Health Informatics Opportunities to Bend the Health Care Cost Curve

In 2017, we crossed the $3.3 trillion dollar mark (17.9% of Gross Domestic Product [GDP]) on health care spend in America. In 2008, we spent $2.4 trillion (15.2% of GDP). The rate of health care spending increase continues to surpass the country’s overall inflation rate. Our most promising payment reforms seem to be limited to specific cases like bundled payments and reference pricing.  Accountable Care Organizations (ACO’s) participating in two-sided risk models appear to save Medicare about three times more money than one-sided ACO’s ($4.5 million vs. $1.3 million) in 2016. During the same period, the total Medicare spend across the patients participating in all ACO’s was over $81 billion. The sum of all ACO interventions saved Medicare less than one percent of its spend. Without radically different approaches, it seems unlikely that payment reform will lead to meaningful reductions in health care spend.

I would support spending more on health care if that increase was linked to better health outcomes. In this country, we have seen a decrease in potential years of life lost (from 4870 years/100,000 population in 2008 to 4580 years/100,000 in 2013), but our performance lags behind countries with similar GDP (2765 years/100,000 in 2013). International comparisons are tricky, but this comparison challenges the belief that Americans are getting more health for their spend.

A different approach is to have the patient pay a higher percentage of their health care costs. Uwe Reinhardt, the Princeton health care economist who died last month, predicted companies would shift health costs from defined-benefit (health care coverage) to defined-contribution (fixed contribution toward health insurance premiums) back in 1989. As health care costs rise annually, patients in “regular” health insurance plans (not high-deductible ones) are liable for a higher amount of health care costs. When patients shoulder a larger portion of their health care costs, they decrease both appropriate and potentially inappropriate health care services. Cost-shifting to patients appears to be too blunt an instrument to increase health care value.

Can information technology save health care? The Health Information Technology for Economic and Clinical Health (HITECH) Act in 2009 included billions of dollars to spur electronic medical record adoption among offices and hospitals across the country. Congress expected technology to revolutionize the health care industry like technology had transformed other industries like finance. Nearly eight years later, the revolution has not arrived yet. Health information technology with existing care delivery settings has addressed some problems (legibility, access to medical records), but exacerbated old problems and created new ones (errors in order entry, clinician burnout). Recent investments in health care startups and mergers among larger companies suggest private-sector interest in changing the way health care is delivered to improve value. None of the press releases or articles I have read make me feel hopeful that these approaches will radically accelerate health care value creation.

The National Library of Medicine defines health informatics as “the interdisciplinary study of the design, development, adoption and application of information technology-based innovations in healthcare services delivery, management and planning.” Here are four* areas where health informaticists within health systems, hospitals and offices could improve value in health care:

Suggest diagnostic and treatment pathways (standardization)

For diagnoses or interventions where there a) is high-quality evidence or strong consensus, b) is clear evidence of higher quality care (e.g., safer, more efficient, more effective) c) are interventions that are easy to monitor electronically, AND d) are few or no financial costs to the health care system, health informatics can help move users toward new ordering patterns. The marked reduction in blood product use over the last decade reflects how the industry can change relatively rapidly. Many health systems, including my organization, have used health informatics tools to guide users toward more appropriate blood product ordering behaviors.

Unfortunately, blood product utilization is an exception. The American Board of Internal Medicine launched the Choosing Wisely campaign in 2012 to help identify tests and treatments of low-value to patients. The campaign suggests health care workers order too many tests and treatments unnecessarily. Choosing Wisely now includes over 70 specialty organizations with nearly 500 recommendations. An analysis of seven tests and treatments on the Choosing Wisely list showed effectively no change in ordering behavior. Presenting new evidence is not enough. Patashnik and Dowling suggest specialty societies and public policy should play a larger role in shifting practice patterns. Dr. Lisa Rosenbaum, a New England Journal of Medicine national correspondent, wrote a piece arguing against our current movement against “less-is-more” in health care. Efforts to reduce unnecessary testing appear to be linked with withholding appropriate testing among high-risk patients.

Rather than focus on reducing waste, health informatics tools could be used to present information to help users determine when specific diagnostic or treatment choices may be more or less effective. For packed red blood cells, the literature quickly converged on managing patients to a specific hemoglobin value in most cases. This allowed health informatics teams to develop interventions to embed that information in ordering processes. For patients with high blood pressure who are not controlled after two months with three antihypertensives, the health informatics team could design interventions to prompt users to consider referring a patient to a hypertension specialist to help consider more aggressive interventions to manage the patient’s resistant hypertension. Health informatics would be positioned to address overuse and underuse.

Provide feedback

In addition to guiding users to make different ordering decisions at the point of care, health informatics can show users how they are performing compared to their peers. This feedback has been shown to be modestly effective in changing ordering behaviors outside clinical encounters.

The performance data can be used by managers to identify outliers. Users who are markedly above (or below) their peers may represent inappropriate behavior or unique expertise on approaching a specific clinical problem. Subject matter experts and clinical councils can use the information to update diagnostic or treatment pathways accordingly. Health informatics in 2017 is not ready to determine which outliers are doing something really poorly or really well without additional expert input. In the absence of clinical consensus, the performance information can help leaders identify behaviors that are associated with better outcomes.

Sharing information across care settings

Health informatics could support more patient-specific information sharing across care settings. A site-specific electronic medical record facilitates data sharing within the site. An enterprise-specific electronic medical record facilitates data sharing across the enterprise. The standard of care today is to transfer information seamlessly within a health system’s inpatient and outpatient care settings including care management when appropriate.

But patients make decisions based on where to get their care based on their value perceptions for a diagnosis or treatment, not the health care provider’s organizational affiliation. Today, transferring information across health care systems is erratic because it depends on multiple processes that are poorly monitored. Many health systems have resisted the federal government’s efforts to facilitate interoperability by limiting what information is available within patient-facing portals and maintaining processes that hinder electronic data sharing.

Beyond sharing clinical data across health systems, health informatics could have a role in empowering patients and their caregivers in determining what the clinical data means. Translating imaging results and discharge summaries into patient-friendly language could help patients and caregivers make determinations about what follow-up care is needed next, regardless of where that care might be provided.

Support shared decision making

Once patients and caregivers understand their health care information, the next step in value creation is to determine how their preferences affect the next diagnostic or treatment decision. A patient who prefers to maximize quality-adjusted life years will make a different decision than someone who prefers to minimize iatrogenic complications. Without explicitly accounting for these preferences at every encounter, the health care system risks delivering inappropriate care. Medication non-adherence could be viewed as a mismatch between a health care provider and patient understanding of an intervention’s costs and benefits (Caveat: this would not apply to self-destructive behaviors [e.g., narcotics for a narcotic-addicted patient]).

There are two levels where health informatics could provide additional value in shared decision making. First, health informatics can expose the relevant options for the next diagnostic or treatment decision. The presentation could include patient-specific preferences to help determine the most appropriate option. Second, health informatics can provide details for where a particular health care good or service could be obtained. A patient who prefers obtaining a hip replacement from a high-volume, low-complication provider regardless of cost or travel may choose a different orthopedic surgeon than a patient who prefers a hip replacement closer to home even if there is an increased risk of complications. Service providers would be motivated to publish outcomes relevant to a patient’s preferences if that information would distinguish them in the marketplace.


There are no clear indications that the American health care system will be taking the necessary steps to deliver significantly more value for the forseeable future. In an effort to control costs, I believe employers and health plans will continue to propose health insurance products that shift risk to patients. In a high-deductible or catastrophic care coverage environment, health informatics can help patients and health care workers take incremental steps toward identifying and implementing a patient-specific interpretation of high-value care. While not as striking as payment reform or as simple as downloading an app, this work could be more meaningful in bending the health care cost curve over the next decade.

*I have not included two opportunities for health informatics to leverage information technology to improve the American health care system because I think the work will be done by others.

First, health systems could use information technology to improve the patient’s experience within the health system. Finance and retail have been quicker to use information technology to add value for their customers. I can interact with my bank or favorite store through a website or mobile app to complete most of my transactions. More unique tasks require a online chat or phone call. The technology and end-user expectations are present today to convert scheduling, registration, payment and communication between health care encounters from face-to-face or telephone interactions to electronic ones. In most cases, I believe health systems will rely on its technology vendors rather than their employed health informaticists to make these changes.

Second, health information technology could be used to help patients making lifestyle changes. Although I believe supporting patients to change their health behaviors (stopping smoking, changing their diet, exercising regularly) would be more effective than any of the other opportunities I outlined above in reducing years of life lost at a lower cost, I expect this work to be led by behavior change experts outside health care. Clinicians can suggest patients lose weight or exercise more, but those patients are more likely to use their own support networks with their own associated technology to make a behavior change and then engage friends and family to sustain that behavior change over time.