The Evidence Crisis: Causal Inference – Don’t be a chicken (Part 3)


Part 1

Part 2

Physicians have been making up numbers longer than people have been guessing weights at carnivals.  How much does this statin lower the chances of a heart attack? How long do I have to live if I don’t get the aortic valve surgery?

In clinics across the land confident answers emerge from doctors in white coats.  Most of the answers are guesses based on whatever evidence about the matter exists applied to the patient sitting in the room.  The trouble is that the evidence base used to be the provenance of experts and anecdotes that have in the past concluded leeches were good for pneumonia.

And so came the randomized control trial to separate doctors from homeopaths.  Random assignment seeks to achieve balance between two groups for everything but the treating variable to isolate the effect of the treatment.  But does randomization really guarantee a balance between groups?  At least the known confounders may be measured in the two groups, but what about unknown confounders?

Consider some examples:

Blood stream infections are currently treated by antibiotics targeted towards the specific organism found in the blood.  The distribution of organisms within the blood stream is as follows:

Vancomycin is active against S.aureus, Coag Negative Staph (CoNS), and other gram (+) organisms.  Cefepime is active against Gram negatives.  Now imagine a world where there was no understanding of the different organisms that caused blood stream infections, and an earnest researcher seeking to do right by patients with blood stream infections ran a trial which randomized all blood stream infections to either Vancomycin or Cefepime.  In the beautiful table that would accompany the published article to demonstrate balance via randomness of age, sex, gender and perhaps some type of frailty index, the most important factor – the type of organism – would be missing.  What we don’t know in this case drives the outcome.  A trial with a high population of homeless patients could have a very different distribution of organisms than a trial performed in suburban patients in the rich Mainline of Philadelphia.  The inability to generalize a trial of mainline moms to other populations means a trial fails to have external validity.  There are other problems as well.  Perform a large trial that combines the homeless with suburban soccer moms and the average treatment effect loses meaning unless you are a clinician that may randomly take care of a mainline mom or a homeless person on any given day.

Continue reading

Self-Driving Cars are Like Most EMRs


Drivers are distracted klutzes and computers could obviously do better. Self driving cars will make all of us safer on he road.

Doctors have spotty knowledge and keep illegible records. EMRs with decision support will improve the quality of healthcare.

The parallels are obvious. And so far the outcomes are disappointing on both fronts of our new war against human error.

I remember vividly flunking my first driving test in Sweden. It was early fall in 1972. I was in a baby blue Volvo with a long, wiggly stick shift on the floor. My examiner had a set of pedals on the passenger side of the car. At first I did well, starting the car on a hill and easing up the clutch with my left foot while depressing and then slowly releasing the brake pedal with my right forefoot and at the same time giving the car gas with my right heel.

I stopped appropriately for some pedestrians at a crosswalk and kept a safe distance from the other cars on the road.

A few minutes later, the instructor said “turn left here”. I did. That was the end of the test. He used his pedals. It was a one way street.

Three times this spring, driving in the dark between my two clinics, I have successfully swerved, at 75 miles (121 km) per hour, to avoid hitting a moose standing in the middle of the highway. Would a self driving car have done as well or better? Maybe, maybe not.

Every day I get red pop up warnings that the diabetic medication I am about to prescribe can cause low blood sugars. I would hope it might.

Almost daily I read 7 page emergency room reports that fail to mention the diagnosis or the treatment. Or maybe it’s there and I just don’t have enough time in my 15 minute visit to find it.

For a couple of years one of my clinics kept failing some basic quality measures because our hasty orientation to our EMR (there was a deadline for the incentive monies to purchase EMRs) resulted in us putting critical information in the wrong “results” box. When our scores improved, it had nothing to do with doing better for our patients, only clicking the right box to get credit for what we had been doing for decades before.

Our country has a naive and childish fascination with novelties. We worship disrupting technologies and undervalue continuous quality improvement, which was the mantra of the industrial era. It seems so old fashioned today, when everything seems to evolve at warp speed.

But the disasters of these new technologies should make us slow down and examine our motives. Change for the sake of change is not a virtue.

I know from my everyday painful experiences that EMRs often lack the most basic functionalities doctors want and need. Seeing a lab result without also seeing if the patient is scheduled to come back soon, or their phone number in case they need a call about their results, is plainly speaking a stupid interface design.

I know most EMRs weren’t created by doctors working in 15 minute appointments. I wonder who designed the software for self driving cars…

Hans Duvefelt is a family doctor in Maine. This piece was first published at his blog A Country Doctor Writes

from THCB

Health in 2 point 00 -Episode 35, Shafi Ahmed takeover edition

Jessica DaMassa’s European tour continues. This week she’s at the #WebIT conference in Sofia, Bulgaria (no, I couldn’t find it on a map either!) and the #HealthIn2Point00 takeovers continue! This time the guest is pioneering British surgeon Shafi Ahmed, who has lots to say about medical education, the future of digital hospitals, what he’s up to in Bolivia and how cool #WebITHealth will be–Matthew Holt

from THCB

Information Blocking–Gropper & Peel Weigh in

Today is the last day for public comments on the proposed CMS regulations regarding Medicare hospital inpatient prospective payment systems (IPPS). While there are several changes proposed, the one that’s raised lots of attention has been the idea that access to Medicare may be denied to those providers guilty of information blocking. Here are the comments submitted by Gropper & Peel from Patients Privacy Rights— Matthew Holt

Executive Summary of PPR Comments on Information Blocking

Information blocking is a multi-faceted problem that has proved resistant to over a decade of regulatory and market-based intervention. As Dr. Rucker said on June 19, “Health care providers and technology developers may have powerful economic incentives not to share electronic health information and to slow progress towards greater data liquidity.” Because it involves technology standards controlled by industry incumbents, solving this problem cannot be done by regulation alone. It will require the coordinated application of the “power of the purse” held by CMS, VA, and NIH.

PPR believes that the 21st Century Cures Act and HIPAA provide sufficient authority to solve interoperability on a meaningful scale as long as we avoid framing the problem in ways that have already been shown to fail such as “patient matching” and “trust federations”. These wicked problems are an institutional framing of the interoperability issue. The new, patient-centered framing is now being championed by CMS Administrator Verma and ONC Coordinator Rucker is a welcome path forward and a foundation to build upon.

To help understand the detailed comments below, consider the Application Programming Interface (API) policy and technology options according to two dimensions:

API Content and Security Institution is Accountable Patient is Accountable
API Security and Privacy
  • Broad, prior consent
  • Patient matching
  • Institutional federation
  • Provider-directed interop
  • Compliance mindset
  • Directed authorization
  • Known to the practice
  • Individual credentials
  • Patient-directed interop
  • Privacy mindset
API Content / Data Model
  • Designated record set
  • FHIR
  • Patient-restricted data
  • De-identified data
  • Bulk (multi-patient) data
  • Designated record set
  • FHIR
  • Sensitive data
  • Social determinants
  • Wearables and monitors

This table highlights the features and benefits of interoperability based on institutional or individual accountability. This is not an either-or choice. The main point of our comments is that a patient-centered vision by HHS must put patient accountability on an equal footing with institutional accountability and ensure that Open APIs are accessible to patient-directed interoperability “without special effort” first, even as we continue to struggle with wicked problems of national-scale patient matching and national-scale trust federations.

Here are our detailed comments inline with the CMS questions in bold:

Additionally, we specifically invite stakeholder feedback on the following questions regarding possible new or revised Conditions of Participation (CoPs) and Conditions for Coverage (CfCs) for interoperability and electronic exchange of health information:

  • If CMS were to propose a new CoP/CfC standard to require electronic exchange of medically necessary information, would this help to reduce information blocking as defined in section 4004 of the 21st Century Cures Act?

Yes it would. The most significant help would be to require patient-directed exchange as the minimum and as a safe-harbor from sanction for information blocking. Patient-directed exchange is well established and common when the medium is paper via postal service or fax. This mode is accessible to all patients and all practitioners with patient consent. Patient-directed exchange also poses no patient-matching problem as has been proven over decades. Patient-directed information sharing spans diverse sources of information, including sensitive data and social determinants of health, and diverse uses of information, including research and public health. Applying the principles and policies of postal / fax release of information to the health records’ API would preserve all of these benefits and eliminate the problems such as delay, cost, and difficulty in machine processing. It’s time to make technology that eases burdens on patents and ensures they can move their data to the right place, at the right time.

Patient-directed information sharing is arguably the only foundation for practical reduction in information blocking. This foundation provides the privacy/consent mechanisms and inherent patient-matching capability onto which more sophisticated functionality such as alerts to subscribed providers upon patient admission can be built.

After 15 years of industry data-blocking, it’s abundantly clear that ONLY patients will assure their health data is used for treatment. The ‘covered entities’ treat patient health data as a corporate asset.

  • Should CMS propose new CoPs/CfCs for hospitals and other participating providers and suppliers to ensure a patient’s (or his or her caregiver’s or representative’s) right and ability to electronically access his or her health information without undue burden? Would existing portals or other electronic means currently in use by many hospitals satisfy such a requirement regarding patient access as well as interoperability?

Yes, but simple access is not enough. Patient access must be defined to include patient-directed sharing without any restriction in order to create an efficient market for consented use of the patient data. Patients need the right to direct their information to any end user in order to avoid loss of provenance, authenticity, delay, and cost. They also know best which diagnoses, medications, and treatments are correct and effective; so they can prevent the propagation of erroneous data and diagnoses. It is especially important because physicians can no longer be counted on to assume responsibility for reconciling wrong and/or out-of-date diagnoses, medications, and data. Recipients of patient-directed information must have the assurance that the information was not tampered-with that can only come from getting it directly from the source. This was/is characteristic of postal/fax access, for example for Social Security Disability determination, where the need for direct transfer of records to avoid tampering seems obvious. This feature of patient-directed sharing must be preserved. As we introduce electronic APIs, it’s not enough to put personal health records in the middle, because these can be tampered-with. We need the APIs to enable direct transfer from source to destination as directed by the patient.

Existing portals play an important role in eliminating undue burden from electronic interoperability but they need to be enhanced to provide the same functionality that patients have when presenting a Release of Information form in-person but without the inconvenience of in-person. In other words, the patient portal must allow the authenticated patient to fill out a Release of Information form and specify the scope and destination of external records access that is equivalent to what’s available in-person today. Policy around this was developed by the 2016 ONC API Task Force and accepted by the relevant committee. Technology standards such as Kantara User managed Access (UMA) that build upon the existing FHIR / OAuth security standards are also available that can be linked into the patient portal enhancement for patient-directed access without undue effort.

It’s finally time to develop health technology systems that serve patients, and make it convenient for them to use, benefit from, allow research queries of their records instead of disclosures, and understand their own health data.

  • Are new or revised CMS CoPs/CfCs for interoperability and electronic exchange of health information necessary to ensure patients and other treating providers routinely receive relevant electronic health information from hospitals on a timely basis or will this be achieved in the next few years through existing Medicare and Medicaid policies, Health Insurance Portability and Accountability Act of 1996 (HIPAA), and implementation of relevant policies in the 21st Century Cures Act?

Even though HIPAA and the 21st Century Cures Act provide the legal basis for routine reception of relevant information, new CMS CoPs/CfCs may be required because interpretation of the applicable regulations leaves too much latitude for information-blocking through censorship of destinations, arbitrary limits to access, delay in access, lack of consent, lack of transparency and accountability, arbitrary transaction obstacles, charging for patient access, and inadequate patient matching.

New CMS CoPs/CfCs should follow the lead of the API Task Force to eliminate the excuses that hospitals use to delay or censor competitive uses and make patient-directed access by patient and other treating providers the rule.

  • What would be a reasonable implementation timeframe for compliance with new or revised CMS CoPs/CfCs for interoperability and electronic exchange of health information if CMS were to propose and finalize such requirements? Should these requirements have delayed implementation dates for specific participating providers and suppliers, or types of participating providers and suppliers (for example, participating providers and suppliers that are not eligible for the Medicare and Medicaid EHR Incentive Programs)?

A reasonable timeframe for compliance with new or revised CMS CoPs/CfCs for interoperability and electronic exchange of health information is the same timeframe that any API access is deployed for provider-to-provider, provider-government, provider-research electronic exchange.

Health records vendors and hospitals have a strong incentive to provide APIs to their selected suppliers through controlled “app stores” and other vendor-controlled tollgates, while blocking information by services they do not approve or that enable competing providers as part of the care team. This natural information blocking tendency is only amplified by compensation mechanisms such as ACOs and narrow networks that seek to limit competition to treat a particular patient. Large provider networks use control over the patient information as a way to get larger at the expense of independent practices and practice innovators, limiting competition and preventing out-of -network physicians from having data for treatment. Large networks are helped by policies such as Stark Law exemptions for health IT. The result has been continued price increase and difficulty in value-based payment innovation.

By linking the timetables for patient-directed access to internal app store access, we ensure that information blocking is not supported through standards manipulation and regulatory capture.

  • Do stakeholders believe that new or revised CMS CoPs/CfCs for interoperability and electronic exchange of health information would help improve routine electronic transfer of health information as well as overall patient care and safety?

As patients and physicians wanting to represent the interest of our patients, we insist on the option for unblocked access to patient information when the patient directs that access. Patient-directed sharing needs to be the first kind of sharing, and accessible to all patients. A physician should not have the power to block patient access and say, “I’m afraid I can’t do that.” (to quote HAL in the movie 2001 A Space Odyssey).

Once this capability is in place a market for innovative services to support treatment and patient safety will be created that is accessible to patients and to physicians. Right now, this market does not exist because patients and physicians have no market power in the face of institutional control over the information technology, ie, because of data-blocking.

  • Under new or revised CoPs/CfCs, should non-electronic forms of sharing medically necessary information (for example, printed copies of patient discharge/transfer summaries shared directly with the patient or with the receiving provider or supplier, either directly transferred with the patient or by mail or fax to the receiving provider or supplier) be permitted to continue if the receiving provider, supplier, or patient cannot receive the information electronically?

Yes. As stated above, patient portals can remove the undue costs, loss of time, missing work, and other burdens of in-person visits patients must bear before initiating non-electronic forms of sharing. Patient portals should serve the equivalent function of in-person authentication and consent regardless of whether the actual transmission medium is mail or fax.

  • Are there any other operational or legal considerations (for example, HIPAA), obstacles, or barriers that hospitals and other providers and suppliers would face in implementing changes to meet new or revised interoperability and health information exchange requirements under new or revised CMS CoPs/CfCs if they are proposed and finalized in the future?

No. As mentioned above, hospitals and other providers have an incentive to enhance their internal systems through APIs for app stores that they control. The management of consent and patient matching remain unsolved problems even after almost a decade of HITECH regulations. At this point, we must conclude that interoperability and health information exchange is only possible and scalable with a patient-directed approach.

  • What types of exceptions, if any, to meeting new or revised interoperability and health information exchange requirements, should be allowed under new or revised CMS CoPs/CfCs if they are proposed and finalized in the future? Should exceptions under the Quality Payment Program including Certified Electronic Health Record Technology hardship or small practices be extended to new requirements? Would extending such exceptions impact the effectiveness of these requirements?

The API Task Force addressed this issue by concluding that restrictions to patient-directed exchange should be allowed only in cases where a patient-directed system could impact the access rights of others using the same API. For example, API access can be denied to a system that makes so many requests in a period of time that it slows the information system as seen for access to other patients.

By leading with patient-directed exchange without significant exceptions an institution is effectively given a safe harbor against liability for privacy or information-blocking. Security is a reason for exception only when the security of other patients is affected. The safe harbor provisions of patient-directed exchange protect the CMS CoPs/CfCs hospital from any security breaches related to a specific patient-directed data use or data processor.

We would also like to directly address the issue of communication between hospitals (as well as the other providers and suppliers across the continuum of patient care) and their patients and caregivers. MyHealthEData is a government-wide initiative aimed at breaking down barriers that contribute to preventing patients from being able to access and control their medical records. Privacy and security of patient data will be at the center of all our efforts in this area. CMS must protect the confidentiality of patient data, and CMS is completely aligned with the Veterans Affairs, the National Institutes of Health, ONC, and the rest of the federal government, on this objective. While some Medicare beneficiaries have had, for quite some time, the ability to download their Medicare claims information, in pdf or Excel formats, through the CMS Blue Button platform, the information was provided without any context or other information that would help beneficiaries understand what the data was really telling them. For beneficiaries, their claims information is useless if it is either too hard to obtain or, as was the case with the information provided through previous versions of Blue Button, hard to understand. In an effort to fully contribute to the federal government’s MyHealthEData initiative, CMS developed and launched the new Blue Button 2.0, which represents a major step toward giving patients meaningful control of their health information in an easy-to-access and understandable way. Blue Button 2.0 is a developer-friendly, standards-based API that enables Medicare beneficiaries to connect their claims data to secure applications, services, and research programs they trust. The possibilities for better care through Blue Button 2.0 data are exciting, and might include enabling the creation of health dashboards for Medicare beneficiaries to view their health information in a single portal, or allowing beneficiaries to share complete medication lists with their doctors to prevent dangerous drug interactions.

To fully understand all of these health IT interoperability issues, initiatives, and innovations through the lens of its regulatory authority, we invite members of the public to submit their ideas on how best to accomplish the goal of fully interoperable health IT and EHR systems for Medicare- and Medicaid-participating providers and suppliers, as well as how best to further contribute to and advance the MyHealthEData initiative for patients. We are particularly interested in identifying fundamental barriers to interoperability and health information exchange, including those specific barriers that prevent patients from being able to access and control their medical records. We also welcome the public’s ideas and innovative thoughts on addressing these barriers and ultimately removing or reducing them in an effective way, specifically through revisions to the current CMS CoPs or CfCs for hospitals and other participating providers and suppliers. We have received stakeholder input through recent CMS Listening Sessions on the need to address health IT adoption and interoperability among providers that were not eligible for the Medicare and Medicaid EHR Incentives program, including long-term and post-acute care providers, behavioral health providers, clinical laboratories and social service providers, and we would also welcome specific input on how to encourage adoption of certified health IT and interoperability among these types of providers and suppliers as well.

PPR welcomes the intent of the MyHealthEData and initiative but, as currently announced, the CMS Blue Button 2.0 API is not developer friendly and poses tremendous burdens to patients that simply want to direct access to their own data. As proposed, the CMS API is incompatible with open source software because it assumes that the entity connected to the patient’s data is a corporation rather than a community. There is no obvious reason to block access to non-commercial technology because the access is still secured by the patient’s credentials. No matter how bad the technology is, be it corporate or community code, the patient’s trust and right of access must not be second-guessed by CMS. The process is also burdensome by including 1-2 week delay and a costly interview. If these restrictions on the patient’s intent as directed through the Medicare patient portal are not an “undue burden”, what is? We hope that CMS Blue Button 2.0 will reduce the burden on patients and reduce administrative costs, by adopting established dynamic client registration standards immediately and by offering a FHIR API that supports User Managed Access (UMA) standard alongside the FHIR/OAuth standard of the current version. UMA is much more patient-friendly because it allows the patient to manage their CMS, VA, All of Us, and private-sector consents without having to routinely access separate patient portals for each system.

MyHealthEData policies, including those that are implemented in CMS Blue Button 2.0, in the VA systems, and in the National Institutes of Health All of Us initiative must be aligned with the 2016 API Task Force recommendations and any restrictions or burdens on patient-directed access to the patient’s own information must be justified by evidence of actual harm or security risk rather than the current practice of making these policies out of the public eye.

Expanding access to patient data beyond Covered Entities without meaningful transparency, accountability, and informed consent will clearly cause even greater, potentially crippling mistrust in physicians and the entire US healthcare system. As of Jan 2016, the lack of trust in US health technology and in health professionals who use it led 89% of 12,090 patients surveyed to withhold data from providers.

Granting thousands or millions of entities such as “suppliers, hidden access to patient data, could cause 100% of patients to withhold data and/or refuse treatment.

The data being accessed under the flag of “analytics” and “research” is more often used to increase patient cost and further discrimination than it is used to advance the public good. Huge corporations inside and outside the US healthcare system spend billions buying massive health databases full of bad data to build proprietary analytics, isolated learning health systems, and black-box AI. Examples; IBM, Palantir, Google, etc.

But can AI and analytics be accurate if the data are bad and the biases hidden? Predictive AI and analytics are misapplied in prison industry; it predicts that the same inmates who are/were incarcerated will be re-incarcerated.

The lack of trust in health IT interferes with making correct diagnoses, and leads to prescribing inappropriate medications and treatment. Lack of trust in health IT interferes with quality and value-based payment initiatives because sensitive social determinants of health are not available or reliable.  The result of this massive deficit of trust is: US EHRs are filled with bad data, omissions and errors. Not only should more hidden entities NOT be given access to patient data, but those with hidden access, the Covered Entities, must earn the public’s trust back. Congress and industry must restore patient control over data via meaningful, transparency, accountability, and informed consent.

Adrian Gropper is CTO and Deborah Peel is President of Patient Privacy Rights


from THCB

India’s Health Insurance Experiment. Who will be the winners?


Though the exact cost of Modicare, the government’s extension of health insurance for poor people, estimated at one lakh crore (a trillion U.S. dollars), is open for debate, what is not disputable is that the cost of insuring India’s poor won’t fall with time. A sure way of accelerating healthcare inflation, that is speeding the rate of increase of healthcare costs, is by subsidizing or paying for health insurance. Insurance is like Newton’s Second Law of Motion – the velocity keeps increasing as long as the force is applied.

Healthcare is a peculiar industry. Cars get cheaper but medical care doesn’t. The Maruti eventually became cheaper than the Ambassador, and more aesthetically pleasing than its Neanderthalic predecessor. Medical care doesn’t get cheaper because a life saved from cancer is a life waiting to be killed by another disease, which needs treating, too. Survivors of cancer get heart attacks and survivors of heart attacks get cancer, and survivors of both get dementia.

It’s like a restaurant where you can’t just pay for lunch – if you pay for lunch you have to pay for breakfast and dinner and may be a few samosas in between the meals. But unlike eating, consumption of medical care is not guarded by satiety. The insatiable medical sciences keep delivering even more expensive ways death can marginally be deferred. For example, the once dreaded stroke which leads to paralysis is now treatable. However, the treatment is not cheap and comprises clot busters, dangerous drugs with fatal side effects. Further, to treat stroke you need rapid diagnosis by modern imaging – that is you need CAT scans and radiologists. If penicillin for pneumonia is like eating at a roadside dhaba, treatment for acute stroke is fine dining at the Taj.

Insuring all citizens is fine but someone must pick up the tab. Since the costs of medical care are shared by everyone – ill or well, which is the whole point of insurance, the market mechanisms for controlling prices disappear. Hospitals become more daring with their charges, as they feel more emboldened charging faceless insurers than patients. If you think the charges for an ICU stay at Apollo Hospitals is too high – wait till the insurance industry expands.

Then there is a well-known phenomenon called moral hazard – when people don’t pay specifically for individual episodes of care, they use more medical services – i.e. the use of both necessary and unnecessary care rises. Think of it this way – if you’re splitting the restaurant bill there’s a tendency for everyone to get the more expensive items on the menu, because you’re not saving money for your own thrift there’s no incentive to be thrifty. The bill is higher for everyone than if they had gone dutch. The total bill is also higher, and restaurants make more money.

Money spent on insurance premiums, rupee for rupee, doesn’t all go to medical care. Inevitably, administration is required to manage insurance claims. Insurers employ staff to ensure that hospitals are billing correctly, and what is being billed for is indeed medically necessary. Hospitals employ staff to make sure insurers pay for what is billed, and that doctors are billing correctly. Far from reducing the medical costs, the administrative costs of insurance increase net costs, which increases premiums, which keep rising.

Insurers may restrict which doctors the insured can see and hospitals may restrict which insurance they accept, diminishing patient choice. As far as jobs are concerned Modi’s stimulus, ostensibly for poor people, will create white-collared jobs – i.e. jobs for the middle class. This creates an intriguing trade-off – looking after poor people creates jobs for the middle class. Modicare is a windfall for doctors, economists, data scientists, talentless box tickers, statisticians, and the pharmaceutical and device industry, but not for school teachers or autorickshaw drivers or chaiwallahs.

Modicare will look after the health of unemployed Ramu Kaka with an acute stroke from rural Bihar. But economists warn us that there’s no free lunch – there is always an opportunity cost. The money spent on treating Ramu Kaka’s stroke, as humane as that is, is money not available for good schools and public works. Thus, unemployed Ramu Kaka gets a CAT scan for stroke, but his clever children still attend poor quality schools and remain working class. Meanwhile, mediocre Raj from urban Patna will become a claims administrator for an insurance company.

If you asked Ramu Kaka about his preferences, maybe he’d forego that free CAT scan for a job and for better education for his kids. Perhaps if he wasn’t unemployed, he’d not be smoking so heavily, and not have gotten that stroke in the first place.

There is only so much that medical care can do to improve the health of a nation. Without infrastructure, medical care is like a rose in a swamp. You can have all the CAT scans and clot busters to treat stroke but if the roads are so congested that it takes ambulances two hours to get to a hospital, all the medical effort is for naught.

Universal healthcare should be the goal of affluent nations. But the cost of Modicare isn’t trivial. And given the state of the air in many Indian cities, the lingering lack of basic infrastructure, and the desiccated state of public education, Modicare is a tad premature.

About the author:

Saurabh Jha is a contributing editor to THCB. He can be reached @RogueRad. This piece originally appeared in The Telegraph, India






from THCB

Adjusting for Risk Adjustment

Risk adjustment in health insurance is at first glance, and second, among the driest and most arcane of subjects. And yet, like the fine print on a variable-rate mortgage, it can matter enormously. It may make the difference between a healthy market and a sick one.

The market for individual health insurance has had major challenges both before and after the Affordable Care Act’s (ACA’s) risk adjustment program came along. Given recent changes from Washington, like the removal of the individual mandate, the market now needs all the help it can get. Unfortunately, risk adjustment under the ACA has been an example of a well-meaning regulation that has had destructive impacts directly contrary to its intent. It has caused insurer collapses and market exits that reduced competition. It has also led to upstarts, small plans and unprofitable ones paying billions of dollars to larger, more established and profitable insurers.

Many of these transfers since the ACA rules took effect in 2014 have gone from locally-based non-profit health plans to multi-state for-profit organizations. The payments have hampered competition not just in the individual market, which has never worked very well in the U.S., but in the small group market, which arguably didn’t need “help” from risk adjustment in many states.

The sense of urgency to fix these problems may be dissipating now that the initial rush for market share under the ACA is over and plans have enough actuarial data to predict costs better. There has been an overall shift to profitability. But it would be a serious mistake to think that just because fewer plans are under water, the current approach to risk adjustment isn’t distorting markets and harming competition.

The Needle and the Damage Done

Numerous problems have been reported and discussed since 2016 when the damage became widely evident, but despite a few tweaks to the model in the right direction, little has changed until now. With the latest CMS guidance, states have two new openings to reduce the harm from the federal risk adjustment process. I’ll describe them below, but first I want to be clear that these criticisms are not intended to do away with attempts to create a fairer marketplace.

The principle of risk adjustment is great in theory: make plans compete on providing the best, most efficient service rather than by “cherry-picking,” or deliberately enrolling the lowest risk members (in other words, the healthiest ones). Cherry-picking doesn’t improve the quality of care and insurers who do it focus on appealing to the healthy at the expense of the sick. Under the ACA, the idea is to remove this problem by having insurers whose enrollees are sicker get subsidized by insurers whose enrollees are healthier.

The reality of risk adjustment has been mixed. In Medicare Advantage (MA) the formula is similar, but it has not precipitated a crisis among insurers because it is not a zero-sum game in which money from one plan is handed to its competitors. Instead, plans are measured against a benchmark and the risk adjustment payments to plans with higher risk populations come from the federal government.

Both ACA and MA plans have an incentive to identify all the risk factors they can for both clinical (to help identify their enrollees’ conditions) and financial reasons. Risk is typically measured by the diagnoses included on medical claims, but active measures can be taken to find risks that haven’t appeared on claims. The insurers that are more aggressive than average at recording risks get an advantage in higher payments. CMS has long recognized that MA plans will report more health conditions than doctors report for a comparable traditional Medicare population, and so CMS deflates the risk adjustment payments for MA plans by a few percentage points every year (5.9% in 2018).

However, there is a difference for an insurer between failing to keep up with a national risk inflation factor (the MA approach) and having to give 5% or 10% or more of the premium it earned to its local competitors (the ACA approach). The latter is inherently more volatile and allows aggressive coding to become a competitive weapon. This weapon can be wielded more effectively by large, well-funded organizations that have sophisticated data systems to analyze potential missing diagnoses, a large field operation to seek out and confirm diagnoses, and a member population that has been with the plan for a long time so that there is a rich claims history to mine. In short, it tends to favor incumbents. This problem could be greatly reduced by using multi-payer claims databases so that a person’s risk score truly reflects all the data and doesn’t change when a person moves from one plan to the next, but that approach runs into legal, regulatory and competitive barriers that are not easy to bridge.

The problems of risk adjustment in the ACA have been more profound and damaging to competition than this data-capture dynamic alone, however. To start, the risk adjustment model makes a small allowance for what’s called induced demand, or the fact that when people pay less out of pocket for their health benefits they tend to use more services, and in the process more diagnoses become known to the system. The model assumes this greater use of care has nothing to do with the underlying population being sicker.

We know, however, that different plan designs attract different people even if the insurers are not trying to cherry-pick healthier populations. A small network plan with less doctor choice is likely to attract healthier and less wealthy people who are more sensitive to premium price differences. Do we want to penalize these plans and the people in them? Insurers must increase premiums to compensate for the money they will lose in risk adjustment. In doing so, we are penalizing one of the main cost control vehicles plans have available to exclude the most expensive providers, as well as an important method of aligning a plan’s network with value-based contract models, and one of the best methods new upstart plans have of breaking into a market and increasing competition.

In general, the model doesn’t distinguish between disruptive innovation and old-fashioned cherry picking. This is no easy matter and I don’t claim to have a solution for how to do it, but the absence of a competition-sensitive distinction should reduce the magnitude of the adjustment amounts out of modesty. Another marker of competition is profitability, which the current model fails to account for. Why are smaller plans that are struggling to survive paying larger plans that are already profitable? Or even if both plans are unprofitable prior to risk adjustment (as occurred frequently in the early years of the ACA), some sensitivity of the potential for risk adjustment to drive an insurer into insolvency should be included in the transfer calculation, as long as the insurer isn’t cherry-picking.

There are many other concerns with the model, but the last I will mention is that it fails to account for policy differences between states. New York, for example, allows multiple children as dependents on a single contract, while the federal risk adjustment formula allows only one dependent in the member count. The higher premium for the entire family is included in the risk adjustment process, so it inflates the premium per allowed member and increases the size of the transfers from one plan to another. This amplifies any advantages a plan may have in risk coding.

These technical issues have real consequences. To stick with New York, two major departures of new entrants were caused in part by risk adjustment. CareConnect conducted a tactical exit by its parent, Northwell Health. Previously, Health Republic, one of the largest co-ops formed in response to the ACA, entered bankruptcy. At the time they left the market, they had grown rapidly and provided serious competition to giant incumbents in the state, including Blue Cross Blue Shield plans and United Healthcare. Both cited past and ongoing losses from risk adjustment as reasons for leaving, having paid hundreds of millions to their largest competitors, and over 20% of their entire revenue in some years. Health insurance is an industry that typically sees only a 3-5% profit margin, so paying out 15-20% or more to risk adjustment in multiple years is a catastrophic loss for plans that are not deliberately cherry-picking the healthiest enrollees and are pricing aggressively to grow market share.

After a great deal of unnecessary harm to the markets, CMS has come to recognize many of these issues and is giving states leeway to modify what plans pay each other as a result of risk adjustment. States have always had the ability to apply for approval from CMS to conduct their own risk adjustment programs but given the numerous constraints and expenses in doing so, only Massachusetts attempted it. And after a couple of years it gave up. New less costly or more immediately accessible options from CMS are welcome.

Option 1: Federally-Managed Risk Adjustment with a State-Specific Discount

In April 2018, CMS created an approach for “state flexibility” in which a state may apply to CMS for a modification to the federal risk adjustment model. On this approach, a state would present its case for why the federal model results in inaccurate and harmful payments between plans, and estimate a percentage by which the federal model’s results should be reduced for the state. The reasoning could refer to unique regulations in the state that distort the model’s measurements and other factors like those mentioned above.

CMS also now recognizes that the individual and small group markets may not need the same sort of risk adjustment. The small group market may have differences in plan design and competitive dynamics that mitigate the need for risk adjustment every time differences in measured risk appear, and so CMS will allow reductions of transfers in the small group market of up to 50%. CMS is making a striking claim here. It is essentially acknowledging that there may be self-correcting features of the market that accommodate differences in the health of members across plans while maintaining vigorous competition, so that these differences don’t all need to be transferred away through risk adjustment.

To paint a picture of how this could work: a small group plan with a very wide network and rich benefits may be attractive to firms with older than average employees who have established doctor relationships over a metropolitan region and who place high priority on being able to keep seeing their docs (older almost always means sicker), but that same insurance product may also attract firms that employ relatively healthy well-paid professionals who can afford to pay the premium for a high-end product, thereby mitigating the increased risk and cost, and creating a stable risk pool and a product that can profitably attract business in the small group market. This product could compete against a smaller network product with higher out of pocket costs at a lower price point, which tends to attract different types of firms with different priorities on cost and access. Employees in the second set of firms could on average be younger (healthier) and poorer than those in the other product, with a risk pool that is also stable, and a slightly lower average risk than the high-end type of product. Why shouldn’t this situation be allowed to persist, rather than effectively have one set of small employers (who have chosen to provide a less costly insurance product insurance out of necessity or priority) subsidize the richer insurance selection of another set of firms? This is a philosophical point as much as it is a point about the health of markets.

Allowing states to apply for a modification to the formula is a good step in the right direction, but any such changes approved by CMS under this option will not to take effect until 2020. While many new entrants and underdogs have become insolvent or left the individual and small group markets, many still remain, and this wait will ensure that they continue to subsidize their competitors in 2018 and 2019. Combined with the instability created by the loss of the individual mandate, more may leave.

Option 2: State Implementation of Supplemental Risk Adjustment

After I first wrote about this in 2016, New York became the only state to take matters into its own hands to stabilize its small group and individual community-rated health insurance markets. It created its own risk adjustment pool, designed in reference to the federal program. The purpose of the pool was to reduce the size of transfers by 30% for the small group market in 2017 and to reduce both small group and individual market transfers by 26% in 2018.  New York referenced some of the issues above in justifying these numbers, as well as the fact that up until 2016 administrative costs were fully baked into the average statewide premium that plugged into the risk adjustment model, meaning that large plans with high administrative costs and profit could skew the average premium upward and with it raise risk adjustment payments (because they are pegged to the average state premium). That particular feature of the federal model has largely disappeared, but the rest of the problems remain.

Neither New York nor any other state to my knowledge has announced its own adjustment to stabilize these ACA markets for 2019. In part the lack of follow-through by other states may be because the 2017 and 2018 New York State adjustments resulted in a lawsuit from the insurer with the biggest winnings under the old model. So the question is: can and should New York and other states act on their own to reduce the damage from risk adjustment, at least until the new Federal process take effect?

Encouragingly, the most recent federal guidance appears to confirm that they can. When asked whether New York’s emergency stabilization action was in conflict with the federal risk adjustment program, HHS responded:

States are the primary regulators of their insurance markets, and as such, we encourage States to examine whether any local approaches under State legal authority are warranted to help ease the transition for new participants to the health insurance markets. States that take such actions and make adjustments do not generally need HHS approval as these States are acting under their own State authority and using State resources. However, the flexibility finalized in this rule involves a reduction to the risk adjustment transfers calculated by HHS and will require HHS review as outlined above.

Given a direct invitation to wag its finger and declare that New York should have received HHS approval, instead HHS provided a general statement about state authority to regulate insurance markets. Specifically, states have the authority to set up their own risk adjustment and stabilization measures using their own resources, which is what New York did. The reference to easing the transition for “new participants” is a bit confusing, since state regulatory authority in insurance extends far wider than this. Even if the focus is on new entrants, a state must consider the possibility that yet more companies may seek to enter a market and the State may issue regulations conducive to such entry to improve competition. New York, and other states where risk adjustment continues to be a problem, should take this opportunity to create a bridge fix for another year until the federal process can be improved.

New York is actively considering the use of this authority for 2019. On April 19th its Department of Financial Services posted instructions to insurers that they “should not include an assumption for a New York market stabilization pool in [developing their] 2019 rates” but did so “solely to assist insurers in using consistent assumptions for upcoming rate submissions for 2019” and retained full discretion to implement market stabilization measures after reviewing the impact of the federal risk adjustment program.

New York’s past corrective actions have contributed to greater market stability in 2018. The scale of payouts to the largest plans was reduced. In the individual market, the state reported modest increases in QHP enrollment during the annual election period despite the turmoil and uncertainty around the ACA. In the small group market one new insurer, Oscar, began offering plans. But without further action the same pattern of the smaller players subsidizing the bigger ones is likely to continue and that does not bode well. Based on all the evidence, New York should take the steps within its authority to correct for the adverse effects of federal risk adjustment in 2019 for the sake of the health and continued viability of their ACA markets.

Other states should take heart from the CMS comments and New York’s experience, and seriously consider whether circumstances in their individual or small group markets warrant similar temporary actions in 2019 before the formal remediation option begins in 2020.

Jonathan Halvorson is with the Sachs Policy Group and was formerly a New York State health plan regulator

from THCB