As part of its value-based care initiative, the College of American Pathologists has pursued research to better understand the role pathologists can have in population health.
To answer the following questions: (1) what is the impact of population health and population health management on pathologists; (2) what roles are pathologists playing in population health management; (3) is population health something that pathologists in both larger and smaller settings can engage in; (4) are pathologists in a position to analyze laboratory data for population health, and, if so, what are the key information sources those pathologists must access; and (5) what steps can a pathologist take to become involved in population health?
We conducted 10 semistructured interviews with pathologists and other medical laboratory leaders who have been active in population health. These interviews were supplemented with a review of the medical literature.
Pathologists have demonstrated that laboratory data can provide unique value-added contributions to improving the health of populations. These contributions are not limited to pathologists in large, integrated settings. However, pathologists need to be proactive to contribute to health systems' population health efforts and may need to both enhance their own skills and the quality of their data to maximize the value of their contributions.
Although not necessarily a definitive summary of the roles that pathologists are playing in population health, this article identifies some of the promising and innovative activities occurring among pathologists and laboratorians.
Changes in health care delivery and payment are altering the economic model for recognizing and rewarding the value of pathology practice and the clinical laboratory. Regardless of whether pathologists are in provider-owned practices or are employed by hospitals or other entities, practices are experiencing a shift toward ensuring payment for quality of health care outcomes rather than volume. The focus of both public and private payers is shifting emphasis from fee-for-service to pay-for-performance. Compounding these pressures are regulatory changes to Medicare, such as the scheduled reductions in payments for laboratory tests resulting from the Centers for Medicare and Medicaid Services (CMS) implementation of the Protecting Access to Medicare Act of 2014, revaluation of Medicare payment rates for physician care, and private sector efforts to reduce laboratory costs.1–3
Along with those challenges, however, come opportunities for pathologists. The changing focus of hospitals and health plans to value-based care includes requirements for health care providers to demonstrate where and how they are adding value, both to the health of individuals and to the health of populations served. There is an increased emphasis on preventive medicine, closing gaps in care, identifying and reducing health risks, and a better understanding of the drivers of health care resource consumption. With laboratory tests accounting for much of the structured data in a patient's medical record, which ultimately drives patient care decisions, pathologists have an opportunity to define new roles for themselves and for laboratory medicine in meeting the changing goals of health plans and systems.
The College of American Pathologists (CAP) Board of Governors recognized those challenges when, in 2015, it formally adopted a 5-pronged approach for implementing its value-based care strategy and further refined that strategy in 2017. That approach included (1) developing tools, training, and education to build pathologist and practice awareness of value-based care business models, such as Medicare's Merit-based Incentive Payment System, advanced alternative payment models, accountable care organizations (ACOs), and bundled payment arrangements; (2) working to ensure that pathologists are fairly compensated in new value-based care models implemented by public and private payers; (3) developing tools, such as a Qualified Clinical Data Registry, to enable pathologists and their practices to demonstrate their value to public and private payers through measurement and documentation of quality-of-care metrics, resource use metrics, clinical practice improvement activities, and use of health information technology; (4) enhancing patient, payer, health care organization, and health community awareness of the pathologist's value in emerging value-based care models; and (5) gaining a better understanding of population health and helping to define the roles of pathologists in this area.
This research is designed to address the fifth prong—gaining a better understanding of population health and helping to define the role of pathologists in this area. In pursuing that line of research, we recognize that “population health” is one part of the broader category of “value-based care” and that both have similar overall goals—to lower the cost of care per episode and overall, to obtain better outcomes for given populations, closing gaps in care between different types of individuals, improving transitions in care, and meeting quality metrics (eg, Healthcare Effectiveness Data and Information Set [HEDIS], CMS Star Ratings) which can determine payment and signify quality of care to patients. Specifically, our research is designed to answer the following questions:
How should we define population health and/or population health management?
What is the impact of population health and population health management on pathologists?
What roles are pathologists playing in population health management today?
Is population health something that pathologists in both large settings and smaller community-based settings can engage in?
Are pathologists in a position to analyze laboratory data for population health, and, if so, what are the key information sources those pathologists must access to do so?
What steps can a pathologist take to become involved in population health?
Our research relies mostly on interviews with 10 pathologists and other medical laboratory leaders who have been active in some aspect of population health. We supplemented that research with a review of the limited amount of literature around activities by pathologists in population health. Our research was overseen by a workgroup of pathologists who are recognized leaders in expanding the role of the medical laboratory in population health management in their own institutions and communities.
POPULATION HEALTH AND PATHOLOGY
Population health is an emerging area about which there are strong opinions but no universally accepted definition. It is often closely coupled with the term population health management. However, many accept the definition of population health (including CMS) developed by Kindig and Stoddart4 in 2003 as “the health outcomes of a group of individuals, including the distribution of such outcomes within the group.”(p380) A 2015 survey conducted by George Washington University's Milken Institute School of Public Health (Washington, DC) asked more than 100 health care leaders to define population health. There was little consensus among the 37 responses they received, with only 2 of the 37 respondents (5%) citing the Kindig and Stoddart definition.5
Part of the problem in defining population health is that there are different perspectives in how to define the “population” in population health. That problem was explicitly recognized by the Roundtable on Population Health Improvement developed by the National Academies of Sciences, Engineering, and Medicine. Although they adopted the Kindig and Stoddart definition as their working definition of population health, they explicitly noted that population health outcomes are the product of both clinical and nonclinical factors (eg, medical care, public health, genetics, behaviors, social factors, and environmental factors), and cite research studies that define population health mostly from the perspective of affecting “population medicine” as well as that which focuses on the overall health of a population in a given geographic region.6 Kindig7 proposed in 2015 that the term population health be used when referring to traditional public health activities that are under the jurisdiction of public health officials, business leaders, and community organizations, and that interventions and policies applied to specific patient populations should be referred to as population health management or population medicine. By contrast, Jacobson and Teutsch8 suggested that the term population health should be abandoned, and that the term total population health be used to refer to the overall health of a geographic population.
For the purposes of this analysis, we define population health as “The overall health outcomes of a defined group of individuals, based on the preservation of health, prevention of disease, and management of acute and chronic disease among members of the group.” We define population health management as “an approach to health care delivery that aims to improve the overall health outcomes of a defined population of individuals, based on the preservation of health, prevention of disease, and management of acute and chronic disease among members of the group. It typically relies on the aggregation of patient data across multiple health information resources and analysis of data and actions through which care providers can improve both clinical and financial outcomes.”
Regardless of the definition, population health is becoming a widely used term that encompasses pathology and laboratory medicine. For example, in the last 9 months, CAP Today featured 2 articles on population health activities by pathologists.9,10 The 2017 Executive War College (May 2–3, 2017; New Orleans, Louisiana) featured several sessions on population health and the development of a new business model, known as Clinical Lab 2.0 (Project Santa Fe, Albuquerque, New Mexico), for the clinical laboratory. Population health was the theme of the 50th Anniversary Annual Meeting of the Association of Pathology Chairs (July 25–28, 2017).
In all of these discussions, pathologists and laboratorians who are working on population health note the importance of population health to the changing priorities in health care, particularly among providers and health care institutions. They note the increasing trends in payment methodologies, which are shifting financial risk to providers through incentives, payment for quality, or penalties based on performance compared with established benchmarks. It is in this context that an empirical meaning of population health arises, comprising those measures that arise from and are applied to patients other than on an individual basis, which enable the providers of health care services to meet the societally endorsed “triple aim” in health care.11
This analysis is not intended to be a definitive summary of the roles that pathologists are playing, or can play, in population health (as variously defined). Rather, it represents a sampling of some—but not all—of the promising and innovative activities occurring among pathologists and laboratorians. If anything, the analysis shows that population health is an evolving area, with practical implementation still early in its development. There is no clear single road map for engaging in population health and there is an absence of accepted roles for pathologists in population health activities. However, the examples provided in this article do offer several perspectives on the opportunities pathologists see in population health. Interestingly, opportunities exist irrespective of whether the pathologist practices in a large, integrated system or in a smaller practice that serves 1 or 2 community hospitals. Consistent with the triple aim, these opportunities have in common a concern across all health care providers about how to reduce costs, increase efficiency, and reduce risk to patients. This article shows different ways that some pathologists have approached these issues.
THEMES
Changes in Health Care Finance and Delivery Are Increasing the Importance of Data Analytics but Do Not Always Take Full Advantage of Clinical Laboratory Data
The ongoing shift to what is known as value-based health care means that payment for health care services—both inpatient care and ambulatory care services—is increasingly tied to patient outcomes and quality of care and that health plans and providers will be asked to bear more and more of the financial risk of caring for populations. Whether payment for health care services is provided through a bundled payment arrangement, capitation payments, or other risk-sharing arrangements, plans and providers are likely to see that some part of the reimbursement will be tied to outcomes and acceptance of some level of financial risk.12,13
This changing environment increases the importance for health plans, institutions, and providers to understand their patient population so that they can understand the risks they face, find ways to mitigate risks (eg, through better patient care), and have available the information necessary to qualify for payment incentives associated with meeting quality standards. The most effective way of obtaining the information needed for this is through analysis of digitalized data on large populations served by the plans, hospitals, and providers. Moreover, just as payment systems are changing, the types of data available are changing as well. Historically, hospitals, health systems, and providers have relied on actuarial data (billing records and/or claims data) as a primary source of information for analyzing patient outcomes and projecting costs and risk. However, as electronic health records (EHRs) and interconnectivity continue to mature, the availability of patient information continues to expand, allowing for improved analysis of patient-related outcomes analyses, risk assessment, and cost and use drivers.14
Actuarial data are relatively limited in their predictive and explanatory power.15 Perhaps most important, these data are retrospective and sometimes have substantial lag times before they become available, and thus do not provide opportunities for “real-time” analysis. Some experts have suggested that health care enterprises can benefit substantially by improving their use of real-time data, such as that associated with clinical laboratory data. Among the ways that real-time clinical data provide substantial benefits over lagged claims data include the following:
Earlier identification of patients with new diagnosis and/or change in health status.
Earlier identification of patients enrolling in specific managed care plans.
Resetting of patient goals for those already enrolled in care-management programs.
Development of models for care-management programs that identify individuals who face a high risk of complications and that can implement a variety of interventions tailored to the needs of specific groups.
Identification of gaps-in-care for targeted patient populations.
Supporting payer quality and accreditation standards by supplementing claims data to provide evidence for higher-quality ratings from HEDIS and CMS measures, and reducing the need for manual review of medical records.
Identifying drug contraindications and the need for alternative therapeutic care plans.
Autopopulation of patients' health portals with laboratory data results.
Enabling improved predictive analysis, for example, using models to project which patients might develop diabetes in the next 12–24 months and/or identifying diabetics with undiagnosed comorbidities.16
As health care plans and health systems become more involved in population health, it is important for them to find ways to identify established patients at risk for incurring higher costs or identifying new populations at risk without intervention. Furthermore, it is in their interest to identify better ways to identify prospective needs and to follow which patients may not be getting the care needed, perhaps secondary to inappropriate or underuse of appropriate diagnostic testing. It still leaves many costs undefined, and to meet these requirements, systems need meaningful data and analytics. Without reliable integrated health care data, systems will find it much more difficult to identify patients who need—or are not getting—care for chronic or worsening conditions. In fact, these data are needed even to identify the people who suffer from, or are at risk for, these conditions. Data are needed to design and measure the effect of interventions meant to improve the care, reduce the costs of care, and improve the outcomes.
Pathologists Can Have Important Roles in Population Health, Regardless of Their Settings
Experts with whom we spoke cited several reasons why laboratory-generated clinical insights can serve as a catalyst for improving care for populations. The first of these reasons is the massive amount of clinical information that is contained in the laboratory information system (LIS). It is widely recognized that laboratory medicine has a substantial effect on clinical decision-making.17 As such, the data in the LIS has great potential as a source of information for patient evaluation. Second, data obtained from the clinical laboratory provides a longitudinal history of care that the patient has received. As Crawford et al13 note, every time the laboratory generates a test result for a patient, it “…creates a specific and indelible record of the state of health of a patient at a given point of time and often over an extended spectrum of disease progression.”(p4) These kinds of data are particularly important when the patient receives care from multiple providers and even multiple health systems, if the same regional laboratory can serve as aggregator and integrator of those data. Finally, clinical laboratory data provide information that is almost in real time, including not only the actual analyte information in the results, but also the site-of-care for the laboratory testing. Providers working on patients targeted by a population health initiative often can get access to testing and test results the same day it was performed, thereby being able to better monitor those patients and, where necessary, quickly identify the need for intervention. In contrast, billing data, claims data, and even EHR data may not be available for some time after the event that it records. In addition, laboratory data (unlike clinical data) are often structured, quantifiable, and classifiable, meaning that inherently, they are more amenable than other clinical data types to various compilation and analytic methods.13
Specifically, pathologists can leverage longitudinal laboratory data that they generally already have—sometimes in combination with other data in the patient's EHR—to address particular problems for a population, to identify interventions that may affect those problems, and to measure the effect of those interventions. In addition, pathologists' access to, and understanding of, a broad range of laboratory data on a particular patient potentially allows them to be real-time “subject matter experts” on those patients. This kind of analysis is not necessarily the sole purview of pathologists. Some private companies that perform data-mining services on behalf of health care systems are developing software to provide population health analysis. Other nonpathologist physicians or analysts, with skills in areas such as data analytics, may also bring valuable skills for combining laboratory and clinical data to support population health. However, pathologists are in a unique leadership role because they are the specific subject matter experts in the testing being performed and because they hopefully already have relationships with the clinicians who need to be involved in any interventions. Thus, pathologists have the potential to expand their own purview by using their skills to implement population health programs and analysis.
However, pathologists' contributions to population health, that is, data analytics, do require expanding their traditional role, from providing a specific diagnosis or result on an individual, to analyzing longitudinal laboratory data both on individual patients and on patient cohorts, to address the needs of the institution to improve population health.
Among the ways that pathologists can show value with population health–related activities are:
Enhancing Risk Management of the Patient Population
As noted by Xu,18 one source of value from laboratory data analytics is to better identify and manage at-risk populations through identification of specific patients in need of intervention. Pathologists can contribute actionable data that informs risk stratification, creating a strategy for dynamically updating risk stratification and guiding the clinical support of those high-risk patient groups. Some of these data analytics may be retrospective, that is, providing a basis to support increased risk-adjustment factors for Medicare payment. Others may be more prospective, such as using laboratory data to identify which patients are at risk of developing disease (eg, diabetes) or complications within the next year or 2.
Enhancing Clinical Care Management
Laboratory data are often key in any clinical management efforts for persons with, or at risk of, chronic medical conditions. Pathologists can improve the ability of health plans to maximize the clinical outcomes and coordination of care by working directly with clinicians to develop algorithms and β-test practices, measure clinician use patterns, deliver educational material to clinicians on proper test use and interpretation, and produce fast test results delivered directly to all clinicians involved in that patient's episode of care. Some laboratories have used predictive medical analytics to develop “medical forecasts” of a population covered by the plan or diagnosed with a specific condition. Other analytics are used to develop targeted interventions that focus on prevention of chronic disease, prevention of complications from chronic disease, and cost avoidance.13
Enhancing Quality Reporting
Pathologists involved in population health management contend that the laboratory will add value by helping payers and health system executives manage patient costs more effectively and provide optimal patient care to large populations. Laboratories may be better suited for identifying health care issues that can contribute to comorbidities or that might be predictors for risk of future chronic conditions. This is particularly true when the patient receives care from several sources (eg, not from a single integrated health care system) because the laboratory's outreach business may have records that are not captured in the payer's claims. In addition, laboratories can help hospitals and health systems quantify the data measures required for Merit-based Incentive Payment System, HEDIS, and other quality metrics—and can set up systems that allow them to improve their metrics. These systems involve working with care coordination programs to ensure that patients enrolled in managed care plans are receiving the full spectrum of care required to meet the required quality metrics.17
Identifying Care Variations
Another way that pathologists can contribute to population health is to identify care variations in different kinds of conditions. Myra Wilkerson, MD, chair of the Division of Laboratory Medicine for the Geisinger Health System (Danville, Pennsylvania), believes that the specialty is in a good position to add value in this area (oral communication, September 7, 2017). The work starts with identifying a condition or precondition (eg, diabetes/prediabetes), identifying the tests needed to identify the condition, and then making sure that people get the follow-up care they need. In addition, pathologists can help identify care variations, a source not only of increased costs of delivering health care but also potentially of reduced quality of that care. This information needs to be tied back to caregivers or nurse managers to achieve optimal improvements.
These interventions are not the same for every practice or every setting. Some laboratories will have data that can address the population of the entire health care system in which they work. In some cases, those systems—or the laboratories serving those systems—so dominate the geographic area that they can effectively address the needs of the population of an entire geographic area. Other laboratories may not have access to such broad data, but have found ways to address population health issues within their institutions with data that are in their EHRs, especially in the inpatient setting and in supporting transitions in care between inpatient and outpatient settings.
A number of interviewees offered examples of their organizations' population health initiatives that involve laboratory data. Samples of these initiatives are presented below:
Pathologists have been among the leaders who established HEALTHeLINK (Buffalo, New York), a not-for-profit regional health information organization (RHIO) in western New York. Indeed, a pathologist, David Scamurra, MD, is the chair of the organization's board of directors and has had an instrumental role in incorporating laboratory data into the system. HEALTHeLINK is a collaboration among physician, hospital, and insurance organizations designed to efficiently share clinical information from different institutions to improve the delivery of care, enhance clinical outcomes, and control health care costs throughout the region. The year 2013 marked the completion of its 3-year Beacon Community project, which was focused on producing real-life clinical improvements for patients with diabetes. The Western New York Beacon Community project (Buffalo, New York) focused on 3 strategic pillars: building and strengthening the health information technology infrastructure; improving health, quality of care and cost; and testing innovative approaches. Forty-seven data sources were used in HEALTHeLink, which accounted for 98% of laboratory results in western New York state. Electronic health records were used to generate diabetes registries to better track laboratory values, vitals, and necessary tests. The focus was on high-risk patients with diabetes toward a goal of reducing preventative emergency room visits and hospital readmissions. Where appropriate, the data were also made available to primary care providers via HEALTHeLINK, giving physicians the ability to adjust their patients' treatment quickly and proactively. Quarterly reports drawn from the registry have helped physicians identify opportunities to improve care and reduce cost. Early adopter Beacon practices prevented 3 hospitalizations for every 100 patients with diabetes in 2012. This translated to a savings of approximately $600 per diabetic patient per year. The rate of hospitalization for patients with diabetes decreased by 26% between 2009 and 2012.
TriCore Reference Laboratories (Albuquerque, New Mexico), the largest medical laboratory in New Mexico, provides an example of a public health intervention that has had dramatic results for the health system. In a pilot program with UnitedHealthcare Group (Minnetonka, Minnesota), TriCore developed a public health intervention to reduce complications and improve outcomes among pregnant women who receive services through New Mexico's Medicaid-managed care system. Medicaid is a major payer of care for pregnant women in New Mexico, accounting for about 72% of all pregnancies in the state. Before the population health intervention, 20% of the births were to women who received prenatal care beginning in the second trimester, and another 8.5% received no prenatal care. A substantial number of women gave birth to newborns with complications whose cost of care averaged around $14 000 in fiscal year 2010 and fiscal year 2011, compared with slightly more than $800 for the average cost of healthy newborns. By analyzing laboratory data and combining it with Medicaid enrollment data, data were available on who prenatal patients were, and through use of a health condition algorithm based on accepted quality standards, TriCore was able to identify those pregnant women who were at high risk for complications or comorbidities in real time, and updated obstetrics and gynecology statuses based on laboratory orders were compiled. Those data were updated weekly, and information was sent to physicians in real time. That approach allowed quicker identification of pregnant women—65% of whom would not have been identified by using claims data alone. It allowed greater monitoring of services, including use of emergency rooms. The higher degree of risk stratification allowed them to quickly identify members with an increased risk of comorbidity and to provide that information to physicians in real time. The project resulted in closing 71% of the care gaps (ie, pregnant women not receiving prenatal care), a substantially lower neonatal intensive care unit occupancy rate compared with members not receiving the intervention, reduced preterm delivery rates, reduced lengths of stay (from an average of 22.7 days for women who received no care to an average of 4.4 days for those who did receive care), and savings to the health care system of millions of dollars.
Geisinger Laboratories, a part of the Geisinger Health System, instituted a pilot project designed to reduce mortality and inpatient costs of patients with sepsis. According to an analysis provided by Geisinger, sepsis occurs at a rate of about 240 cases per 100 000 population, and accounts for 1 in 5 admissions (20%) to intensive care units.19 Sepsis has a 20% to 25% mortality rate, including a 30% to 50% mortality rate for cases defined as being severe.20,21 Geisinger Laboratories designed a quasiexperimental study to test the hypothesis that adverse outcomes—both clinical and financial—could be reduced through more timely and actionable blood culture results. Their intervention resulted in a 21% reduction in the mean mortality ratio, and a 13% reduction in mean length of stay.
Northwell Health (New Hyde Park, New York) is a not-for-profit health care network with 23 hospitals and more than 600 outpatient facilities. The overall work of Northwell in the population health informatics field focuses on extensive clinical integration, RHIO connectivity and health information exchange interoperability, EHR optimization and normalization, a broad portfolio of quality programs, and driving business intelligence data analytics and reporting. Northwell Health Laboratories has undertaken multiple initiatives in each of the areas discussed above to add value to the organization. Particularly productive initiatives have involved (a) using laboratory data proactively to identify attributed patients (“members”) in risk-based managed care plans and to track their coordination of care; (b) using laboratory data to identify patients with high-risk comorbidities, both to guide care of those patients and to ensure that the health system properly documents their conditions for risk adjustment; and (c) supporting health system reporting of HEDIS and value-based purchasing metrics to health plans through real-time aggregation of the relevant laboratory reporting data. An additional benefit of those activities is the laboratory's ability to link patients receiving care from different providers, both within and outside the health system, to support better coordination of care, especially for chronically ill patients. In terms of risk stratification, this work has shown that the LIS is a rich, independent source of relevant data for patient risk stratification and coordination of care, providing data that is not otherwise being extracted from the EHR or claims databases.
Another project at Northwell Health Laboratories uses real-time laboratory serum creatinine data to identify early stage acute kidney injury (AKI) in hospitalized patients. Tarush Kothari, MD, MPH, a pathology informaticist, has led this project in close partnership with clinical champions and hospital administrative leadership. A key issue is that AKI is often underdiagnosed and underrecognized in inpatient settings because of a lack of awareness among nonnephrologists and the absence of effective clinical decision support tools in the EHR leading to variable standards of care. A real-time LIS electronic alerting system, founded on evidence-based guidelines for AKI diagnosis was implemented. It also emerged that hospital administrative personnel were not capturing the true incidence and severity of AKI because of a lack of understanding of diagnostic criteria and inherent limitations in administrative data (based on International Statistical Classification of Diseases and Related Health Problems [ICD] 10 codes [World Health Organization, Geneva, Switzerland]) for disease phenotyping. Laboratory creatinine data were aggregated to provide more-granular information on AKI onset and severity. In addition to providing alerts, the laboratory partnered with health information management professionals and administrative personnel to emphasize the value of laboratory data for early diagnosis of AKI and to increase awareness among physicians and nurses. This clinical program has led to a significant increase in detection, treatment and clinical documentation of early AKI before the condition can became more severe. This project demonstrated a clear role for laboratory data—and pathologist leadership—in providing evidence-based clinical decision support, reducing variability and latency in diagnosis, and preventing disease progression. From a population perspective, laboratory data have allowed a better understanding of the true risk of AKI in the hospital population, better documentation of disease comorbidity and severity of illness, and better risk-adjusted reimbursement to the hospital. An ongoing extension of this project is to use laboratory creatinine data for real-time adverse drug event surveillance of nephrotoxic medications.
Standardizing anticoagulation care for patients on warfarin presents a major clinical and operational challenge in outpatient settings, and laboratory international normalized ratio time-in-therapeutic range data provide actionable information to improve disease monitoring for such patients. Northwell Health Laboratories has been an early contributor to a systemwide interdisciplinary disease management program to improve warfarin outpatient anticoagulation. Using a standardized anticoagulation management protocol, installation of clinical decision support software, provider education, standardization of laboratory prothrombin time/international normalized ratio point-of-care testing equipment, standardization of operating procedures in laboratory patient service centers and outpatient practices, and a continuous quality-improvement methodology, the mean time-in-therapeutic range for high-volume internal medicine and cardiology practices has been consistently greater than the CMS 65% time-in-therapeutic range benchmark. These results have been sustained for a period of more than 3 years and have improved chronic care management for more than 5000 unique patients on long-term warfarin therapy every year. This project demonstrates that laboratory data are essential for monitoring variability in standards of care for patients on warfarin anticoagulation and that pathologists can have a crucial role in population health efforts by aggregating and analyzing data, creating actionable metrics, and being part of disease management teams.
Effective Population Health Interventions Can Occur Even Outside of Integrated Health Systems
The activities described above can be effectively implemented by a laboratory that has access to patient data throughout a large health care system. Those data provide a comprehensive, or nearly comprehensive, set of information on all the medical care received by patients. Thus, when looking at all the people with a certain condition (eg, pregnancy or acute kidney failure), the laboratory is able to identify the care that the patients with those conditions have received over time, making it easier to identify them, to assess risk, and to implement interventions.
Most pathologists, however, do not practice in that kind of environment. According to the 2017 CAP Practice Characteristics Survey, nearly 60% of pathologists work in practices of 10 or fewer full-time pathologists. Slightly more than 40% of pathologists work in practices of 5 or fewer full-time pathologists. Although these pathologists likely have access to a broad range of patient data within their local institution (ie, hospital), they may not have access to a patient's record in other parts of the health system. Moreover, to the extent that the patient receives laboratory testing from outside the system (eg, outpatient care or through admissions to other hospitals), the pathologists may not have access even to all of the patient's laboratory data.
Nevertheless, some pathologists who work in systems such as those have developed interventions to improve health outcomes for populations with specific, suspected diagnoses. One example of this kind of work has been performed by the pathology department at the University of Texas Health Science Center (UTPath, Houston, Texas). Pathologists there have proposed a new role for pathologists as providers of integrated diagnostic consulting services for problem areas that are contributing to the high health system costs. The pathologists at UTPath identified a problem of information overload—too many tests for a primary care provider to know well—arcane ordering procedures, and difficulties in interpreting results or following up correctly.22 After meeting with clinicians to identify their biggest frustrations and needs for help, the pathologists developed a pilot project designed to ease testing for rheumatology referrals. Before the intervention, the hospital in which these pathologists worked had a 6-month wait for new appointments to the rheumatology clinic. Half of the new patients did not have conditions that required treatment by a rheumatologist, and most of the rest of the patients had not had adequate workups performed at the time of their first appointment. As a result, the population of patients with suspected rheumatology diagnoses was waiting too long for appointments and often needed more than one workup. Patients who didn't need to see the rheumatologist had to wait 6 months to find that out. To address that problem, UTPath worked with primary care physicians and rheumatologists to develop a lupus/antinuclear antibodies preconsultation algorithm. Patients suspected of having systemic lupus erythematosus would be referred to the pathology department. The pathology laboratory would order appropriate tests to determine whether an appointment was necessary and communicate the results to the primary care provider and the patient. Those patients with positive results were referred to the rheumatology clinic, whereas those with negative reports would have reports sent to the primary care physician. By serving as a clinical consultant, the pathology department was able to achieve substantial improvements and cost savings. By standardizing the testing, they were able to reduce the mean number of phlebotomy visits among patients suspected of needing rheumatology services from 2.7 to 1.0, for a savings of $1300 per patient. The pathology department eliminated about 90% of unneeded referrals and reduced the wait time for a rheumatology clinic visit from 6 months to 1 to 3 months. The number of referrals rejected for incomplete testing dropped from 40% to less than 3%. This treatment regimen has been adopted system-wide, and UTPath pathologists are now developing a similar program for treatment of anemia.
Another intervention was performed at the University of West Virginia (Morgantown) to improve adherence to treatment guidelines and reduce redundant testing for patients in the emergency department suspected of suffering a myocardial infarction. Ducatman et al23 observed that only 22% of cases were following the testing guidelines to use only troponin l to rule out myocardial infarction, whereas 64% were using troponin l and creatinine kinase-MB in some combination—possibly because the hospital's upgraded EHR included an order set for the more detailed testing. Despite the guideline, physicians felt that the more-involved testing was necessary to accurately rule out myocardial infarction. Using laboratory data, Ducatman et al23 reviewed retrospective records of patients who had received the combined test. They identified 4785 order sets in which the troponin l (the guideline test) was within reference range, and in only 60 of those order sets (1.3%) in 43 patients was the creatinine kinase-MB or MB isoenzyme elevated—making that extra test redundant in 99.2% of the orders. Only 1.3% of the order sets had the theoretical possibility of finding a heart attack that was not identified by the troponin test. However, further assessment of the clinical data on the 43 patients showed that none of them had myocardial infarctions. The pathologists found that the redundant tests were, in their words, “futile,” and also that they resulted in potentially unnecessary admissions for 42 of those 43 patients (98%) for no reason other than the false-positive creatinine kinase/creatinine kinase-MB index.23
The Most Effective Interventions Involve Collaboration Outside the Laboratory
Universally, people with whom we spoke agreed that pathologists need to be proactive in identifying ways that their laboratory data can inform population health and add value to their institutions. Health system leaders and physician colleagues will use laboratory data in their analysis but may not think of the pathologist as a valued partner in those studies. They may need to be informed that pathology can work directly with clinicians to develop algorithms and β test practices, to measure clinician use patterns, to deliver educational material to clinicians on proper test use and interpretation, to provide the most efficient up-to-date laboratory instrumentation available, and to produce fast test results delivered directly to all clinicians involved in that patient's episode of care.
One important recommendation is for pathologists to build relationships within their institutions to raise awareness of laboratory value and also to identify and understand the pressure points that they are facing. The project at the University of Texas Health Science Center, for example, came from pathologists directly asking their physician colleagues about the most pressing issues they were facing.
Another recommendation is to proactively involve all key internal and external stakeholders. In addition to other physicians, this could include other types of providers (eg, pharmacists, nurses), PhD scientists, informaticians, information technology specialists, billing and finance departments, and others. Executive buy in is important to obtain needed resources, encourage support from other providers, and explicitly identify—up front—the value you expect any project to bring to the institution and how you will be quantifying that value. This conversation could be particularly important in defending the value of the laboratory and medical expertise that an in-house laboratory can offer as compared with the services that may be offered by external data-mining companies.18
Finally, experts with whom we spoke identified the need for strong leadership in the clinical laboratory to lead population health work. Those individuals are more likely to be members of institutional committees and to have built the clinical relationships needed to develop the kind of multistakeholder interventions associated with population health projects. These leaders can also build the relationships and trust needed with the executive suite. An overarching objective of these interactions should be demonstrating that the clinical laboratory is a core asset of the health system and can bring high value to a vast array of enterprise strategies. Quite simply, “How can Laboratory help?” should be an overarching objective of laboratory leadership across the enterprise, starting at the executive suite level.
BARRIERS AND LIMITATIONS
Data Analytics Are Key, but the Data Are Not Always Where You Need Them—and May Not Be Telling You What You Think
To assess the health of a population (eg, to identify trends, care gaps, and risk), it is necessary to be able to analyze aggregated (and deidentified) data of individual patients. To close “care gaps” and guide coordination of care, those data must be identified. Clinical data—both laboratory data and other clinical information—must be in a form that is accessible for the intended purposes.
Ideally, a health system would have a data warehouse that contains data from throughout the system, including patient tests results, claims, clinical history, and pharmacy use. A good data warehouse allows for an easier and more complete analysis of the population. However, our research suggests that data warehouses are not in a state where they can be widely used. One expert with whom we spoke suggested that there are only “a handful” of integrated health networks and academic institutions that have true data warehouses. Other ones may still be primitive and may not have good data-mapping capabilities that allow a researcher to conduct population health queries. Moreover, because many institutions do not have good data warehouses, it is difficult for most pathologists—especially those not associated with integrated systems or academic institutions—to access the integrated kind of data contained in warehouses.
Furthermore, there are data problems associated with some warehouses. Critical challenges include standardization of data fields, normalization of data entered into those fields, and the validation of that standardization and normalization process. Specifically, a data warehouse is usually composed of disparate data sets. Substantial work is required to organize the warehouse so that data can be pulled from multiple data sets at the same time and be aligned in a standardized fashion. The data from those disparate data sets must be normalized so that all data fields are comparable, regardless of the source of the entered data. This challenge arises because common terms may not have common meanings across data sets, even within the same system, so it is important to spend time understanding the source data sets and ensuring that they are successfully converted into standard language and usage when imported into the warehouse. An even greater challenge for data warehousing is the enterprise master patient index, without which patients cannot be uniquely identified and tracked across multiple care settings. Absent enterprise master patient index, automated algorithms can resolve many, but not all, issues of patient duplication or misidentification. Manual review of unresolved duplications can close the identity gaps further. However, the risk of erroneous merging of patient identities remains, with very negative consequences if clinical management is based on an erroneous merge.
Members told us that just having a data warehouse is not necessarily enough, especially in larger institutions. In addition to standardizing, normalizing, and validating the data, there have to be data scientists and statisticians who understand the meaning of those data and the conclusions that are being made from interpretation of the integrated data. Then, health economists may be needed to determine the actual value to the institution of acting on such data. There is a new field called value assessment framework that economists are using today that can be applied to health care systems.
A number of interviewees talked about data integrity, and some mentioned it as one of the most important tasks for pathologists. Among the factors to consider are how diligent the data mappings are, what is available to query (and what is left as a data “gap”), and the validity of the data that are extracted. Numerous terminologies and classification systems are used to store data (eg, Logical Observation Identifiers Names and Codes [Regenstrief Institute, Indianapolis, Indiana], SNOMED CT [US National Library of Medicine, Bethesda, Maryland], ICD) codes, Current Procedural Terminology [American Medical Association, Chicago, Illinois] codes) and serve different purposes. A nontrivial task is to ensure data integrity within each classification system and to ensure correct “mapping” between different systems to achieve interoperability within the integrated data warehouse.
To address data validity and interoperability, good data dictionaries are needed. It was suggested by several interviewees that perhaps one major contribution from pathologists would be the development of good data dictionaries to standardize laboratory information. Cleaning and normalizing the data and creating a data dictionary take a substantial amount of effort but are necessary first steps in population health. Resources will be needed to standardize as well as maintain those data dictionaries but, in the end, those dictionaries will help ensure that any analysis that uses laboratory results will be of high value because of its integrity.
In some communities, providers and health care systems have coordinated to establish RHIOs that create the equivalent of a data warehouse for the entire community. An example of such a system is the HEALTHeLINK, a RHIO that covers 8 counties in western New York State. The HEALTHeLINK contains data from 27 hospitals, 5 independent laboratories, as well as home health care agencies, long-term care facilities, and regional radiology providers. It allows providers to access health data from wherever in the system patients receive their care. Substantial work has been performed on normalizing those data and making them useful for analysis.24
Many experts with whom we spoke suggested that pathologists can—and should—be proactive in working with data warehouses and RHIOs in their area. Even if the system is not yet ready for population health, the pathologist can start getting ready by cleaning and normalizing their own laboratory data to make sure that it is relatable for analysis and, in so doing, establish working relationships with the data scientists running those warehouses and RHIOs. This latter point is particularly important because our experts have found that the people running data warehouses and RHIOs are not always aware of the subtleties around the needs for standardizing the laboratory data and the contribution that pathologists can make as subject matter experts.
Pathologists Still Need Training and Skills to Take Full Advantage of Data
It does indeed take pathologists to advise on the clinical and institutional impact of potentially acting on data analytics. It is precisely the clinical subject matter expertise of pathologists that enables both the insights—and appreciation of opportunity—to come forth.
Our experts did have the concern that, although pathologists have access to a great deal of data and may be in a situation in which they can use those data to contribute to population health, many lack the skills and training to take advantage of those data, whether the data were “mined” by themselves or by a data scientist. These analytic and interpretive skills are important for identifying sources of illness or trends in a population. The data are there, but getting to the data—and producing meaning out of it—is the hard part.
Alternatively, other experts contend that not every pathologist needs to be an informatician to be involved in population health but that they do need to be a steward of the data their laboratory creates and to make it as useful as possible for physicians. Some institutions have obtained the services of nonpathologists who are experts in data analysis and work closely with pathologists.
Another important aspect that should not be forgotten is ongoing education of pathology trainees and medical laboratory scientists. These people will need a fundamental understanding of clinical data standards and common methods of querying and analyzing data (eg, structured query language) and how to draw statistical inference from such analyses. This will allow them to be active participants and well-equipped laboratory data experts for population health initiatives within their environments. The Association of Pathology Chairs (Wilmington, Delaware), the Association for Pathology Informatics (Pittsburgh, Pennsylvania), and the College of American Pathologists (Northfield, Illinois) have created resources, such as Pathology Informatics Essentials for Residents training.25,26 Similar resources need to be created and developed on a much wider scale for the broader laboratory workforce to gain necessary competencies in data analytics.27
Data Management Technology for Anatomic Pathology Is Still Evolving
Most of the pathologist activities around population health have been focused on the clinical laboratory. Clinical laboratories have the advantage in that almost all the information is already structured (digitalized) and therefore in a form that allows for large data analysis.
By contrast, most of the information obtained in anatomic pathology (AP) is in unstructured formats (eg, in text format) and not easily standardized for data mining. However, there are some ways in which standardized data can be used in population health for AP-type services. The first is cancer registries, which are large databases that collect and aggregate data to monitor cancer trends over time, show cancer patterns in various populations and high risk groups, guide planning and evaluation of cancer control programs, and advance clinical, epidemiologic, and health services research.28 Similarly, Qualified Clinical Data Registries, such as that recently developed by the College of American Pathologists, are also emerging as a new way to collect, record, and analyze the care patients receive. Clinical data registries can extract data from multiple sources, such as providers' information systems, and transform the data into a common format for the registry. This ability to aggregate and report on data from multiple sources is the promise of clinical data registries, and their advantage over any EHR or LIS.
However, the heart of anatomic pathology is the “report”: the legal document that contains diagnostic and interpretive information from surgical pathology, cytopathology, molecular pathology, and cytogenetics and, when available, autopsies. To date, these reports are “blobs”—binary linear objects containing alphanumeric information—and are not structured.
The interface of the analog legacy of anatomic pathology and the digital world that drives population health includes the following:
SNOMED CT coding.—SNOMED CT has been in effect for several decades and provides a detailed classification of AP diagnoses. Ideally, for those institutions that are using SNOMED CT coding, the data entry occurs at the time of case sign-out. However, it is possible to retrospectively classify AP reports by SNOMED CT. Unfortunately, the potential for SNOMED CT informing population health analytics is limited by the incomplete use of SNOMED CT by health care institutions.29
Synoptic Reporting.—Achieving a high compliance rate for inclusion of the College of American Pathologists' synoptic reporting in the pathology reports of surgically resected specimens is now a requirement of the American College of Surgeons (Chicago, Illinois) for “cancer center” designation. The format allows interested parties (eg, state cancer registries and researchers) to extract and compare the important data elements and thus obtain valuable population health data. The data obtained can help improve the care rendered to patients with cancer. Although synoptic reporting provides opportunity to enter the content as structured data fields, the most common use of synoptic reporting is for that analog data to be embedded in the otherwise “blob” surgical pathology report.
Natural Language Processing.—New promise is offered by natural language processing, which can be used to parse AP reports into structured formats.30 Although the fidelity with which reliable structured data can be extracted from analog “blob” reports will need further investigation, this is a burgeoning field.
Biobanking.—We should not overlook the continuing importance of biospecimen repositories to the science of population health.31 Those research institutions that have biorepositories, resources for the molecular analysis of those tissue resources, and perhaps, most important, robust clinical data repositories for determining the effect of molecular and morphologic features of human tissues on clinical outcomes and response to therapies, contribute enormously to the medical science of population health.
Digital Pathology.—Digital pathology is rapidly emerging as the potentially transformative—and disruptive—event in AP. Although attention can be given to “machine-assisted diagnosis” and even “artificial intelligence” in the interpretation of AP case material,32 those applications of machine-assisted digital pathology still devolve into a “report,” with the provisos given above. However, moving beyond rendering of a diagnosis through a report, digital pathology provides opportunities to create new structured approaches to the quantification of AP diagnostics, prognostics, and therapeutic optimization. The extent to which the quantitative data emanating from digital pathology both transforms the practice of AP and, potentially, informs population health is now the subject of intense investigation.33
Compensation for Population Health Management Activities Still Has Not Been Totally Established
Although many innovators contend that the laboratory needs to provide value-added services, such as population health activities, to survive in a changing and more competitive health care system, they also agree that there is no established blueprint for linking these activities to increased compensation for these services.
To the extent that some laboratories and pathologists are receiving compensation for value-added services, most are doing so through agreements negotiated on an institution-by-institution basis. For example, the Northwell pilot projects were able to demonstrate both cost savings and improved coding of patient acuity for diagnosis-related group reimbursement, which justified both the initial hiring of a pathologist informatician and the acquiring of additional staff resources. The leaders at Northwell's laboratory were able to demonstrate to corporate leadership that their interventions more than paid—at the enterprise level—for the increased pathology staff. Other institutions, such as Geisinger, have many years of experience of providing payments in a coordinated care environment that incentivizes provision of value over quantity of services. At Geisinger, all providers are salaried and are eligible for substantial incentive payments for areas such as cost savings and development-of-care innovations.34 An exception to that approach is the University of Texas Health Center Laboratory, which billed Medicare for services it provided in its intervention. They billed under Medicare codes that required a written request for the pathology report relating abnormal laboratory values and the provision of a written report that requires medical judgment.
WHAT PATHOLOGISTS CAN DO TO BE INVOLVED
Our interviewees offered a number of ideas when asked how pathologists can be involved in population health initiatives. Participation levels can vary depending on the institution and pathology department. Several overarching themes evolved from our discussions.
Understand Your Health Care System Data Environment
One of the first steps in becoming involved in any type of population health management activities is to understand the philosophy of your institution's chief executive officer and senior management in terms of population health. Given the wide array of definitions, it is important to understand how your institution defines population health and strategizes its implementation. As part of that discovery process, it is also important to understand whether leadership views laboratory data as essential data to their population health management goals.
Our experts agreed that it is imperative to understand the health care system's landscape and the role of laboratory data in that landscape. For example, is the health care system part of an ACO or RHIO, and if so, how is the laboratory data sent and stored there? Is it cloud based? Who can retrieve the data? If you are not part of an ACO or RHIO, how are data collected and stored across your institution and other institutions within your health system? For value-based payment managed care, what laboratory data inform the metrics for the value-based payment performance? Beyond that, what are your opportunities for influencing what kind of information technology system is used in your laboratory and at your institution? How do those information technology platforms facilitate—or impede—your ability to provide meaningful data to your institution's population health programs?
What EHR systems are used by your institution? If the entire enterprise, which may be multiple hospitals, uses one EHR platform, is the LIS provided by the same vendor or are different vendors integrated into the EHR? If multiple hospitals use the same vendor, not all hospitals may be on the same version of that EHR or LIS for that matter. Whatever the circumstance, it is important to understand whether data are accessible across the hospitals of an enterprise or are limited to single institutions. It is also important to understand what information in your LIS is also in the EHR and what information remains only in the LIS.
Get Your Own Laboratory Data in Order
Our experts pointed to the need to ensure integrity of laboratory data. A good way to accomplish that is to standardize the laboratory data within your own network. This is especially important (and challenging) if there are multiple EHR and/or LIS platforms in your organization. Creating and maintaining a data dictionary that standardizes test names, units of measure, terminology, and associated coding systems across the enterprise will greatly simplify analysis when analytics are used to compare outcomes. Some of our experts went so far as to mention that they thought normalizing data and creating a data dictionary is the most important task a laboratory can perform for population health-management studies.
Be a Catalyst Toward Population Health Management
Laboratory data are actionable and have both analytic and predictive value. This can be a valuable message to leadership. Our experts suggested focusing on the pre-preanalytics—the decision-making of ordering the right test at the right time—as well as the post-postanalytics—defining and deriving maximum value from analysis using laboratory data. Pathologists are in unique positions to make new associations for different types of data that the organization may not have insight into in its typical analyses.
It is also important to realize that the contribution of the laboratory may extend beyond what is obtainable from laboratory data only. Many of the interventions described in this research rely on other clinical data from the EHR, from the pharmacy, or from other sources. For example, the use of laboratory data in conjunction with pharmacy data enables the system to identify adverse drug events in real time for patients at high risk (such as when monitoring for anticoagulation risk by patients taking warfarin). The laboratory data become more powerful when used in conjunction with those other data sources.
Show Leadership/Reach Out to Clinical Colleagues
Our 2012 ACO report said “many clinicians do not understand the analytic role that pathology plays or the expertise of pathologists in understanding the most effective applications of laboratory medicine. As a result, it is easy for pathologists and for laboratory medicine to be overlooked during the development of ACOs.”34(p12) This finding is also relevant when discussing population health management. Pathologists who want to be involved need to be proactive in their efforts—the words intrusive and persistent have also been used by our experts. Few administrators may initially reach out to pathologists unless pathologists have inserted themselves into the decision-making process, not only of the data analysis but also of the overall population health program design and execution. As previously mentioned, determine whether there is a RHIO in your area and how the laboratory can become involved, at least to ensure the integrity of the laboratory data that transmits from the LIS and/or EHR into the RHIO database.
The examples given in this manuscript did not occur in isolation. Pathologists involved in population health initiatives have reached out to the leadership and other clinical colleagues to identify major problems and pain points (real or perceived problem needing solutions). By working together, specific problems were identified and prioritized, interventions were created, and implementations were put into place. A number of experts stated that the future of health care will be driven by data integration and who controls that data.
Realize That Involvement May Be Critical to Your Laboratory's Survival
Laboratory data can be used to identify care gaps, to help stratify risks, and to create new opportunities. Laboratory data can improve and inform the quality reporting that is becoming an increasing influence on reimbursement. Once an intervention is in place, it is important to demonstrate the success of that intervention and the role the laboratory data had in the process. Documentation of those results is an important activity for the laboratory to perform. It is a way to demonstrate the laboratory as a value center and not just a cost center. With the rapid increase in the number of clinical laboratories being bought by large reference laboratories, hospital and health system leaders may be viewing the laboratory as a commodity and not recognizing the value of the laboratory (and pathologist) in providing vital information and insights on populations.
It is important for pathologists to understand that the environment is rapidly changing and that a culture change may need to occur. There is no question that pathologists have always participated in a form of “population health.” However, our experts emphasized the need for pathologists to communicate the value of the data that the laboratory collects and its value to the organization. Not all pathologists need to be actively involved in these types of activities, but there should be a champion within the institution who can represent the value of laboratory data to administration. Instead of asking “how are pathologists relevant to population health,” pathologists should be assertive in highlighting the value of laboratory data to ensure their work is visible to their administrations and colleagues.
References
Author notes
All authors, except Dr Crawford, are current members of the College of American Pathologists Policy Roundtable's Population Health Workgroup. Dr Crawford is an advisor to the workgroup. The authors received no financial support for the research presented in this manuscript, other than the salaries paid by the College of American Pathologists to those coauthors who are employed by the college (Dr Gross and Ms Kennedy). The authors have no other relevant financial interest in the products or companies described in this article.