Two BIDS faculty members, Drs. Ahmed Hassoon and Harold Lehmann, led the development of a computable phenotype for Symptom-Disease Pair Analysis (SPADE) that was used by the Agency for Healthcare Research and Quality (AHRQ) to develop a tool for diagnostic excellence. The tool is designed to support research related to potentially avoidable diagnostic safety events using administrative claims data.
Dr. David Newman-Toker at the Armstrong Institute Center for Diagnostic Excellence led the development of the SPADE framework in 2018, and his team worked further on statistical methods and validations studies for the framework. Dr. Lehmann guided the team to create the information model to enable standard SPADE application in EHR data or claims data at scale.
In 2025, AHRQ, part of the U.S. Department of Health and Human Services, added the SPADE Research Tool to their Diagnostic Excellence toolkit. The SPADE methodology offers a systematic approach for identifying hospital admissions that may follow ambulatory visits (such as to the Emergency Department) for related symptoms.
When AHRQ informed the Hopkins team that their research would help develop the diagnostic excellence package, they provided resources and met with their technical team to explain the method in detail. Because of the prior framework design, statistical modeling, the computable phenotype, information model, and case studies to prove that the tool worked, the process of working with AHRQ was smooth.
The other Johns Hopkins team members who developed SPADE include Drs. Hetal Rupani, Susan Peterson, Michelle C Johansen, Kathy McDonald, J. Matthew Austin, and David E. Newman-Toker.
According to Dr. Lehmann, “With AI and large language models taking center stage, it is important that we retain the specificity and exactitude needed to translate from data to clinical action. Activities like information models and computable phenotypes ensure that shortcuts are not taken and that patients are not harmed. In the domain of patient safety and diagnostic excellence, this assurance is even more important.”
The SPADE Research Tool includes four pre-defined SPADE symptom-disease pairs focused on acute myocardial infarction (AMI) and stroke, applying both look-forward and look-back approaches. For each disease, these specifications estimate two “observed” rates:
1. The rate of inpatient admissions for the given condition (AMI, stroke) occurring within 7 or 30 days1 after ED visits for potentially-related symptoms (look-forward approach)
2. The rate of inpatient admissions for the given condition (AMI, stroke) that were preceded by ED visits for potentially-related symptoms within the prior 7 or 30 days (look-back approach).
The SPADE Research Tool may help identify cases in which there was a previous visit to the emergency department for a symptom that is associated with the diagnosis. In many cases, a prior ED visit with a symptom from a symptom-disease pair may represent an unrelated diagnosis or condition, but in some cases, an AMI or stroke diagnosis may have been missed at a prior visit. This tool will help identify the frequency with which these symptom-disease pairs have occurred in a health system as well as the specific cases in which these symptom-disease pairs occurred for further analysis. With this information in hand, health systems may be able to develop guidelines and intervention solutions that can reduce the potential for missed diagnoses.
Dr. Hassoon stated, “Research translation takes time and patience. We knew since the start that our work had translation potential, so we worked on providing the key ingredients to enable translation. Unfortunately, the research area of diagnostic errors is still hugely underfunded in comparison to the level of harm it can cause. We hope that this work will bring more resources to further push our research to eliminate diagnostic errors and subsequent patients’ harm. We are very excited but also cautious about how AI will impact diagnostic errors. My current goal is to build AI systems that can detect and flag diagnostic errors.”
The SPADE tool can now be used by hospital systems, researchers, and others to identify events that could potentially represent a missed opportunity for diagnosis. It is intended primarily for research within individual hospitals or health systems, and for identifying cases that may or may not be instances of diagnostic error.
To read more about the SPADE Research Tool including the technical specifications, user guide, and quick start guide, visit the AHRQ Diagnostic Excellence tools website.