Eduvest – Journal of Universal Studies Volume 1 Number 12, December, 2021 p- ISSN 2775-3735- e-ISSN 2775-3727
INTEGRATION OF MACHINE LEARNING IN CLINICAL DECISION SUPPORT SYSTEMS
Moazzam Siddiq Independent Researcher, Manchester, United Kingdom Email: [email protected]
ABSTRACT
Clinical decision support (CDS) systems offer healthcare professionals real-time, evidence-based assistance for the diagnosis, treatment, and monitoring of medical disorders, which has the potential to enhance patient outcomes. The application of machine learning algorithms in CDS systems has increased the reach and precision of these systems, enabling the examination of intricate patient data and the discovery of as-yet-unrecognized connections and patterns. This review article focuses on machine learning algorithms' uses in diagnosis and disease categorization, treatment selection and optimization, and patient monitoring and prognosis in order to examine the advantages and drawbacks of utilizing them in CDS systems. It also looks at the moral and legal issues surrounding the use of these systems, such as privacy issues and responsibility for choices made utilizing CDS technologies. A discussion of potential paths and difficulties for applying machine learning algorithms in CDS systems finishes the review. The incorporation of real-time data streams, the creation of more understandable algorithms, and the inclusion of patient preferences and values in decision-making are all possible ways to enhance CDS systems using machine learning. The necessity for thorough validation and regulatory control, as well as worries about the possible impact on clinician-patient relationships, are obstacles and difficulties to widespread implementation in clinical practice. The potential advantages of applying machine learning algorithms to CDS systems are highlighted in this paper, but it also stresses the need to address moral and legal issues and make sure that these systems are used in a responsible and open manner.
KEYWORDS	Clinical decision support systems, healthcare, diagnosis, treatment, patient monitoring, prognosis, ethical considerations, legal considerations, privacy, liability, challenges, opportunities
	This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International

INTRODUCTION

Clinical decision support (CDS) technologies are being used more frequently in the healthcare sector than ever before to assist physicians in choosing the best course of therapy for their patients. These systems offer real-time information on potential diagnosis and treatment options using patient-specific data, such as medical history, results of diagnostic tests, and vital signs [1]. Traditional CDS systems frequently rely on pre-established rules and norms, which could not always account for the full complexity of a patient's particular circumstances. This may result in treatment choices that are less than ideal or even detrimental [2].

In order to personalize treatment choices and enhance patient outcomes, academics and healthcare professionals have resorted to machine learning algorithms to address this issue. Large-scale patient data analysis using machine learning algorithms allows for the prediction of potential diagnosis, available treatments, and prognosis [3]. Many stakeholders, including the National Institutes of Health (NIH) and the Food and Drug Administration (FDA), have recognized the potential of machine learning to revolutionize healthcare and have committed significant resources to the development and assessment of machine learning-based CDS systems. We intend to present an overview of the current state of the art for machine learning-based CDS systems in this review paper. We will talk about the need for personalised medicine, the different kinds of CDS systems that are already in use, and the advantages and difficulties of incorporating machine learning algorithms into these systems. We will also go through the various machine learning techniques frequently applied in CDS systems, as well as their benefits and drawbacks. Following that, we'll talk about some specific uses of machine learning in CDS systems, like disease categorization and diagnosis, treatment selection and optimization, and patient monitoring and prognosis [4]. We will talk about the moral and legal difficulties surrounding the application of machine learning in healthcare, such as liability concerns. In order to wrap up, we will talk about the obstacles and future directions for this fast developing sector.

The Need for Personalized Medicine

Personalised medicine is a method of treating patients that takes into account the particular biological, environmental, and behavioral elements that affect each person's health and susceptibility to disease. Instead of taking a generalized strategy, it tries to customize medical treatment and prevention techniques for each patient. Traditional treatment strategies frequently rely on recommendations and clinical trials that are founded on averages at the population level and do not account for individual variability [5]. According to research, genetic, environmental, and lifestyle factors can have a major impact on how individuals respond to the same medication. By choosing therapies that are more likely to be successful for particular patients, limiting the risk of negative side effects, and avoiding treatments that are unlikely to be successful, personalised medicine offers the potential to enhance patient outcomes. Since they can examine vast volumes of patient data to find patterns and relationships that human experts might not notice, machine learning algorithms are especially well suited for personalised medicine [6]. Machine learning algorithms can find personalised treatment solutions that may not be obvious using conventional approaches by learning from data. Utilizing genomic sequencing to find potential genetic cancer-causing mutations is one example of personalised medicine in action. Doctors can choose targeted medicines that are more likely to be effective while avoiding treatments that are unlikely to work by determining the precise mutations that are causing the cancer. Additionally, machine learning algorithms can be used to identify patients who are more likely to contract specific diseases, enabling earlier intervention and prevention measures [7].

In order to identify patients who are more likely to acquire heart disease, for instance, machine learning algorithms can analyse electronic health information. This enables tailored therapies to lower the risk of heart disease. By limiting the usage of therapies that are unlikely to work, avoiding unnecessary medical procedures, and lowering the risk of side effects, personalised medicine can not only improve patient outcomes but also result in cost savings [8]. Personalised medicine also faces difficulties, such as the necessity for vast volumes of high-quality data, the possibility of privacy issues, and the requirement for meticulous validation of personalised treatment alternatives. Nevertheless, the science of personalised medicine can advance and patient outcomes can be improved with the help of machine learning algorithms [9].

Types of Clinical Decision Support Systems

Systems for clinical decision support (CDS) can take many different shapes, from straightforward rule-based systems to intricate machine learning algorithms. The exact clinical task and the data that are available determine the sort of CDS system that is employed. Here are a few illustrations of several CDS system types:

Rule-based systems: The clinical direction provided by rule-based CDS systems is based on pre-established norms and criteria. A rule-based system, for instance, can advise a specific course of treatment for a patient with a certain diagnosis or set of symptoms. Even though rule-based systems are sometimes useful for simple therapeutic activities, they might not be able to fully account for the complexity of a patient's unique circumstances [10].

Knowledge-based systems: In order to provide clinical assistance, knowledge-based CDS systems require a knowledge base, which is a collection of rules, guidelines, and clinical knowledge. Knowledge-based systems can incorporate more complicated clinical knowledge and take into consideration unique patient features than rule-based systems can. The calibre and comprehensiveness of the knowledge base that is now available, however, may potentially be a limitation for knowledge-based systems [11].

Machine learning-based systems: Using algorithms, machine learning-based CDS systems learn from massive volumes of patient data and then offer therapeutic advice based on the patterns and links found in the data. Clinical tasks including diagnosis, choosing a course of therapy, and keeping track of patients can all be accomplished using machine learning algorithms. Compared to rule-based or knowledge-based systems, machine learning-based systems have the potential to offer more individualised and precise therapeutic assistance. They may be constrained by the algorithms' interpretability, though, and need a lot of high-quality data [12].

Hybrid systems: Hybrid CDS systems combine many CDS system types to offer more thorough clinical recommendations. For instance, a hybrid system might employ a rule-based approach to provide early clinical guidance before refining the advice based on unique patient features using a machine learning algorithm. Although hybrid systems can benefit from both rule-based and machine learning-based systems, they can also be more difficult to build and put into use [13]. The precise clinical task at hand and the data that are readily available determine which CDS system is employed. To provide individualised therapeutic recommendations, a straightforward rule-based system might work in some situations, while a more sophisticated machine learning-based system might be required in others. It's important to note that in order to make sure that CDS systems are accurate, dependable, and suit the demands of patients and healthcare professionals, collaboration between doctors, data scientists, and other stakeholders is required during the development and implementation process. We will go over specific uses of machine learning in CDS systems in the following sections of the review paper, covering diagnosis and disease classification, treatment selection and optimization, and patient monitoring and prognosis [14].

Benefits and challenges of using clinical decision support (CDS) systems in clinical practice

Clinical decision support systems offer the ability to improve clinical decision-making, boost efficiency, and decrease medical errors, among other clinical practice advantages. However, there are a number of difficulties with using CDS systems, including problems with data dependability and quality, the complexity of the algorithms, and healthcare professionals' aversion to change. The capacity of CDS systems to enhance clinical decision-making is one of their key advantages. CDS systems can assist healthcare professionals in making better educated decisions about diagnosis, therapy, and patient management by giving real-time feedback based on the most recent medical research and patient data [15]. Better patient outcomes and more effective use of healthcare resources may result from this. By automating specific healthcare operations like prescribing medications or conducting diagnostic tests, CDS systems can also improve efficiency. As a result, healthcare professionals may have less work to do and more time to devote to providing direct patient care. The ability of CDS systems to lower medical errors is another advantage. CDS systems can aid in preventing adverse events and enhancing patient safety by sending out real-time notifications for possible pharmaceutical interactions or dosing errors [16].

The employment of CDS systems in clinical practice, however, is not without its difficulties. Data dependability and quality are one of the primary issues. For CDS systems to deliver precise clinical guidance, patient data must be accurate and complete. Data entry errors, missing or incomplete data, and differences in data gathering practices amongst healthcare facilities can all have an impact on data quality. The intricacy of the algorithms utilized in CDS systems presents another difficulty. Healthcare professionals may find it challenging to comprehend and have faith in the recommendations made by CDS systems due to the complexity and difficulty of machine learning algorithms. When integrating CDS systems in clinical practice, healthcare practitioners' resistance to change might be a problem [17]. It may be challenging to incorporate CDS systems into current clinical systems because some healthcare professionals may be reluctant to accept new technology or alter their clinical practices. Finally, I can state that although CDS systems have a lot of potential for use in clinical practice, there are a lot of difficulties involved. In order to ensure that CDS systems are accurate, dependable, and suit the needs of patients and healthcare professionals, coordination between healthcare providers, data scientists, and other stakeholders is necessary [18].

RESEARCH METHOD

Clinical decision support (CDS) systems rely on machine learning algorithms to deliver real-time assistance to healthcare professionals based on the most recent scientific research and patient data. Various machine learning methods, like as supervised learning, unsupervised learning, and deep learning, can be applied in CDS systems. A collection of inputs and related outputs are provided to supervise learning algorithms, which then learn to map the inputs to the outputs [19]. These algorithms may be applied to tasks like forecasting patient outcomes or advising a course of therapy based on the patient's characteristics. On the other hand, unsupervised learning algorithms are trained on unlabeled data, which means they are given a set of inputs without corresponding outputs and are taught to find patterns or links in the data. These algorithms can be applied to projects like grouping patients based on shared traits or locating probable risk factors for certain ailments. An example of a machine learning algorithm is the deep learning algorithm, which is made to learn many levels of representation in complex data [20]. These algorithms can be incorporated into CDS systems to analyse complex patient data and give real-time recommendations to healthcare workers. Typically, these algorithms are employed for tasks like picture recognition or natural language processing. However, there are a number of difficulties with using machine learning algorithms in CDS systems, including as problems with bias and poor data quality. Bias is another potential issue with the use of machine learning algorithms in CDS systems. Machine learning algorithms require large amounts of high-quality data to be trained effectively, and the quality and reliability of this data can be affected by factors such as missing or incomplete data, data entry errors, and variations in data collection across different healthcare settings [21]. For some patient groups, machine learning algorithms may learn to generate recommendations based on biased or inadequate data, which may be unjust or erroneous [22]. Healthcare organizations are investigating strategies like explainable AI, which tries to make machine learning algorithms more transparent and understandable to healthcare providers, to address these issues. They are also investigating how to employ heterogeneous datasets and algorithmic fairness frameworks to enhance data quality and lessen bias in machine learning systems. Last but not least, I would like to point out that machine learning algorithms are an important part of CDS systems, giving healthcare providers real-time advice based on the most recent research and patient data [23]. Healthcare organizations must take action to guarantee that machine learning algorithms are precise, dependable, and impartial because the usage of these algorithms also comes with a number of issues

RESULT AND DISCUSSION

Applications of Machine Learning in CDS Systems

There are several applications which are described below.

Diagnosis and disease classification:

In order to assist in the diagnosis and categorization of diseases, machine learning algorithms can be employed to analyse patient data, such as electronic health records or medical imaging. Deep learning algorithms, for instance, can be used to analyse medical pictures like CT or MRI scans to find anomalies that might be signs of a particular disease. Similarly, using parameters like symptoms, medical history, and genetic data, machine learning algorithms can be used to analyse patient data and classify people into various illness categories. Machine learning algorithms can assist healthcare professionals in creating more efficient treatment plans and enhancing patient outcomes by supplying more precise and faster diagnoses [24]. Additionally, by identifying patients who could be at risk of contracting specific illnesses, these algorithms can aid in earlier intervention and treatment.

Treatment selection and optimization:

Based on details including a patient's medical history, genetic information, and response to previous therapies, machine learning algorithms can also assist medical professionals in choosing the best course of action for their patients. To identify probable side effects of a medication, for instance, or to forecast which treatments are most likely to be helpful for a certain patient, machine learning algorithms can be used to analyse patient data [25].

By modifying dosages or schedules depending on current patient data, machine learning algorithms can also be utilized to optimize treatment programs. Machine learning algorithms can aid in improving patient outcomes and lowering the likelihood of negative occurrences by offering more individualised and data-driven therapy alternatives.

Patient monitoring and prognosis:

The prognosis of patients can also be predicted using machine learning algorithms based on a variety of variables, including vital signs, medical history, and therapy response. For instance, wearable device data from heart rate monitors or glucose sensors can be analyzed by machine learning algorithms to look for trends that might indicate a condition that is getting worse. Additionally, based on patient data, machine learning algorithms can be used to forecast patient outcomes, such as the propensity for readmission or mortality. Machine learning algorithms can help to enhance patient outcomes and lower the likelihood of adverse events by giving healthcare practitioners real-time patient information [26]. Clinical decision support systems can use machine learning algorithms for a variety of tasks, such as disease classification, treatment selection, patient monitoring, and prognosis. These algorithms can enhance patient outcomes, lower the incidence of adverse events, and advance the area of precision medicine by giving healthcare practitioners more precise and personalised information about their patients [27].

Ethical and Legal Considerations:

Privacy Concerns

Regarding patient privacy in particular, the use of machine learning algorithms in clinical decision support systems raises a number of ethical and legal questions. The privacy of patients must always be safeguarded, and healthcare professionals must make sure that patient data is acquired and handled in a secure and private manner. Data privacy is one of the main issues. To be effective, machine learning algorithms need a lot of patient data, which must be safeguarded against unauthorized access, theft, or misuse. Healthcare providers are required to follow laws and regulations controlling the use and storage of sensitive patient data and to be open and upfront about their data privacy policies [28].

The possibility for prejudice is another problem. Large datasets can be analyzed by machine learning algorithms to find patterns and make predictions. But these algorithms could also unintentionally reinforce pre-existing prejudice and discrimination, like racial or gender biases. Healthcare organizations must take action to prevent bias in machine learning algorithms against any patient population. Additionally, patients' consent must be sought before their data can be used in clinical decision support systems once they have been told about how it will be used. This makes it necessary for healthcare providers to explain to patients their data privacy policies and the benefits and hazards of using these systems [29]. The security of patient data may also be threatened by data leaks and cyber-attacks. To prevent unauthorized access to or theft of sensitive patient data, healthcare providers must have adequate security measures in place, such as encryption and firewalls. Healthcare providers must put patient privacy and data security first because machine learning algorithms are used in clinical decision support systems. Healthcare providers must make sure that their data privacy policies adhere to all relevant laws and regulations, and patients must be informed about how their data is used. Additionally, healthcare professionals must take action to prevent bias or discrimination from being perpetuated by machine learning algorithms [30].

Liability and Responsibility for Decisions Made Using CDS Systems

Liability and responsibility for judgments made with clinical decision support systems are another crucial ethical and legal factor. Healthcare professionals must use these systems responsibly and ethically, and they must be prepared to defend the choices they make based on the results of these systems. The potential for unexpected and illogical findings from machine learning algorithms makes deploying them in clinical decision support systems challenging. In these situations, medical professionals must be able to assess the algorithm's output and determine whether it is accurate and dependable. The healthcare provider must use their clinical judgment and knowledge to make decisions if the algorithm's output is unreliable [31]. Another problem is that if machine learning algorithms are taught on incomplete or biased data, they may become biased or yield inaccurate findings. Healthcare providers must make sure that the data utilized to train these algorithms is impartial, complete, and representative of the population being serviced. It is also necessary to consider liability and accountability for choices made with the use of clinical decision support technologies. Healthcare practitioners can be held responsible for any negative outcomes if the system yields inaccurate or damaging results. When using machine learning algorithms in clinical decision support systems, healthcare practitioners must be ready to take responsibility for any unfavorable outcomes and take the necessary precautions to protect patient safety [32].

Healthcare professionals must make sure they are using clinical decision support systems responsibly and ethically, and that they can defend the decisions they make based on the results of these systems, in order to deploy machine learning algorithms. Furthermore, healthcare professionals need to take measures to guarantee that the algorithms employed in these systems are trustworthy, objective, and yield correct findings. Last but not least, healthcare professionals need to be ready to take responsibility for any unfavorable outcomes that may come about as a result of the application of machine learning algorithms in clinical decision support systems [33].

Future Directions and Challenges:

Clinical decision support systems that use machine learning algorithms are a fascinating field of study with a lot of potential to enhance patient outcomes. Future research and development in this field has a lot of exciting potential, including the creation of more precise and efficient machine learning algorithms, the blending of numerous data sources, and the inclusion of patient-generated data in clinical decision support systems [34]. The creation of individualised clinical decision support systems is a significant area of research. These systems can be customized to each patient's unique traits and can offer recommendations based on that patient's particular requirements and circumstances. Further investigation is also required into the long-term impact of applying machine learning algorithms to clinical decision support systems, particularly with regard to patient outcomes and healthcare expenditures [35]. Despite the potential advantages of utilizing machine learning algorithms in clinical decision support systems, a number of difficulties and obstacles still need to be overcome. The necessity for a lot of high-quality data to train these algorithms is one of the main obstacles. Over fitting is another possibility, where the algorithm gets too narrowly focused on the training data and is unable to generalize to new data. The algorithm must also be open and understandable so that healthcare providers may comprehend the rationale behind its predictions and take appropriate action [36].

Opportunities for improving CDS systems with machine learning:

Clinical decision support systems could become much more accurate and efficient thanks to machine learning algorithms. These systems can be enhanced using machine learning in a variety of fascinating ways, including the creation of more precise and efficient algorithms, the blending of numerous data sources, and the inclusion of patient-generated data in clinical decision support systems. The creation of individualised clinical decision support systems is a significant field of research [37]. These systems can be customized to each patient's unique traits and can offer recommendations based on that patient's particular requirements and circumstances. More interpretable algorithms must also be created in order to give healthcare professionals knowledge of the foundations for their suggestions. Integration of many data sources is a significant opportunity for machine learning to enhance clinical decision support systems. Healthcare professionals can have a better picture of a patient's health status and make better treatment decisions by merging data from many sources, such as medical records, test findings, and patient-generated data [38].

Challenges and barriers to widespread adoption in clinical practice:

Although machine learning-based clinical decision support systems may have advantages, there are still a number of obstacles that need to be removed in order for them to be widely used in clinical practice. The lack of standardization and interoperability among healthcare information systems is a significant issue. As a result, integrating machine learning algorithms into current workflows and information systems is challenging. The requirement for thorough evaluation and validation of machine learning algorithms presents another difficulty. Before these algorithms are extensively used in clinical practice, it is crucial to make sure they are accurate, dependable, and successful, just like with any new technology [39]. Large-scale clinical trials as well as exacting standards for validation and evaluation are necessary for this. Concerns exist over the cost-effectiveness of implementing clinical decision support systems based on machine learning. These systems need a lot of money to set up, train, and maintain, even if they have the potential to enhance patient outcomes and lower healthcare costs. Before implementing these systems, healthcare providers must carefully balance the costs and possible advantages to make sure they are cost-effective [40]. To ensure the appropriate and ethical use of clinical decision support systems based on machine learning, there are legal and ethical issues that need to be taken into account. These include issues with privacy, liability and responsibility for choices made using these systems, and making sure that patients are fully aware of and have consented to their usage. While there are many prospects for machine learning to enhance clinical decision support systems, there are also many obstacles that must be overcome in order to assure widespread implementation in clinical practice.

These include interoperability and standardization-related technical difficulties as well as problems with evaluation, validation, and cost-effectiveness. Collaboration between healthcare practitioners, researchers, politicians, and industry partners will be necessary to meet these issues, as will a dedication to the ethical and responsible use of these technologies [41].

CONCLUSION

The use of clinical decision support (CDS) systems in healthcare has both potential advantages and disadvantages, with a focus on the uses of machine learning algorithms in these systems. In conclusion, this review paper has addressed both of these aspects. The review has demonstrated how machine learning algorithms can boost CDS systems' precision and effectiveness, resulting in better patient outcomes and lower costs. These algorithms can be used in a number of healthcare settings, including as disease categorization and diagnosis, choice and optimization of treatments, and patient monitoring and prognosis.

The review has, however, also noted a number of difficulties related to the application of machine learning algorithms in CDS systems. These include the necessity for a lot of high-quality data to be collected in order to train these algorithms, the danger of over fitting, and the requirement for transparency and interpretability in order to give healthcare providers the information they need to make judgments. The assessment has also drawn attention to a number of moral and legal issues that should be taken into account when utilizing machine learning algorithms in CDS systems, including liability for choices made using these systems and privacy issues. Future potential for machine learning-enhanced CDS systems include the creation of more complex algorithms that can manage complex data sources and the incorporation of technology for image and natural language processing. Widespread implementation in clinical practice is hampered by a number of issues, including the need for better data exchange and interoperability as well as the need for more thorough validation and assessment of these systems. There are various implications for future research and practice in light of these findings. The accuracy and interpretability of machine learning algorithms for use in CDS systems must first be improved. This calls for ongoing development and validation of these algorithms. Second, there is a need to enhance the usability and user experience of CDS systems by better integrating them into clinical workflows and decision-making processes. Third, in order to make sure that these systems are secure and efficient in clinical practice, they need to be subjected to more thorough examination and monitoring. Healthcare could be transformed by using machine learning algorithms in CDS systems, but there are also considerable hurdles and ethical issues to be taken into mind. In order to ensure that these systems are secure, efficient, and advantageous for patients, it is crucial for healthcare professionals and researchers to keep collaborating on their development and application

REFERENCES

[1] Afshar, M., Arain, E., Ye, C., Gilbert, E., Xie, M., Lee, J., Churpek, M.M., DurazoArvizu, R., Markossian, T. and Joyce, C. 2019. Patient Outcomes and CostEffectiveness of a Sepsis Care Quality Improvement Program in a Health System. Journal of the American Medical Association. (Jul. 2019).

[2] Ahmad, M.A., Eckert, C., Teredesia, A. and McKelvey, G. 2018. Interpretable Machine Learning in Healthcare. IEEE Intelligent Informatice Bulletin. 19, (Aug.2018), 1–7.

[3] Angus, D.C., Seymour, C.W., Coopersmith, C.M., Deutschman, C.S., Klompas, M., Levy, M.M., Martin, G.S., Osborn, T.M., Rhee, C. and Watson, R.S. 2016. A Framework for the Development and Interpretation of Different Sepsis Definitions and Clinical Criteria. Journal of the American Medical Association. 44, 3 (Mar. 2016), e113–e121.

[4] Bedoya, A.D., Clement, M.E., Phelan, M., Steorts, R.C., OBrien, C. and Goldstein, B.A. 2019. Minimal Impact of Implemented Early Warning Score and Best Practice Alert for Patient Deterioration. Critical Care Medicine. 47, 1 (Jan. 2019), 49–55.

[5] Berwick, D.M. 2003. Disseminating innovations in health care. JAMA. 289, 15 (Apr. 2003), 1969–1975.

[6] boyd, D. and Crawford, K. 2012. Critical Questions for Big Data. Information, communication, and Society. 5 (May 2012), 662–679.

[7] Burrell, J. 2016. How the machine “thinks”: Understanding opacity in machine learning algorithms. Big Data & Society. 3, 1 (Feb. 2016), 205395171562251–12.

[8] Caruana, R., Lou, Y., Gehrke, J., Koch, P., Sturm, M. and Elhadad, N. 2015. Intelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30- day Readmission. KDD. (New York, New York, USA, 2015), 1721–1730.

[9] Choi, E., Schuetz, A., Stewart, W.F. and Sun, J. 2016. Using recurrent neural network models for early detection of heart failure onset. Journal of the American Medical Informatics Association. 61, 4 (Aug. 2016), ocw112–10.

[10] Citron, D.K. and Pasquale, F. 2014. The Scored Society: Due Proccess for Automated Predictions. Washington Law Review. (Mar. 2014).

[11] Collins, G.S., Reitsma, J.B., Altman, D.G. and Moons, K.G.M. 2015. Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): The TRIPOD Statement. Annals of Internal Medicine. 162, 1 (Jan. 2015), 55–11.

[12] Davis, S.E., Lasko, T.A., Chen, G., Siew, E.D. and Matheny, M.E. 2017. Calibration drift in regression and machine learning models for acute kidney injury. Journal of the American Medical Informatics Association. 24, 6 (Mar. 2017), 1052–1061.

[13] Doshi-Valez, F. and Kim, B. 2017. Towards a Rigorous Science of Interpretable Machine Learning. (Mar. 2017), 1–13.

[14] Downing, N.L., Rolnick, J., Poole, S.F., Hall, E., Wessels, A.J., Heidenreich, P. and Shieh, L. 2019. Electronic health record-based clinical decision support alert for severe sepsis: a randomised evaluation. BMJ Quality & Safety. (Mar. 2019), bmjqs–2018–008765–7.

[15] Elish, M.C. 2018. The Stakes of Uncertainty: Developing and Integrating Machine Learning in Clinical Care. EPIC Proceedings. (Nov. 2018).

[16] Evans, L. 2019. A Closer Look at Sepsis-Associated Mortality. JAMA Network Open. 2, 2 (Feb. 2019), e187565.

[17] Ferryman, K. and Pitcan, M. 2018. Fairness in precision medicine. Data & Society. 2018. Data & Society.

[18] Futoma, J., Hariharan, S. and Heller, K. 2017. Learning to Detect Sepsis with a Multitask Gaussian Process RNN Classifier. International Conference on Machine Learning. (Jun. 2017), 1–9.

[19] Futoma, J., Hariharan, S., Heller, K., Sendak, M.P., Brajer, N., Clement, M., Bedoya, A. and OBrien, C. 2017. An Improved Multi-Output Gaussian Process RNN with Real-Time Validation for Early Sepsis Detection. Proceedings of Machine Learning for Healthcare. (Aug. 2017).

[20] Gawande, A. 2010. The Velluvial Matrix. The New Yorker. (Jan. 2010).

[21] Green, B. and Chen, Y. 2019. Disparate Interactions. (New York, New York, USA, 2019), 90–99.

[22] Grote T, Berens P. On the ethics of algorithmic decision-making in healthcare. J Med Ethics. 2020;46:205-211. https://doi.org/10.1136/ medethics-2019-105586.

[23] van Baalen S, Carusi A, Sabroe I, Kiely DG. A social-technological epistemology of clinical decision-making as mediated by imaging. J Eval Clin Pract. 2017;23:949-958. https://doi.org/10.1111/jep. 12637.

[24] Svenaeus F. The Hermeneutics of Medicine and the Phenomenology of Health: Steps Towards a Philosophy of Medical Practice. Vol 5. Dordrecht: Springer Science & Business Media; 2013.

[25] Leder D. Clinical interpretation: the hermeneutics of medicine. Theor Med. 1990;11:9-24.

[26] Gadamer H-G. Truth and Method. New York: Crossroad; 2004 /1975.

[27] Ihde D. Philosophy of Technology: An Introduction. New York: Paragon House; 1993.

[28] Verbeek P-P. What Things Do: Philosophical Reflections on Technology, Agency, and Design. University Park, PA: Pennsylvania State University Press; 2005.

[29] Tschandl P, Rinner C, Apalla Z, et al. Human–computer collaboration for skin cancer recognition. Nat Med. 2020;26:1229-1234. https:// doi.org/10.1038/s41591-020-0942-0. KUDINA AND de BOER 535 13652753, 2021, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jep.13535 by INASP/HINARI - PAKISTAN, See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

[30] Yoon S-W. Korea's third AI-based oncology center to open next month. The Korea Times. March 16, 2017. http://www.koreatimes.co. kr/www/tech/2017/03/129_225819.html. Accessed August 4, 2020.

[31] de Boer B, Kudina O. What is morally at stake when using algorithms to make medical diagnoses? Expanding the discussion beyond risks and harms. Theor Med Bioeth. In press.

[32] Chung J, Zink A. Hey Watson - Can I sue you for malpractice? Examining the liability of Artificial Intelligence in medicine. Asia Pac J Health Law Ethics. 2018;11:51-80. Symptomate. Available online: https://symptomate.com/ (accessed on 25 April 2021).

[34] Hanover Project. Available online: https://www.microsoft.com/en-us/research/project/project hanover/ (accessed on 25 April 2021).

[35] Schaaf, J.; Sedlmayr, M.; Schaefer, J.; Storf, H. Diagnosis of Rare Diseases: A scoping review of clinical decision support systems. Orphanet J. Rare Dis. 2020, 15, 1–14. [CrossRef]

[36] Walsh, S.; de Jong, E.E.; van Timmeren, J.E.; Ibrahim, A.; Compter, I.; Peerlings, J.; Sanduleanu, S.; Refaee, T.; Keek, S.; Larue, R.T.; et al. Decision Support Systems in Oncology. JCO Clin. Cancer Inform. 2019, 3, 1–9. [CrossRef] [PubMed]

[37] Mazo, C.; Kearns, C.; Mooney, C.; Gallagher, W.M. Clinical decision support systems in breast cancer: A systematic review. Cancers 2020, 12, 369. [CrossRef] [PubMed]

[38] Velickovski, F.; Ceccaroni, L.; Roca, J.; Burgos, F.; Galdiz, J.B.; Marina, N.; Lluch-Ariet, M. Clinical Decision Support Systems (CDSS) for preventive management of COPD patients. J. Transl. Med. 2014, 12. [CrossRef] [PubMed]

[39] Durieux, P.; Nizard, R.; Ravaud, P.; Mounier, N.; Lepage, E. A Clinical Decision Support System for Prevention of Venous Thromboembolism Effect on Physician Behavior. JAMA 2000, 283, 2816–2821. [CrossRef]

[40] Lakshmanaprabu, S.; Mohanty, S.N.; Sheeba, R.S.; Krishnamoorthy, S.; Uthayakumar, J.; Shankar, K. Online clinical decision support system using optimal deep neural networks. Appl. Soft Comput. 2019, 81, 105487. [CrossRef]

[41] Mattila, J.; Koikkalainen, J.; Virkki, A.; van Gils, M.; Lötjönen, J. Design and Application of a Generic Clinical Decision Support System for Multiscale Data. IEEE Trans. Biomed. Eng. 2012, 59, 234–240. [CrossRef]