Artificial intelligence (AI) is becoming ubiquitous in health care, largely through machine learning and predictive analytics applications. Although many of these tools have been available for decades, their recent application to common health care scenarios, such as screening and diagnosing, has fueled optimism about the use of advanced analytics to improve care. For example, researchers recently demonstrated that a deep convolutional neural network (CNN) may enable automated screening and diagnosis for retinopathy of prematurity with high accuracy and repeatability. Their algorithm diagnosed 91 of 100 images (91.0 percent) correctly, whereas 8 experts had an average accuracy of 82.0 percent [1]. A CNN also outperformed the majority of 58 dermatologists tested in accurately diagnosing melanoma, with a median area under the receiver operating characteristic curve of 0.86 compared with 0.79, P < 0.01 [2]. Similarly, researchers from Johns Hopkins University developed a novel machine-learning approach that provides rapid, remote, frequent, and objective assessment of Parkinson’s disease symptom severity using smartphones [3]. These tools are important contributors to consistent decision quality in treatment planning.
Perhaps the most common example of advanced analytics in health care, however, is predictive models to identify high-risk patients [4, 5, 6, 7, 8]. Potential benefits of these analytic tools include the early identification of patients at high-risk for 30-day readmission or mortality. In addition, the knowledge derived from these predictive models holds promise as a tool to target limited resources and address pressing public health issues such as preventing suicide attempts in adolescents [9].
Without a doubt, advanced analytics is transforming health care at a rapid pace. This transformation has been propelled by advances in computational power; the spreading of algorithms capable of simplifying complex tasks; the increasing availability of large amounts of data; and the number and diversity of innovators in this area. Advanced analytics are used in domains ranging from electronic health records to imaging and diagnostics, remote monitoring, drug discovery, billing and fraud prevention, and molecular profiling.
The potential benefits of applying advanced analytics in health care seem to be indisputable. However, evidence that these advanced analytics tools improve care and outcomes in a cost-effective, measurable way is limited. Many of the advanced analytics articles focus on the methods used, the assumptions made, the validation efforts undertaken, and the limitations. Few efforts have been made to include details about real-world performance of analytic tools, the clinical interventions that accompanied computer-generated knowledge, the implementation strategy adopted, the disruption to existing workflows, the adoption of technology, the education provided to both clinicians and patients about the use of advanced analytics in their care interaction, or cost-effectiveness. The promise of advanced analytics in health care is tremendous; however, ignorance of these challenges may overshadow its potential clinical impact and could result in waste. In this article, we offer basic practical considerations to individuals in health care settings who are involved in developing, buying, piloting, implementing, and evaluating advanced analytics solutions.
A comprehensive list of factors to consider is outside the scope of this commentary, but we offer some basic requirements based on many years of experience developing, implementing, and evaluating advanced analytics tools:
There is no question that advanced analytics will become an essential solution for improving health care. Arming our clinical experts with the best advanced analytics will enhance the delivery of care by allowing them to focus on issues directly related to patient care. However, careful and objective considerations need to be made before implementing a solution. Critically evaluating any advanced analytics solution before, during, and after its implementation will ensure safe care, good outcomes, and the elimination of waste.
The authors have no competing interests to declare.
Brown, JM, Campbell, JP, Beers, A, Chang, K, Ostmo, S, Chan, RVP, et al. Automated Diagnosis of Plus Disease in Retinopathy of Prematurity Using Deep Convolutional Neural Networks. JAMA Ophthalmol. Published online May 02, 2018. DOI: https://doi.org/10.1001/jamaophthalmol.2018.1934
Haenssle, HA, Fink, C, Schneiderbauer, R, Toberer, F, Buhl, T, Blum, A, et al. Reader study level-I and level-II Groups. Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists. Ann Oncol; 2018 May 28. DOI: https://doi.org/10.1093/annonc/mdy166
Zhan, A, Mohan, S, Tarolli, C, Schneider, RB, Adams, JL, Sharma, S, et al. Using Smartphones and Machine Learning to Quantify Parkinson Disease Severity: The Mobile Parkinson Disease Score. JAMA Neurol. Published online March 26, 2018. DOI: https://doi.org/10.1001/jamaneurol.2018.0809
Evans, RS, Burke, JP, Classen, DC, Gardner, RM, Menlove, RL, Goodrich, KM, et al. Computerized identification of patients at high risk for hospital-acquired infection. Am J Infect Control. 1992 Feb; 20(1): 4–10. DOI: https://doi.org/10.1016/S0196-6553(05)80117-8
Benuzillo, J, Caine, W, Evans, RS, Roberts, C, Lappe, D and Doty, J. Predicting readmission risk shortly after admission for CABG surgery. J Card Surg. 2018 Apr; 33(4): 163–170. DOI: https://doi.org/10.1111/jocs.13565
Evans, RS, Benuzillo, J, Horne, BD, Lloyd, JF, Bradshaw, A, Budge, D, et al. Automated identification and predictive tools to help identify high-risk heart failure patients: pilot evaluation. J Am Med Inform Assoc. 2016 Sep; 23(5): 872–8. DOI: https://doi.org/10.1093/jamia/ocv197
Evans, RS, Kfoury, AG, Horne, BD, Lloyd, JF, Benuzillo, J, Rasmusson, KD, et al. Clinical Decision Support to Efficiently Identify Patients Eligible for Advanced Heart Failure Therapies. J Card Fail. 2017 Oct; 23(10): 719–726. DOI: https://doi.org/10.1016/j.cardfail.2017.08.449
Woller, SC, Stevens, SM, Evans, RS, Wray, DG, Christensen, JC, Aston, VT, et al. Electronic Alerts, Comparative Practitioner Metrics, and Education Improves Thromboprophylaxis and Reduces Thrombosis. Am J Med. 2016 Oct; 129(10): 1124.e17–26. DOI: https://doi.org/10.1016/j.amjmed.2016.05.014
Walsh, CG, Ribeiro, JD and Franklin, JC. Predicting suicide attempts in adolescents with longitudinal clinical data and machine learning. J Child Psychol Psychiatry; 2018 Apr 30. DOI: https://doi.org/10.1111/jcpp.12916
Shah, ND, Steyerberg, EW and Kent, DM. Big Data and Predictive Analytics Recalibrating Expectations. JAMA. Published online May 29, 2018. DOI: https://doi.org/10.1001/jama.2018.5602