Improving Clinical Outcomes in the Era of Information Ubiquity

• clinical outcomes
• life expectancy
• precision medicine

A “fourth industrial revolution”¹ that blends biologic, physical, and data sciences is currently reshaping biomedical research and our larger understanding of the spectrum of human health. However, although these developments are expanding important capacities, the convergence of an aging population and multiple lifestyle-related risk factors is expected to drive an increase in the prevalence of kidney disease in coming years.² In the United States, these trends are taking place against a background of declining average life expectancy and growing disparities in health outcomes.^3,4 In contrast, throughout much of the rest the world, reductions in infant mortality and deaths from infectious diseases are contributing to significant improvements in life expectancy.⁵ However, this positive trend is partially offset by major increases in the incidence of noncommunicable diseases, including kidney disease, cardiovascular diseases, cancer, and chronic lung disease, that mirror the United States experience.

From the perspective of the American Society of Nephrology, the trends observed in the United States are particularly concerning. As we near the end of a third consecutive year marked by reductions in average United States life expectancy,⁶ the collective impetus for this retrograde motion is coming into focus: significant increases in “diseases of despair” that include drug overdose (predominately opioids), suicide, and cardiometabolic disease.⁷ This decline in health status is not uniform. We see substantial variation in outcomes among individuals as a function of geographic location (rural settings doing worse than urban settings, and pockets of urban regions have noticeable poorer health outcomes), wealth, education, sex, and race.^8–10 Recent data show dramatic differences in life expectancy as a function of location by county as well as rapid deteriorations in health status among white persons living in a belt reaching from Oklahoma to West Virginia (Figure 1),¹⁰ deteriorations driven by the “diseases of despair” described above. The “social determinants of health” related to wealth, education, and neighborhood manifest through the prevalence of traditional risk factors: lipid measurements, blood glucose levels, obesity, cigarette smoking, lack of exercise, and hypertension. These poor outcomes are accompanied by escalating medical costs in the United States, making it a unique outlier versus other high-income countries.¹¹ In short, among 18 peer nations, we have achieved the worst life expectancy at almost twice the cost.¹²

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

(A) In 2014, counties in South Dakota and North Dakota had the lowest life expectancy at birth, and counties along the lower half of the Mississippi and in eastern Kentucky and southwestern West Virginia also had very low life expectancy compared with the rest of the country. Counties in central Colorado had the highest life expectancies. (B) Compared with the national average, counties in central Colorado, in Alaska, and along both coasts experienced larger increases in life expectancy between 1980 and 2014, whereas some southern counties in states stretching from Oklahoma to West Virginia saw little, if any, improvement over this same period. Modified from ref. 10, with permission.

However, despite these dire circumstances, new tools offer us the means to stem and even reverse this tide. The digitization of previously disparate information offers the chance to reorganize our approach to preventing and treating noncommunicable diseases and to use high-quality information to focus on reducing the effect of chronic diseases, including renal disease, at individual and population levels. Ongoing and profound changes in computing are revolutionizing the ability to measure the human condition. Human biology, including the whole genome, proteins, metabolites, and the immune system, can be characterized and measured at previously impossible depth and speed.

At the same time, expanding capacity to collect, organize, curate, and store this complex information enables the use of this complex, multidimensional information for research—and soon, for clinical purposes. The ubiquity of electronic health records (EHRs) and curated billing claims data provides a rich resource of clinical information. Behavioral measures known as digital phenotyping¹³ can be assessed using sensors, wearable devices, and smartphones,¹⁴ providing profiles of behaviors that contribute to health or disease. In the past, the measurement of changes over time was constrained by the need for periodic assessments during clinic visits or reliance on human recall, and continuous monitoring was impractical due to limitations affecting the capture and storage of information. Now, streaming of information and two-way communication via cellular technology enable both a detailed longitudinal view of outcomes and intervention on the basis of “just-in-time” data, whereas robust data pipelines combine these disparate data within a common analytic framework.

For the individual, the integration of biologic, clinical, behavioral, social, and environmental information with data from similar patients should enable support for an increasing number of decisions. These detailed data will be further augmented by external information from clinical trials and outcomes studies and placed within a decision-making framework through clinical practice guidelines and care pathways (Figure 2).¹⁵

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Data and measurements from sources, including genetics, genomics, integrative physiology, electronic health records, digital phenotypes, and the environment, can be analyzed, combined with external data, interpreted in the context of clinical treatment guidelines, and integrated into clinical pathways to support decision making tailored to the specific needs of patients, caregivers, and clinicians. EHR, electronic health record; HR, heart rate. Modified from ref. 15, with permission.

At both individual and population levels, we will increasingly be able to produce detailed mapping and modeling that allow us to estimate the effect of proposed interventions and policies with more confidence. A frequently unappreciated aspect of this evolving picture is the ongoing reduction in data latency caused by the modern capacity to automate the organization and curation of data while flagging uncertain items for human attention. The ubiquity, liquidity, and short latency of data availability are creating a new era of evidence generation that enhances our ability to identify effective approaches, understand the balance of risk and benefit in alternative treatments, and implement proven interventions for both individuals and populations.

Successful examples of large-scale aggregation and use of information can be seen in innovative studies conducted by the National Patient-Centered Clinical Research Network (https://pcornet.org/) and the National Institutes of Health (NIH) Healthcare Systems Research Collaboratory (NIH Collaboratory; http://rethinkingclinicaltrials.org/). These large programs, characterized by collaborations among health systems, their clinicians, patients and their families, and payers, are conducting prospective studies that use data from EHRs, billing claims, and disease and quality improvement registries. Although the methods continue to be developed, early successes point to potential for substantial reductions in cost and an acceleration of evidence generation to guide practice. Furthermore, because the research is being conducted within existing health systems, translation of findings into practice can be expedited.

Given these broad examples, it may be useful to consider four large populations of patients with renal disease.

First, the designation of patients with ESRD—who have markedly reduced life expectancy and more frequent medical observation compared with almost any other patient group—as a special population with Medicare as a single payer, combined with prescient leadership by the clinical community has resulted in an impressive data system for tracking these patients and their outcomes. However, little progress has been made in improving survival, major disparities exist in health outcomes, and new therapies have not developed at rates seen in areas such as oncology. Furthermore, the clinical culture has adopted practices on the basis of experience and observation rather than randomized trials, leading to uncertainty about the actual risk-benefit balance of common practices. However, the voluminous data available from patients with ESRD through clinical data systems already in place can be combined with a growing array of digital information about quality of life to enable detailed characterization of these patients. Applying randomization to these rich data sources should resolve many uncertainties of practice, increase understanding of the biology of ESRD, and stimulate development of new technologies and therapies. The Time to Reduce Mortality in ESRD Trial¹⁶ has provided proof of principle that trials can be conducted with high-quality data at a low cost, and the ongoing Hi-Lo Trial¹⁷ is taking the learnings into a new era of ESRD trials that promise to answer the many questions about dialysis care.

Second, the burgeoning population of people with CKD is now amenable to much more aggressive systematic early intervention. Because health systems are increasingly able to aggregate data and conduct needed research much less expensively and more rapidly, our knowledge about which interventions are effective will correspondingly advance. The creatinine values and estimated creatinine clearance values are available for the vast majority of Americans as shown in the Veterans Administration population (Figure 3),¹⁸ and the ability to use decision support to suggest interventions and identify people for trials to define how to prevent progression of renal disease provides a scale that has not been possible until recently. However, in contrast to the ESRD population, with its dedicated access to nephrologists, much of this primary and secondary prevention research will need to be done in conjunction with broad health system constituents. At the individual level, a major change will be the integration of digital medicine into behavioral interventions. At the population level, rules-based guidelines should provide a basis for effective intervention.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

(A) Crude prevalence of rapid eGFR decline. Prevalence represents the number per 100 United States veterans. (B) Spatial cluster analysis reveals geographic clustering of rapid eGFR decline prevalence. Adjusted for age, race, sex, diabetes, and hypertension. Median (confidence interval) prevalence represents the number per 100 United States veterans. Modified from ref. 18, with permission.

Third, persons with AKI make up a significant population. However, although substantial uncertainty remains about optimal approaches for preventing and treating AKI, the vast majority of these patients have detailed EHRs with data that should enable prospective, interventional clinical trials at a much lower cost than traditional trials, facilitating more rapid progress in preventive and therapeutic strategies.

Fourth, experts estimate that 10% of CKD has a biologic basis that is not among the common noncommunicable disease variants.¹⁹ The ability to measure the genome and the full range of its phenotypic expression is leading to targeted therapies for these patients. As the dimensionality of biologic data grows richer and our ability to analyze it improves, we are likely to see faster breakthroughs with precision therapies in small populations, offering hope to families who hitherto have lacked options. The large-scale health system data described above should enable aggregation of patients with similar genomic etiologies for their diseases so that definitive trials can be done with adequate sample sizes.

Activation of patients and their families will be critically important in realizing this bright future. As patients become involved in the generation of evidence and the delivery of high-quality measurable care, effective therapies will be used more reliably, ineffective therapies will be rapidly retired, and new therapies will be developed more efficiently and with greater assurance that their use is appropriately supported by high-quality evidence. The Kidney Health Initiative,²⁰ a public-private partnership comprising patients and their advocates, government, clinicians, and industry, offers a template for effective patient organizations that can accelerate learning and therapeutic development.

This perspective may seem unrealistically optimistic about the capacity of the newly emerging field of digital medicine to provide a substrate for substantially improving United States health outcomes at a lower cost, but the direction of this new information era—this “fourth industrial revolution”¹—is clear, and the rate of progress in the collection, curation, and analysis of digital data is accelerating. However, no amount of digital information and analysis can substitute for an informed and motivated consortium of patients, families, and clinicians supported by policy makers who wisely guide efforts to provide incentives and enable systems that reinforce effective behaviors and interventions at reasonable cost.

• Copyright © 2019 by the American Society of Nephrology