Healthcare Town Hall

Over the past several years there has been substantial interest in reducing avoidable emergency department (ED) visits. A wide variety of strategies have been employed to achieve these reductions including:

1. Benefit design changes such as increasing visit copays or putting limits on the reimbursement of number of un-necessary ED visits by a single patient.
2. Provider incentives through programs, such as Patient Centered Medical Homes (PCMH) to reduce the avoidable ER rate.
3. Structural delivery system changes to emphasize Urgent Care Facilities and after hours primary care.

Many of these interventions rely on analytics based on NYU’s avoidable ED algorithm which uses a probabilistic algorithm based on primary diagnosis code to identify the likelihood of avoidable ED visits within populations. Several analyses have now been done that analyze the effectiveness and/or the safety of these interventions.

The first analysis was done by the Washington State Health Care Authority (HCA). It cites an over 10% decrease in ED utilization and ED PMPM costs in the first six months of a program instituting seven best practices for Medicaid enrollees in the State. The best practices included the electronic exchange of information between emergency departments, patient education of ED utilizers, sharing of lists of frequent ED utilizers, development of ED care plans, guidelines and monitoring of narcotic prescribing and the periodic review of feedback reports. For more information on this program, read HCA’s report, Emergency Department Utilization: Assumed Savings from Best Practices Implementation.

The second is a peer reviewed study by ED physicians, whose conclusion was that the NYU ED algorithm did a relatively poor job in identifying an individual patient’s need for an ED visit. In this study they compared presenting complaint data with ED discharge diagnosis run through the NYU ED algorithm. They found that the presenting compliant predicted poorly whether the visit should have been avoided and that doing so could have safety consequences. While arguably the NYU ED algorithm wasn’t designed to guide individual patient decisions, the article is thought provoking and undoubtedly can be cited as an argument against ED Visit interventions. Read recent article in JAMA, Comparison of Presenting Compaint vs. Discharge Diagnosis for Identifying “Nonemergency” Emergency Department Visits for more information.

I’d expect that many more articles to be published about these interventions in the coming months and years. It will be important for informatics to be aware of these evaluations.

This article first appeared at Milliman MedInsight.

Quality of Care avoidable events, healthcare analytics, MedInsight, Rich Moyer

In the late 1980s and early 1990s, managed care plans frequently featured offerings with provider networks of limited size, based on the idea that using the most cost-effective and efficient providers would result in lower healthcare expenses. As a result, these plans, which usually had lower premiums than larger network counterparts, were hoped to funnel a greater number of patients to smaller networks, resulting in an additional “volume” of traffic to the providers. Advocates of this approach also argued that a smaller network would produce a more favorable risk profile, because members willing to choose from a smaller list of providers were less likely to have an existing condition already in treatment.

A variation on this theme is the “tiered network,” in which the highest benefits are paid when members visit the most efficient providers. However, a “narrow network” is not necessarily a tiered network because the concepts involved represent two different methods for reducing costs while improving access and quality (although “narrow networks” and “tiered network” concepts are often utilized in tandem).

The narrow network approach, which was often combined with other payment methodologies such as capitation or staff-model network design, did produce significant cost savings, but several market factors, including members’ demand to see specific providers, and provider contracts that made inclusion in the most favorable tier a requirement for participation, caused typical network size to steadily increase in the years since.

Another factor that affected the success of this approach was the diverse methods payors used to define the network’s composition; in addition to fee negotiation, providers were often analyzed using a variety of measures, to determine those with the best quality outcomes. However, because these analysis methods varied between payors and were usually not completely disclosed, providers often challenged the results, arguing that important factors such as the health status of a particular group of patients had been overlooked.

Now, with the Patient Protection and Affordable Care Act (PPACA) and its associated exchange dynamics, as alternative reimbursement methodologies and risk adjustment are fundamentally changing the way health plan business is conducted, the narrow network concept is being revisited. Several plans have introduced narrow network offerings in hopes that such plans will be attractive in the exchange environment.

MedInsight has always offered a variety of innovative ways for payors to measure provider quality. The platform includes provider network management capabilities, which enable organizations to analyze, compare, and manage the performance of providers and provider networks. These analytic techniques include the ability to understand both the overall and relative cost performance of provider contracts, analyze how well specific disease and healthcare conditions are managed by providers, compare efficiency within provider peer groups, and identify best practice patterns, all of which can assist in developing and administering “narrow networks.” In addition, MedInsight supports a variety of analytic tools, both proprietary and from third parties, which assist in the quality measurement process. Table 1 below provides an example, derived using MedInsight sample data:

Table 1: Sample Provider Measurement Report

Because many new contracting methodologies, including accountable care organizations (ACOs), rely on quality measures, and because federal and state risk adjustment will incorporate payor-provided claims and electronic health records (EHR) data submitted for audit, these abilities will continue to be of increasing value to MedInsight customers. Finally, these tools can also be used to help provider organizations participating in “narrow networks” to create a better relationship with members.

This article first appeared at Milliman MedInsight.

Quality of Care Healthcare Reform, MedInsight, Mike Bush, provider profiling

In physician profiling initiatives, risk adjustment is often employed as part of the profiling process. The use of a risk adjuster can adjust for some of the effects of patient characteristics that may vary across providers. Using a risk adjuster can be a helpful advantage when reviewing and presenting data to physicians in a meaningful, credible manner. In most risk adjustment tools, the models offered present two perspectives. There is a concurrent model and a prospective model. Each offers different advantages and their uses will vary based on the business question or need to be addressed. This post will take a brief look at the use of concurrent model results in calculating a provider efficiency score.

The concurrent model describes the health status of a physician’s panel of patients based on the patients’ claim and enrollment experience during an assessment period. The assessment period is often the most recent 12 months. The concurrent model is particularly helpful in provider profiling when evaluating patterns or outcomes of practice.

In the following table, the populations enrolled with three hypothetical physician panels were compared to calculate efficiency scores; this process addresses a common provider concern that “my patients are sicker.” Efficiency scores are typically calculated as a ratio of physicians’ actual allowed claim costs and the expected allowed claim costs, based on the concurrent risk scores of the population for which a physician is responsible.

Without risk adjustment, one may draw incorrect conclusions, because the physician practice or panel that appears to have the worst outcomes may simply have the sickest patients. In the example above, while Provider B has the lowest concurrent risk score and lowest actual per-member per-month (PMPM) cost, its efficiency score is the highest at 1.09, or 9% higher than the average for the total population.

Depending on your decision support tools and risk adjustment tools, risk scores are also available by service breakouts, such as inpatient, outpatient, and pharmacy. In this example, a further drill down to the service categories to monitor the distribution of the costs across the service categories would provide additional insight—for example, a comparison of providers’ efficiency in managing inpatient costs by comparing a population’s actual inpatient costs to the costs predicted by the concurrent inpatient risk score.

This article first appeared at Milliman MedInsight.

Quality of Care Barb Ward, healthcare analytics, MedInsight, provider profiling, risk adjustment

The UnitedHealth Center for Health Reform & Modernization has published a new report offering modernized healthcare approaches aimed at reducing Medicare and Medicaid spending.

Here is an excerpt from the report’s introduction:

A consensus emerged during the recent debates on national health care reform that fee-for-service payment mechanisms are at the root of the U.S. health care system’s problems with quality and efficiency. Yet of the roughly $1 trillion spent today on Medicare and Medicaid by federal and state governments, about 75 percent is funded in that way – including over two-thirds of Medicaid’s spending and nearly 80 percent of Medicare’s spending.

The structural problems in these programs are well documented: disparate funding streams; an inability effectively to influence geographical and other inappropriate variation; and a one-size-fits-all approach to managing costs through the crude lever of administered price controls.

We have over the last several years sought to contribute to the debate on how to modernize those programs in a series of Working Papers. The approaches we discussed were potential “win-win” options which would benefit both their enrollees and the taxpayers who fund them.

This working paper updates and combines those approaches in a single volume. In some cases, we have updated our original estimates for new developments in the policy arena.
In designing these options, we have made use of our data and insights from serving one in five seniors nationwide and our overall experience serving more than 75 million Americans, many of whom work for large employers who have been at the forefront of efforts to modernize health care. We have therefore been able to contrast some of their care patterns and programs with those currently available to seniors while incorporating the external research evidence on effective cost-containing strategies and techniques. For Medicaid, the estimates also draw on the track record of some of the most innovative states, as well as our own experience as America’s largest Medicaid health plan. Some approaches presented in this paper would require beneficiary participation in new models of care while some alternative options are based on voluntary and incentive-based designs.

…Taking into account overlapping effects, we estimate a strategic combination of these initiatives could yield $542 billion in federal savings over the 2013 to 2022 period, helping to reduce Medicare and federal Medicaid spending by about 4.4 percent. Of that amount, $437 billion would represent reductions in Medicare spending. States would also see savings from reduced Medicaid spending of $69 billion over the decade.

While much of the recent debate on Medicare and Medicaid savings has centered on either cutting consumers’ benefits or providers’ payments, the options we assess favor a different approach: better care coordination and support for beneficiaries so as to unleash greater value from the health care system.

Read the entire report here.

Quality of Care, Reform Cost, Healthcare Reform, Medicaid, Medicare, UnitedHealth

In a growing trend, insurance companies have begun appending critical illness (CI) riders as part of traditional life insurance policies. Insureds can accelerate receipt of a share of their policy if they are diagnosed with a critical illness such as heart attack, stroke, or cancer.

In this Best’s Review article (subscription required), Milliman’s Darrell Spell discusses how CI riders function. Here is an excerpt from the article:

Proceeds from the rider are paid directly to an insured, with no restrictions on how the funds are spent, noted Darrell Spell, principal and consulting actuary for Milliman.

“The remainder becomes payable upon death, so if you accelerate 50% of your $100,000 face-amount term life policy, your beneficiaries would receive the remaining $50,000 upon the insured’s death,” Spell said.

Similar to stand-alone CI products, insureds can use the riders’ benefits for virtually anything–mortgage, credit card bills, travel expenses, experimental drugs, child care. “The choice is theirs,” Spell said.

Quality of Care critical illness, Darrell Spell, life insurance

Many organizations struggle with the business decision of whether to build their own healthcare data warehouse and decision support solution or to license a solution from an organization that specializes in healthcare data analytics.

Most healthcare payor organizations have some form of an operational data store (ODS) that serves as a storage site for their claims adjudication system data. The temptation to expand the ODS data sources to include non-claims adjudication system data sources and layer data tools on top is strong. But is this the right course of action? Should the ODS be transformed into a full healthcare business intelligence (BI) solution? If not, what role should a third-party BI solution play?

In my opinion, the answer is a bit of both—an ODS for operational reporting and a third-party business intelligence tool for cutting edge business analytics. I postulate that the following structure optimizes the strengths of both models:

1. The payor organization builds and maintains an ODS that has frequent (or real-time) updates of claims adjudication system data. The data is subjected to little or no transformation or enhancement.
2. The ODS has a limited number of “operational” reports written against it. The defining metric on whether a report should be written against the ODS is the “currency” of the data. If you need a near real-time list of open claims, this is a report that should populate from the ODS.
3. Layer on top of the ODS a “best in class” healthcare decision support system. This system is typically characterized by periodic data updates, typically monthly, and a number of advanced data analytic enhancements. Enhancements include methods such as risk scores, service classification grouping, episodes of care, quality metrics, completion factors, attribution methods, benchmarks, etc.
4. Included as part of the decision support system are user tools such as dashboards, online analytical processing (OLAP) cubes, standard reports, and user portals. These different interfaces provide for access to a wide variety of users in the organization.
5. The decision support system also allows for a wide variant of different data sources to be combined together in a standardized format. By combining data from pharmacy benefit managers, third-party carve-outs such as vision or mental health, lab results, wellness programs, or even administrative data, powerful new analysis can be accomplished.

Where do you draw the analytic line between your ODS and BI solution?

Electronic Health Records, Quality of Care Brian Studebaker, business intelligence, EHR, healthcare analytics, MedInsight, operational data store

The implementation of the Patient Protection and Affordable Care Act (PPACA) makes risk adjustment an increasingly important element of regulation-driven health insurance products. Medicare Advantage, Medicare Part D, and many Medicaid programs are already leveraging risk adjustment for plan payments, and healthcare reform mandates risk adjustment for exchanges, as well as Pioneer Accountable Care Organizations (ACOs).

Risk adjustment was first introduced as a compliance component in the Balanced Budget Act of 1996, which mandated a phased-in approach to set budgets, as well as payments to Medicare Advantage plans. Subsequently, the federal Centers for Medicare and Medicaid Services (CMS) created a risk adjustment methodology called “Hierarchical Condition Categories” (CMS-HCCs).

Risk adjustment (or risk assessment) methodologies calculate risk for each individual, which supports comparisons of illness burden between or among health plans. This methodology is designed to “level the playing field” between insurers, and requires a two-step process: Assessing the illness burden of each enrollee (risk assessment), then moving funds from plans that enroll lower-risk populations to those that cover higher-risk members (risk adjustment).

Read more…

Quality of Care, Reform Healthcare Reform, MedInsight, Mike Bush, risk adjustment, severity adjustment

Care coordination is a critical success factor in the broader spectrum of improved outcomes and effective cost management. There are many considerations when evaluating care coordination successes and opportunities.

Potentially avoidable events have been identified as a means of opportunity savings and improved access to care, including, but not limited to:

• Preventable hospitalizations
• Avoidable emergency department (ED) visits

These are critical components to analyze because:

• They may be an indication of access difficulties to the appropriate primary care
• When evaluating these events, it is also important to understand the geographic and socioeconomic factors to help identify opportunities for improved care coordination within a given population
• From a financial perspective, these events can also contribute to higher costs, presenting an opportunity for savings if better care coordination is in place
• The ability to identify and monitor occurrences of these types of events depends on the ability to identify events that are potentially preventable and events that are not preventable

Preventable hospitalization: Prevention quality indicators (PQI) of the Agency for Healthcare Research and Quality (AHRQ) provide a consistent and industry-accepted basis for objective measurement and analysis. Based on the AHRQ literature, hospitalization for an ambulatory care sensitive condition (ACSC) is considered to be a measure of access to appropriate primary healthcare. As a reminder, ACSCs are medical problems that are potentially preventable. While not all admissions for ACSCs are avoidable, it is assumed that appropriate ambulatory care could prevent the onset of this type of illness or condition, control an acute episodic illness or condition, or manage a chronic disease or condition. For example, hypertension (high blood pressure) is a condition that can be treated outside of a hospital. With proper medication and management of care, most people should not need to be hospitalized for hypertension. When interpreting the data, a disproportionately high rate is presumed to reflect problems in obtaining access to appropriate primary care.

The identified conditions include angina, asthma, chronic obstructive pulmonary disease (COPD), diabetes, grand mal status and other epileptic convulsions, heart failure and pulmonary edema, and hypertension.

Avoidable ED visits: The use of algorithms that provide bucketing of ED visits based on diagnosis codes supports the analysis of preventable ED visits. Some examples include the New York University algorithm and the Medi-Cal algorithm. Below is a sample from a large data set for a one-year period showing the percent of visits identified as preventable/avoidable and primary care treatable at a high-level rollup by line of business. Further analysis would include population-specific information to identify factors contributing to these types of visits. Risk adjusting these populations and evaluating by more specific demographics such as age and gender will provide further detail to drive an action plan to reduce these types of visits.

In closing, a better understanding of costs and benefits—how and to whom to target incentives, at which levels of risk—is essential for care coordination and other improvement initiatives to be economically viable and sustainable.

An interesting strategy to address readmissions: Nurse Led Clinics Battle Readmissions.

This article first appeared at Milliman MedInsight.

Cost, Quality of Care avoidable events, Barb Ward, care coordination, healthcare analytics, MedInsight

Using benchmarks for comparative analysis is essential (e.g., weighing the risk of a new reimbursement structure, determining performance targets, or measuring outcomes). In fact, when creating performance-based benchmarks for a population, one must consider how to account for the disease burden. The question becomes, how does one adjust for circumstances outside of the control of a healthcare system (e.g., genetics and environment) or where the health system has limited influence (e.g., lifestyle choices)? The objective is to try and remove all uncontrollable factors, thereby identifying any variance from the benchmark as an opportunity that could be addressed by either medical management or contractual change. Variance from the benchmark would then illuminate potential calls to action.

In order to create this ideal condition benchmark adjustment, we have found the following characteristics to be essential in the adjustment factor development process:

1. Broad vs. narrow disease categories:
• Cohorts should present in credible amounts for a condition in order to reflect the overall disease prevalence
• A broad definition captures all stages of a disease’s progression
• Broad condition adjustments can be used for population predictive modeling

2. Condition definitions should not be weighted too strongly by bursts of healthcare that are due to short-term events.

3. The methodology assumptions should not restrict the data sets used in benchmark construction. Traditional large medical and prescription claim data sets can be used, yet we can also leverage credible electronic health records (EHR) or billing/charge master data sets as well if these are available

In MedInsight, we have created several condition-based benchmark adjustment factors, each targeted at providing a solution for a specific, often narrow, need, yet we feel the best overall method for condition adjustment (adj) is our Chronic Condition Hierarchical Groups (CCHG) product. The CCHGs assign all individuals to a single category (100% of the population, both the sick and the healthy), the cohorts are clinically similar, the number of cohorts are small or manageable, and, most importantly, the hierarchy algorithm stems from the clinical point of view; physician decision making is at the heart of the system.

Traditionally, we combine the CCHG condition factor with our Milliman Health Cost Guidelines™ (HCGs) benchmarks as a means to improve our trend management benchmarks. As seen below, we have a traditional factor table for some of our inpatient HCG-level adjustments:

In the HCG benchmarks above, the CCHG condition factors are meant to resolve or tie out to the HCG benchmarks to adjust for uncontrollable disease-related factors. Later in 2012, we hope to apply the CCHG methodology as an optional benchmark factor itself and not a part of the HCG benchmark. We believe this will improve our risk adjustment process and better explain results in both clinical performance projects and health plan marketing activities. Using the CCHG-based condition adjustment factors will allow plans to better quantify their value propositions while identifying new ways to improve their overall efficiencies.

This article first appeared at Milliman MedInsight.

Quality of Care benchmarking, healthcare analytics, Marcos Dachary, MedInsight

The innovations and opportunities resulting from healthcare reform—accountable care organizations (ACOs) and patient-centered medical homes—rapidly expand the roles of nurse care managers. These models require population-based analysis and prioritization, proactive care gap closure and care coordination, and credible assessment of impact to determine if cost and quality are improving.

At the core of all these models and the transformative direction of health reform, are key design principles that include:

• Population-based analysis used to target and prioritize interventions
• Accountability for achieving quality and financial goals for defined populations
• Patient-centered care design that addresses medical as well as psychosocial needs
• Education and engagement strategies that directly involve the patient and family in achieving better health and preventing exacerbation of chronic care
• Medical homes and medical “neighborhoods” that integrate primary care, specialty care, and community programs in a manner that provides better resource use and easier access
• Care coordination that assists patients as they maneuver the complexities of the healthcare delivery system
• Proactive outreach that encourages timely preventive care and supports care compliance
• Transparent discussions and thoughtful choices made based on evidence-based options and costs of alternatives

Read more…

Quality of Care, Reform ACOs, healthcare analytics, MedInsight, nurses, nursing