Refining Substance Use Classification: An Ontological Framework for Enhancing Large-Scale Data Collection

Abstract

Substance use disorders (SUD) remain prevalent in the United States. The Office of Addiction Services and Support plays a critical role in tracking SUD trends in New York State and reports data to the federal system. However, ambiguities in substance classification pose challenges to data accuracy and consistency. To address these issues, we developed the foundations for the Addiction Substance Ontology (ASO) using Basic Formal Ontology principles. Definitions in the ASO are expressed in terms of genus and differentiae which form the backbone for a taxonomy of substances in function of their chemical composition and certain other characteristics essential for tracking their acquisition and use. While 143 classes have been developed thus far based on a specific program admission use case, pilot testing and stakeholder collaboration are necessary to refine the ASO and validate its application in real-world settings. These efforts aim to improve data reliability, enhance tracking of SUD patterns, and support effective public health interventions.

Introduction

In the United States, the prevalence of SUDs (see Table 1 for all acronyms used in this study) is high and varies across drug types, age groups, and demographic populations. An analysis of the National Epidemiologic Survey of Alcohol and Related Conditions–III, a nationally representative cross-sectional study of the non-institutionalized adult population, revealed that men exhibit higher prevalence rates of AUD, TUD, and CUD compared to women across most ages.(1) The prevalence of SUDs generally peaks between the ages of 20 and 30 before declining with age. For AUD, TUD, and OUD, White individuals exhibit higher rates than Black and Latino individuals. In contrast, the prevalence of CUD is higher among Black individuals compared to White and Latino individuals.

Table 1:

Acronym list

A recent review article highlighted the growing use of large administrative databases to access real-world evidence for monitoring interventions and trends in SUDs, particularly OUD and opioid overdose.(2) Since these databases are collected as part of routine clinical practice, they might provide valuable insights into real-world outcomes. The article also emphasized the potential for creating representative cohorts by linking data from multiple sources and defining samples based on specific types of encounters such as hospitalizations or overdose incidents. These encounters could then be mapped to different settings such as SUD treatment programs or mortality records. This approach might capture a broader spectrum of individuals with OUD, including those who may not seek formal treatment services, and might support more comprehensive analyses for risk factors and outcomes. In 1989, the SAMHSA in the United States established the TEDS, a national data system for substance use treatment.(3) Data collection for admissions (TEDS-A) began in 1992, followed by discharge data (TEDS-D) in 2000. TEDS primarily aims to collect and analyze data on substance use treatment episodes nationwide, supporting the monitoring of treatment trends, evaluation of intervention effectiveness, and development of informed policy decisions regarding substance use treatment services. TEDS data are compiled from state agency systems that collect episode-level information from substance use treatment facilities, which are typically state-licensed or certified and receive public funding. These facilities report certain details about each admission and discharge, including demographic information, substances used, treatment services provided, and outcomes. States standardize their data before submitting them to SAMHSA. In NYS, the OASAS oversees the collection and submission of TEDS data and monitors over 900 treatment programs.(4) The agency utilizes multiple forms, including admission, discharge, and assessment forms, to collect data from clients admitted to and discharged from OASAS-certified programs for substance use issues.(5) The admission form captures a broad range of demographic and socioeconomic information about clients, including age, gender, marital status, living situation, education level, and income.(4, 6) Currently, OASAS collects data on 32 types of substance use in NYS and maps this information to corresponding fields in the TEDS, which tracks 18 types of substance use (see Table 2).(6, 7)

Table 2:

Substance use data collected on the OASAS client admission forms and the TEDS

Emerging evidence highlights the growing use of data linkage across multiple administrative databases to advance research in the field of SUDs.(2) It is believed that inking diverse data sources, including medical claims, clinical records, and social or family history, may provide a comprehensive and continuous view of an individual’s health, disease progression, and socioeconomic changes. This approach might offer a broader perspective beyond treatment records only, allowing researchers to better understand risk factors such as substance use patterns and healthcare access to identify specific health outcomes such as opioid overdose. However, data in different administrative databases may not be readily interchangeable or linked directly. For example, Harron et al. identified several challenges in linking administrative data for research, including ensuring privacy preservation within the data linkage environment, addressing complexities in the linkage process such as data preparation and managing the quality of the linkage while mitigating potential biases in the source data.(8)

A potential solution to address data heterogeneity across systems is the use of ontologies. When developed lege artis, ontologies provide a formal description of the entities within a domain and the relationships among them. Advocates of this view are, for example, Hoehndorf et al. who described desirable features for ontologies, including the use of standard identifiers for classes and relations, a standardized vocabulary for the domain, metadata describing the intended meaning of classes and relations, and machine-readable axioms and definitions that enable computational access to the semantics of classes and relations. (9) An ontology that exhibits all these features can be a powerful tool for enhancing data standardization and interoperability amongst databases that are created in its terms. An example in the field of SUD research is the DAO which represents concepts and relationships related to prescription drug abuse, including substances such as buprenorphine and opioids.(10, 11) The current version of the DAO consists of 315 classes, 31 permissible relationships between them, and 814 instances among these classes. Although the DAO includes a schema and an instance base of assertions, it does not use a standard upper-level ontology and is therefore hard to reuse for other purposes than the one for which it has been designed, and not suitable for data linkage in combination with other ontologies. Indeed, although two distinct ontologies may each one exhibit the desired features described by Hoehndorf et al., they may still be incompatible or only partly compatible with one another, simply because these features can be satisfied in distinct ways, for example by resorting to different sorts of formalisms or because of inconsistencies in definitions about the same sort of entities described. For instance, the term “cocaine” has distinct definitions in the ChEBI and SNOMED CT.(12, 13) Such differences can lead to misclassification or missing data, particularly when researchers need to integrate or use both ontologies simultaneously. The purpose of an upper-level ontology, such as the Basic Formal Ontology (BFO), is to provide a unified and formal representation of all terms independent of application context.(14, 15) BFO organizes all types of entities that exist in reality in a hierarchy of 35 categories two of which are disjoint top-level categories: continuants and occurrents, whereby the latter include changes that the former can undergo. Although there is a vast number of ontologies that import terms from the BFO, only a limited number of ontologies adhere fully to BFO principles, highlighting the need for further development and standardization in this area, especially in the SUD domain.

The OASAS admission form is a comprehensive and essential tool for clinical and administrative purposes, facilitating the collection of critical data on substance use among individuals entering SUD treatment programs.(4) While the form is extensive and valuable, potential refinements could enhance its utility to better support the classification accuracy of data collected in its terms, and pave the way for integrating artificial intelligence (AI) applications for SUD research. The aims of this paper are to: (i) identify areas for refining the OASAS admission form to enhance substance classification and reduce misclassification, (ii) to describe how the ASO builds on the BFO principles to standardize definitions, and (iii) demonstrate how ASO integration improves data interpretation, minimizes ambiguity and supports AI-driven insights.

Methods

Source materials. We reviewed the most recent versions of four documents relevant for this study: (1) the NYS OASAS Client Admission Report (also referred to as ‘the admission form’ in this study), along with (2) the Client Admission Report Instructions, (3) the OASAS Client Data System Batch Transaction File Specification, and (4) the Code Table Documentation.(5) These documents are publicly available on the OASAS website. The Client Admission Report is used to collect clients’ personal health information related to substance use treatment in OASAS-certified programs.(6) It captures a wide range of data, including individual sociodemographic details, primary diagnoses related to SUDs expressed using the International Classification of Diseases, 10th Revision, types of problem substances (primary, secondary, and tertiary), routes of administration, frequency of use, and age of first use. The Client Admission Report Instructions provide additional guidance for completing the admission form.(16) For instance, in the “Problem Substances” section, the instructions offer explanations on when and how to select specific types of substances (see Table 3). The Batch Transaction File Specification details the data validation rules for submissions.(17) For example, for the field labeled [PRIMARY_SUBSTANCE_CODE]: ‘If the program is an opioid program, [PRIMARY_SUBSTANCE_CODE] must be one of the following: 1 (Heroin), 2 (Non-RX Methadone), 3 (Other Opiate/Synthetic), 29 (OxyContin), or 22 (Buprenorphine)’. The Code Table Documentation lists current active codes.(18) Each type of substance is assigned a specific value within the OASAS system, ensuring consistency in data entry. Adhering to these data validation rules minimizes errors during data entry and improves the reliability of the collected information. The streamline data entry process is shown in Figure 1.

Table 3:

Selected examples of problem substances with explanations from the OASAS instruction file

Figure 1.

Illustration of the data entry process into the OASAS database and the potential role of ASO integration

There are several ontologies, such as the ChEBI and SNOMED CT, that provide definitions related to drugs and substances.(12, 13) However, Ochs et al. raised concerns regarding the reuse of content from other ontologies, highlighting that many ontologies in, for example, the BioPortal exhibit issues such as inconsistent utilization of reused properties and redundant, unused class hierarchies.(19) Additionally, aligning these ontologies with the BFO hierarchy presents significant challenges since they are based on terminological, rather than ontological principles. Consequently, we opted not to import classes from other ontologies unless they also use the BFO as upper ontology, choosing instead to develop the ASO from the ground up to mitigate potential misclassifications of these substances.

Overall ontology design. The ASO development process followed key BFO principles to ensure a consistent design.(15) The ASO was therefore guided by 2 core objectives: (1) realism, to describe entities as they exist in reality and (2) perspectivism, to acknowledge multiple accurate descriptions of reality. In the ASO, each chemical used in substances of interest is defined by its specific and unique chemical structure at the molecular level. To achieve this, the PubChem CID was used to standardize and aggregate data.(20) This is motivated by the fact that each record submitted to PubChem is assigned a SID, and multiple SID records sharing the same structure are aggregated into a single CID. The CID can thus serve as a comprehensive summary of information for a given pure chemical structure. Additionally, the International Union of Pure and Applied Chemistry name was employed to ensure precise identification of each compound’s molecular structure. By leveraging both identifiers, the ASO explicitly defines a pure molecule while excluding non-active ingredients found in formulated drug products.

When formatting terms, singular count nouns (or terms) were used to refer specifically to what the BFO considers a universal or a defined class, rather than to a plural or mass noun (e.g. ‘portion of alprazolam’ rather than ‘alprazolam’).(15) In documentation, type terms are written in lowercase italics to distinguish them from terms for individuals in the domain of discourse. Acronyms were avoided to promote clarity to create standardized terminologies that could be universally understood and reliably used across disciplines, both now and in the future. When defining terms, all non-root terms were provided with Aristotelian definitions in the format: S :=def. a G that Ds. The “G” (genus) represents the immediate parent term of “S” (species) satisfying the isa-relation (sometimes also called ‘subclass’ or ‘subsumption relation) in the ontology, and “D” (differentia) specifies the characteristics that distinguish certain Gs as S. By leveraging Aristotelian definitions, the process facilitated fully unpacked definitions that trace back to the top node of the ontology. This approach avoided circularity, as circular definitions offer no additional information about the nature of the entities. Additionally, essential features were used in defining terms to maintain precision and alignment within the ontology.

Core entity types. In line with the BFO, the ASO distinguishes “molecule” as subtype of “BFO:object”, from “portion of molecules of type X” as subtype of “BFO:object aggregate”, where the latter has as “member-parts” only molecules of type X. Using “alprazolam” as an example, the entity “alprazolam molecule” in our ontology represents thus the universal encompassing all and only those molecules with that exact chemical structure, and “portion of alprazolam” as the “BFO:object aggregate” where the “member-parts” (a formal relation used in the BFO to describe what aggregates are composed of) consist exclusively of alprazolam molecules.(15) When the OASAS admission form provided a group name to classify a type of substance, such as “Barbiturate” or “Synthetic Stimulant” (see Table 2), we first searched PubChem to determine whether the existence of the drug class is motivated by the presence in all relevant molecules of some specific arrangement of atoms in their molecular structure. As an example, all molecules belonging to the barbiturate-drug class have as part of their molecular structure a specific arrangement of atoms which is the same arrangement of atoms which makes up a large part of a barbituric acid molecule. Barbiturate drugs are synthesized via chemical substitutions at specific positions on a barbituric acid molecule. For instance, the distinction between phenobarbital and barbituric acid arises from the presence of ethyl and phenyl functional groups, which replace hydrogen atoms originally located at the C5 position of the barbituric acid molecule (see Figure 2).(21, 22) Since barbituric acid is an existing and well-defined compound with its own PubChem CID 6211, our ontology includes the entities “barbituric acid molecule” to represent the pure barbituric acid molecule(21), “barbiturate core” to represent such common structure, “barbiturate core molecule” to represent any molecule which has as proper-continuant-part a barbiturate core, and “portion of barbiturate” to represent any aggregate of barbiturate core molecules (see Table 4). However, it is important to acknowledge that the substances self-reported by clients on the OASAS admission form may not always align with their actual chemical identity.(4) For example, a client may report using phenobarbital, but the substance consumed could be counterfeit or adulterated, containing different active ingredients. This gap between self-reported use and actual composition poses a potential challenge, not for ontology-based classification grounded in chemical definitions and molecular structures, but for substance use registries independent of any classification used. Ultimately, registries reflect the client’s perception of their substance use rather than a verified chemical analysis.

Figure 2.

Illustration of barbituric acid and phenobarbital molecules. Adapted from Reference (21,22)

Table 4:

Barbiturate related class definitions

In BFO, a “proper continuant part” is a formal relation to express parthood between continuants axiomatized according to the theory of Minimal Extensional Mereology.(15, 23) If no common molecular structure exists for the sort of substance class, for instance when expressed in terms of a function or route of administration, the term “BFO:disposition” was used.(15) For instance, we used “stimulant disposition” to describe a drug that has the disposition to increase under appropriate conditions central nervous system activity and to elevate alertness, mood, and energy when introduced into the human body. Indeed, drugs from several different classes can be classified as stimulants. Additionally, when the OASAS admission form requires tracking a substance based on a specific acquisition method, such as “Non-Rx Methadone” to monitor methadone obtained illegally, or “Synthetic Stimulant” to track stimulants manufactured through synthetic processes, the term “BFO:process” was applied (see Table 2).(15) For instance, “chemical compound acquiring process” is defined as an occurrent entity that exists in time by occurring or happening, has temporal parts, and always depends, inter alia, on substances and individuals who acquire them in one or other way. Subclasses such as “with prescription”, “illegal”, and “over-the-counter” were created to describe specific methods by which substances can be obtained.

Results

A total of 143 classes have been developed in the ASO to date. Of these, 128 fall under the category of BFO:continuant, while 15 are categorized as BFO:occurrent. Among the 126 continuant classes, 117 belong to either BFO:object or BFO:object aggregate, and 4 are classified as BFO:disposition. The disposition classes include stimulant disposition, sedative substance disposition, tranquilizer disposition, and hallucinogen disposition. The 15 occurrent classes are all categorized as BFO:process. These processes are used to (1) describe the methods of acquiring substances; (2) distinguish between naturally extracted and synthetic manufacturing processes, and (3) define the routes of administration. Figure 3 presents a snapshot of the hierarchical structure of ASO, while Table 5 showcases the first-order logic axioms that illustrate its underlying design. Table 6 provides selected examples from the ontology

Figure 3.

Snapshot of the ASO hierarchical structure

Table 5:

ASO axioms in Common Logic Interchange Format

Table 6:

Examples of ASO classes

Table 7 shows some key issues identified in the current OASAS admission form during the ASO development. These issues, spanning categories such as synonyms, brand names, group names, acquisition methods, synthetic substances, and the use of broad categories like “Other”.

Table 7:

Potential concerns identified in the current OASAS admission form and instruction file

Discussion and Conclusions

During the development of the ASO, 6 areas of concern were identified (see Table 7) in the current OASAS admission form and its instruction file, all involving ambiguities and variations that might impact the clarity and consistency of data collection.

Synonymy. The first area for potential improvement is a more coherent use of synonymy and pseudo-synonymy. As an example, the OASAS admission form lists “cocaine” and “crack” as separate entries, with the OASAS instruction file explaining that crack is a more purified form of cocaine.(6, 16) However, the form lacks precise guidelines on how “purified” a substance must be to qualify as “crack” rather than “cocaine”. Additionally, “crack” is just one of many street names for cocaine, yet the form does not specify whether other street names should also be recorded and categorized under “crack” or if OASAS intends to track only the term “crack” as a distinct entry for cocaine use.(24) The ASO provides a mechanism to distinguish “cocaine” and “crack”. At the molecular level, “ASO:cocaine molecule” represents the pure cocaine molecule and, similarly as with other molecules, “ASO:portion of cocaine” represents a pure aggregate of only cocaine molecules. Here, however, an extra category “ASO:portion of cocaine containing compound” is defined as an aggregate of which at least some member parts are cocaine molecules (see Table 8). As a result, if a client specifically reports “crack”, it would not fall under “ASO:portion of cocaine”, but under “ASO:portion of cocaine containing compound”. If a client is not sure whether what he uses is pure cocaine, then the same category is applicable. In real-world settings, clients reporting substance use on the OASAS admission form often use street names, which may not accurately reflect the chemical composition of the substance consumed. For instance, a significant portion of street cocaine is contaminated with fentanyl, a fact of which many users are unaware.(25) This imprecision in self-reported substance use has important implications for clinical interpretation and data classification. Beyond the ontology-based classification, further clarification from OASAS on how the OASAS form currently differentiates between “cocaine” and “crack,” including the specific criteria used would help refine the ASO to better align with real-world practices. Additionally, accounting for contamination issues might be essential to ensure that ASO reflects the complexities of real-world substance use when tested in practice. Therefore, because ASO is grounded in chemical definitions, it should be used in a framework that accounts for practical uncertainties in OASAS data. Addressing these discrepancies might improve data accuracy, intervention strategies, and build the foundation of AI-driven insights within the OASAS system.

Table 8:

Cocaine related classes definitions

Brand names. The use of brand names requires further clarification. The form includes “OxyContin” as a specific entry but does not specify whether generic versions of oxycodone should also be reported under this category.(6, 16) Providing clear guidance could help ensure consistency in reporting opioid use, thereby helped by “ASO:portion of oxycodone” defined as representing pure oxycodone, enabling tracking data related to the oxycodone molecule itself, regardless of brand-name or generic labeling. Furthermore, the ASO includes an “ASO:illegal chemical compound acquiring process” class, which provides a framework for distinguishing between substances obtained legally through a prescription and those acquired through illicit means. For example, a client might purchase OxyContin from someone who originally obtained it legally via prescription. In this scenario, the client’s method of acquisition is illegal, but the current OASAS form does not capture this distinction, which only lists “OxyContin” as a reporting option. While reporting the high prevalence of OxyContin use might inform interventions, it does not directly address the root problem of illegal redistribution of legally prescribed substances, potentially limiting the effectiveness of targeted measures. By contrast, the ASO approach explicitly defines and categorizes acquisition processes, such as “ASO:chemical compound acquiring process with prescription” and “ASO:illegal chemical compound acquiring process” to capture both the source and context of substance acquisition. This structured representation eliminates ambiguity, enabling a more accurate analysis of OxyContin use patterns while enhancing data interpretation and ensuring consistency when OASAS transmits data to TEDS.

Group names, such as “Barbiturates” and “Benzodiazepines,” provide an area for refinement in the form’s design. (6, 16) While the instruction files include examples for these categories, the examples mix brand-name and generic drugs, which could introduce variability during data collection. For example, “Alprazolam (Xanax)”, is listed separately from other benzodiazepines on the form, even though alprazolam both as a brand-name drug (Xanax) and in generic formulations. It is unclear whether the form is intended to track all alprazolam use, regardless of brand or specifically refers to Xanax. Additionally, the OASAS form separates “Alprazolam (Xanax)” and “Rohypnol” from the broader “Benzodiazepine” category, which could also create another classification challenge. For instance, if a client reports using flunitrazepam, the generic name for Rohypnol, and the individual completing the form is unaware of its classification, they might incorrectly categorize it under “Benzodiazepine” rather than “Rohypnol.” This could lead to underestimating the prevalence of Rohypnol use. It is here that, once more, the hierarchical structure of the ASO might prove to be useful. A broader category, “ASO:portion of benzodiazepine” includes substances that have as continuant part only “ASO:benzodiapzepine molecule”. It allows ASO to include subclasses such as “ASO:portion of alprazolam” and other benzodiazepine related substances listed under this category. Substances not explicitly listed on the current OASAS form would also automatically fall under this broader class, supporting the ontology update when the OASAS is updated as well as tracking specific drug types if needed.

Mode of substance acquisition. The OASAS admission form includes categories for substances obtained through specific methods, such as “Non-Rx Methadone,” and substances produced synthetically, such as “Synthetic Stimulant”.(6, 16) The “Non-Rx Methadone” category is currently the only entry on the form to explicitly track substances obtained and used without a legal prescription. However, this raises a question about whether similar tracking should be considered for other prescription drugs obtained illicitly. Additionally, if a client reports methadone use but does not specify the method of acquisition (e.g., via prescription or illicitly), the form’s design could lead to misclassification. Specifically, a client using legally prescribed methadone should select “Other Opiate/Synthetic” based on the source, but they might incorrectly select “Non-Rx Methadone”. It would inflate the reported prevalence of “Non-Rx Methadone” in NYS. The ASO addresses this issue by leveraging defined classes such as “ASO:illegal chemical compound acquiring process” and “ASO:chemical compound acquiring process with prescription” to capture both the source and method of acquisition rather than separating the methadone data entry.

Synthetic drugs. The OASAS currently lists “Synthetic Stimulant” as a problem substance category, with “Bath Salt” provided as an example.(6, 16) According to the United States DEA fact sheet, synthetic stimulants, often referred to as “bath salts”, belong to the synthetic cathinone class of drugs.(26) While “Bath Salt” is a notable example, it does not encompass all synthetic stimulants, leaving gaps in capturing utilization patterns. Additionally, the OASAS form includes “Khat” as a separate category.(6, 16) According to the DEA, khat is a flowering shrub with stimulant-like effects, containing the active ingredients cathine and cathinone, illustrating that cathinone could exist naturally and synthetically.(27) Therefore, the ASO defines “ASO:stimulant disposition” class to encompass all substances, including those categorized as “Synthetic Stimulant” and “Other Stimulant”, capturing shared stimulant properties regardless of their origin. Additionally, the ASO includes separate classes such as “ASO:portion of khat alkaloid” and “ASO:portion of amphetamine-like substance” to classify specific substances. For instance, consider dextroamphetamine, the generic form of Dexedrine, a synthetic drug that should belong to the “Other Amphetamine” category. However, due to the form’s current design, which lists Dexedrine but not dextroamphetamine, the substance might be misclassified under “Other Stimulant,” as it is neither “Bath Salt” nor explicitly mentioned in the instructions. The ASO framework raises awareness for such overlaps by systematically categorizing substances under appropriate classes, ensuring that all problem substances are accurately represented.

The dreadful ‘other’. Finally, the use of the term “other” for certain substances warrants attention. The OASAS form includes 7 categories utilizing “Other,” such as “Other Sedative/Hypnotic”, “Other Opiate/Synthetic”, “Other Tranquilizer”, “Other Amphetamine”, “Other Stimulant”, “Other Hallucinogen”, and “Other”.(6, 16) The effectiveness of these categories depends on the clarity and specificity of the primary substance categories on the form. For example, fentanyl, an opioid that has significantly contributed to the opioid epidemic since 2015, does not have a specific category on the OASAS admission form.(28) The most relevant option currently available is “Other Opiate/Synthetic.” However, if there is uncertainty about whether fentanyl should be classified under “Other Opiate/Synthetic,” it is possible that “Other” might be selected as a default. This could result in fentanyl being inadequately tracked if it is not explicitly classified under the opioid category.

Limitations. Our ontology has several limitations related to both its design and practical implementation. First, it was developed based on the most recent version of the OASAS admission form, while earlier versions may include substances no longer listed.(5) This discrepancy could lead to gaps if the ontology does not account for substances removed from previous versions. Second, ASO has not yet been tested in real-world settings so its practical effectiveness remains unverified. Pilot testing will be necessary to assess its alignment with OASAS’s workflow and data entry process and identify areas for further refinement. One potential approach is to conduct a comparative analysis of data from previous studies before and after integrating ASO into the analysis.(29-31) As shown in Figure 1, ASO cannot prevent errors made prior to the data entering the OASAS system. However, ASO may detect and help to explain errors post-hoc using logic-based axioms. Certain issues can only be addressed through improvements in user interface design, validation protocols, or staff training, rather than through the ontology itself. For instance, ASO specifies that the correct route of administration for alprazolam tablets is by mouth; if a different route is entered, the system could flag and report the discrepancy for review by OASAS staff, rather than automatically rejecting the input. Third, the current ontology structure may require ongoing refinement to resolve logical inconsistencies within type-term definitions, maintain internal coherence, and ensure interoperability for broader application. Achieving interoperability with external systems such as SNOMED CT and RxNorm will involve applying bridge axioms using established methods for linking BFO-based ontologies.(32, 33) Additionally, some assumptions made during ASO development, such as those related to data structure or naming inconsistencies, may not fully align with internal practices already in place at OASAS. Collaboration with OASAS personnel and public health researchers will be important to validate the ontology and guide further improvements.

The development of the ASO has provided insights into areas for refinement in the OASAS admission form and its data collection processes. In contrast to the DAO, the ASO follows the BFO principles for each substance listed on the OASAS admission form. By doing so, the ASO introduces a structured framework to improve clarity and consistency in substance classification. Its hierarchical structure facilitates comprehensive representation and analysis of substance use data. These advancements highlight the potential of ontological approaches to enhance data collection and classification processes, enabling more accurate tracking of substance use patterns and supporting effective public health interventions. As the ASO continues to evolve, it is expected to become a valuable tool for advancing data standardization and improving substance use trend analysis, ultimately contributing to informed policy development and programmatic strategies.