Improving diagnostic triage for patients with abdominal pain as a possible symptom of cancer

Patient experiences of testing

Test-related factors

Patient-related factors

Test purpose (e.g. screening, diagnosis)

Type of test

Test protocols

Duration of [image] acquisition

Physical test environment

Socio-demographics

Age

Gender

SES

Race/ethnicity

Comorbidities

Noises

Narrowness of tunnels (e.g. MRI)

Ambient temperature

Previous experiences of testing

Symptomatic/asymptomatic

Degree of pain/discomfort

Health/illness beliefs

Health literacy

Verbal literacy

Ideas, concerns, expectations

Tolerability of procedures

Adverse reactions

Extent to which they feel informed about test

Cultural norms/beliefs

Health system-related

Staff

Manner/interpersonal skills

Cultural/personal sensitivity (e.g. gender discordance, gowning)

Physical handling/manouvreing skill

Extent to which details of test have been communicated to patients & how

Educational attainment

Conceptual Models

IOM 'diagnostic process'

Safer Dx Framework/Instrument (presence/absence of Dx error determined through EMR)

'Soft intelligence' [Mary Dixon-Woods]
The mining/use of 'soft' EHR data (e.g. patient letters) or in-person clinician accounts to generate 'intelligence' on system performances for patient safety

Core (identifies opportunities for Dx error arising from communication between GP-patient & MDT)

Personalising the diagnostic process to align with patients needs, values & preferences

Questions if enough info has been collected to make a diagnosis or optimally guide decisions

Identifies importance of clinicians communicating iterative working diagnoses or health explanations to patients

Identifies patients as possessing vital information that can foster shared decision-making to inform the diagnostic decisions

Identifies strategies to help patients articulate health info, and for clinicians to improve communication and patient understanding

Model for defining diagnostic error and ide opportunities for diagnostic error arising from GP-patient interaction and the broader healthcare team (e.g. specialists, pathologists)

Core (identifies standardised methods of measuring & monitoring diagnostic error through automated EMR alerts)

Consider role of workflow, HCPs, organisational policies & computing infrastructure towards Dx safety & seeks to improve organisational awareness of patient safety & measurement/monitoring of Dx error

Encourages leveraging Safer Dx Instrument to develop tools for measuring/monitoring Dx error, but currently does not incorporate a patient communication component strategy

C. The question is what additional information could be collected from patients that is either not currently, or collected sub-optimally that could lead to improved/safer decision-making?

C. Could it be that patients are unconsciously withholding pertinent information because they have normalised them or attribute them to other benign issues e.g. menstruation, meaning GPs are only getting part of the history?

C. Are patients offering cues (verbal/non-verbal) indicating they wish to share more information that GPs are not following-up on, meaning they are potentially missing relevant context/detail?

C. Could patients' beliefs/preconceived ideas on what they believe to be wrong with them be influencing GPs hypothesis generation/decisions? e.g. are patients offering diagnoses? if so, does GPs subsequent line of questioning suggest they are following patients verbal cues or do they ask additional questions suggesting they are also open to the possibility of other diagnoses?

This information could be used to develop the Safer Dx Instrument, into the Safer Communication Instrument that could be used by GPs in real-time

Safety netting

C. Could a better understanding of the evolution of abdominal Sx & how patients describe Sx improve decisions. Are there key 'milestones' or changes in pain quality or severity that could be important? How do GPs interpret patients descriptions & how do they reconcile them with standard clinical descriptors of pain (e.g. dull ache)

'Safety netting' poorly standardised. GP variation in how it is executed

C. Do patients understand safety netting advice as it is currently executed?

C. Could this information be used to standardise monitoring or 'safety netting' of patients with abdominal pain & how? Do we want to be aiming to identify key features patients should look out for?

C. Challenge is how to standardise safety netting in such a way that it is understood by GPs/patients, protects patient safety, but allows some flexibility for it to be patient/context driven

C. Does safety netting continue indefinitely? Is there a point when a GP can be reasonably reassured cancer is ruled out and from which test or series of 'normal' test results?

Organisational behaviour approach to patient safety. Identifies where processes are breaking down at the grass-roots level & how organisations can alter their processes to mitigate system breakdowns through clear channels between doctors & managers

C. Seems to be centred on sharing intelligence & service innovation, rather than breakdowns at the GP-patient intersection

Deep learning/neural networks

Use of varied data sources to guide decision-making by predicting likelihood of outcomes (predictive modelling)

Offsets against errors in cognition/implicit biases which could lead to erroneous decision making

C. Abdominal pain susceptible to cognitive error due to the vagueness of the Sx. GPs may fail to search for additional information in some scenarios (e.g. female patient <40yrs) automatically attributing a benign based on a single risk factor without properly considering the full clinical picture

Cognitive errors attempted to be overcome may be perpetuated due to the quality of EMR data (e.g. missing data) from which model would be derived

C. What data could be incorporated that would strengthen decision making? Addressed by either identifying data not collected or currently reported in EMR that should be included in EMR (this would facilitate future data mining for development of AI tools) OR by mining existing data in EMR to identify data types that could be leveraged?

C. Can/how can narrative data in EMR be translated into data units that could enhance algorithms/incorporated into AI learning? e.g. word-embedded models

Diagnostic error begins to be made manifest during GP-patient exchange

Patient-centred outcomes (PCOs)
'any outcomes valued by patients, caregivers & clinicians)

Both patients & clinicians contain knowledge, values etc. that are important for making decisions that are in the best interest of the patient

Symptom types

C. What are these outcomes with regards to testing for patients with possible symptoms of cancer?

C. Where are current areas of tension in GP/+patient decision making for tests among patients with non-specific abdominal pain?

e.g. Patients understanding of testing decision + test selection: patients not ordered a test when they think they should be, or sent for a test when they think they should have been ordered a different test...

e.g. GP challenges/perceptions of patients understanding of risk

Testing

Optimizing decisions

C. For patients with abdominal pain, what was the optimal first test ordered & what were the associated patient & clinical characteristics?

Verbal/oral literacy (the ability of patients to describe their health problems/needs, understand spoken health information & respond appropriately)

Patients often don't understand what is being communicated to them verbally due to use of medical jargon/unfamiliar language. To be able to engage with GPs in shared decision-making the ability to make sense of information is a pre-requisite

Patients with low verbal literacy = less likely to engage in shared decision-making

Oral literacy demand (conceptualising key aspects of doctor communication that may strain patient understanding)

Roter et al. 2010:

  1. Use of medical jargon
  2. General language complexity
  3. Contextualized language
  4. Structural characteristics of dialogue

C. Compare testing for patients with abdominal pain in US & UK to determine what the testing strategies were (e.g. single, sequential, multiple etc.); are they the same/different? given patient & clinical characteristics which was the most effective strategy?

Over-testing

C. How much over-testing is required to safely rule out cancer and after which tests?

-ve/normal test results may subconsciously/prematurely reassure GPs (which supposes that the initial working hypothesis was correct). GPs may be reluctant to order further tests as reassured or perhaps concerns about over-testing (i.e. -ve test results often signify overuse)

C. After normal results, what happens next? Do GPs stop testing? What is the plan? Do they order further tests? What factors influence these decisions? What is/is not communicated to patients? What are patients experiences? How is this interpreted by patients?

C. Of the tests that are ordered do GPs receive an alert when no test results are returned? If no alert what happens to these patients and what are the outcomes? What are GP/patient experiences of this? [could be linked to Mary Dixon-Woods 'soft intelligence' and Safer Dx]

For patients with normal results, what happens next: is a follow-up plan communicated to patients, when, via which medium?

How much time should you leave after a normal test result until the next test is ordered?