The rate of growth in data and human knowledge is staggering. Already-dated estimates are that the doubling time for medical knowledge will be just 0.2 years in 2020; learnings acquired by students in the 2020 graduating class in their first 3 years of medical school will amount to just 6% of those amassed from 2010 to 2020.¹ A typical primary care physician would need to spend about 627.5 hours per month to read all available articles in their field²; the average general internist devotes only 4.4 hours per week.³

On a separate but related front, health information from various sources is being gathered at an increasing rate. An estimated 153 exabytes (1 exabyte = 10⁹ gigabytes) of health care data were generated in 2013; the projected output is 2,314 exabytes in 2020.⁴ Widespread adoption of remote monitoring devices, among other emerging technologies, will fuel the surge in health care data beyond what prior projections could have estimated.

These trends have brought challenges yet hold great promise. The challenges relate to the enormous quantity and variety of data being collected for a given patient, much less an entire medical practice, that no health care provider can readily access, let alone digest. Software is required that operates in the background, collating such information and processing it into practical and individualized clinical algorithms. Fundamentally, it would need the capability to evolve over a patient's lifetime and deal simultaneously with multiple conditions on numerous levels. The promise, in terms of supporting optimal care, cannot be overstated. For the provider, advantages would include quicker and more accurate disease diagnosis, targeted and effective management support with less potential for error, and rapid notification of changes in patient status through remote and ongoing surveillance. For the patient, monitoring would facilitate health maintenance through continuous and real-time feedback regarding lifestyle habits and treatment support.

Computer technology capable of processing and effectively translating large amounts of data into decision aids is already available, and computational ability is itself increasing at an exponential rate.⁵ Computers were predicted to match human brain capability by around 2020 and be able process about 10²⁶ calculations per second by about 2040.⁵ Accordingly, processing capability should be able to cope easily with data growth. Healthcare has lagged behind other sectors in adopting electronic systems; wide implementation of electronic health record software has only recently occurred, whereas other industries have long used similar applications.

One notable field of technological advance in medicine is artificial intelligence, with systems already capable of accurately diagnosing physical and psychological illness, interpreting tests such as medical imaging, and directing robotic surgery.⁶ Another quickly developing area, which would benefit from an artificial intelligence interface, involves computerized decision support systems (CDSS). These tools have been commonplace in business management, and the health informatics industry is increasingly developing such applications for use in clinical care. Those currently available have predominantly been bundled with large-scale health information technology platforms intended for hospital-based use. Too commonly, their functional scope has been limited to simple actions such as suggesting when to initiate disease screening, ordering diagnostic tests, or assisting drug prescribing. Importantly, in view of their potential impact on patient well-being, few have undergone rigorous evaluation. A review of 162 randomized controlled trials of CDSS found that only 52-64% significantly improved process of medical care and just 15-31% positively impacted patient outcomes, which were predominantly surrogate in nature.⁷

There have been CDSS specifically developed and assessed regarding atrial fibrillation (AF). A retrospective, observational cohort analysis of 6,123 Ohio Medicaid patients examined a CDSS aimed at improving warfarin prescribing through recommendations regarding anticoagulant eligibility based on an automated, patient-specific risk-benefit assessment.⁸ Only 9.9 % of patients for whom the CDSS recommended warfarin treatment actually received it. The only influence on outcomes was a higher risk of gastrointestinal bleeding in patients prescribed warfarin where the CDSS recommended against its use. An observational study with controls evaluated the efficacy of a CDSS intended to improve guideline-concordant warfarin use in 268 patients with newly diagnosed AF at a tertiary care setting.⁹ The system had no impact on physician practice. An observational study in three primary care trusts in England assessed a three-component approach aimed at increasing appropriate anticoagulant use.¹⁰ This included the following:

• Altering professional beliefs through clinical guidance and education
• Using computer software to support clinical decisions and patient review
• Motivating change through evaluative feedback regarding practice performance, both individually and relative to peers

Appropriate increase in anticoagulant use (from 52.6% to 59.8%, p < 0.001) and decrease in aspirin use (from 37.7% to 30.3%, p < 0.001) resulted.

Expert-AF (Improving stroke prevention in atrial fibrillation), a cluster randomized trial involving 781 Dutch patients, assessed a CDSS aimed at optimizing stroke prevention therapy in general practice by providing actionable prescribing recommendations based on patient clinical profiles.¹¹ It could not demonstrate effectiveness, largely because the tool was under-used. Another cluster randomized trial, involving 1,857 patients with newly diagnosed AF in 43 primary care clinics in Sweden (the CDS-AF [Clinical Decision Support for Stroke Prevention in Atrial Fibrillation] study), evaluated a CDSS designed to increase appropriate, guideline-based prescribing of anticoagulant therapy.¹² After 12 months, a modest but significant increase in anticoagulation prescriptions over baseline was observed (70.3-73.0% in the CDSS group; 70.0-71.2% in the control group; p = 0.013). There was no effect on stroke, transient ischemic attack, or systemic embolism, but fewer significant bleeding events occurred in the CDSS group¹² versus controls¹⁶ (p = 0.04).

A few studies have been patient focused. An open exploratory randomized trial of 109 patients from 2 general practices in Northern England evaluating the impact of a computerized decision aid in shared decision-making consultation versus direct doctor-led advice based on paper guidelines showed mixed results.¹³ The tool had a beneficial effect on decision conflict, but intervention patients not already on warfarin were much less likely to start it versus those in the guidelines group (25% vs. 93.8%, respectively; relative risk 0.27; 95% confidence interval, 0.11-0.63). A study from primary care practices in the Veterans Affairs Connecticut Healthcare System randomized 135 patients with AF to interaction or no interaction with a computerized decision aid intended to provide education about the association between AF and stroke, the different antithrombotic treatment options, and the choice such treatment entails.¹⁴ Whereas patient knowledge increased and physician-patient communication was enhanced, the decision support tool had no influence on treatment plan. Guo et al. tested a mobile app aimed at patients with AF that incorporated clinical decision support tools.¹⁵ This cluster randomized trial of 113 Chinese patients found that the app improved patient knowledge, therapeutic adherence, and quality of life in terms of reducing levels of anxiety and depression. Sheibani and colleagues evaluated a tool that aimed to improve patient adherence to guideline-directed anticoagulation treatment in 373 patients with AF.¹⁶ Drug adherence increased from 48% at baseline to 65.5% after implementation of the CDSS (p-value < 0.0001). A before-and-after study of 44 patients from a centralized clinical pharmacy anticoagulation service found that a decision support tool that recommended warfarin dose to patients based on their international normalized ratio had no impact on percentage of time in therapeutic range.¹⁷

For the most part, the reviewed studies focussed narrowly on individual objectives, such as anticoagulation prescribing or patient education. Managing AF in its totality, including risk modification, rate and rhythm control, and appropriate and effective antithrombotic therapy use, especially in more challenging cases (e.g., a patient with renal impairment or recent ischemic cardiac event), is far more complex. Moreover, few patients have AF in isolation. Many have concomitant conditions in which adverse drug-drug or other interactions might occur (e.g., patients with arthritis managed with nonsteroidal anti-inflammatory drugs). IMPACT-AF (Improve Treatment With Anticoagulants in Patients With Atrial Fibrillation), a randomized controlled trial conducted in the province of Nova Scotia, Canada, examined just such a comprehensive tool. Preliminary findings were presented at the 2018 American Heart Association's Scientific Sessions; the full results are pending publication. There were 204 primary care physicians and 1,145 patients with AF enrolled, representing about 25% and 10%, respectively, of these groups in the province. The composite of AF-related emergency department visits or unplanned CV hospitalization over 12 months did not differ between study groups; however, patients in the usual care arm sustained more events than those managed with the CDSS (24.3% vs. 18.8%, incident rate ratio 0.78; 95% confidence interval, 0.51-1.18). Possible reasons for the lack of outcome difference include a 93% baseline use of antithrombotic therapy in patients ≥65 years or CHA₂DS₂ ≥1 and the practical challenges of using a tool that could not be integrated into the proprietary electronic medical record software used by participating physicians.

In summary, physicians already struggling with a massive amount of data will only find themselves further inundated. Tools to help them with its assimilation and application are overdue but are becoming increasingly available. However, the utility and benefits of specific products should not be assumed. Applications that focus on a single disease or process of care will have little utility. Instead, systems must be capable of managing patients across their entire lives and the variety of diseases they may acquire. Fundamentally, as with any intervention with a potential impact on patient outcomes, they will require objective evaluation.

References

1. Densen P. Challenges and opportunities facing medical education. Trans Am Clin Climatol Assoc 2011:122:48-58.
2. Alper BS, Hand JA, Elliott SG, et al. How much effort is needed to keep up with the literature relevant for primary care? J Med Libr Assoc 2004;92:429-37.
3. Saint S, Christakis DA, Saha S, et al. Journal reading habits of internists. J Gen Intern Med 2000;15:881-4.
4. Stanford Medicine 2018 Health Trends Report (Stanford Medicine website). 2019. Available at: https://med.stanford.edu/school/leadership/dean/healthtrends.html. Accessed February 5, 2019.
5. Kurzweil R. The Singularity is Near: When Humans Transcend Biology. New York: Viking Penguin; 2005.
6. Loh E. Medicine and the rise of the robots: a qualitative review of recent advances of artificial intelligence in health. BMJ Leader. 2018;2:59-63.
7. Roshanov PS, Fernandes N, Wilczynski JM, et al. Features of effective computerised clinical decision support systems: meta-regression of 162 randomised trials. BMJ 2013; 346:f657.
8. Wess ML, Schauer DP, Johnston JA, et al. Application of a decision support tool for anticoagulation in patients with non-valvular atrial fibrillation. J Gen Intern Med 2008;23:411-7.
9. Cook DA, Enders F, Caraballo PJ, Nishimura RA, Lloyd FJ. An automated clinical alert system for newly-diagnosed atrial fibrillation. PLoS One 2015;10:e0122153.
10. Robson J, Dostal I, Mathur R, et al. Improving anticoagulation in atrial fibrillation: observational study in three primary care trusts. Br J Gen Pract 2014;64:e275-81.
11. Arts DL, Abu-Hanna A, Medlock SK, van Weert HC. Effectiveness and usage of a decision support system to improve stroke prevention in general practice: A cluster randomized controlled trial. PLoS One 2017;12:e0170974.
12. Karlsson LO, Nilsson S, Bang M, Nilsson L, Charitakis E, Janzon M. A clinical decision support tool for improving adherence to guidelines on anticoagulant therapy in patients with atrial fibrillation at risk of stroke: A cluster-randomized trial in a Swedish primary care setting (the CDS-AF study). PLoS Med 2018;15:e1002528.
13. Thomson RG, Eccles MP, Steen IN, et al. A patient decision aid to support shared decision-making on anti-thrombotic treatment of patients with atrial fibrillation: randomised controlled trial. Qual Saf Health Care 2007;16:216-23.
14. Fraenkel L, Street RL Jr, Towle V, et al. A pilot randomized controlled trial of a decision support tool to improve the quality of communication and decision-making in individuals with atrial fibrillation. J Am Geriatr Soc 2012;60:1434-41.
15. Guo Y, Chen Y, Lane DA, Liu L, Wang Y, Lip GYH. Mobile Health Technology for Atrial Fibrillation Management Integrating Decision Support, Education, and Patient Involvement: mAF App Trial. Am J Med 2017;130:1388-96.
16. Sheibani R, Sheibani M, Heidari-Bakavoli A, Abu-Hanna A, Eslami S. The Effect of a Clinical Decision Support System on Improving Adherence to Guideline in the Treatment of Atrial Fibrillation: An Interrupted Time Series Study. J Med Syst 2017;42:26.
17. Simmons BJ, Jenner KM, Delate T, Clark NP, Kurz D, Witt DM. Pilot study of a novel patient self-management program for warfarin therapy using venipuncture-acquired international normalized ratio monitoring. Pharmacotherapy 2012;32:1078-84.

Keywords: Ischemic Attack, Transient, Warfarin, International Normalized Ratio, Decision Support Systems, Clinical, Anticoagulants, Atrial Fibrillation, Aspirin, Physicians, Primary Care, Mobile Applications, Tertiary Healthcare, Medicaid, Retrospective Studies, Confidence Intervals, Control Groups, Diagnostic Tests, Routine, Quality of Life, Electronic Health Records, Decision Making, Decision Support Techniques, Stroke, Embolism, Primary Health Care, Artificial Intelligence, Algorithms, Patient Compliance, Risk Assessment, Anxiety, Monitoring, Physiologic, Life Style, Drug Prescriptions, Diagnostic Imaging, Arrhythmias, Cardiac

< Back to Listings