When it comes to patient identification in health information systems, there is a clear statement of fact: A uniform, standard method of identifying patients does not exist in the United States at this time, said Indranil Ganguly, vice president and chief information officer at Centra State Healthcare System in Freeport, N.J.
Mr. Ganguly discussed the different aspects of patient identification techniques in health IT at AHIMA's 2011 Legal EHR Summit on Aug. 15 in Chicago and what challenges hospitals and healthcare systems are having today without a standardized way of identifying patients in their information systems. He gave an example of a county hospital in Texas with a database of 3.4 million patients. Roughly 250,000 of those patients had the same first and last names, and 70,000 patients shared both names and had the same date of birth. To go a step further: There were 231 Maria Garcias with the same date of birth.
With so much patient information, some of which could be confused for other patient data, Mr. Ganguly said managing the information and maintaining the accuracy of that information is no easy task. He added that unique patient identifiers could alleviate the problem, but patient matching is currently the only method of patient identification.
Unique patient identifiers
Unique patient identifiers are non-changing, alphanumeric keys that uniquely link patients to their data. In 1996, HIPAA required a unique identifier for each individual, employer, health plan and healthcare provider, but after privacy concerns regarding potential data breaches were raised two years later, an appropriations rule prohibited the Department of Health and Human Services from actually exploring a nationwide patient identity solution. As of last year, the Office of the National Coordinator for Health Information Technology required state health IT plans to address health information exchanges — but did not address unique patient identifiers within exchanges. Consequently, Mr. Ganguly said patient matching was left as the only alternative to identify patient data in health IT systems.
Patient matching
Patient matching is a method to identify a patient based on multiple elements, including name, Social Security number, date of birth, gender, address and others. There are three types of methodologies:
• Deterministic: The patient's key data must match exactly. It relies on accurate and consistent data, which is not always the case.
• Probabilistic: The patient's key data is weighted and scored using sophisticated algorithms. While it adjusts for minor data entry errors, it involves complex implementation.
• Fuzzy logic: The patient's key data must match the established logic set within the system.
This is the only way for hospital and health information management staff to achieve patient identification, but the methods have led to significant errors, Mr. Ganguly said. They can lead to false positives, which link a patient to the wrong record, or they can lead to false negatives, which involves the creation of a duplicated record because there was a missing link between a patient and some part of his or her record. The error rates range from 8 to 12 percent in medical databases, Mr. Ganguly said, and that number trends higher in databases with millions of records.
Challenges and future
The challenges with patient matching are fairly obvious, Mr. Ganguly noted. All data elements within an EHR are not always accurate due to human error or methodology variance, and that could lead to bigger problems as systems will look to HIEs in future steps of meaningful use. "We are very concerned about how we are going to positively match our records in the HIE, and they haven't necessarily been resolved," he said.
However, there are challenges for unique patient identifiers as well. A lack of congressional support to revoke the ban on the HHS and ONC to do studies on unique patient identifiers is a major hurdle, as well as the costs associated with assigning every U.S. resident with a unique patient identifier and overcoming privacy concerns.
Mr. Ganguly said the public is currently disengaged and not fully educated on the benefits of patient identifier solutions, but a coalition for an informed patient identity integrity solution was established in the fall of 2010 with the hope that a study could be proposed to and conducted by the Government Accountability Office to look at the effects of a unique patient identifier system. "[Unique patient identifiers] are a very expensive initiative, but the reason we want to implement this system is to improve patient care," Mr. Ganguly said.
3 Steps to Save Lives, Improve Quality from Ascension Health's Dr. Pryor
Portland VA Medical Center Missing 75 Patient Identification Cards
Mr. Ganguly discussed the different aspects of patient identification techniques in health IT at AHIMA's 2011 Legal EHR Summit on Aug. 15 in Chicago and what challenges hospitals and healthcare systems are having today without a standardized way of identifying patients in their information systems. He gave an example of a county hospital in Texas with a database of 3.4 million patients. Roughly 250,000 of those patients had the same first and last names, and 70,000 patients shared both names and had the same date of birth. To go a step further: There were 231 Maria Garcias with the same date of birth.
With so much patient information, some of which could be confused for other patient data, Mr. Ganguly said managing the information and maintaining the accuracy of that information is no easy task. He added that unique patient identifiers could alleviate the problem, but patient matching is currently the only method of patient identification.
Unique patient identifiers
Unique patient identifiers are non-changing, alphanumeric keys that uniquely link patients to their data. In 1996, HIPAA required a unique identifier for each individual, employer, health plan and healthcare provider, but after privacy concerns regarding potential data breaches were raised two years later, an appropriations rule prohibited the Department of Health and Human Services from actually exploring a nationwide patient identity solution. As of last year, the Office of the National Coordinator for Health Information Technology required state health IT plans to address health information exchanges — but did not address unique patient identifiers within exchanges. Consequently, Mr. Ganguly said patient matching was left as the only alternative to identify patient data in health IT systems.
Patient matching
Patient matching is a method to identify a patient based on multiple elements, including name, Social Security number, date of birth, gender, address and others. There are three types of methodologies:
• Deterministic: The patient's key data must match exactly. It relies on accurate and consistent data, which is not always the case.
• Probabilistic: The patient's key data is weighted and scored using sophisticated algorithms. While it adjusts for minor data entry errors, it involves complex implementation.
• Fuzzy logic: The patient's key data must match the established logic set within the system.
This is the only way for hospital and health information management staff to achieve patient identification, but the methods have led to significant errors, Mr. Ganguly said. They can lead to false positives, which link a patient to the wrong record, or they can lead to false negatives, which involves the creation of a duplicated record because there was a missing link between a patient and some part of his or her record. The error rates range from 8 to 12 percent in medical databases, Mr. Ganguly said, and that number trends higher in databases with millions of records.
Challenges and future
The challenges with patient matching are fairly obvious, Mr. Ganguly noted. All data elements within an EHR are not always accurate due to human error or methodology variance, and that could lead to bigger problems as systems will look to HIEs in future steps of meaningful use. "We are very concerned about how we are going to positively match our records in the HIE, and they haven't necessarily been resolved," he said.
However, there are challenges for unique patient identifiers as well. A lack of congressional support to revoke the ban on the HHS and ONC to do studies on unique patient identifiers is a major hurdle, as well as the costs associated with assigning every U.S. resident with a unique patient identifier and overcoming privacy concerns.
Mr. Ganguly said the public is currently disengaged and not fully educated on the benefits of patient identifier solutions, but a coalition for an informed patient identity integrity solution was established in the fall of 2010 with the hope that a study could be proposed to and conducted by the Government Accountability Office to look at the effects of a unique patient identifier system. "[Unique patient identifiers] are a very expensive initiative, but the reason we want to implement this system is to improve patient care," Mr. Ganguly said.
Related Articles on Patient Identification:
Colorado Healthcare Agency Loses Personal Data3 Steps to Save Lives, Improve Quality from Ascension Health's Dr. Pryor
Portland VA Medical Center Missing 75 Patient Identification Cards