs to be found in another medical record.
Cases with substantial complexity are more likely to need detailed and unstructured information along with hands-on underwriter analysis. Figure 1 lists examples of impairments that may be found in complex cases.
Figure 1: Common diagnoses that may require more information than an EHR provides |
Alcohol abuse / addiction | Drug abuse / addiction |
Cancer | Multiple sclerosis |
Coronary artery disease | TIA (transient ischemic attack) |
Depression | Valvular heart disease |
EHR Utility
Right now, it is most effective to use digital health data to underwrite cases rated standard or better or cases that are likely declines. Figure 2 provides examples of impairments that are good candidates for underwriting using the EHR. For these cases, EHRs can add value in many ways, as they contain information that can supplement, complement, or even take the place of basic evidence items. For example, vital sign information in an EHR, such as build and blood pressure, could be used in lieu of a paramedical exam for certain age and face amount bands. An EHR could also speed verification of application disclosures by validating clean applications or highlighting nondisclosures.
EHRs can be particularly helpful when verifying applicant MIB codes. Certain MIB codes might require additional investigation, and the ICD or SNOMED codes contained in an EHR could provide the needed information, eliminating the necessity for (or at least reducing the time needed) for a follow-up applicant tele-interview.
An EHR that provides information pertaining to specialist referrals can help underwriters target requests for additional information from specialist medical records. Also, effective dates and statuses in an EHR can provide a good timeline, which can help an underwriter determine if additional information is needed. As an example, a diagnostic code for depression in full remission last noted 10 years ago may need no additional information, whereas that same code, if noted one year ago, may require additional information.
Finally, EHRs can be part of an underwriting department’s “Heads up” or “Triage” program, identifying cases that could be processed with no or minimal underwriter review or cases that could go to a junior underwriter for processing.
Figure 2: Common applicant disclosures that may be underwritten with only an EHR |
Anxiety | Gastroesophageal reflux disease (GERD) |
Asthma | Hypertension |
Basal cell carcinoma | Hernia |
Benign prostatic hypertrophy or prostatitis | Hypercholesterolemia |
Build (overweight) | Hypothyroidism/Hyperthyroidism |
Cholecystitis | Osteoarthritis |
Room for Improvement
As we have studied the use of digital data found in EHRs, we have found that there is room for further refinements.
Sometimes a medication is indicated as having been prescribed, but a diagnosis code matching that medication’s purpose does not appear. For example, an EHR might indicate a prescription for metformin that instructs the patient to take one tablet daily by mouth with evening meals for diabetes, but a diagnosis code indicating the purpose of the prescription is not included.
Also, some medical providers use very general diagnostic codes in EHRs that makes it difficult for underwriters to assess the risk. For example, an EHR might contain SNOMED code 41368006, indicating urethral disease, but not the additional codes that would provide the specificity an underwriter would need. Figure 3 shows the range of types of codes that can appear in one person’s EHR.
Indications of active or inactive statuses can also be helpful but they are not always updated, and these indications need to be compared with a date. For example, one EHR contained a status of “generalized enlarged lymph nodes” with an onset date of November 13, 2017, which was still marked “active” in a 2019 record. However, no update had been provided since the initial date of diagnosis. In another instance, a condition was diagnosed three years ago and never mentioned again, yet medications were still being prescribed for it. This is where an APS, with details and dates for every visit, can be helpful.
Doctor instructions on prescriptions can help an underwriter as well. For example, one EHR indicated three drugs had been prescribed: trazadone, an anti-depressant, with instructions to take half a tablet by mouth at bedtime for sleep; hydrocodone, with instructions to take as needed for post-surgical pain; and Revatio, a treatment for pulmonary hypertension, with instructions to take three tablets one hour prior to sexual activity. The doctor’s instructions clarify why a medication is prescribed, which can be especially helpful when medications have more than one use or are prescribed for off-label uses.
Figure 3: Examples of Codes, Code Descriptions, and Dates in an EHR |
Applicant Demographics: 55 Y/O Female, Non-Smoker; Height 5’3”; weight 155; BP 125/85 |
Code Value | Code set | Code Description | Date Reported |
14760006 | SNOMED | Constipation (disorder) | 11/20/2018 |
23595009 | SNOMED | Gastroesophageal reflux disease (disorder) | 1/29/2018 |
59621000 | SNOMED | Essential hypertension (disorder) | 1/29/2018 |
267434003 | SNOMED | Mixed hyperlipidemia (disorder) | 1/29/2018 |
1076151100019101 | SNOMED | History of pulmonary embolism on long-term anticoagulation therapy (situation) | 1/29/2018 |
61582004 | SNOMED | Allergic rhinitis (disorder) | 10/27/2017 |
711150003 | SNOMED | Long-term current use of anticoagulant (situation) | 10/27/2017 |
193462001 | SNOMED | Insomnia (disorder) | 10/27/2017 |
00822004 | SNOMED | Hyperlipidemia (disorder) | 10/27-2017 |
2788600009 | SNOMED | Chronic low back pain (finding) | 2/17/2017 |
239873007 | SNOMED | Osteoarthritis of knee (disorder) | 2/17/2017 |
414916001 | SNOMED | Obesity (disorder) | 3/29/2016 |
48694002 | SNOMED | Anxiety (finding) | 12/3/2015 |
81576005 | SNOMED | Closed fracture of phalanx of foot (disorder) | 9/8/2015 |
064.00 | ICD-9 | Constipation, unspecified | 9/12/2013 |
21897009 | SNOMED | Generalized anxiety disorder (disorder) | 9/12/2013 |
415.19 | ICD-9 | Other pulmonary embolism and infarction | 8/28/2013 |
280.0 | ICD-9 | Iron deficiency anemia, unspecified | 3/6/2013 |
626.8 | ICD-9 | Other disorders of menstruation and other abnormal bleeding from female genital tract | 4/18/2012 |
724.2 | ICD-9 | Lumbago | 12/17/2010 |
49218002 | SNOMED | Hip pain (finding) | 12/17/2010 |
698.3 | ICD-9 | Lichenification and lichen simplex chronicus | 12/17/2010 |
631.81 | ICD-9 | Esophageal reflux | 4/3/2007 |
564.1 | ICD-9 | Irritable bowel syndrome | 4/3/2007 |
692.9 | ICD-9 | Contact dermatitis and other eczema, unspecified cause | 2/8/2007 |
35489007 | SNOMED | Depressive disorder (disorder) | 11/14/2006 |
717.9 | ICD-9 | Unspecified internal derangement of knee | 8/30/2006 |
718.31 | ICD-9 | Recurrent dislocation of joint, shoulder region | 8/30/2006 |
780.52 | ICD-9 | Insomnia, unspecified | 8/30/2006 |
38341003 | SNOMED | Hypertensive disorder, systemic arterial (disorder) | 8/30/2006 |
472.0 | ICD-9 | Chronic rhinitis | 8/30/2006 |
715.90 | ICD-9 | Osteoarthrosis, unspecified whether generalized or localized, site unspecified | 8/30/2005 |
724.2 | Icd-9 | Lumbago | 9/10/2005 |
727.43 | ICD-9 | Ganglion, unspecified | 3/23/2004 |
599.0 | ICD-9 | Urinary tract infection, site not specified | 3/23/2004 |
466.0 | ICD-9 | Acute bronchitis | 10/7/2003 |
478.1 | ICD-9 | Other diseases of nasal cavity and sinuses | 9/21/2003 |
477.8 | ICD-9 | Acute rhinitis due to other allergies | 8/4/2003 |
461.9 | ICD-9 | Acute sinusitis, unspecified | 8/4/2003 |
719.46 | ICD-9 | Pain in joint, lower leg | 12/3/2002 |
726.90 | ICD-9 | Enthesopathy of unspecified site | 3/30/2001 |
278.00 | ICD-9 | Obesity, unspecified | 11/24/2000 |
Ready for Prime Time? Not Quite…
EHRs are here to stay. Although they may be long and repetitive and may contain gaps and digital noise, they are also deep, rich sources of applicant medical data. However, EHRs should not yet routinely be substituted for APSs. It will be the insurance industry’s challenge to discover how best to access the information nuggets in EHRs and stitch them together so that repetition is deleted and digital noise quieted, enabling the information to be digested and well-utilized in underwriting.