One of the most promising uses of Electronic Health Records (EHR) is research.
As EHRs become more and more widespread and the clinical data previously held in silos of paper charts becomes fluid, exchangeable and duly collected, large clinical repositories should emerge and be made available to those engaged in research, presumably medical research. The results of such research are expected to help us identify cost effective therapies, health care trends and a myriad other quality improving, cost reducing strategies. With such lofty goals in mind, it may be beneficial to look at the actual contents of an EHR as they are today and as they may be tomorrow.
The clinical data captured in many of today’s EHRs, and in all future EHRs, contains a very rich structured and codified data set.
Diagnoses complete with dates of onset and resolution (if any) and characteristics such as improving or worsening over time
Medications and Allergies, including all historical changes and adjustments
Procedures and Immunizations, including diagnostic procedures, surgeries and treatments, all with dates and some with outcomes
Hospitalizations, ED and Office Visit dates and durations
Diagnostic Tests and Screenings, including labs, radiology, cognitive screenings, genetic screenings, etc. all with results and dates of service.
Complete vitals over time, such as height, weight, blood pressure, etc.
Family Histories of disease and outcomes all the way to aunts, uncles, cousins, nephews and nieces
On top of the clinical data, EHRs also contain demographics and socio-economic data sets.
Names, addresses, emails, phone numbers for patients and sometimes family members if appropriate
History of insurance coverage over time with effective coverage dates
Pharmacies and testing facilities used in the past
Ethnicity, preferred language, level of education, occupation and employment history
Academic performance, preferred areas of study, schools and camps attended
Safety measures in the home (fire safety) and on the road (seatbelts, helmets)
Family circumstances, including number of children, miscarriages, sexual orientation, habits and abusive or predatory history
Travel, hobbies, diet, exercise, alcohol, coffee and tobacco consumption over time
Very limited financial information, such as credit card numbers, bill paying promptness and necessity of payment plans or discounts
By all accounts, this is an exceedingly rich data set and should facilitate almost any type of research one can dream of. We could examine the effects of various therapies on disease progression and estimate cost effectiveness and various correlations with socio-economic circumstance. We could study disparities. We could assess quality of outcomes, sliced and diced by region and institution type. The sky seems to be the limit, or is it?
As anyone engaged in research knows, data points are only one side of the story. Data quality is another. The most publicized assault on the contents stored in our EHRs came over a year ago from the, now very famous, e-Patient Dave who attempted to download his medical records from a hospital into his Google Personal Health Record (PHR), only to discover that his EHR was laden with inaccuracies. There were diseases he never had, missing dates, missing meds and tests results, visits that never occurred and all that from one of the most advanced hospital EHR systems in existence. If you look at smaller hospitals and ambulatory practices, you will find that some medical records are electronic but some portions of care are still documented on paper. Patients sometimes ask their doctors to not document certain things and, given the option, some choose to “hide” sensitive information. Physicians sometimes dispense sample medications or call in a script and make only a cursory note in the EHR, or advice patients to see a specialist with no particular follow up documentation. Clerical errors are a given. The possibilities of incomplete, corrupt and misleading, data sets are endless. If that’s not enough, EHRs barely communicate with each other and creating an accurate picture of the continuum of care for any given patient shuttled from doctor(s) to hospital(s) to specialist(s) is nearly impossible.
One solution suggested by so called e-patients is the self-maintained PHR. Patients who want their “damn data” in a computable format seem to think that they could clean it up and make it usable for all sorts of innovative applications. The question here is of course whether the average patient has enough knowledge and enough information to correct the data set and accurately supplement it. For personal use, it is plausible to assume that some of us would be motivated enough and capable enough to create a decent longitudinal medical record in a PHR. For clinical research purposes, such PHR edited data sets would not pass muster any more than the deficient and erroneous original EHR data sets would.
The inescapable conclusion is that, currently, EHR data is ill suited to the customary rigor of clinical research, and has very little to offer above and beyond claim data which is already in use for general studies on broad trends, such as the famed Dartmouth Atlas Medicare expenditures studies. Looking ahead, it is feasible to assume that data quality will be improving and as paper is phased out and EHRs become more robust, user friendly and widespread, and as privacy and security issues are resolved, we may indeed end up with the proverbial mountain of pure gold for equally pure clinical research.
With lots of patience, determination and proper stewardship, the long term societal rewards from Electronic Health Records are very clear. Right now, we just need to do the right thing and we all, physicians, patients, entrepreneurs, government regulators, know exactly what the right thing is.