Secondary Logo

Journal Logo

Building Bridges Across Clinical Registries

Glance, Laurent G., MD*†; Wanderer, Jonathan P., MD, MPhil‡§; Dick, Andrew W., PhD; Dutton, Richard P., MD, MBA

doi: 10.1213/ANE.0000000000002005
The Open Mind: The Open Mind

Published ahead of print April 3, 2017.

From the *Department of Anesthesiology, University of Rochester School of Medicine, Rochester, New York; RAND Health, RAND, Boston, Massachusetts; Departments of Anesthesiology and §Biomedical Informatics, Vanderbilt University School of Medicine, Nashville, Tennessee; and United States Anesthesia Partners, Dallas, Texas.

Published ahead of print April 3, 2017.

Accepted for publication January 23, 2017.

Funding: This project was supported with funding from the Department of Anesthesiology at the University of Rochester School of Medicine.

The authors declare no conflicts of interest.

Reprints will not be available from the authors.

Address correspondence to Laurent G. Glance, MD, Department of Anesthesiology, University of Rochester Medical Center, 601 Elmwood Ave, Box 604, Rochester, NY 14642. Address e-mail to

We often hear that Big Data will usher in an era of precision medicine and predictive analytics. But for precision medicine and predictive analytics to become mainstream, we need an infrastructure that connects the massive number of digital breadcrumbs—Big Data—generated by individual patients over their lifetimes.1 Although current clinical registries do not contain Big Data,2 we expect that as registries expand to become more population based and capture increasingly more granular and longitudinal data, the explosion in health care data will offer unique opportunities to improve health care outcomes. There are now >80 clinical registries listed in the National Quality Registry Network.3 Unfortunately, the depressing reality is that we have yet to harness the power of existing Small Data because most clinical registries exist in stand-alone silos that are not connected to one another.

Building the digital infrastructure for a learning health care system4 requires that data from these registries, and other sources of patient information such as electronic medical records (EMRs), are linkable. Suppose an anesthesiologist wants to deliver the anesthetic most likely to lead to a good 30-day outcome for a patient with heart failure who is undergoing colorectal surgery. The physician could use one of the many available prognostic scoring systems, such as the American College of Surgeons (ACS) Universal Risk Calculator,5 to risk stratify the patient and use that information to help formulate the anesthetic plan. But prognostic scoring systems do not identify best practices. Instead, imagine a query tool that would allow anesthesiologists to input patient and procedure characteristics and would then apply machine learning to Big Data6 to recommend the best anesthetic plan. This tool could be embedded in the clinical decision-support system7 of the institution or be available online for practice settings without an EMR. To succeed, this query tool must have access to the right data: (1) information on patient risk, surgical procedure, and 30-day outcomes from the ACS National Surgical Quality Improvement Program (NSQIP) registry; and (2) information on what was done for the patient in the operating room and the patient’s physiologic response to those interventions in the American Society of Anesthesiologists National Anesthesia Clinical Outcomes Registry (ASA NACOR). Today, the NSQIP registry and NACOR exist as digital silos. Together, these registries could serve as the digital platform for “precision surgery.”

NACOR and NSQIP were designed to improve patient outcomes using a data-driven approach for performance improvement and clinical decision making.8,9 However, the true potential of these and other clinical registries to help improve patient outcomes will not be achieved until we find a way to bridge the political divide between national organizations that currently limit opportunities for data sharing across registries. The first step is to encourage dialogue between member organizations. Our proposal to link the ACS NSQIP registry and NACOR was recently presented and favorably received at the 2016 National Surgical Patient Safety Summit—a collaboration among ACS, American Academy of Orthopedic Surgeons, ASA, and other key organizations to lay the groundwork to improve patient safety in surgery.

To connect existing data silos, we need to foster a culture of information sharing to encourage medical societies, big health systems, and EMR vendors to build bridges across clinical registries and EMRs. In pursuing this goal, we must recognize and respect the interests of registry owners that led them to invest significant resources to create and maintain robust high-quality data repositories. However, because most of the cost of collecting clinical data is borne indirectly by patients, they deserve to receive the full benefit of a learning health care system built on complete data, instead of the fragmented information that is currently employed. Organizations such as ASA and ACS may derive important benefits from data sharing. Although the NSQIP registry includes very detailed clinical data on baseline patient risk and 30-day outcomes, it includes very little information on intraoperative processes of care. In comparison, NACOR includes comprehensive Anesthesia Information Management System (AIMS) data describing what was “done” to patients, but it lacks data on baseline patient risk and outcomes beyond the first 24 hours (NACOR currently only includes AIMS data on a limited subset of patients. Given the rapid pace of EMR adoption, it is not unreasonable to assume that NACOR will include more comprehensive AIMS data in the future). By linking AIMS data to NACOR with ACS NSQIP data, it may be possible to identify best intraoperative practices and allow surgeons and anesthesiologists to jointly develop evidence-based care protocols. ASA currently lacks the granular clinical data on patient risk factors that it needs to develop risk-adjusted outcome measures. Linking NACOR to NSQIP would fill this gap and allow ASA to develop risk-adjusted outcome measures to use for the Centers for Medicare & Medicaid Services (CMS) Merit-Based Incentive Program. Furthermore, if ASA and ACS were to actively collaborate to improve patient care, together, they would be in a stronger position to represent patient interests and to influence federal government efforts to redesign health care. In the end, the purpose of health care data is to improve patient health. One of the most powerful ways to improve population outcomes is to learn from the collective experiences of tens of millions of patients. This can be realized only by connecting their digital breadcrumbs.

To make better use of data that are currently housed in clinical registries and to harness the potential of the Big Data that are being collected in EMRs, we need a unique patient identifier (UPI) that can be attached to each patient encounter in the health care system. Many registries lack common patient identifiers, such as name, social security number, date of birth, and date of surgery, which could be used to link data across registries. Although the Health Insurance Portability and Accountability Act mandated the creation of a UPI, Congress subsequently prohibited, and continues to prohibit, the use of funds from the Department of Health and Human Services to implement a UPI.10 Meanwhile, the Department of Health and Human Services roadmap envisions a health information technology infrastructure in which “multi-payer claims databases, clinical data registries, and other data aggregators will incrementally become more integrated as part of an interoperable technology ecosystem.”11 But without a UPI, patient records can only be linked by matching them across different data sources using protected patient health information (PHI). Statistical matching uses patient identifiers, such as name and birth date, instead of a UPI to link data records across different data sources.12 According to RAND, the privacy concerns that led Congress to ban funding for a UPI may have the unintended consequence of fostering the “repetitive use and disclosure of an individual’s personal identifying information” through data linkage.12 Nevertheless, using protected PHI remains the de facto approach for data linkage because it does not require a national debate on security issues and can be used without public understanding of the privacy issues of using protected PHI to link patient data.12

Using protected PHI to link registries is not the best solution for bridging the divide across data silos. Instead, registry owners should consider creating an encrypted UPI (eUPI) for use in clinical registries. This UPI would not contain identifiable patient information and would be created by mapping existing patient identifiers (name, date of birth, sex, and social security number) to a new identifier using a 1-way cryptographic hashing function, which creates a nonsense string of characters unique to each patient.13 This hashed identifier could then be encrypted along with unique facility data using public-key cryptography, a proven technology that underpins much of modern day electronic commerce. This would allow any facility to create an eUPI using a published public key but would restrict decryption to clinical registries that would hold the private key. It would be extremely difficult to reverse this process and map the eUPI back to the original protected PHI.14 Each facility contributing data to a clinical registry would use identical hashing and encryption algorithms. The eUPI would be generated at the time of data transfer to the registry and would not reside in the local EMR of any facility. Protected PHI data would not have to be transferred to the registry.13 Clinical registries could determine which research projects merited linkage of clinical data across facilities. This approach makes it possible to avoid including protected PHI in a central database while still enabling linkage of registries. Hashing has been validated for this application using data from the National Health Services in the United Kingdom13 and data from a large metropolitan area in the United States using a hashing function from the SHA-2 family developed by the National Security Agency.15

There are major barriers to the implementation of our proposal. First and foremost, the ACS has expressed, at best, very limited support for our proposal to create the digital infrastructure for data linkage. Second, NACOR is a relatively young registry and does not yet have comprehensive AIMS data on most patients. Because access to AIMS data is likely to be the most attractive incentive for ACS and other surgery registries to consider data linkage with NACOR, ACS is not likely to see a strong short-term benefit to laying the groundwork for data linkage at this point in time. Third, no encryption algorithm is completely secure, and adding encrypted patient identifiers to NACOR and NSQIP, which do not include common patient identifiers such as name and date of birth, would increase the privacy risk. We believe that the potential benefits of the eUPI will, in the future, outweigh the privacy risk of adding an encrypted identifier. However, this risk is not easily quantifiable and must be carefully considered, especially in light of recent breaches of US government databases.16 However, unlike government databases, under our proposal, NACOR and NSQIP would contain only an encrypted identifier, not patient names and social security numbers. Finally, is it ethical to add an encrypted patient identifier to registry records without patient consent? To answer this question, we should consider that patient records are routinely transmitted, with direct patient identifiers, to CMS and other third-party payers for reimbursement purposes. We believe that the privacy risk to patients from our proposal is no greater, and probably less, than the risk of privacy intrusion encountered by most individuals in the course of their daily lives. The use of Big Data to uncover best practices and guide clinical decision making is aspirational and could be transformative if it delivers clinically useful information by leveraging the clinical experience of thousands of anesthesiologists in one electronic portal readily available to practicing clinicians. The decision to move ahead with the creation of a UPI will have to be carefully considered with input from all major stakeholders, including patient representatives. Also, like any other therapeutic interventions, the use of machine learning to recommend best practices in perioperative medicine will need to be tested and validated before it is implemented on a large scale.

The underlying premise of this editorial, that Big Data coupled with predictive analytics will lead to better health care outcomes, is largely unproven. Outside of surgery, the promises of machine learning to increase diagnostic accuracy in ophthalmology and radiology provide a hint of the potential value of Big Data in surgery.17–19 In theory, precision surgery will use a patient’s clinical characteristics to recommend treatment strategies based on the real-world outcomes of thousands of similar patients. The potential of Big Data to transform health care delivery is based on the success of other industries to personalize consumer services in real time.20 Amazon and Netflix use predictive analytics to make recommendations for online shopping and television viewing. The biggest limitation of our proposal is that it assumes that the successful use of Big Data and machine learning in business can be replicated in health care. Although aspirational, the Institute of Medicine sees Big Data as the backbone of future learning health care systems.6 In surgery, the first challenge to operationalizing precision surgery is to create the data infrastructure to take advantage of the data that are currently collected by clinical registries and of the increasing amount of digital data that are digitized in electronic health records.

Creating a system to assign a unique encrypted patient identifier to each person in the United States is a potential solution for connecting the many data silos that currently exist in our health care system, while minimizing privacy risk. Big Data in health care is becoming a reality. The next step will be to draw on the expertise of the bioinformatics community, along with that of registry owners and health policy experts, to test the feasibility of implementing an eUPI in a small group of academic centers and large community hospitals in a single registry. The long-term goal will be for registry owners to launch a broad-based initiative to create an electronic key to begin to unlock the potential of Big Data in health care for the benefit of us all. If successful, this “incremental step” may ultimately help convince Congress to reverse its funding ban on the UPI and would place physicians at the vanguard of efforts to reimagine the US health care system.

Back to Top | Article Outline


Name: Laurent G. Glance, MD.

Contribution: This author helped conceive and design the study, and write and revise the manuscript.

Name: Jonathan P. Wanderer, MD, MPhil.

Contribution: This author helped conceive and design the study, and write and revise the manuscript.

Name: Andrew W. Dick, PhD.

Contribution: This author helped conceive and design the study, and revise the manuscript.

Name: Richard P. Dutton, MD, MBA.

Contribution: This author helped conceive and design the study, and revise the manuscript.

This manuscript was handled by: Nancy Borkowski, DBA, CPA, FACHE, FHFMA.

Back to Top | Article Outline


1. Heitmueller A, Henderson S, Warburton W, Elmagarmid A, Pentland AS, Darzi A. Developing public policy to advance the use of Big Data in health care. Health Aff (Millwood). 2014;33:1523–1530.
2. Levin MA, Wanderer JP, Ehrenfeld JM. Data, big data, and metadata in anesthesiology. Anesth Analg. 2015;121:1661–1667.
3. National Quality Registry Network. Clinical Registries in the United States. Available at Accessed February 1, 2016.
4. Olsen LA, McGinnis M. IOM Roundtable Evidence-Based Medicine: The Learning Healthcare System. 2007.Washington, DCThe National Academies Press.
5. Bilimoria KY, Liu Y, Paruch JL, et alDevelopment and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons. J Am Coll Surg. 2013;217:833–842.
6. Angus DC. Fusing randomized trials with Big Data: the key to self-learning health care systems? JAMA. 2015;314:767–768.
7. Bright TJ, Wong A, Dhurjati R, et alEffect of clinical decision-support systems: a systematic review. Ann Intern Med. 2012;157:29–43.
8. National Anbesthesia Clinical Outcomes Registry. Available at Accessed February 1, 2016.
9. ACS National Surgical Quality Improvement Program. Available at Accessed February 1, 2016.
10. Slabodkin G; Congress Continues to Block Nationwide Unique Patient Identifier. Office of the National Coordinator for Health Information Technology, U.S. Department of Health and Human Services (HHS). Available at Accessed February 1, 2016.
11. Connecting Health and Care for the Nation: A 10-Year Vision to Achieve an Interoperable Health IT Infrastructure. Office of the National Coordinator for Health Information Technology, U.S. Department of Health and Human Services (HHS). Available at Accessed February 1, 2016.
12. Hillestad R, Bigelow JH, Chaudhry B, et alIdentity Crisis: An Examination of the Costs and Benefits of a Unique Patient Identifier for the U.S. Health Care System. 2008.Santa Monica, CA: RAND Corporation.
13. Hipisley-Cox: Validity and Completeness of the NHS Number in Primary and Secondary Care Electronic Data in England 1991–2013. 2013.Nottingham, UKUniversity of Nottingham.
14. Churches T, Christen P. Some methods for blindfolded record linkage. BMC Med Inform Decis Mak. 2004;4:9.
15. Kho AN, Cashy JP, Jackson KL, et alDesign and implementation of a privacy preserving electronic health record linkage tool in Chicago. J Am Med Inform Assoc. 2015;22:1072–1080.
16. Nakashima E. Hacks of OPM databases compromised 22.1 million people, federal authorities say. Washington Post; 2015.
17. Beam AL, Kohane IS. Translating artificial intelligence into clinical care. JAMA. 2016;316:2368–2369.
18. Jha S, Topol EJ. Adapting to artificial intelligence: radiologists and pathologists as information specialists. JAMA. 2016;316:2353–2354.
19. Gulshan V, Peng L, Coram M, et alDevelopment and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA. 2016;316:2402–2410.
20. Parikh RB, Kakad M, Bates DW. Integrating predictive analytics into high-value care: the dawn of precision delivery. JAMA. 2016;315:651–652.
© 2017 International Anesthesia Research Society