Data-Driven Approaches to Assessing the Value of Innovations in Health Care: A Q&A with Professor Amol Navathe

Data are critical to health care innovation – not only on the clinical side but also for assessing the value of new technologies, methods, and policies.

Today, the amount and kinds of data available on how patients receive and pay for medical treatments is extensive. Such data can be used to develop innovative technologies, methods, and policies that have the potential to influence clinical practice and improve the quality and cost effectiveness of care. Importantly, these same types of data can also be used to assess and estimate the value created by such innovations.

To explore this topic, Managing Principal Alan White and Vice President Michael Carson sat down with Analysis Group affiliate Amol Navathe, a health economist, practicing physician, and professor of medical ethics and health policy and health care management and economics at the University of Pennsylvania. Professor Navathe is the vice chair of the Medicare Payment Advisory Commission (MedPAC), an agency that advises Congress on Medicare policy. He co-founded the software company Embedded Healthcare, which applies behavioral economics to the development of value-based care model design in clinical practices. His published research covers a wide range of topics, including the impact of financial incentives offered to health care providers for delivering higher-quality patient care.

Dr. White and Mr. Carson, two of several Analysis Group consultants who have supported Professor Navathe on expert engagements, spoke with him about the multifaceted role of data in assessing the value of innovations in health care.

What are some of the unique challenges faced when assessing the value of innovations in health care?

The value associated with a health care innovation varies by context. For example, innovative approaches to analyzing health care data may provide clinical value insofar as they help maximize patient safety and reduce adverse health events. Separately, the launch of new virtual health care delivery platforms may provide commercial value to the extent that they increase the capacity of providers to offer additional or higher-quality care. Further, the introduction of novel drugs or treatments could lead to a net increase in societal value, including in terms of cost savings for insurance policy holders and payers.

In short, there can be a variety of different types of value created by an innovative health care technology, method, or policy. Moreover, assessments of value can have wide-ranging implications. For instance, they could be used by companies to determine appropriate levels of physician compensation, medical services pricing, and projections of expected profitability. Such assessments could also be leveraged to help justify investments or support policy recommendations. In a litigation context, assessing the value or impact of an innovation can be a central issue when evaluating liability and damages, including in breach of contract or licensing disputes, trade secret misappropriation cases, or False Claims Act matters.

How could the application of data-driven approaches to assessing the value of innovations in health care help stakeholders overcome the challenges you have outlined?

Anecdotal evidence or generalization can lead to imprecise estimates of value. When assessing the value of an innovation, it is critical to analyze relevant data to assess how it might impact patients, providers, payers, manufacturers, or other stakeholders in the health care system. For example, analyses of patient-level data or insurance payer–level data could be performed to evaluate whether the price paid for a novel health care technology was justified based on its clinical, commercial, or societal impact.

A data-driven approach to an assessment of value often represents the most rigorous means of identifying and estimating value. Such an approach is also becoming more feasible given the increasing availability of data and tools that can be used to analyze these data. This type of approach can also open the door to analyses of data on non-traditional elements of value, which can provide additional assurances on the validity of an assessment’s findings. For example, an assessment of the overall value of a treatment could be reinforced by augmenting that assessment with an additional analysis of the value that some patients may place on treatments that offer more certain outcomes, as compared to treatments associated with more variation in expected outcomes.

 

What about the clinical side? In what ways could the increasing collection and availability of health care data affect how patients are treated?

Health care delivery is complex, and the varied presentations and courses of diseases and their comorbidities can give rise to uncertainty in identifying effective treatments. It is challenging for health care professionals to actively monitor all the potential harms that patients might experience. As a result, it is possible that new adverse health events could occur, even while care is being delivered. For example, if a provider fails to identify that an admitted patient has a pre-existing condition, that condition could be left untreated, potentially leading to significant harm to the patient.

As a physician, I have observed that my clinical decisions have been increasingly tailored to the needs of individual patients due to the availability and use of patient-level data. To take one example: Recommendations for treatment of a wide range of cancers can depend on which (if any) mutations are present in a patient’s cancer cells. Since certain medications are more effective for treating patients with specific mutations, patient-level data can be used to efficiently match patients to the treatments that they need.

By tailoring health care delivery and medical treatments to individual patients, we are getting much closer to realizing the promise of “precision medicine”– the tailoring of care to an individual’s particular genetic makeup, personal characteristics, and environment. Precision medicine has the potential to unlock greater health benefits for patients than approaches based solely on analyzing population-level data. Such an approach to care will require the collection and analysis of an even greater amount of patient-level data and could allow physicians to be more targeted in the treatments that they offer, potentially leading to improved health care outcomes.

Can you provide an example of when you used patient-level data to assess the value of a health care innovation?

In one engagement, I evaluated a software algorithm that uses dozens of logical rules to analyze patient-level data from electronic health records (EHRs). That software algorithm was used to identify patient safety concerns and adverse health events and communicate them to health care professionals. Increasingly, government and private insurance payers are assessing provider performance based on quality and safety metrics, such as improving patient safety and reducing patient harm, and tying financial incentives for providers to treatment outcomes related to those metrics. Since providers interact with patients in medical settings ranging from routine physicals to emergency room visits, recording data in EHRs on patients, including their vital signs, diagnostic tests, age, and biological sex, can be critical in delivering high-quality and cost-effective health care.

To assess the software’s impact, I reviewed technical documentation that described the logical rules it employed and the user interface that communicated the algorithm’s results. With this understanding in mind, I reviewed published literature that described the benefits offered by similar types of software algorithm systems. I ultimately identified from my research that the algorithm supporting this software has the potential to generate significant cost savings for patients, providers, and payers.

Could insurance payer-level data be used similarly to assess the value of an innovation in health care?

Yes, but analyzing such data to assess the value of an innovation can be a complex exercise and typically requires extensive expertise and in-depth reviews. That is because payer-level data not only reflect health care services provided, but also financial transactions between payers, providers, and patients. As a result, these data likely include large amounts and varied types of detailed information. 

An example of my use of payer-level data to assess a health care innovation is a study that I led to evaluate an innovative care model for patients with advanced kidney disease. In general, such patients typically receive kidney care in a traditional facility setting. However, for this innovative care model, certain patients would receive additional treatment visits outside of that setting. Using payer-level data and my clinical expertise, I analyzed patients receiving dialysis in varied settings, and examined health care resource utilization and outcome patterns related to their treatment. I found that offering patients additional visits outside of the traditional facility setting, as enabled by this innovative model, was correlated with longer survival for patients and cost savings for payers.

Does the emergence of artificial intelligence (AI) have implications on how health care data can be used to generate value for stakeholders?

The use of AI-powered computer processors and machine learning techniques to rapidly perform analyses and identify insights from an increasing amount and variety of real-world data, could, in fact, generate significant value for health care system stakeholders. For instance, AI can be used to compare and assess data on patients, providers, or payers against other data to discover previously unidentified patterns. Such patterns could then be used to support and enhance the development of innovations.

An example of this is the ability of AI algorithms to identify groupings of clinicians who practice medicine in different ways and assess how such differences could have implications on the cost and quality of care available to patients. These findings can be used to modify and improve clinician practice styles, thereby helping to promote safer and more cost-effective methods of providing patient care. Similarly, AI can also look at patterns of behavior exhibited by patients and lend insights into those patients’ risks of comorbidities that might otherwise go undetected by providers in clinical settings. ■