While advocates of health information technology (HIT) emphasize its role in achieving the Triple Aim of lower cost, improved health, and better care (Berwick 2008, Sheikh 2015), insurers and providers are increasingly asking just how HIT translates into greater business efficiencies, better clinical outcomes, and enhanced customer loyalty.
Thanks to the rapid pace of change in HIT, the moving targets of health reform, and the particularities of local health care delivery, “it depends” is really the best answer. However, that doesn’t mean there aren’t some lessons from HIT implementation so far that can be applied across many settings.
Jaan Sidorov, MD
In our roles as founders of a health tech company, we’ve observed a number of trends that may inform the collaboration between HIT service providers and buyers. Below, we offer some insights and practical suggestions about HIT and its relationship to evidence-based medicine, clinical workflows, physicians, cost savings, and big data. We also look at the future, sounding a note of caution about whether artificial intelligence will not be ready for widespread adoption any time soon while expressing optimism about the potential for gamification to improve patient engagement.
Evidence-based vs. innovative. While businesses in every economic sector conduct research, health care’s large investment in evidence-based medicine is noteworthy. History, training, and culture have led generations of physicians, health services researchers, academics, and regulators to routinely apply considerable scientific scrutiny as well as skepticism to reports of new diagnostics or treatments. It’s no surprise, then, as providers, insurers, vendors, pharma, and other stakeholders compete for market share, claims of new HIT-based innovations continue to prompt old questions on whether the underlying data are tainted by bias, poor design, limited generalizability, questionable effectiveness, and unknown or unintended long-term effects.
To the frustration of investors and management teams, the pace and proprietary nature of much innovation in HIT is ill-suited to the pace of traditional peer review. Fortunately, the number of biomedical publishers has expanded, and many have expedited their review process and offer online publication. Two telling examples of the risk of failing to take advantage of this are lab provider Theranos and some direct-to-consumer skin care smartphone apps, which rushed to market with little peer-reviewed evidence to back up their claims (Ioannidis 2016, Resneck 2016).
As a result, HIT service providers and their customers should routinely contemplate investing in gathering, interpreting, and reporting their impact on Triple Aim-based outcomes with every customer in a peer-reviewed forum. Innovative health care entities that neglect health care’s reliance on evidence-based medicine and go to market without the benefit of any peer review do so at their peril. While time-consuming and expensive, promotion without accompanying proof can ultimately be far more costly.
This isn’t to say that a fast pace of innovation is bad for the health care industry. However, a careful consideration of the scope of “rollouts” and design of supporting studies can achieve getting the latest in the market quickly while verifying that it is, indeed, the greatest based on well-designed studies. Moreover, as health care evolves so individuals can make treatment choices based on quality, price, and convenience, researchers can adopt some of the practices used in the retail industry, such as quick parallel studies, to develop evidence that has traditionally come from randomized trials, observational studies, and case series.
The expedient adoption of health technology can be additive, not substitutive. Absent modification of existing job descriptions or workflows, frontline employees who are asked to adopt new technology will have to grapple with additional roles, new policies, unfamiliar procedures, extra oversight, and unforeseen problems. For them, the addition of more HIT—no matter how innovative it might be—inevitably leads to more work. Absent the concurrent assignment of incumbent low-value duties to machines or the dustbin, layering more HIT on top of a group of busy employees turns them into even busier employees. The introduction of HIT often means copying and pasting across multiple applications scattered across two or more devices rather than added efficiency.
Outside the travails of the electronic health record (Mandl 2012), this additive downside to HIT has gone largely unexamined in the peer-reviewed medical literature. Other reports describing this problem have noted the importance of obtaining frontline worker input and fostering collaboration across multiple vendors (Perna 2014). In our experience, importing an HIT solution also should serve as an opportunity to review which legacy tasks can be modified or discarded. Buyers also should be wary of “middleware” and quick fixes, which tend to add complexity over the long term. Instead, they should consider implementing staged rollouts of HIT-based innovations using a plan-execute-evaluate-adjust (“PLEXEVAD”) strategy.
Additive work often happens when new technology is not accompanied by adoption of the new ways of doing things by the end users of the technology. Without those changes, the promise of technology will often go unrealized. For a health care organization, that can mean wasted time, money, and employee goodwill.
Whither physicians? While physicians have seen their roles expanded to leading a team of providers overseeing complex care episodes, their expertise remains diagnosis and treatment. Yet, as health reform continues, physicians also are being relied on to fix its unintended consequences. Counting on physicians to make up for the shortcomings of the electronic health record should serve as a warning as HIT’s role expands in other sectors of health care delivery (Kesselheim 2011, AHRQ 2016). While HIT’s architects mean well, the impact of even a single small item added to a physician’s task list can range from a distraction to a cascade of new tasks with a cumulative impact on clinical workflows. Some systems ask doctors to identify the inevitable false-positives of risk-stratification algorithms and repeatedly embed blanket “talk to your health care provider” disclaimers in multiple patient or beneficiary interactions. Buyers and vendors instead should seek technology solutions that facilitate the functioning of the health care team by matching, when possible, any necessary human oversight to the appropriate level of non-physician expertise.
Grappling with insurance risk-transfer. One value proposition for HIT includes the reduction of avoidable health care utilization or costs. However, successfully decreasing claims expense often translates to a combination of 1) a very real loss of provider revenue, and 2) an abstract calculation avoiding health care utilization. Both are reduced further by the direct and indirect costs of HIT’s associated personnel and capital. As the art and science of shared-risk arrangements continue, it should be recognized that as the parallel role of HIT expands, gauging just how it “bends the cost curve” involves multiple care settings and has wide confidence intervals (Asch 2016).
To effectively navigate this challenge, both HIT providers and customers need to account for the health burden of the insured population being served, the impact of social determinants, baseline expenses, insurance claims trends, background cost inflation, and local provider network performance. Without this knowledge, calculations of the economic value of a particular HIT-based innovation in a particular setting for a particular population may not be a question that can be precisely answered.
The irony of less is more. Health care correlations derived from the analysis of huge, multisource datasets need to not only render meaningful insights, but be accompanied by actionable opportunities that can be scaled to available resources. In other words, the “big” of “big data” needs to be boiled down to a manageable number of achievable interventions for a manageable number of patients. For example, while the clinical issues and social determinants underlying an increased risk of rehospitalization have been the subject of considerable research, far less is known about prioritizing and modifying these determinants so that the few patients who are most likely to benefit are selected for the right intervention.
As population health and care management spreads to more and more consumers, the value proposition of HIT will include access to insights that give the greatest impact on cost and clinical outcomes. Once that is achieved, the experience can inform additional interventions for additional numbers of patients in a virtuous cycle of continuous improvement.
The ecosystem and gadgetry. Early versions of HIT were single source, end-to-end, and complex. This has given way to a networked and decentralized array of smaller and interchangeable billing, claims, health record, laboratory, imaging, data warehousing, and analytics components (Surviving 2016). As a result, health care’s chief technology officers increasingly preside over complex “ecosystems” of local and remote software and hardware. As with other systems, the whole becomes greater than the sum of its parts. The growth of HIT means the merits of upgrading, swapping, or supplementing any part is dependent not only on their individual functionality, but also on their interdependent compatibility and synergy.
Given this reality, the incorporation of consumer apps and monitoring devices into the HIT ecosystem is less revolutionary than evolutionary. While the consumer allure is undeniable, the ultimate value proposition of these apps and gadgets will depend more on their ability to enhance the consumer and provider experience by supporting, for example, patient-centeredness and shared decision making (Moore 2006). Their potential in these and other areas of the Triple Aim has only just begun to be documented in the peer-reviewed literature. The health care app and device vendors that can prove they create value in this ecosystem will have a key competitive advantage.
Humans plus health tech will still beat either alone. While artificial intelligence promises to completely outsource much complex decision-making to machines, the experience in many nonhealth care settings with established robotics is that computers plus human insight make for greater efficiency and effectiveness than either alone (Automation 2015). In addition, the art and science and associated cost management of health care delivery are still a matter of limited knowledge, insufficient evidence, myriad logic exceptions, and very human irrationality (Eichner 2010). While expedited access to scientific databases and the generation of potential diagnoses and treatments are well within reach, it remains unlikely that medical diagnostic and treatment guidelines will be translated into accurate computer code in the near future (Semigran 2016). The superiority of having live subject matter experts enter the HIT loop for even “simple” clinical tasks, such as giving dietary advice for the treatment of obesity (Ross 2016), suggests that for now, HIT will remain a decision support tool rather than a decision substitution tool.
This is not to say that the change to a more automated approach to management of diseases should be discouraged, but a note of caution about how implementation is warranted. Adopters of cutting-edge HIT should be skeptical about claims that we are on the cusp of HIT that is independent of any human oversight. If implemented prematurely, the result could be a limited menu of one-size-fits-all care options or a high frequency of exceptions. For now, the artificial intelligence version of “Dr. Watson” that is fully independent remains experimental. It is best to leave it in the labs of researchers or to your competitors.
The advent of gamification. Until now, mainstream efforts to improve diet, exercise, medication adherence, or provider appointments have had limited success (Dixon-Fyle 2010). Largely based on 1) educational appeals to improving personal health status or 2) using economic incentives to change behavior, the former has had a disappointing track record, while the latter have substantial cost and regulatory limitations.
Enter the alternative of health care “gamification,” in which consumers pursue healthful behaviors by competing for noneconomic and symbolic awards. This is emerging as a surprisingly effective tool in motivating behavior change, and its science is still evolving. The phenomenon of millions of Pokémon Go users increasing their physical activity levels in the pursuit of virtual avatars is just the latest, if very public, example of the potential low-cost synergies of gamification and HIT (Althoff 2016).
Gamification has been the subject of a considerable amount of applied research (King 2013) and, in contrast to artificial intelligence, may be ready for adoption in many health care settings. Once this tipping point is achieved, the disruptive technologies that support gamification for health promotion and disease management are likely to transform patient education and engagement. As a result, we predict early adopters will have a competitive advantage.
As HIT service providers rush to provide innovative solutions in the health care marketplace, they will need to manage multiple challenges all at once. They and their customers will need to meet the expectations of evidence-based medicine, deliver on the substitutive promises of innovation, avoid burdening physicians with additional tasks, grapple with risk-transfer calculations, leverage big data in the service of achievable outcomes, and serve as one of many components of an informatics ecosystem that also includes patient apps and gadgetry. While artificial intelligence holds great promise, the even greater complexity of health care decision making means its adoption is likely to be delayed for several years. In the meantime, the limitations of traditional education and incentives and the surprising appeal of handheld games makes “gamification” the next frontier of consumer engagement. HIT vendors and customers that succeed in these key areas will be the most likely to succeed in achieving the Triple Aim.
AHRQ (Agency for Healthcare Research and Quality). Alert fatigue. July 2016. Accessed Nov. 19, 2016.
Althoff T, White RW, Horvitz E. Influence of Pokémon Go on physical activity: study and implications. Microsoft Research. Oct. 6, 2016. https://arxiv.org/pdf/1610.02085v1.pdf. Accessed Nov. 19, 2016.
Asch DA, Pauly MV, Muller RW. Asymmetric thinking about return on investment. N Engl J Med. 2016;374(7):606–608.
Automation angst. Economist. Aug. 15, 2015. Accessed Nov. 19, 2016.
Berwick DM, Nolan TW, Whittington J. The triple aim: care, health, and cost. Health Aff (Millwood). 2008;27(3):759–769.
Dixon-Fyle S, Kowallik T. Engaging consumers to manage health care demand. McKinsey & Co. January 2010. Accessed Nov. 19, 2016.
Eichner J, Das M. Agency for Healthcare Research and Quality. Challenges and barriers to clinical decision support (CDS) design and implementation experienced in the Agency for Healthcare Research and Quality CDS demonstrations. March 2010. Accessed Nov. 19, 2016.
Ioannidis JP. Stealth research and Theranos: reflections and update 1 year later. JAMA. 2016;316(4):389–390.
Kesselheim AS, Cresswell K, Phansalkar S, et al. Clinical decision support systems could be modified to reduce ‘alert fatigue’ while still minimizing the risk of litigation. Health Aff (Millwood). 2011;30(12):2310–2317.
King D, Greaves F, Exeter C, Darzi A. ‘Gamification’: influencing health behaviours with games. J R Soc Med. 2013;106(3):76–78.
Mandl KD, Kochane IS. Escaping the EHR trap—the future of IT. N Engl J Med. 2012;366(24):2240–2242.
Moore LG, Wasson JH. An introduction to technology for patient-centered collaborative care. J Ambul Care Manage. 2006;29(3):195–198.
Perna G. Healthcare information system integration comes to life in Michigan Healthcare Informatics. Healthc Informatics. Jan. 23, 2014. Accessed Nov. 19, 2016.
Resneck JS Jr, Abrouk M, Steuer M, et al. Choice, transparency, coordination, and quality among direct-to-consumer telemedicine websites and apps treating skin disease. JAMA Dermatol. 2016;152(7):768–775.
Ross KM, Wing RR. Impact of newer self-monitoring technology and brief phone-based intervention on weight loss: a randomized pilot study. Obesity. 2016;24(8):1653–1659.
Semigran HL, Levine DM, Nundy S, Mehrotra A. Comparison of physician and computer diagnostic accuracy. JAMA Intern Med. 2016 Oct 10. doi: 10.1001/jamainternmed.2016.600. [Epub ahead of print].
Sheikh A, Sood HS, Bates DW. Leveraging health information technology to achieve the “triple aim” of healthcare reform. J Am Med Inform Assoc. 2015;22(4):849–856.
Surviving seismic change: Winning a piece of the $5 trillion US health ecosystem. September 2016. PwC Health Research Institute. Accessed Nov. 19, 2016 [Registration required].