7/7/21, 9:13 PMPrint

Page 1 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

Part I An Overview of Analytics in Healthcare and Life Sciences

7/7/21, 9:13 PMPrint

Page 2 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

1 An Overview of Provider, Payer, and Life Sciences Analytics

Thomas H. Davenport and Marcia A. Testa

The healthcare industry is being transformed continually by the biological and medical sciences, which hold considerable potential to drive change and improve health outcomes. However, healthcare in industrialized economies is now poised on the edge of an analytics-driven transformation. The field of analytics involves “the extensive use of data, statistical and quantitative analysis, explanatory and predictive models, and fact-based management to drive decisions and actions.”1 ( Analytics often uses historical data to model future trends, to evaluate decisions, and to measure performance to improve business processes and outcomes. Powerful analytical tools for changing healthcare include data, statistical methods and analyses, and rigorous, quantitative approaches to decision making about patients and their care. These analytical tools are at the heart of “evidence-based medicine.”

Analytics promises not only to aid healthcare providers in offering better care, but also more cost-effective healthcare. Several textbooks have been written on the cost-effectiveness of health and medicine, and health economics and the methods described can be used in healthcare decision making.2 ( , 3

( , 4

( Moreover, as healthcare spending rose dramatically during the 1970s and 1980s in the United States, an increased focus on “market-driven” healthcare developed.5 ( Today, as the amount spent on healthcare has risen to nearly 20% of GDP in the United States, analytic techniques can be used to direct limited resources to areas where they can provide the greatest improvement in health outcomes.

Analytics in healthcare is an issue for several sectors of the healthcare industry involving patients, providers, payers, and the healthcare technology industries (see Figure 1.1 ( ). As shown, the patient is the ultimate consumer within the healthcare system. This system consists of several sectors, including providers of care; entities such as employers and government that contribute through subsidized health insurance; and life science industries, such as pharmaceutical and medical device companies.

7/7/21, 9:13 PMPrint

Page 3 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

Figure 1.1 The healthcare analytics environment

Provider Analytics A key domain for the application of analytics is in healthcare provider organizations—hospitals, group practices, and individual physicians’ offices. Analytics is not yet widely used in this context, but a new data foundation for analytics is being laid with widespread investments—and government subsidies—in electronic medical records and health outcomes data. As data about patients and their care proliferate, it will soon become feasible to determine which treatments are most cost-effective, and which providers do best at offering them. However, to maximize their usefulness, analytics will have to be employed in provider organizations for both clinical and business purposes and to understand the relationships between them.

Tom Davenport and Jeffrey Miller in Chapter 2 ( , “An Overview of Analytics in Healthcare Providers ( ,” make the case that analytics for healthcare providers is poised to take off with the widespread digitization of the sector. They describe the current maturity level of provider analytics as low and describe current analytical applications along the continuum of descriptive, predictive, and prescriptive for both clinical and financial business purposes. And they address future areas for analytics contributions including meaningful use, accountable care organizations, taming the complexity of the clinical domain, increased regulatory requirements, and patient information privacy issues.

7/7/21, 9:13 PMPrint

Page 4 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

Payer Analytics Payers for healthcare, including both governments and private health insurance firms, have had access to structured data in the form of claims databases. These are more amenable to analysis than the data collected by providers, who have relied largely on unstructured medical chart records. However, historically payers focused on collecting data that ensure efficiencies in billing and accounting, rather than healthcare processes and outcomes. Even with limited administrative databases, payers have, at times, been able to establish that some treatments are more effective and cost- effective than others, and these insights have sometimes led to changes in payment structures. Payers are now beginning to make inroads into analytics-based disease management by redesigning their information databases to include electronic medical records. However, there is much more to be done in developing medical information databases and systems and employing analyses within payer organizations. In addition, at some point, payers are likely to have to share their results with providers, and even patients, if systemic behavior change is to result.

Kyle Cheek in Chapter 3 ( , “An Overview of Analytics in Healthcare Payers ( ,” concentrates on analytics as a value driver to improve the business of health insurance and the health of its members. He provides a framework of the types of analytics that can add value, and he reviews the current state, which he describes as “analytical sycophancy.” He concludes with paths to maturity and best practice examples from leading organizations.

Life Sciences Analytics Life sciences companies, which provide the drugs and medical devices that have dramatically changed healthcare over the past several decades, have also employed analytics much more than providers. However, their analytical environment is also changing dramatically. On the R&D and clinical side, analytics will be reshaped by the advent of personalized medicine—the rise of treatments tailored to individual patient genomes, proteomes, and metabolic attributes. This is an enormous (and expensive) analytical challenge that no drug company has yet mastered. On the commercial analytics side, there is new data as well—from marketing drugs directly to consumers, rather than through physicians—and new urgency to rein in costs by increasing marketing and sales effectiveness.

Dave Handelson in Chapter 4 ( , “Surveying the Analytical Landscape in Life Sciences Organizations ( ,” starts off with the contextual reality that it is no longer “business as usual” in the life sciences industries, which has resulted in a heightened focus on analytics. He describes the potential analytical contributions related to the primary business functions, including research discovery, clinical trials, manufacturing, and sales and marketing. He notes that healthcare reform and the emphasis on cost containment place more reliance on analytics that includes new reimbursement strategies and the need to use comparative effectiveness results in assessing the value of therapies.

7/7/21, 9:13 PMPrint

Page 5 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

Patients Analytics Patients are, of course, the ultimate consumers of healthcare and will need also to become better informed consumers of analytics—at least to some degree. They will need analytics to decide which providers are most effective, whether the chosen treatment will work, and in some payment structures, whether they are getting the best price possible. These consumer roles are consistent with the “consumer health informatics” and “Health 2.0” (use of web-based and e- technology tools by patients and physicians to promote healthcare and education) concepts. Of course, complex biostatistics and the results of comparative effectiveness studies are unlikely to be understood by most patients and will have to be simplified to be helpful.

Collaboration Across Sectors Each of the sectors that participate in healthcare progressively adds analytical capability, although at different rates. For true progress, analytics must be employed collaboratively across the various sectors of the healthcare system. Providers, payers, and pharmaceutical firms must share data and analyses on patients, protocols, and pricing—with each other and with patients—and all with data security and privacy. For example, members of each sector had data that might have identified much earlier that COX-2 drugs (Vioxx, Celebrex, and Bextra) were potentially associated with greater risk of heart disease.

Barriers to Analytics Healthcare organizations desiring to gain more analytical expertise face a variety of challenges. Providers—other than the wealthiest academic medical centers—have historically lacked the data, money, and skilled people for analytical projects and models. Even when they are able to implement such systems, they may face difficulties integrating analytics into daily clinical practice and objections from clinical personnel in using analytical decision-making approaches. Payers typically have more data than providers or patients, but as noted above the data are related to processes and payments (administrative databases) rather than health outcomes (research databases). Moreover, many payers do not now have cultures and processes that employ analytical decision making.

Life sciences firms have long had analytical cultures at the core of their research and clinical processes, but this doesn’t ensure their ongoing business success. Clinical trials are becoming increasingly complex and clinical research more difficult to undertake given the restrictions imposed by Institutional Review Boards, ethics committees, and liability concerns. Drug development partnerships make analytics an interorganizational issue. And the decline of margins in an increasingly strained industry makes it more difficult to afford extensive analytics.

While statistical analyses have been used in research, analytics has not historically been core to the commercial side of life sciences industries, particularly in the relationship with physicians’ practice patterns. Life sciences firms must normally buy physician prescribing data from a third-party source, and the data typically arrive in standard tables and reports rather than in formats suitable for further analysis. The firms increasingly need to target particular physicians, provider institutions, and buying groups, but most do not have the data or information to do so effectively.

Despite these obstacles, healthcare organizations have little choice but to embrace analytics. Their extensive use is the only way patients will receive effective care at an affordable cost.

7/7/21, 9:13 PMPrint

Page 6 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

Notes 1 ( . Thomas H. Davenport, Jeanne

G. Harris, Competing on Analytics (Boston, MA: Harvard Business Press, 2007) p. 7.

2 ( . Marthe R. Gold, Joanna E. Siegel, Louise B. Russell, and Milton C. Weinstein, Cost-Effectiveness in Health and Medicine (New York, NY: Oxford University Press, 1996).

3 ( . Rexford E. Santerre and Stephen P. Neun, Health Economics: Theories, Insights, and Industry Studies, 5th Edition (Mason, OH: South- Western Cengage Learning, 2010).

4 ( . Sherman Folland, Allen Goodman, and Miron Stano, The Economics of Health and Healthcare, 6th Edition (Upper Saddle River, NJ: Prentice Hall, 2009).

5 ( . Regina E. Herzlinger, Market Driven Healthcare: Who Wins, Who Loses in the Transformation of America’s Largest Service Industry (Cambridge, MA: Basic Books, A Member of the Perseus Books Group, 1997).

7/7/21, 9:13 PMPrint

Page 7 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

2 An Overview of Analytics in Healthcare Providers

Thomas H. Davenport and Jeffrey D. Miller

For the most part, analytics is just beginning to be employed in most healthcare provider organizations. However, it offers a high degree of potential to immediately improve patient safety and the financial and operational performance of the healthcare enterprise. Given the rapid rise in adoption of electronic medical records (and reimbursement based on their “meaningful use”), we expect to see considerable advances in the use of analytics by providers over the next five years as analytics provide the foundation for improved clinical quality and reduced cost of care.

There are two major areas of analytical activity in provider organizations:

• Financial and operational—These areas include analytical applications to monitor, predict, and optimize facilities, staffing, admissions, reimbursements, and other key factors driving the performance of a provider institution.

• Clinical and patient safety—This domain includes analytics related to evidence-based medicine and clinical decision support, comparative effectiveness, patient safety, survival rates, and compliance with care protocols.

In both areas, descriptive analytics (reporting, scorecards, dashboards, and so on) have been the primary focus (as compared to predictive analytics or prescriptive analytics). In many cases, reporting has been driven by increased regulatory requirements or narrowly defined performance improvement programs (e.g., reduced supply costs). Predictive analytics have largely been employed for forecasting patient admissions (in some cases by service line) on the business and operational side, and for scoring patients likely to need intervention on the clinical side—but largely in academic medical centers and on a pilot basis.

Overall, analytics at providers are highly localized, generally within department or functional boundaries. More sophisticated providers have pockets of analytical activity in various places, but little coordination. One leading academic medical center, with a strong history of clinical decision support and some degree of business analytics as well, had never had any contact between the two groups until a recent task force on “meaningful use.”

In terms of the analytical capability framework in Analytics at Work (see Figure 2.1 ( ), few provider institutions go beyond Stage 3, “Analytical Aspirations.” Most community hospitals are probably at Stage 1 (“Analytically Impaired”), because they lack electronic medical record (EMR) data (as of 3Q 2010, only 50% of hospitals with fewer than 200 beds had achieved the third of the seven stages of EMR adoption), and because they lack the human and technology resources to analyze financial and operational data. Mission-focused (religious or charitable) providers may find it particularly difficult to marshal the resources to perform serious analytics given the many demands on the limited capital budgets of these institutions.

7/7/21, 9:13 PMPrint

Page 8 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

Figure 2.1 Five levels of analytical capability

The larger for-profit hospital chains and integrated delivery networks (IDNs) are largely at Stage 2 (“Localized Analytics”). Because of their increased focus on financial and operational performance, analytics are used primarily on the business side in this type of provider. However, the rise of EMRs and pressure for cost-effective treatments will create a more balanced clinical/business analytics environment over the next several years.

Academic medical centers are also typically at Stage 2, although their analytical focus is more likely to be on the clinical side. By virtue of their missions they are more likely to do clinical research, comparative effectiveness studies, clinical decision support implementations, and so forth. Many have chief medical information officers or medical directors with a focus on clinical decision support. Some may have a degree of descriptive (and some predictive) analytics on the business side as well. However, the analytics remain largely localized, and for the great majority have not been viewed as an enterprise resource.

Only a few healthcare organizations are at Stage 3, “Analytical Aspirations.” Those that qualify are primarily integrated care organizations (Kaiser Permanente, the Mayo and Cleveland Clinics) and a few academic medical centers (Intermountain, Partners, Johns Hopkins). These organizations are still largely localized in their analytical focus, but they have several distinguishing characteristics:

• Their localized analytics groups are beginning to communicate and collaborate.

• They have high degrees of analytical activity and personnel (particularly on the clinical side), some of which are world-class.

• Their senior management teams recognize that analytical capabilities are key to their success.

However, because of organizational boundaries, resource limitations, disparate technologies, and other constraints, they cannot yet be classified as “analytical organizations” or “analytical competitors” compared to standard-bearers for those

7/7/21, 9:13 PMPrint

Page 9 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

titles in other industries.

Analytical Applications in Healthcare Providers In this section we describe the key analytical applications used in provider organizations and classify them as to whether they involve descriptive or predictive/prescriptive analytics (see Table 2.1 ( ). The latter types use past data to construct predictive models (predictive analytics), or use past data to suggest action (prescriptive analytics), such as the most effective treatment to employ, or the optimum level of staffing to use. Where appropriate, we describe each application as to how granular it typically is in terms of data. As more granular levels of analysis are possible—for example, moving from institutional data to service line data to procedure data—analytics are easier to translate into action. Highly granular data may also help lead to behavior change; for example, reporting certain data at the individual clinician level tends to bring out a sense of competition and a desire to excel against peers.

Table 2.1 Key Analytical Applications in Provider Organizations

Descriptive Analytics Predictive Analytics Financial/Operational

Occupancy analytics—Models reporting on or (much less common) forecasting and predicting bed occupancy, either across the institution or within particular service lines

Revenue cycle analytics—Reporting on billing, collections, and accounts receivable, most frequently at the institutional level, but occasionally at the payer or service line level

Quality and compliance analytics— Reporting models assessing compliance with regulations and payer requirements, usually at the institution level

Cost management—Reporting models of staff productivity and cost, typically by service line, and comparisons to industry benchmarks

Operational benchmarks—Reporting on basic measures of operational performance, such as length of stay, percentage discharged by particular time—usually at institutional level, but increasingly at service line and even

Population analytics—Predictive models (usually provided by third parties, and often based on Medicare claims data) of population demand in a particular service area used for physician and facility planning, either for overall population health or by disease/service line

Financial risk—Predictive models of patients’ likelihood of self-payment or reimbursement, usually at admission; granular to the patient level

Supply chain analytics—Reporting on, and in some cases, optimizing levels of inventory for hospital supplies—sometimes in partnership with an external supply company or group purchasing organization (GPO)

7/7/21, 9:13 PMPrint

Page 10 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

individual physician level

Clinical Clinical problem tracking—Reporting and tracing particular clinical problems, such as post-operative infections, which caregivers touched the patient, which other patients were touched by those caregivers, and so on

Patient safety analytics—Reporting on common safety problems, such as central line infections, generally at entire hospital level

Prediction of likely treatment results—Based on attributes of the patient’s biodata, symptoms, and metabolic data, analytical predictions of the likelihood of acquiring certain diseases and of successful treatment can already be made, although they primarily are done only in experimental mode by academic medical centers at this time

Scoring of patients at risk—Using analytics to identify predictive variables for risk of particular problems (e.g., cardiovascular failure), and then scoring patients to identify those most at risk; presently primarily in pilot mode

Care protocol analytics—Controlled research studies determining which treatment protocols are most effective; usually only in academic medical centers

Evidence-based medicine analytics—Translation of clinical research results (at the particular hospital or elsewhere) into order sets and rules for provider order entry (CPOE) systems; may be incorporated into EMR and CPOE systems from external vendors

Personalized genetic medicine—Predictions of medication choice and dosing based on genetic profile, presently in pilot mode at some academic medical centers

The Future of Analytics in Provider Organizations As we have suggested, the future of analytics in healthcare provider organizations is much brighter than in the past. A number of factors will drive considerably greater activity in the future. Several are described in this section.

The “meaningful use” criteria for EMR reimbursement are clearly going to lead to more analytical activity.1 ( This is one of the first times in any industry that reimbursement for a system demands actually using the data contained in the system! The criterion involving clinical decision support rules, for example, presumes some analytical underpinnings. Others involve specific reports, such as, “Generate lists of patients by specific conditions to use for quality improvement, reduction of disparities, research, and outreach.” While most of the 25 Stage 1 meaningful use criteria are more transactional in nature than analytical, future meaningful use criteria are likely to be more analytically focused.

The move to accountable care organizations (ACOs) will clearly drive providers toward greater reliance on descriptive

7/7/21, 9:13 PMPrint

Page 11 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

analytics. Since reimbursement for physicians and hospitals will depend on their ability to control costs and meet quality-of-care goals, metrics of those factors will be highly prized. It is likely that providers will move toward highly granular reports—down to the individual clinician level—to motivate the desired (and reimbursed) behaviors. ACOs will need to blend financial, operational, and clinical analytics in a way that they have rarely been integrated in the past.

Analytics will also be required to address growing complexity in the clinical domain. There is arguably already too much medical knowledge for humans to hold in their brains, so clinical decision support based on analytics will be necessary to deliver the best care. Personalized genetic medicine presents an even more daunting amount of content to master. Providers will need to transform the roles of their clinicians toward greater use of analytics and decision support, as opposed to relying primarily on experience and intuition.

Analytics will also be advanced through increased regulatory requirements for reporting and transparency. There are already substantial institutional compliance issues from organizations like Centers for Medicare and Medicaid Services (CMS), Food and Drug Administration (FDA), and so on. As more data at the patient and clinician levels become available, CMS and other regulatory bodies will undoubtedly require more data. Caregivers at every level will be assessed on whether they followed a prescribed protocol, whether they have documentation for protocol deviations, and so forth.

Patient information privacy issues won’t necessarily advance analytics, but they will certainly influence them. This has not yet been an issue at the aggregated data level, but with more atomic data, privacy issues will be much more difficult to address. It will also make outsourcing of analytical services more challenging. Questions about the use of patient data for analytical purposes abound. For example, if an institution sends out a patient profile for cancer screening, can the organization receiving it guarantee that it will remain private? Will patients have to opt in to sharing data for analytical purposes, even when it has been anonymized? Sharing data across payer and provider networks may be impossible without the patient’s permission to use the data.2 (

In short, the future analytical world in which healthcare providers will live will be very different from that of today. Many of the analytical initiatives and programs started in the current environment will have to be exploratory, but they will undoubtedly confer both experience and advantage on early adopters. Uncertainty is no excuse for inaction.

Notes 1 ( . For a concise list of meaningful

use objectives, see the Healthcare IT News list at meaningful-use-criteria ( .

2 ( . For additional reading, see “Security, Privacy and Risk Analytics in Healthcare” published in the HLSARC compendium.

7/7/21, 9:13 PMPrint

Page 12 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

3 An Overview of Analytics in Healthcare Payers

Kyle Cheek

For every complex problem there is a solution that is simple, neat, and wrong.

—H. L. Mencken

Quantitative analytics have long been utilized by healthcare payer organizations, tracing back to the actuarial origins on which the payer industry is based. In spite of its quantitative origins, though, the deployment of analytics as a competitive driver beyond the traditional actuarial pricing function has been slow and uneven, even as healthcare payers have evolved from their traditional indemnity role to one of trusted health advocate—a transition that is inherently information-centric and demands a greater reliance on deeper analytics. Payers do generally recognize that there is latent value in the large volumes of data they accumulate through healthcare transaction processing, especially in the mountains of healthcare claims data that contain rich financial and clinical information and, when properly aggregated, provide rich longitudinal views of patient experiences.

Historically, though, efforts to exploit that value have largely disappointed, either because of a failure to employ the deep analytics that drive competitive advantage or because deep analytics have been adopted without being fit holistically into the program they support. The failure to employ deep analytics is seen in the ubiquitous reporting applications—from employer group reporting applications to executive dashboards—that are driven by simple summary views of the underlying information. That focus on reporting applications and other less sophisticated uses of payer data resources has tended to be accompanied by a concomitant focus on technology as the delivery mechanism, rather than the sophistication of the actual analytics delivered. Similarly, the adoption of more sophisticated analytics in the payer space has often occurred within programs that are not properly designed to avail themselves of deep quantitative results or, conversely, that produce value across the longer time horizon necessary to realize value from efforts to effect changes in patient behavior, but concurrently making efforts to measure that value more difficult.

While the adoption of analytics as a competitive catalyst has met with mixed success industry-wide, there do exist payer organizations that have made a principled commitment to exploit analytically derived value and stand above the industry norm. These organizations typically have fact-driven leadership and a comprehensive strategy for analytics as a value driver, and a commitment to the infrastructure necessary to support an enterprise analytical vision. The successes these organizations have experienced and the means by which they have realized those successes serve as a bellwether for the industry to follow toward more effective adoption and deployment of analytics as a competitive driver and as a real source of change in patient outcomes and population wellness.

Payer Analytics: Current State

Adoption That healthcare payers have struggled to realize significant value from analytics does not mean that there have not been important forays into analytics within the payer space. Rather, payers have made a concerted effort to leverage their formidable data assets in several distinct areas. Examples of those solutions by analytical subject include

• Patient analytics—Many payers have adopted solutions that are focused on the analysis of patient outcomes as a means of predicting and reducing avoidable costs. These disease management (DM) tools typically focus on identifying those patients at greatest risk of poor (and costly) health outcomes and prioritizing clinical

7/7/21, 9:13 PMPrint

Page 13 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

intervention to those whose predicted costs are most actionable (often by a nurse-administered telephone outreach call or by direct mail).

• Provider analytics—Another class of analytic solutions employed by payers to promote improved outcomes and cost containment includes evidence-based medicine (EBM) and pay for performance (P4P). These analytically based programs focus on provider efficacy and rely on empirical analytics to determine which treatments provide superior outcomes at lower cost. In turn, providers are encouraged to employ more cost- effective treatments, with reimbursement levels depending on the extent to which a provider complies with EBM standards. Similarly, provider transparency programs promote informed patient choice by helping patient populations assess which providers are most compliant with EBM standards and whose patients experience the best health outcomes. All of these analyses require the extraction of deeper insights from the data than simple reporting or basic analyses can provide.

• Customer analytics—Employer group reporting and directed marketing are examples of payer efforts to utilize information assets to better allow the purchaser of benefits to manage costs. In addition, some payers have even adopted analytics heavily utilized in direct marketing to address attrition and churn, especially within the individual products segment.

• Financial/operations analytics—The final class of analytics adopted by payers is largely inward-facing, and comprises that class of analytics commonly referred to as “business analytics.” Solutions widely adopted by payers within this category include financial forecasting, actuarial rating, operations monitoring, and fraud detection.

As the continued proliferation of the preceding examples illustrates, there is awareness among payers of the potential value of analytics. The maturity ascribed to the payer segment must be tempered, though, by a consideration of the effectiveness of these analytic endeavors as a competitive differentiator. Put differently, the general trend among payers to champion analytical solutions has not translated into a corresponding record of clear added value. In one conspicuous example, payers have widely adopted disease management (DM) solutions that rely on analytics to identify patient segments likely to incur significant future costs. Those patients identified by the analyses are then targeted for some sort of intervention—ranging from preventive care service reminders distributed by mail to telephone engagement by clinicians to coach behaviors that promote wellness—designed to reduce overall patient costs. However, in spite of wide adoption and a highly competitive vendor space, there is a body of literature from both public1 ( and private payers that maintains there is little empirical evidence that these programs deliver a significant return. Instead, it is argued that payers view these programs as “table stakes”—a cost that must be borne to compete for certain attractive customer segments (typically large group customers).

Another example of widespread analytics adoption is the trend among payers to leverage analytics to combat the problem of fraudulent claims and billing schemes. Solutions that leverage analytics (of varying levels of sophistication) to identify fraudulent claims, providers, and patients have become commonplace in payers’ Special Investigation Units. Despite the availability (and large-scale adoption) of tools designed to support the detection of fraud, most authorities still estimate losses to fraud, waste, and abuse at greater than 10% of claims costs. An accepted rule of thumb is that only 10% of those losses are detected through current fraud detection capabilities, and only 10% of those detected dollars are recoverable, suggesting that only about 1% of the total fraud problem is addressed through current antifraud programs and their supporting analytics.

These examples illustrate what might be described as an analytical sycophancy across much of the payer space—a clear recognition that the market demands analytical value and even a demonstrated willingness to adopt, but a marked failure to deploy analytics to meaningfully drive value. Specific to the previous examples, the failure to realize value

7/7/21, 9:13 PMPrint

Page 14 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

from disease management programs is often attributed not to shoddy analytics, but rather a failure to build programs that effectively intervene with those patients identified by analytics. In the case of fraudulent claims, the failure to significantly reduce losses was once regarded as a function of inferior analytics, but now is seen more as a product of the deployment of fraud analytics retrospectively—after payments have already been made. In short, any evaluation of the analytical maturity of the payer space should recognize payers’ efforts to adopt value-enhancing analytics. However, it is also important to recognize that many such efforts have failed to generate clear-cut successes. In turn, the driver behind the lack of clear analytical success appears to be attributable less to the analytics themselves and more to the efficacy of the programs and processes that are designed to act on analytical results.

Maturity Model The Davenport-Harris Analytics Maturity Model articulates five levels of analytical competency to describe organizations’ capabilities in this domain. From highest to lowest, the levels of Analytical Capability are

• Stage 5—Analytical Competitors

• Stage 4—Analytical Companies

• Stage 3—Analytical Aspirations

• Stage 2—Localized Analytics

• Stage 1—Analytically Impaired

Most payer organizations are best classified at Stage 2 on this model. They have some localized capabilities, as described previously, but have not yet articulated analytical aspirations in a holistic strategy. They probably also have not developed data warehouses or a common analytical toolset at the enterprise level.

Impediments to Maturity Health payers offer an interesting case study of the evolution of an industry and the capability of its constituent members to adapt to changing competitive dynamics. Considering the evolution of the healthcare industry, and especially the central role played by payer organizations in that evolution, it is not surprising that payers have experienced an uneven transition toward analytical competency. Payers historically operated as transaction-oriented companies, focused on the timely processing of claim, membership, and network transactions. It is of recent vintage that payers have assumed the more analytically-dependent and information-centric role of trusted advisers for the health and wellness of their policyholders. Delivering that evolving focus on information-centric business processes is complicated, though, by the decades of infrastructure that have been built to facilitate the timely processing of transactions.

The vast amounts of data created and warehoused by payers—and upon which any analytical evolution relies—are also a direct artifact of payers’ transaction-centric history. As a result, the data that are most readily available to support forays into outcome and wellness analytics (disease management, provider transparency, etc.) are in fact naturally suited for financial analysis rather than for clinical analysis. The manipulation then required to properly condition claims transaction data for use in clinically-oriented studies is not insignificant and is a component of analytical infrastructure that can be prohibitively resource-intensive for smaller payers that cannot make large investments in their data infrastructure.

Despite the seemingly large amount of transaction data available at an individual payer organization, the amount of data may be too small or too narrowly focused to make analytical generalizations to national populations. This problem

7/7/21, 9:13 PMPrint

Page 15 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

is being addressed by the rise of national-scope insurers and databases (e.g., United Healthcare and Ingenix), and collaborations among regional or state-based payers (e.g., Blue Health Intelligence). However, these large-scale data aggregations have only recently been compiled, and most of the analytics on them have yet to be performed.

Conclusion: Paths to Maturity The healthcare payer industry is perhaps best conceived as an earnest occupant of Stage 2 status in the analytical maturity model referenced previously—solidly committed to localized analytics and beginning to develop sincere analytical aspirations. Fortunately, while the payer space as a whole is only beginning to formulate analytical aspirations, there do exist leaders in the space that point to different pathways to maturity for other payers. Among the leading organizations are

• Aetna—Aetna is a true analytic competitor, with a rich tradition of fact-based leadership and deep analytics embedded across business processes and products. The Aetna model is best thought of as bottom-up, tracing its origins to localized analytics that grew organically over time to eventually support a wide array of business operations.

• Humana—Whereas Aetna’s analytics evolution was bottom-up, Humana’s commitment to analytics has been top-down. Humana’s experience reflects a concerted effort by its leadership to commit to analytics as a value driver. Its analytics organization was created specifically to nurture an enterprise analytics competency that was not previously recognized within the organization.

• UnitedHealthcare—United, primarily through its Ingenix subsidiary, has both acquired and developed an impressive array of analytical capabilities. It is one of the few payer organizations with analytical groups serving not only its own needs, but those of other payers and providers as well.

While it requires a major strategic commitment to resources, process, and often even culture change for a payer organization to embark on a path toward analytical excellence, the prospects are good for this segment of the healthcare industry to succeed in this endeavor over time. Significant potential value opportunities have already been identified, for example, in the disease management and fraud detection spaces. The regulatory environment is also supportive of the necessary infrastructure investments and, in general, of a more effective use of analytics to promote better outcomes and efficiencies.

Note 1 ( . Nancy McCall, Jerry Cromwell,

Shulamit Bernard, “Evaluation of Phase 1 of Medicare Health Support Pilot Program Under Traditional Fee-for- Service Medicare,” Report to Congress by RTI International, June 2007. Online at ( .

7/7/21, 9:13 PMPrint

Page 16 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

4 Surveying the Analytical Landscape in Life Sciences Organizations

Dave Handelsman

The life sciences industries are collectively engaged in discovering, developing, and commercializing new therapies. Historically, this market segment has included just the pharmaceutical industry when, in actuality, life sciences organizations encompass pharmaceutical, biotechnology, medical device and vaccine developers, as well as supporting organizations such as contract research (service) organizations and the federal government. Each of these organizations is engaged at overlapping, and multiple, points along the spectrum of bringing new therapies to market, and this chapter puts the analytics landscape in the context of the overall life sciences ecosystem without straying into the nuances of each organizational segment.

For many years, the life sciences industries were viewed as both socially-conscious enterprises and financial powerhouses. These companies developed therapies to treat illness, while at the same time rewarding their shareholders. In recent years, however, this perception is changing due to safety concerns regarding approved therapies, diminishing research pipelines, patent expirations, and a dwindling number of new therapy approvals each year, all of which is coupled with an estimated cost of $1.2B to bring a new therapy to market.1 ( Because of these changes, it is no longer “business as usual” in the life sciences industries. This changing business environment, coupled with the added challenges associated with healthcare reform, is forcing life sciences organizations to revisit their decision-making processes. This has led to an increased focus on analytics across the different functions of life sciences firms, although as yet there has been little collaboration or coordination across the functional groups. The most significant analytical functions are reviewed in this chapter.

Discovery Perhaps no place else in the life sciences industries have analytics been more aggressively pursued than in the discovery process. When the human genome was mapped in 2000, itself a sophisticated analytics project, there was widespread belief that this event would usher in a new collection of cures and treatments. Instead, it ushered in an ever- increasing level of sophisticated analytics, as each new layer of knowledge regarding the fabric of life was studied in complex detail. Microarrays, proteomics, Single-Nucleotide Polymorphisms (SNPs), copy number, and other emerging sciences were studied in detail in an attempt to identify patterns in the spectrum of data that would serve to identify causes or predictors of disease and, ultimately, complementary treatments to address those diseases.

The intensity of analytics associated with this exciting field of research has resulted in not only dedicated analytical software solutions targeted at this market segment, but bespoke hardware strategies to provide the high performance computing required to complete these assessments on a more industrialized basis. The Archon X prize ( ( ), like other similar groundbreaking awards, will be awarded to the company that can sequence 100 human genomes, in ten days, at a cost of no more than $10,000 per genome. Parallel efforts are under way to develop advanced analytics to support the $1,000 genome.

While great progress has been made in the genomic and associated sciences, it has had limited impact on real-world issues facing the life sciences industries today. Arguably the most success can be seen in the commercialization of Herceptin by Genentech. This treatment, perhaps at the forefront of personalized medicine, has shown great success in treating a specific form of cancer characterized by the HER2 receptor. This type of analytics-driven result, where a specific diagnostic is paired with a targeted therapy, holds great potential, but has not (yet) been frequently replicated or

7/7/21, 9:13 PMPrint

Page 17 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

industrialized despite the sophisticated analytics routinely being applied to the fields of discovery sciences.

Development For therapies that progress beyond discovery, the development process has routinely employed analytics to determine the safety and efficacy associated with a particular clinical (or preclinical) trial, or an aggregation of the trials that point to the overall safety and efficacy of a new treatment. While the vast majority of the analytics applied in documenting safety and efficacy are descriptive in nature, inferential statistics are used to draw statistically valid conclusions from the available clinical trial data.

The application of these descriptive and inferential analytical approaches is mature and well-understood by both life sciences companies and the various regulatory agencies. The acceptance of this analytical approach by both the manufacturer and the approving agency streamlines the approval process and further supports a mature review process. Additional analytical approaches, such as Bayesian methodology, are emerging as alternative analytical tools in assessing clinical trial outcomes, but these approaches have not yet achieved a similar level of acceptance. In February 2010, the Food and Drug Administration (FDA) issued the Guidance for the Use of Bayesian Statistics in Medical Device Clinical Trials, which should further encourage using Bayesian methodology by life sciences organizations. Bayesian approaches, however, are not without controversy, and life sciences companies need to recognize that not all regulatory reviewers are aligned with this emerging analytics approach.

Despite the maturity associated with applying analytics to determine the outcome of a research trial or program, analytics have rarely been applied to other aspects of clinical trial operations. As shown in Figure 4.1 ( , analytics has a strong role to play in the planning and execution of clinical trials. However, this role has typically been limited to less valuable parts of the decision-making process where historical data are summarized (standard reports, query drilldown, etc.), but not used to drive more valuable, optimized, decision-making objectives. Some analysts have suggested that the absence of clinical trial optimization capabilities, either in human experts or analytical software, has been one factor explaining the relatively limited success that many pharmaceutical firms have had in bringing new therapies to market.

7/7/21, 9:13 PMPrint

Page 18 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

Figure 4.1 Analytical maturity model for clinical trials

When life sciences companies can implement optimized trial designs, as an example, this has a cascading effect throughout the organization. Fewer trials are required to get a therapy approved, fewer resources are required for those trials that are conducted, and the overall risk and expense associated with bringing a new therapy to market are reduced.

Manufacturing Pharmaceutical manufacturing has been a headline-grabbing issue in recent years as the industry addresses significant product recalls due to manufacturing issues and widespread drug shortages. The problem is sufficiently important that a Drug Shortages Summit was held in late 2010 to address this emerging issue.

These shortages are being caused by a series of related issues including manufacturing difficulties, unexpected interruptions in the supply chain, increased demand, and both voluntary and involuntary recalls. Additionally, some manufacturers have determined that it is simply no longer profitable to produce certain drugs based upon direct costs, revenue opportunity, and risk of litigation.

It is interesting to note that these manufacturing problems are occurring despite the broad use of manufacturing-based analytics for many years. Most notably, Six Sigma methodology has been implemented in the life sciences as one key way to reduce defects in the manufacturing process. Lean manufacturing is another analytics-based discipline focused on reducing waste in the manufacturing process. Hybrid strategies, such as Lean Sigma, focus on reducing both defects and waste, and enable manufacturers to achieve and maintain a high degree of quality in their manufacturing process.

7/7/21, 9:13 PMPrint

Page 19 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

More recently, Quality by Design (QbD) has taken an increasingly important role in life sciences manufacturing. As its name implies, quality is designed into the manufacturing process from the beginning with the goal of ensuring that the issues associated with product and process are carefully considered to ensure quality in the end product.

In May 2007, the FDA issued a report titled Pharmaceutical Quality for the 21st Century: A Risk-Based Approach Progress Report. This report specifically identifies QbD as a key strategy for ensuring quality in manufacturing and has shaped the Agency’s approach to not only monitoring manufacturing activities but also in paving the way for manufacturers to submit information based on QbD principles. In this same report, the agency outlined the benefit that Process Analytical Technology (PAT) brings to further understand and control the manufacturing process, where “analytical in PAT is viewed broadly to include chemical, physical, microbiological, mathematical, and risk analysis conducted in an integrated manner.”2 (

As the FDA looks to assess and understand why manufacturing quality continues to be a problem despite the widespread use of analytics, new processes and strategies need to be put in place at the federal level. The number of product recalls by the FDA Center for Drug Evaluation and Research (CDER) continues to grow, and increased by more than 50% in 2009.3 ( While it would seem logical for the FDA to increase the number of inspections to identify manufacturing issues more effectively, such an increase is unlikely given the increased costs and the difficult economic climate. The FDA, instead, is turning to a more targeted risk-based approach to identify those manufacturing sites most likely to be at risk for problems.

Sales and Marketing The sweeping changes affecting the life sciences industries extend far and deep in sales and marketing organizations. Manufacturers are dramatically restructuring their sales teams to address the financial challenges facing the industry, and are no longer deploying a commercialization strategy that relies upon ever-growing armies of sales representatives to sell their products. Instead, life sciences companies are dramatically reducing the size of their sales teams, while at the same time determining how best to “do more with less” in an ongoing struggle to retain market share and grow revenue, despite the changing marketplace.

These changes include the often-mentioned patent cliff, diminishing numbers of new products to sell, and the enormous changes expected to be realized through healthcare reform. Analytics, especially in light of the diminishing number of “feet on the street,” will grow in importance for manufacturers. Historically, analytics has meant little more than basic query, rank, and report, where sales analytics were primarily historic reports of past effectiveness, usually sourced from third-party data sources such as IMS Health.

Increasingly, manufacturers are deploying predictive analytics to drive (and, in some cases, recapture) revenue. These predictive analytics span a broad range of the commercial segments, covering such diverse areas as physician targeting, marketing mix, managed care, and rebate optimization. In many ways, the life sciences industries are being forced to adopt mature strategies used in other industries that have not had the luxury of year-over-year revenue growth that has supported their inefficient and expensive sales and marketing practices. Unlike those industries, however, commercial life sciences approaches are more complex, and must carefully consider the world of formularies, reimbursement strategies, and government regulations associated with marketing activities.

In the case of physician targeting, for example, legions of sales representatives would typically be assigned to visit prescribers in their territory that met only the most basic of prescription profiling behaviors—typically those that ranked at some basic level of prescriptions. Instead, predictive physician targeting identifies those physicians that are low-level prescribers but have the potential to not only grow but become high-prescribers with the right level of

7/7/21, 9:13 PMPrint

Page 20 of 21…nt=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

education from the biopharmaceutical sales team. While the existing high-prescribers can’t be ignored, their opportunity for growth is limited, and an intelligent targeting approach can drive increased revenue.

Similarly, new analytical approaches are being applied to more intelligently deploy marketing resources. A wealth of options is available to life sciences manufacturers, and the right mix of media—online, print, radio, and so on—is critical for informing not only consumers, but also physicians and other medical personnel. There is no doubt that direct-to-consumer advertising has an extraordinary impact on driving prescriptions, and it remains increasingly important to be able to monetize marketing investments as the industry gets progressively more competitive.

Unlike other industries, the mechanics of payer reimbursements have a significant effect on purchasing behavior. Life sciences manufacturers frequently structure rebate relationships with providers that share the joint goal of increasing the manufacturers’ revenue while reducing the payers’ expenses. These relationships are complex and must take into account Medicaid pricing constraints, perceived pricing differences between competitive drugs, and profitability goals. The availability of historic data can be used to construct analytical models that optimize the rebate scenarios for the manufacturer, while taking into consideration other various constraints that must be observed.

Health Reform The emergence of health reform will weigh heavily on how best to apply new forms of analytics to the business of life sciences development and commercialization. With reform comes new reimbursement strategies, evolving relationships with payers and providers, and the emerging need to more carefully and thoroughly understand the “comparative effectiveness” of new and existing therapies. “Me-too” drugs that provide little-to-no additional therapeutic benefit when compared to older, lower-priced, drugs will present new challenges in terms of driving revenue. Development plans for such drugs are most likely to be deprioritized in favor of differentiated therapies. Such differentiation may be addressed through targeted patient populations, or through renewed focus on novel therapies more likely to both improve patient health as well as the bottom line. In these cases, advanced analytics becomes even more critical, as choices regarding which therapies to pursue and ultimately commercialize will have critical implications regarding the financial viability of life sciences companies. Healthcare reform may also mean that the diverse and siloed analytics groups within life sciences firms will need to collaborate to a greater degree.

Conclusion Analytics, which have always had a position in the world of the life sciences, are at the forefront of enabling life sciences companies to take their organizations to the next level in today’s economy. The issues faced today span discovery through commercialization, and the larger health-care community is struggling to rein in costs while still improving health. Critical scientific and business decisions can only be confidently made by leveraging historical and current data, and bringing advanced analytics to bear on the complex problems facing the life sciences industry today.

Notes 1 ( . Tufts Center for the Study of

Drug Development ( ( ).

2 ( . (

3 ( . Parija Kavilanz, “Drug Recalls

7/7/21, 9:13 PMPrint

Page 21 of 21…ent=all&clientToken=e4d29f23-e1b8-4f9d-522a-4c850d6e4f73&np=part02

Surge,”, August 16, 2010, ( .