Critical situations in hospitals arise especially during interdisciplinary cooperation and the use of complex medical technology, for example during surgical procedures or in the perioperative area. In particular, overlooking medically relevant treatment data or an incomplete medical history can lead to incorrect treatment. Within the OntoMedRisk project (funded by the German Federal Ministry of Education and Research), methods and tools were developed to identify and analyse risks in perioperative processes. The developed software suite is based on the Risk Identification Ontology (RIO), which provides a framework for risk specification and reasoning. RIO utilises the ontological model of the risk notion defined in the publication and is founded on the General Formal Ontology (GFO).
The starting point of the modelling is the treatment process, which may consist of different treatment phases (gfo:has_part). The whole treatment as well as the phases are complex processes (gfo:Situoid). The treatment has (gfo:projects_to) a temporal extension, the treatment time (gfo:Chronoid). At each point in time (gfo:Time_boundary) of the treatment, exactly one treatment situation (gfo:Situation) exists (gfo:exists_at). A point in time of the treatment is a boundary of the treatment time (gfo:boundary_of), whereas the corresponding treatment situation is a boundary of the treatment itself. For each treatment phase, specific points in time of risk detection can be defined. The treatment situations existing at these points in time are analysed for the presence of risks. Such situations are called potential risk situations because they do not necessarily contain risks. Situations, and in particular treatment situations, possess different properties (gfo:Property). In the context of risk identification, these properties are called Key Performance Indicators (KPI). They can belong to the situation itself, but also to the participants, such as doctors, medical instruments and patients. Such properties are also considered properties of the current treatment situation (gfo:has_property). According to the definitions in the publication, a particular combination of properties (a composite property) of a potential risk situation (e.g., patient age = 3 months, meningitis vaccination = false and antibiotic prevention = false) may indicate a risk, i.e., could lead to an adverse situation.
The current version of RIO, is founded with the new lightweight version of GFO, called GFO-light, which is designed as a framework for the efficient development and foundation of domain/application ontologies. By using GFO-light, some parts of RIO could be simplified, i.e., modelled with fewer entities (classes, properties, individuals).
Fig. 1. RIO - GFO-light embedding (overview)
The main difference to the original version of RIO is a new interpretation of the notion of situation. Whereas in the original RIO, the treatment situations possessed (gfo:has_property) some properties (including composite risk properties), i.e., were bearers of properties, a situation in GFO-light (and correspondingly also in the new RIO) is understood as a combination of properties (gfo-light:Attributive), i.e., a situation consists of (gfo-light:hasSituationPart) attributives.
The class Risk could be omitted. The individual properties do not have to be encapsulated in composite risk properties, but can be assigned directly to the treatment situation. Each subclass of Risk_situation (e.g., Dura_mater_infection_risk_situation) represents a risk type by restrictions of the properties determining the risk of this type (e.g., patient age ≤ 5 months, meningitis vaccination = false and antibiotic prevention = false). Potential risk situations containing these properties in the defined ranges are classified automatically as risk situations of corresponding types.
A further simplification is that in GFO-light points in time of a process can be assigned directly to the process itself (see temporalPartOf in Fig.1), without a detour via its temporal extension (gfo:Chronoid).
In addition, some object properties from GFO-light (e.g., gfo-light:processPartOf, gfo-light:situationPartOf and gfo-light:leadsTo/resultsFrom) could be reused so that it was no longer necessary to define similar properties (risk_in_phase, risk_for_adverse_situation and succeeding_situation) in RIO.
A small example (examples/rio-example.ttl) illustrates the identification (reasoning) of risks for a bacterial infection during cochlear implantation in infants. Five KPIs are defined (as subclasses of the class KPI), each of which has a subclass with a property restriction (Tab. 1).
| KPI class | KPI subclass | KPI restriction |
|---|---|---|
| Age_in_months | Age_in_months_l_5 | Age_in_months and (decimalValue some xsd:decimal[>= 0 , < 5]) |
| Antibiotic_prevention | Antibiotic_prevention_e_false | Antibiotic_prevention and (booleanValue value false) |
| Ear_structure | Ear_structure_e_abnormal | Ear_structure and (stringValue value "abnormal"^^xsd:string) |
| Skull_bone_thickness | Skull_bone_thickness_le_2 | Skull_bone_thickness and (decimalValue some xsd:decimal[>= 0 , <= 2]) |
| Vaccination | Vaccination_e_false | Vaccination and (booleanValue value false) |
Tab. 1. Example KPIs with restrictions
The class Dura_mater_infection_risk_situation (a subclass of Treatment_situation/Risk_situation) has the following restriction (risk identification rule):
(
(hasSituationPart some Age_in_months_l_5) or
(hasSituationPart some Ear_structure_e_abnormal) or
(hasSituationPart some Skull_bone_thickness_le_2)
)
and (hasSituationPart some Antibiotic_prevention_e_false)
and (hasSituationPart some Vaccination_e_false)
Situations that fulfil this rule (i.e., required KPIs are within the defined ranges) are automatically classified in the class Dura_mater_infection_risk_situation by the reasoner. The treatment situation instance rs1_pat1, for example, has the assertions shown in the Tab. 2, fulfils the risk rule/restriction and is therefore classified in the class Dura_mater_infection_risk_situation.
| Object property assertion | Data property assertion |
|---|---|
| hasSituationPart vaccination_false_pat1 | booleanValue false |
| hasSituationPart age_4_pat1 | decimalValue 4 |
| hasSituationPart antibiotic_false_pat1 | booleanValue false |
Tab. 2. Example instance (rs1_pat1) of the class Treatment_situation. In the first column are object property assertions of rs1_pat1, in the second column - the data property assertions of the corresponding KPI instance.
Uciteli A, Neumann J, Tahar K, Saleh K, Stucke S, Faulbrück-Röhr S, Kaeding A, Specht M, Schmidt T, Neumuth T, Besting A, Stegemann D, Portheine F, Herre H. Ontology-based specification, identification and analysis of perioperative risks. Journal of Biomedical Semantics. 2017;8:36. https://doi.org/10.1186/s13326-017-0147-8