Home | Maturity Model for ML Systems in Clinics

Organization

For each question, please select the attribute most accurately describing the situation in your clinic.

Medical directors' ML support

Level of medical directors' ML support

No ML support of medical director

Low ML support of medical director

Medium ML support of medical director

High ML support of medical director

Extensive ML support of medical director and dedicated ML representatives (e.g., Chief AI Officer)

ML strategy

Stage of ML strategy development

No ML strategy in planning

ML strategy planned but not yet developed

ML strategy is under development

ML strategy is under implementation

ML strategy is largely implemented and supported by all stakeholders (e.g., medical directors, physicians)

Clinic's resources for ML

Clinic's technical infrastructure

Unconnected, outdated software systems

Solutions for connecting outdated software systems for data exchange within the clinic

Up-to-date software systems connected for data exchange within the clinic

Up-to-date software systems connected for data exchange within the clinic and with selected partners (e.g., other clinics, academic institutions)

Up-to-date software systems deeply integrated in data networks for regular and extensive data exchange within the clinic and external institutions (e.g., other clinics, academic institutions)

Clinic's financing structure

Strict budgetary constraints exclusively for medication and large medical equipment

Exceptions to budget constraints

Partially free allocation of budget

Free allocation of budget

Dedicated ML budget

Clinic's medical and ML methods expertise

Medical expertise only and no ML methodological expertise

Medical expertise and first medical experts with self-developed ML methodological expertise in data science

Medical expertise and first dedicated ML method experts in data science

Medical expertise and dedicated ML method experts in data science and engineering

Medical expertise, dedicated ML methods experts in data science and engineering, and all have insights into each other’s area of expertise (e.g., through trainings, academic courses)

Adopter system

For each question, please select the attribute most accurately describing the situation in your clinic.

Physicians' ML acceptance

Physicians' attitude towards ML

Scepticism of physicians towards ML

Conservative attitude of physicians towards ML

Pragmatic attitude of physicians towards ML

Openness of physicians towards ML

Enthusiasm of physicians towards ML

Patients' ML acceptance

Patients' attitude towards ML

Scepticism of patients towards ML

Conservative attitude of patients towards ML

Pragmatic attitude of patients towards ML

Openness of patients towards ML

Enthusiasm of patients towards ML

Patient data

For each question, please select the attribute most accurately describing the situation in your clinic.

Availability of patient data

Digitization of patient data

Non-digitized data

Low level of data digitization (e.g., only central organizational process)

Medium level of data digitization (e.g., central diagnostic values)

High level of data digitization (e.g., time series data on the course of diseases)

Fully digitized data

Unified formats

Many different proprietary data formats that are not convertible

Many different proprietary data formats, some of which are convertible

Many different but common data formats that are convertible

Many common data formats

Unified standards for medical data formats (e.g., DICOM, FHIR)

Data quality standards

No uniform quality of data (e.g., handwritten, synonym-biased diagnoses)

First requests to unify quality of data within clinic

Clinic-internal quality standards (e.g., for key terminologies)

Inter-clinical quality standards (e.g., for key terminologies)

International quality standards (e.g., UMLS)

Degree of information value despite anonymization

Non-anonymized personal data

Anonymized data, with almost complete loss of diagnostically relevant information value

Anonymized data, with low level of diagnostically relevant information value

Anonymized data, with medium level of diagnostically relevant information value

Anonymized data, with preserved diagnostically relevant information value

Representativeness of training data

Nearly no match of demographics, region, and work culture in training data related to patient base and clinic

Low match of demographics, region, and work culture in training data related to patient base and clinic

Medium match of demographics, region, and work culture in training data related to patient base and clinic

High match of demographics, region, and work culture in training data related to patient base and clinic

Full match of demographics, region, and work culture in training data related to patient base and clinic

Determine your clinic's readiness for ML-supported diagnostics

Organization

Medical directors' ML support

ML strategy

Clinic's resources for ML

Adopter system

Physicians' ML acceptance

Patients' ML acceptance

Patient data

Availability of patient data