As I spend time in the world of healthcare claims data, I repeatedly see the Streetlight Effect. This effect adds tens of billions per year in costs to the healthcare system (at least). What is the Streetlight Effect?
Let me explain: A policewoman sees a man searching for something under a streetlight and asks what the man has lost. He says he lost his keys, and they both look under the streetlight together. After a few minutes, the policewoman asks the man if he is sure he lost them here, and he replies, no, and that he lost them in the park. The policewoman asks why he is searching here, and the man replies, “This is where the light is."
How does this relate to healthcare? Healthcare stakeholders focus on metrics that are readily available and are often easy to measure, but these may not be the most useful metrics for achieving the goals of effective health plan management — better care at lower cost.
Here are some examples of the types of metrics in most healthcare claims reports, and they don't always draw a straight line to lower costs:
Context-less analysis of disease category:
What this is: A sum of the costs of claims by the primary diagnosis category. For example, an MRI procedure code with a diagnosis of osteoarthritis will be grouped into "Musculoskeletal Spending."
This leads to: Focusing on disease management, when we need to focus more on the overall economics of the relevant health plan.
Example: "Our MSK spend is high and has increased significantly" → "We need to address MSK spending through a disease management application."
…When, in fact, the recent MSK increase may have been caused by a single, unpreventable catastrophic injury. Lack of relevant context limits decision-making when summarizing by disease categories. Managing the pricing of the network and providers would lead to more direct savings than dividing by disease category.
Procedure category and place of service spend totals:
What this is: A sum of the cost of claims by the procedure code's category. For example, a procedure code for the facility fees of an outpatient knee replacement will be grouped into, "Outpatient Surgery."
This leads to: Focusing on specific procedures and places of service, when we need to take a broader view of that service category and healthcare in general.
Example: "Our Outpatient Surgery Costs* are High" → "We need to put in place a Center of Excellence program to steer surgeries to different providers"
...When in fact, prices for the surgeries the plan is buying are reasonable and the plan has many surgeries in the outpatient settings that could have gone inpatient at higher prices. To evaluate actionable cost-saving opportunities, it is necessary to understand the plausible alternatives to a given type of care. Steering surgeries to Centers of Excellence may drive less value than, say, implementing a utilization management program for hospital-administered drugs.
*In fact, those “outpatient” surgery measures often don’t include the anesthesia, pathology, or radiology fees that are included in the relevant surgery because they have a different “procedure category." “Outpatient Surgery” may not even be measuring the entire surgery.
CMS population health metrics used as cost management strategies:
Aside: Often, I’ve found that if a metric isn’t mandated by Medicare, it doesn’t exist in the analytics of various stakeholders – ACOs, Health Plans, etc. If a metric is mandated by Medicare in their value-based care programs, commercial plan sponsors and other payers often assume that they are the defining metrics for running a lower cost plan.
What this is: For various "Value-Based Care" initiatives, CMS either endorses or creates metrics that they want providers be evaluated on. Provider compensation is partly dependent on performance on these metrics. Examples: Gaps-in-Care, readmission rates, ED visits, etc.
This leads to: Many health plans' main strategies to manage costs focus on population health programs that have a limited impact on the cost of care.
Example: "We have low gaps-in-care closure rates for preventative mammograms and colonoscopies." → "To lower spending in the future, we need to make sure everyone has a PCP and is encouraged to get a screening."
...When in fact, it is unlikely that these interventions reduce plan costs in the long term, let alone a term short enough for self-insured health plans to capture savings. These preventative procedures may be important and should be part of the health plan’s strategy for managing quality, but there is limited evidence for many of these interventions driving hard savings (they can be a good value for the patient, however).
Percent of Medicare – not as prevalent but can be equally misleading:
What this is: A comparison of a health plan's current claim level reimbursement to the rates paid under the Medicare payment methodologies. For example, Medicare reimburses an MRI at a doctor's office under the Physician Fee Schedule. To derive a percent of Medicare we would compare the payments of a health plan's MRIs in the office setting to the reimbursement for that service under Medicare's Physician Fee Schedule.
This leads to: Relying on a proxy for price as the price and running into cases where we increase costs by moving to a lower percent of Medicare.
Example: "Hospital A is 180% of Medicare and Hospital B is 190% of Medicare." → "We should use Hospital A more."
...When in fact, Hospital A is 15% more expensive because it has a higher DRG base rate for inpatient care (i.e. Medicare pays Hospital A more), has no ASCs (Medicare pays ASCs about 30% less for the same things), is in a more expensive neighboring metro with a higher cost of living adjustment, and moves PT, imaging, and labs into the hospital when they could be done at the office (Medicare can pay office or independent facilities 20-40% less per service than equivalent hospital care).
This focus on metrics that often do not lead to savings creates issues for measurement, prioritization, and strategy. It leads to us leaving a lot of money on the table.
What’s the solution? Identify and derive the metrics that matter.
A health plan is a collection of decisions: provider decisions, patient decisions, and administrative decisions. To be able to move beyond the Streetlight Effect, we need to get better at looking at those healthcare decisions, assemble data that measures those decisions accurately and contextually, and then use that data to find ways that those decisions can be changed or improved.
We need to see healthcare problems the same way that we manage purchasing decisions: by creating and using a framework around the following metrics.
Price – How much did a thing cost?
Example metric: The price of a non-emergent, diagnostic MRI of knee without contrast. It can be generalized to all procedures in a price index.
Efficiency – Did the thing need to happen where it happened in that quantity of where it happened?
Example metric: The rate at which diagnostic colonoscopies happen at the ASC vs. the Hospital.
Accuracy – Was the claim paid correctly?
Example metric: The rate at which a bundled lab (e.g. a general health panel) is unbundled and billed by its constituent components at higher costs.
Quality – What are the outcomes and adherence to evidence-based guidelines of the providers that a population uses?
Example metrics: The observed vs. expected rate of post-procedural complications by unique procedure.
If we measure decisions with these metrics, we can understand:
- The true drivers of cost – did the prices change or did people use more healthcare?
- Who is responsible for value or lack there of – is there a high-priced and low quality provider in network driving our spending?
- Whether our vendors had their claimed impact – did the care navigator lower the price of care or reduce utilization in a way that is related to their functioning?
- What to do next to impact costs and quality – does a network or program help us purchase better healthcare for less?
As I’ve discussed before, the data isn’t set up well to measure these decisions. At Careignition, we are leveraging new data science methods to create new clinical data models. Our data models produce more meaningful metrics for understanding these healthcare decisions, their context, their impact, and how we can fix them.
At Careignition, we are helping to overcome the Streetlight Effect by giving everyone who manages a health plan an AI-powered search light. We do this by applying AI to healthcare data so that it reflects the decisions stakeholders make, evaluate true performance, and chart a path forward. We can deliver these insights in various formats from a web application to an API to an easy-to-understand health plan credit report.
Here’s an example benefits consultant that we work with who used these metrics to verifiably save his client $2.5MM ($2400 per employee) in the last year:
There is a drug being administered at hospital for 5x the normal cost and is going to cost his client's plan $500,000 this upcoming year → Consultant knows the drug does not need to cost that much or even be administered at the facility. He reaches out to the medical management team at the administrator and arranges for the drug to be procured for Walgreens, delivering a $400,000 annual savings.
Prices across geographic markets are shockingly variable on an apples-to-apples basis, and there are two markets that are significantly damaging the health plan. However, there are more cost-effective health plans available the provide comparable care → Consultant selects a new network that provides more efficient healthcare consumption. Verifiably saves $2,000,000 per year.
Despite that recent network change and the claims of the vendor, the prices went up dramatically in one region → He is pursuing a multi-network strategy with that region. Savings opportunity of up to $1,000,000 per year.
We know what we need to look for, finding ways to buy the same or better things for less. All we need to do is get the metrics in place that help us do that, giving us a search light so we can stop looking under the streetlight.
Learn how Careignition can illuminate your health data.