CAT Project

The Harvard ME/CFS Collaboration at Harvard-Affiliated Hospitals

Given meager research funding and the absence of a consensus on disease mechanism, there has been no definition or single set of criteria that has been validated to make a ME/CFS diagnosis. Therefore, we propose to develop a patient-driven tool named “Personalized Automated Symptom Summary (PASS)” that is intended to aid a clinician more efficiently to define the character and priorities of a patient’s current symptoms of ME/CFS, Post-treatment Lyme Disease (PTLD), or Fibromyalgia (FM). 

The overall strategy is to reduce the time spent by a clinician to perform a patient’s evaluation. The intent is to provide a clinical quantitative metric that can be most efficiently administered, be highly reproducible, and address each of the many possible symptomologies. This approach will provide a highly valid means to support the diagnosis, follow the progression or remission of an individual patient’s disease, as well as provide a metric to assess whether treatment interventions are effectively improving symptoms within an individual patient or within a particular cohort in a clinical trial. 

Read more


Modern science now can alleviate ambiguity and irrelevancy in questions using computer adaptive tests (CAT) with artificial intelligence powered by machine learning to streamline the task of precisely determining those symptoms from the eyes of an  individual patient. This is much like what the individual clinician does in their review with the patient of symptoms when determining a diagnosis.

Using PASS, the patient can spend 30 minutes or more (in one or more sessions) to create a symptom profile that accurately describes their individual current symptoms (including the symptom characteristics and its priorities from their own perspective) in preparation for their upcoming clinician visit. From the patient’s perspective, there is a multitude of different symptom categories in chronic, complex diseases. This patient summary (and also a clinician version) will then be readily available for their clinician to review when the patient arrives for his or her visit. This will assist the clinician to focus their Review of Symptoms (ROS), which is part of every history and physical examination, on those symptoms most important to the patient. The clinician might choose to corroborate the symptom profile reflected by PASS and emphasized in the summary as well as ensure that other relevant symptoms may or may not be present. The product will be a picture of the symptoms important to the patient for use by the clinician. 

By usage of artificial intelligence with an item response theory (IRT) based approach coupled with machine learning to develop the algorithms for the CAT, PASS will perform much like a clinician would in their ROS. For example, if a patient does not have headaches, then neither the clinician nor PASS would require the patient to answer subsequent questions about headaches but simply move on to other possible symptoms. In addition and importantly, the symptom based uni-dimensional domains that are present will reflect metrics that are determined through a few questions posed among many available in the item bank. Each item responded to will inform the system as to what is the next item to be posed by the computer CAT system using artificial intelligence. In most cases of chronic, complex diseases, an ROS can take up to an hour or more if performed exhaustively. Often because of throughput performance, time constraints, and other efficiencies of today’s medical practice, the ROS becomes truncated and can miss important features of chronic, complex diseases. This can lead to frustration for both the patient and their clinician. The PASS is intended to focus the ROS from the patient’s perspective and to reduce the possibility for frustrations that might be introduced because the patient feels that they have not had the opportunity to share their full symptom complex with their clinician. 

The utility of this patient tool, PASS, will be to facilitate a ROS as efficiently and completely as possible by the clinician, particularly one who is not familiar with chronic, complex diseases like ME/CFS. There will be a principal version of the PASS. The CAT is based upon a large item bank across many symptoms requiring a selected subset of items that are posed by the computer. The CAT is the complete profile that can be computer administered on the basis of an extensive item bank but by posing only a few questions from each of the active symptom domains and then can assist the clinician to accurately make an appropriate diagnosis. The CAT is administered by computer focusing only on those current, active symptoms and in the order of the patient’s symptom priorities. For follow up clinician visits, the CAT can be used to follow the patient’s disease progression and any treatment effectiveness or ineffectiveness. It can also be used as the quantitative outcome assessment tool for clinical trials. The system will begin with items posed for those symptoms that were present on prior visits. For symptoms not present on a prior visit, the computer will pose screening questions that indicate their absence. While for other symptom categories, they will be introduced for those emerging for the first time. The uniform usage of the PASS will facilitate comparisons between the results of the multiple clinical trials by providing a consistent and reproducible symptom outcome metric tool. This tool can be easily automated for applications on laptops, hand-held devices, and cellular phones.

Development of the PASS Tool

The development of this tool requires many consensus meetings (maybe a dozen or more) that include 30 or more patients, clinician, and scientists all of whom are familiar with the diseases and the tools required to create the PASS. These meetings are designed to review all existing questionnaires, criteria, and other tools currently used in evaluation of ME/CFS, PTLD, and FM. The intent is to adapt questions from all currently available tools and items developed de novo from focus groups into an extensive database whose content has been fully developed and supported by clinicians and patients. These extensive questions need to be field tested by thousands of ME/CFS, PTLD, and FM patients using online methodologies and the results evaluated statistically to find those questions or combinations of questions that display accurate and precise psychometric performance. 

Further consensus meetings are required to narrow down to only those questions with superior performance to be included in the version 1.0 of the PASS profile. This beta version (1.0) will have been field tested in a calibration phase using advanced psychometric IRT based approaches guided by machine learning artificial intelligence techniques with thousands of patients online. These results will then be used to create a version 1.0 that can be considered the first major product of PASS. Subsequently, the item bank can be tested in a separate online test again with thousands of subjects for reliability and validity testing for future versions of the PASS profile. This will represent an important milestone for the establishment of the official version. We all recognize that subsequent versions of PASS will be required as various updates will continue to improve the product.


The time to beta version 1.0 depends on the resources committed to this project as well as the level of commitment of clinicians, patients, and scientists to accomplish the initial consensus meetings within a concise time frame. With sufficient resources and commitments, it is possible that the initial beta version could be available as early as 24 ⎼ 30 months after the project begins. The official tested beta version 1.0 could then be expected within the 36 ⎼ 48 month time frame.


The Institute of Medicine (IOM), now the National Academy of Medicine (NAM) created a new clinical diagnostic criteria in 2015 as shown in the figure to the right. The new diagnostic criteria by the NAM requires three required symptoms: 

A substantial reduction or impairment in the ability to engage in pre-illness levels of activity (occupational, educational, social or personal life) that: lasts for more than 6 months; and is accompanied by fatigue that is: often profound; of new onset (not life-long); not the result of ongoing or unusual excessive exertion; not substantially alleviated by rest

Post-exertional malaise (PEM) – worsening of symptoms after physical, mental or emotional exertion that would not have caused a problem before the illness.PEM often puts the patient in relapse that may last days, weeks, or even longer

Unrefreshing sleep – patients with ME/CFS may not feel better or less tired even after a full night of sleep despite the absence of specific objective sleep alterations. 

In addition to those three required symptoms, at least one of the following manifestations must also be present:

1) Cognitive impairment – patients have problems with thinking, memory, executive function, and information processing, as well as attention deficit and impaired psychomotor functions;

2) Orthostatic intolerance – patients develop a worsening of symptoms upon assuming and maintaining upright posture as measured by objective heart rate and blood pressure abnormalities during standing, bedside orthostatic vital signs, or head-up tilt testing. The diagnostic algorithm shown above is to be used as an assistive device to determine whether a patient’s condition can be labeled as a ME/CFS diagnosis.

Fibromyalgia & ME/CFS

Fibromyalgia is a similar condition, if not the same, that causes widespread muscle and joint pain, sleep and memory problems, fatigue, and numerous other somatic symptoms. Similar to ME/CFS, the etiology of fibromyalgia is unknown, making it difficult to diagnosis and treat. Unlike ME/CFS, the American College of Rheumatology (ACR) has created a list of diagnostic criteria that is frequently considered to be more accurate than any for ME/CFS and has been validated. Other questionnaires, most notably the Fibromyalgia Impact Questionnaire (FIQ), have been developed that assess the health status of individuals who have been clinically diagnosed with fibromyalgia.

Post Treatment Lyme Disease (PTLD) and ME/CFS

Unlike ME/CFS and fibromyalgia, PTLD frequently has a clearer pathology and in many cases can be serologically diagnosed.  However, like these other conditions, symptoms of PTLD have been reported to be chronic. Notably, patients with PTLD are often diagnosed with chronic fatigue syndrome, fibromyalgia, and/or rheumatoid arthritis since the symptoms are similar. 

Given that fatigue is a prominent symptom in each of these diseases, these patients are often misdiagnosed.

Questionnaires for ME/CFS

In addition to the NAM diagnostic criteria and algorithm, several questionnaires have been developed to measure symptoms and assess ME/CFS patients, perhaps the most inclusive of which is the DePaul Symptom Questionnaire (DSQ). The reliability and validity of the DSQ has been tested in three studies and is also supported by another study that investigated which patient-reported outcomes measurement information system (PROMIS) measures are useful in patients with ME/CFS. However, the PROMIS measures for fatigue are uni-dimensional and are not disease or symptom specific as they are less granular in what they are measuring. Despite using different clinical criteria and addressing a number of major symptoms of ME/CFS, Murdock et al concludes that an improved PROMIS tool for ME/CFS needs to be developed.

Participation in the Studies

This project has not yet begun. Once the study is initiated, a consensus group will be formed including clinicians knowledgeable in ME/CFS, PTLD, and FM, scientists familiar with Item Response Theory, Computerized Adaptive Testing, Machine Learning, ME/CFS, PTLD, and FM patients, as well as care providers, nurses, social workers, therapists, and many other stakeholders. Extensive data collection will be pursued to create a calibrated item pool to begin the process. Once the program evolves, thousands of patients will be recruited for online participation in the field testing and pilot testing for the various phases of the study.