As the nation’s changing healthcare environment continues to challenge even the most advanced and progressive-minded healthcare providers, the financial viability of every provider will depend on its response to these changes. At one level, providers will need to evolve to meet health care’s challenges—making the gradual transition from fee-for-service to fee-for-value payment, for example. Other challenges, however, present an imperative for a more immediate transformation, where a failure to act places a provider at risk of being left behind amid rapid change.
One such imperative is the need for providers to transform how they analyze and use data to make important healthcare decisions and choices, improve financial health, and—most critically—deliver high-quality care to their patients.
Consider a sampling of recent trends challenging providers:
- Healthcare spending is expected to increase at an average rate of 5.6 percent between now and 2025.
- Drug spending increased 12.2 percent in 2015, the highest rate of increase in more than a decade, and prices for more than 3,500 generic drugs at least doubled from 2008 to 2015, and for nearly 400, they increased 1,000 percent.
- Continued consolidation among health plans could increase their scale and bargaining power, potentially reducing payment rates and increasingly shift the risk to hospitals.
- Increasing supply costs and possible wage pressures from the improving economy may add margin pressure (Moody’s).
- With the growing consumerism trend in health care, patients have increased access to information that can help them make choices in the healthcare marketplace.
- The increasing use of high-deductible plans, with higher premiums and out-of-pocket/deductible levels, is adversely affecting consumer finances, with one study by the finding that 59 percent of people who reported they had been contacted by a debt collector said it was for medical services.
And, of course, a continuing concern is the fate of the Affordable Care Act and what new legislation might replace it.
Cost and Care
With all these changes and uncertainties, providers require improved data analytics to maintain high-quality care and reduce costs.
For example, cost-reduction efforts often target supplies and services. It can be much less stressful to find lower-cost gloves or syringes than make workforce reductions (which can adversely affect patient care and employee morale). Many organizations consistently review their costs, and compare them with benchmarks or with what others are paying for the same items. Vendor contracts may be reviewed in the hopes of renegotiating pricing and terms. Yet reducing the cost of individual items may have little impact on the overall cost of care. Products can be overused, used incorrectly, and even thrown away or wasted. How does an organization understand the true cost of providing care and, ultimately, determine where meaningful changes can be made?
Efforts to improve outcomes and the quality of patient care cannot proceed meaningfully without knowledge of the total cost to deliver a service and understanding of the value each component brings. Costs can be compared among practicing physicians and among other hospitals and providers. Understanding the total cost per patient encounter and its impact on profitability (and on patients) can help identify where opportunities for improvement may exist. Moreover, it is the starting point for more sophisticated data analytics.
Comparison of Average Cost and Average Length of Stay (LOS) per Joint Replacement Case for 12 Physicians
Consider, for example, that various cost components that can be collected and analyzed for a single DRG, such as joint replacement. The exhibit above shows such an analysis, reporting the cost per case by physician and the associated length of stay (LOS). In this instance, there is considerable variance among the physicians, which may be due to different products used, different protocols and types of care provided, and different levels of patient acuity. In general, this type of analysis is useful for uncovering such variances, thereby helping to determine where it might be best to focus more-detailed comparative analysis of products, supplies, pharmaceuticals and other patient-related costs and choices in care.
Sample Cost-per-Case Breakdown by Component for 13 Physicians
Such a more detailed cost-per-case breakdown is depicted in the exhibit above, which shows the variance across a few of the key supply cost categories. For each category, a profile can be developed of:
- What products are being used
- The source vendors
- How much of each item is being used per case
- The line-item and total cost for each category
Reports presenting such data can be used in meetings with surgeons, surgical coordinators, and other clinical staff to review variances found, identify the best practices, and standardize to the best-practice level of care. An agreed-upon approach can be developed for each category, such as standardization of product to one or fewer vendors, changes in utilization of certain drug therapies, and even elimination of some products altogether. A target cost or range can be established and agreed upon by all, and then savings can be estimated and measured. In this example, implants clearly make up by far the largest proportion of the cost, but other supplies and practices should be considered and discussed as well.
Changes in approach may be diagnosis-specific,
while other ideas may be applicable across a wider spectrum, potentially throughout the surgical suite. It also is important to review supporting processes that may impair successful transitions, and ensure that changes are made throughout the system in areas such as:
- Product and process updates to physician preference cards
- Item master updates (e.g., deletion of noncontracted items, pricing updates)
- Staff training for changed processes and procedures
This level of review can be the starting point for other detailed analysis, such as reviews of performance under bundled and value-based payment.
Predictive Analytics: The Next Step
With the ready availability of large quantities of data and the advancement in computing capability, organizations can go beyond the basics of cost accounting (addressed in the previous example) to organize and analyze information to make progressively better business decisions. According to IBM, 2.5 quintillion bytes of data are created every day. The use of predictive analytics—the process of using data and tools to identify patterns that can help predict outcomes—can provide additional information and insights using these data points to manage and reduce costs and patient risk.
In the healthcare environment, electronic health records (EHRs) provide a wealth of information, typically capturing everything regarding the patient encounter from the general admission process down to each pill given and at what minute of the day, as well as personal data about the patient. Other data can be collected, such as demographics, health history, and even measurements collected by mobile devices. Trends and patterns of use may be discovered that can provide a basis for developing other types of predictive models. Here are two examples of how data and predictive analytics can be applied.
Readmissions. In October 2012, the Centers for Medicare &Medicaid Services (CMS) began reducing Medicare payments to inpatient prospective payment system (IPPS) hospitals with excess readmissions. (CMS calculates and uses a hospital’s excess readmission ratio, which is a measure of a hospital’s readmission performance compared with the national average for the hospital’s set of patients with that applicable condition.) Over the past several years, CMS has reduced payments to hospitals with excess readmissions, making this a financial as well as a medical concern for hospitals.
Using patient data to understand when readmissions occur and why, providers can identify the patient populations that pose a higher risk for readmissions, develop protocols and practices to reduce the readmissions, and then begin to predict upon admission when a patient may be at risk.
Taking one hospital’s readmissions over a set time period and determining which diagnoses are associated with the highest numbers of readmissions during that period is a start. But just reviewing that data set will not provide all the answers, however. Using advanced analytics and other machine intelligence, patterns can be identified specific to specific groups, and then particular actions can be taken to alleviate those incidences.
Number of Readmissions, November 2016
For example, the readmission analysis for a hypothetical hospital in the exhibit above indicates that one DRG under which the organization has seen a high number of readmissions over a one-month timeframe is diabetes. A deeper dive into the demographics of the population, depicted in the exhibit below, shows that, in one week, the highest number of readmissions occurred among females 18 to 35 years of age.
Readmissions for Diabetes, Nov. 6-12, 2016
Let’s assume that the physicians reviewing the data in this study discovered that the readmissions for this group were mainly due to noncompliance with prescribed insulin regimens. This finding would go against expectations, because such noncompliance tends to occur more often in aging populations. But in this case, the physicians eventually discovered that the women in the younger age group were avoiding the insulin to control and lose weight (given that weight gain is a common side effect for people who take insulin because the hormone regulates the absorption of glucose by cells).
The example shows how analyzing and dissecting the data can highlight an area of concern (e.g., the high incidence of readmissions among younger women) regarding which physicians then can engage in further discussion and analysis to identify previously undiscovered connections and causes. In the case example, the physicians could decide to focus on developing specific counseling and care procedures for the patient group, and begin to stem the readmissions.
Supply chain applications. Supply chain management focuses on having that the right products when and where they are needed to meet customer demand, and doing so with the greatest efficiency and at the lowest cost possible. Manufacturers have refined these processes over the years, employing data analytics to better forecast the “when” and the “where.” Forecasting models traditionally have been based on historical data, but with predictive analytics, models can be built that include many other factors to better predict future demand.
For example, a manufacturer of car headlights that wants to plan production schedules and raw material purchases can best do so if it can predict its customers’ demand, based on factors such as fluctuations in new car sales and headlight replacements. The company could review its past sales and look for patterns and trends, but it can gain greater insights and create even better models to predict future demand by looking at additional data, such as car sale trends and the factors that influence them (e.g., economic conditions, fuel prices) and replacement factors (e.g., car accidents, age of cars on the road and lamp life).
Hospitals can apply these same techniques to manage their own supply chains. Many supply chain teams calculate the amount of supplies to keep on hand based on historical usage; each stocking area may have different par levels established based upon how much has been used, plus some additional buffers for safety stock.
Supply and Inventory Management
For example, the common practice for supplies of the type shown in the exhibit above would be to look at the average use per day of the supply and then simply order to ensure that amount is on hand plus some additional in case of fluctuations. (Typically, items are stocked to a potential maximum needed per day, and supply chain managers often revisit these established levels and adjust them to minimize excesses and waste.)
Adding other factors to the analytics, however, as shown in the exhibit, can make the forecasting more accurate and allow inventory to be further minimized to reduce spend. Supply usage can be tied back to specific patients and events—understanding birth rates in the area, for example, to predict the need for baby blankets. Other supplies may be more related to specific types of patient diagnoses or illnesses (e.g., IVs, implants, tubing) and may require more detailed information about practicing physicians and their patient populations.
In short, by identifying factors that affect the use of supplies in different ways and analyzing them to predict future usage, an organization can more closely match its purchase of supplies to when they will actually be needed, thereby avoiding ordering in excess or “just in case.” As a result, costs can be reduced and there is less chance of overstocking, potential damage, waste, and expired products.
The Data Analytics Continuum
Analyzing costs is not new. What is new is the unprecedented availability of vast amounts of data and advances in computing capabilities. These advances have presented hospitals with an important opportunity—even an imperative—to transform how they analyze costs to gain tremendous new insight into patients and the costs of care delivery, including many factors not considered in the past. The transformative process is simple: By starting with fundamental analytics and then adding data points to develop predictive models, a hospital can begin to tap previously unimagined opportunities not only to reduce costs and improve operations, but also to improve patient care and eliminate unnecessary risks.
Caroline M. Kolman, PE, is managing director, AArete LLC, Chicago, and a member of HFMA’s Western Pennsylvania Chapter.