Keywords
Nursing informatics - environmental health - standardized nursing terminology - systematized
nomenclature of medicine - artificial intelligence
1 Introduction
Our health is interconnected with the air, water and land in which we live and work.
However, environmental risks are not evenly distributed, as vulnerable populations
may have a higher risk of exposure to air pollution, noise pollution and hazardous
chemicals due to pre-existing health conditions, poor nutritional status and other
environmental risk factors. In 2015, United Nations (UN) Member States formally adopted
17 Sustainable Development Goals (SDGs) to achieve by the year 2030. These goals provided
a shared plan for the health, well-being, and prosperity of people and the planet.
For example, SDG 3.9.1 considers the mortality rate attributed to ambient and household
air pollution. Further SDG indicators consider the intersection of sustainability
and human existence related to health, poverty, pollution, water, education, gender
equality, energy, and economic growth [[1]]. To further these goals, various disciplines have come together to collaborate
on the concept of One Health.
One Health is a collaborative, multisectoral approach which highlights the absolute
interconnectedness between human health and the complex ecosystems in which we live
and share [[2], [3]]. This transdisciplinary approach is an important consideration as human populations
are expanding into new geographic areas which has led to potential opportunities for
diseases to pass between animals and people. Each year, millions of people are impacted
by antimicrobial-resistant germs, vector-borne diseases, and ill health related to
the contamination of water. Weather changes impacting temperature and humidity are
also associated with adverse outcomes related to respiratory, cardiovascular, and
neurological conditions and infectious diseases [[4]
[5]
[6]
[7]].
Environmental sustainability further informed this project. Morelli [[8]] describes environmental sustainability “…as a condition of balance, resilience,
and interconnectedness that allows human society to satisfy its needs while neither
exceeding the capacity of its supporting ecosystems to continue to regenerate the
services necessary to meet those needs nor by our actions diminishing biological diversity.”
The question remains how well healthcare is advancing towards meeting environmental
sustainability, One Health and the SDGs. One such area where healthcare clinician
contributions can be leveraged in this evaluation is through secondary data reuse
of clinical documentation. Clinicians assess, diagnose, provide interventions and
evaluate the care of people and communities affected by environmental impacts. In
many settings, this documentation is captured in electronic format with embedded standardized
clinical terminologies. Standardized clinical terminologies are databases of clinical
terms that have computer-readable codes, are arranged in a polyarchy with defined
relationships, have conceptual uniqueness and permanence, and can expand through multiple
granularities [[9]]. As a mechanism to support data reuse, standardized clinical terminologies can
be used to facilitate the aggregation of clinical data across disparate settings and
applied in larger-scale evaluations. Two healthcare terminology standards often used
in this capacity, are the Systematized Nomenclature of Medicine Clinical Terms (SNOMED
CT), a multilingual clinical health terminology [[10]] and the International Classification for Nursing Practice (ICNP), a classification
of nursing diagnosis, interventions, and outcomes specifically describing nursing
practice [[11]]. These standardized terminologies are systematically organized and are computer-readable
collections of medical terms that are defined, codified, and have synonyms associated
with easier cross-mapping and comparison across systems and practice settings [[12]]. In 2021, the International Council of Nurses (ICN) and SNOMED International announced
an agreement whereby ICNP would be managed and distributed by SNOMED International
while content ownership would remain with ICN [[13]]. To support this work, reference sets and tables have been developed to describe
the mapping equivalence between these two terminologies.
While much work has been advanced to represent healthcare concepts in SNOMED CT and
ICNP, it is unknown if the suite of associated health concepts, represented in these
SDGs, exists in these systems. Therefore, this study aimed to evaluate the representation
of environmental concepts associated with health impacts in SNOMED CT and ICNP.
2 Objectives
The objective of this study was to evaluate the representation of environmental concepts
associated with health impacts, in standardized clinical terminologies.
3 Methods
This study used a descriptive methodology informed by Block et al. [[14]] procedural framework for standardized clinical terminology mapping ([Figure 1]). As per the Canadian Tri-Council Policy Statement for Ethical Conduct for Research
Involving Humans-Chapter 2 [[15]], this research study was excluded from requiring research ethics board approval,
and as such, no formal ethics application was conducted.
Fig. 1 Procedural Framework for Standardized Clinical Terminology Mapping (inspired from
[[14]]).
3.1 Source Concepts
The Global Indicator Framework for the Sustainable Development Goals and Targets of
the 2030 Agenda for Sustainable Development document was chosen as the source reference
material to perform concept extraction [[1]]. Developed by the Inter-Agency and Expert Group on SDG Indicators, it contains
a descriptive list of targets and indicators. To define source concept inclusion,
this research group focused on the Good Health and Well-Being SDG and those SDG indicators
whose custodial agency was, or, in discussion to be, the World Health Organization
[[16]]. The remaining indicators were excluded from this study.
3.2. Source Concept Extraction Protocol
A total of 41 SDG indicators were identified as health-related from the source document
of the Global Indicator Framework document. These indicators were then divided among
five nursing researchers (established & emerging) for concept extraction. For this
paper, a concept targeted for extraction was defined as the identification of unique,
explicit concepts which correspond to a singular meaning and term/phrase [[9]]. The researchers a) reviewed the assigned indicator; b) identified clinical concept(s);
c) categorized the concepts into a domain type (as per SNOMED CT) [[17]]; and d) added possible synonyms or extensions as interpreted by their own clinical
knowledge and expertise. These were then reviewed by the research team for verification
of completeness and accuracy. Disagreement was discussed until consensus was achieved.
3.3 Conceptual Modeling
The researchers assigned domain type(s) to facilitate an understanding of the orientation
and descriptive meaning of the extracted concepts (e.g., the extracted concept was “tuberculosis” and the domain type was “diagnosis”). The
purpose of this step was to embed a dimension of meaning, awareness, and boundary
to the conceptual phenomena the symbolized ‘word' was interpreted to portray [[18], [19]]. The domain types were informed by the conceptual domains used in SNOMED CT and
ICNP to arrange health concepts into the terminologies of polyhierarchical ontology
structures [[20]].
3.4 Target Terminology
The target terminologies were: 1) SNOMED CT International Edition (manual: 2022-09-30
version; automated 2022-07-31 version), and 2) ICNP 2019 Edition. All the extracted
source concepts were prepared for SNOMED CT mapping. Given the scope of ICNP, only
source concepts represented as diagnosis, interventions, and outcomes were included.
3.5 Mapping
The following mapping procedures were sequentially applied in this study:
-
Mapping Coordination: Pre-coordination;
-
Mapping Style: Manual and Automated (SNOMED CT); Manual (ICNP);
-
Hierarchical Awareness: All extracted concepts had potential alignment to existing
hierarchies in SNOMED CT and were therefore candidates for SNOMED CT mapping. All
extracted concepts aligned to diagnosis, intervention and outcomes had potential alignment
to existing hierarchies in ICNP and were therefore candidates for ICNP mapping;
-
Systematic Search Strategy (see below);
-
Cardinality (see below).
3.5.1 Systematic Search Strategy
Both manual and automated methods were used. These methods were chosen as a means
to leverage the strengths of human and computer mapping selection, as well as, control
for possible selection bias [[19], [21], [22]]. During the manual mapping phase, extracted concepts were uploaded into an excel
spreadsheet and divided among four researchers where each concept was independently,
manually mapped twice. [Table 1] outlines the systematic search strategy used in each terminology online browser.
Table 1 Systematic Search Strategy.
During the automated mapping phase, extracted concepts were uploaded into an excel
spreadsheet. All punctuation was removed to facilitate CSV upload to the automated
mapping tool Snap2SNOMED CT [[23]]. All concepts were tasked as an ‘automap', whereby the Snap2SNOMED CT software
automatically assigned candidate matches.
3.5.2 Cardinality
The types of manual mapping matches were defined as:
-
Direct match means that the concept matches exactly to the concept found in SNOMED
CT or ICNP;
-
Broader than match means that the concept found in SNOMED/ICNP was conceptually broader
(more general) than the concept found in the source concept list;
-
Narrower than match means that SNOMED/ICNP was conceptually narrower (more specific)
than the concept found in the source concept list;
-
No match means there were no matches in the target terminology (i.e., SNOMED CT, ICNP).
The cardinality of matches produced through the automated mapping tool did not include
comments of match type, and was, therefore, not applicable.
3.6 Evaluation
The candidate lists, from manual mapping and automated mapping, were presented to
two researchers who were not part of the manual or automated mapping phases. The researchers
were tasked to review the extracted concept, and the candidate mapping matches, and
then choose the ‘best' fit mapping and cardinality match(es). The two team members
took note of any concepts or mapping matches which were ambiguous or required further
review. The results of this work were then reviewed by several members of the researcher
group over three meetings, where fit results, alignment, and ambiguity were discussed.
This discussion iterated, refined and verified the final mapping list.
4 Results
A total of 169 concepts were initially extracted from the 41 source SDG indicators
and used as candidates for SNOMED CT mapping. Of these 169 concepts, 31 were given
a domain related to a diagnosis, intervention, or outcome and were therefore also
mapped to ICNP. Through iterative review and group consensus, a total of 119 concepts
with 133 mapping matches were added to the final SNOMED CT mapping list ([Table 2]). Similarly, a total of 26 concepts with 27 mapping matches were added to the final
ICNP mapping list. Of these 133 SNOMED CT mapping matches, 53 (39.8%) were direct
matches, 37 (27.8%) were narrower than matches, 35 (26.3%) were broader than matches,
and 8 (6%) had no matches. Of these 27 ICNP mapping matches, 8 (29.6%) were direct
matches, 4 (14.8%) were narrower than, 7 (25.9%) were broader than, and 8 (29.6%)
were no matches. Of note, some of these concepts were repeated through the extraction
phase of the 41 SDG indicators (e.g., mortality rate) and for the purposes of reporting
SNOMED CT mapping results, these repeats were counted. When no matches and repeated
concepts were removed, a total of 107 unique concepts were represented in SNOMED CT.
Table 2 Final Mapping Results between UN Sustainability Concepts, SNOMED CT, and ICNP.
5 Discussion
Informed by the principle of One Health, whereby health is linked to the ecosystems
with which we live, this study sought to explore the representation of environmental
concepts associated with health impacts, in standardized clinical terminologies. We
chose to extract concepts described in the UN Global Indicator Framework for the Sustainable
Development Goals and Targets and mapped them using a structured process, to SNOMED
CT and ICNP. The multiple methods used added strength to our approach (e.g., manual and automated mapping). As well, our international research team added a
depth of meaningful interpretation through the lenses of varied clinical nursing,
research, and lived experiences, aiding in the progression of concept extraction,
synonym development, mapping selection, and iterative evaluation.
5.1 Mapping Results
Overall, our mapping processes found that the extracted concepts were moderately well
represented in SNOMED CT and ICNP. Of the 133 SNOMED CT mapping matches found, 53
(39.8%) were direct matches, 37 (27.8%) were narrower than matches, 35 (26.3%) were
broader than matches, and 8 (6%) had no matches. Of those concepts which were direct
matches, 14 (10.5%) were of the disorders (diagnosis) ontological axis. In contrast,
only 1 (0.7%) broader than and 4 (3%) narrower than mapping matches were of the disorders
axis. This strong coverage may be related to the generally well-developed diagnosis
content of SNOMED CT [[21], [35], [36]]. Of the 27 ICNP mapping matches, 8 (29.6%) were direct matches, 4 (14.8%) were
narrower than, 7 (25.9%) were broader than, and 8 (29.6%) were no matches. Of these,
4 (15%) of the no matches were related to specific medical diagnoses (e.g., tuberculosis, malaria, hepatitis B, and HIV) and would unlikely be candidates for
specific ICNP inclusion (i.e., this was an anticipated ICNP mapping finding). One of the ICNP matches was equivalent
to that found in SNOMED CT (i.e., substance abuse). This was expected given the work to cross-reference ICNP in SNOMED
CT. However, there were several concepts chosen during the manual mapping process,
which were different. For example, for the source concept “chronic respiratory disease”,
the mapping match in SNOMED CT was “chronic disease of respiratory system (disorder)”,
while the ICNP match was “Impaired Respiratory System Function”. This difference may
relate to the specificity of using the nursing ICNP browser versus the larger cohort
of concepts delivered through the SNOMED CT browser. This difference may also relate
to specific nursing knowledge and interpretation of the concepts extracted from the
source document (i.e., nursing diagnosis versus medical diagnosis). It also demonstrates the variance in
manual mapping when two browsers, with related content, are used. Overall, we found
value to explore both the browsers as a mechanism to crosswalk the extracted concepts
between a large interdisciplinary terminology and a nursing-specific terminology.
It also highlights an opportunity to further explore the UN Global Indicator Framework
for the Sustainable Development Goals and Targets, for the inclusion, comprehensiveness
and interpretation of nursing care delivery.
For concepts relating to cost expenditures, affordability, and community engagement,
gaps were noted and SNOMED CT did not have extensive concept representation. For example,
SDG 3.9.1 evaluates the proportion of the population with large household expenditures
on health as a share of total household expenditure or income. The concepts of large
household expenditure and health expenditure both had no matches in SNOMED CT. These
concepts require further development in SNOMED CT as they relate to broader implications
of sustainability as well as topics associated with social determinants of health.
For example, researchers have reported on the relationship between large out-of-pocket
health expenditures, poverty [[24]] and poor infant and maternal outcomes [[25]
[26]
[27]]. Further, researchers have demonstrated clinicians do document concepts related
to expenditure and health costs, suggesting opportunities exist to use standardized
terminology thoughtfully, as a way to report on and improve patient health and equity
[[28]].
Finally, through the mapping process, we found many concepts had narrower than and
broader than matches. Several of these concepts were related to water, air and sanitation.
For example, the concepts of domestic wastewater and industrial wastewater were mapped
to the broader than SNOMED CT concept: Waste water (substance); household air pollution
and ambient air pollution concepts were mapped to the broader than SNOMED CT concept:
Air pollution (event); sanitation services was mapped to the narrower than SNOMED
CT concept: Access to toileting facilities (observable entity); and using safely managed
sanitation services was mapped to the broader than SNOMED CT concept: Adequate sanitation
(finding). The nuance and specificity of these source concepts are lost as these direct
matches do not exist in SNOMED CT. Point-of-care clinicians, such as homecare and
public healthcare nurses, interact with patients and observe community settings impacted
by these environmental factors (e.g., household wood-burning stoves used for cooking and heating leading to the exasperation
of upper respiratory conditions or having managed wastewater disposal in remote communities
as a means to protect drinking water) [[29]
[30]
[31]]. Therefore, the evaluation, expansion, and inclusion of these concepts in SNOMED
CT are necessary to advance the clinical reporting of these environmental and sustainability
indicators.
5.2 Concept Representation for Artificial Intelligence and Data Reuse
Standardized clinical terminologies have long been used to provide data for population-based
analytics, including trend analysis and public health surveillance, as well as clinical
research [[32]]. In the context of capturing climate change impacts on health, capturing standardized
environmentally-specific clinical data permits the contextualization of vast amounts
of data to predict service demands and population needs during natural disasters and
climate-related events, health promotion and disease prevention and benchmarking of
any subsequent interventions. Researchers have also demonstrated the use of these
terminologies in developing or training models based on methods broadly defined as
artificial intelligence (AI) [[33], [34]]. SNOMED CT, for example, has been used in natural language processing tasks, utilizing
machine learning methodologies to search and analyze free text clinical documentation
entries [[35]]. Standardized clinical terminologies add to these approaches as a way of providing
sharable and comparable data across settings; thus, helping to produce AI models that
are transferable across settings. As environmental concepts encoded in standardized
terminologies expand and ontological relationships associating these to health impacts
develop, using AI methods could provide additional insights into trends in data and
health conditions, as well as for predictive analytics to support better practice,
research, education, and policy-level decision-making. While large-scale use of standardized
clinical terminologies in advanced clinical AI applications has yet to be realized,
its ability to formally represent healthcare language and its ability to be used in
automatic encoding (e.g., natural language processing) suggests its use will be increasingly valuable as these
technologies advance [[36]].
6 Limitations
This study has noted limitations. While the scope of this study phase modeled a component
of a concept by defining a domain type, the explication of relationships amongst the
set of concepts was not completed (e.g., using Unified Modeling Language to develop a conceptual model) [[12]]. Future researchers may wish to further develop these linkages. This study approached
the extraction and mapping of environmental and sustainability health concepts from
a descriptive, small-group, perspective. The intention was to begin exploring the
representation of these types of concepts in two (SNOMED CT and ICNP) international
standardized clinical terminologies. Ongoing work to expand, develop, and test these
representations is recommended to strengthen, add rigour, and expand the conceptual
nuances of this topic in these codified knowledge systems. As well, we limited our
concept extraction to specific concepts associated to health and whose SDG indicator
was linked to the World Health Organization. Work to expand and explore other SDG
indicators for possible inclusion in clinical standardized terminologies is needed.
7 Conclusions
To track the impacts of climate change on human health and to develop climate-resilient
systems able to respond to public health needs during climate-related events such
as heatwaves, droughts, wildfires and floods, it is essential to have standardized
data. By evaluating the representation of environmental concepts associated with health
impacts, in standardized clinical terminologies, we provided a view into current content
coverages as well as opportunities to expand. This study highlighted the availability,
overlap, and gaps found for SDG terms in SNOMED CT and ICNP. However, future work
is needed to more clearly define and model all possible environmental impacts from
a One Health perspective. The availability of codified terminologies is important
as it can enable clinicians, researchers and public health officials to manage vast
amounts of clinical data in a way that can be easily retrieved, analyzed, related
to the context in which the data was obtained, and embedded in AI applications. This
permits policymakers and leaders to model demands on health services as a function
of time and climate change pressures on the environment and prepares communities to
adapt to these pressures whilst supporting the One Health approach.
