This could be a significant step forward in computation. Years ago I read an article on what was required by…
Readers Write: What’s More Useful Than Hospital Pricing Data?
What’s More Useful Than Hospital Pricing Data?
By Data Nerd
An HIStalk reader challenged my recent post, “Hospital Pricing Data: Another Step Down the Rabbit Hole” by asking what healthcare data should be publicly available to help consumers make better choices, not just from CMS, but from providers and private insurers.
I cannot fault anyone for their enthusiasm. Trust me when I say I know how demoralizing it is to come up with a data solution that just doesn’t fit the need. That’s precisely why I felt compelled to speak out on the subject. After setting high hopes and expectations of the analytical possibilities from data in CMS’s pipeline, the solution fell drastically short of what I had hoped it would accomplish when it was finally released.
Having said that, the ideal data solution for me as a consumer would use the same or similar claims data sources, but aggregate the data two different ways to come up with a predictive solution that can be tweaked to assist the patient in their own cost containment efforts. This type of solution would involve:
- Risk-adjusted cohorts. Grouping the data not just by DRG, but by patients with similar risks (age, co-morbidities, etc.) to chart out the most likely course of treatment for someone of my age and health facing the same diagnosis. Ideally, this dataset would include all payer types, but the next-best offering that is within CMS’s reach is to combine Medicare and Medicaid datasets to account for a broader age distribution. Data would not be aggregated by hospital, simply by patient characteristics across the country.
- Once we have an idea of possible treatment routes, we can then couple that with charge data. And, yes, I want that broken down by procedure at each hospital. Like there is no such thing as bad data, there is no such thing as too much data. I’ve never seen OSCAR’s backend, so I’m not sure if it’s possible to break apart every claim and get a procedure-level charge, but I do know with the data as it is today, claims with only one procedure can be isolated and charges or reimbursements tend to have low standard deviations. Since I am not insured by Medicare or Medicaid, knowing what hospitals charge or are reimbursed by CMS does me very little good, though. I would need my own insurance company’s network rates with the hospital to analyze how soon I’ll meet my annual deductible, etc. Or, if I have the luxury of time to make a decision, evaluate if I’d be better served investing in an HSA and initializing treatment in the next fiscal year. But, for the millions covered under Medicare and Medicaid, such an analysis based on the data today would assist in forecasting when deductibles will be met and/or what other amounts will not be covered during the course of treatment.
- In the event that I have a long-term illness or a more drawn-out treatment plan, I would want an analysis of whether or not it would behoove me financially to have different procedures performed at different facilities. Outcomes data would be useful here as well.
All of these data components would need to be dynamically updated and processed, probably using software to evaluate each step of the way, much like the way a simple tax form is completed online. Play with one number and see how it affects the final bill, and in this case assess the risk factor involved in hospitals with poorer outcomes. Ideally, the solution would also interject preventative challenges over time to help the patient meet their health goals in a way that saves the health system money as well, but that is more the quantified-self realm than the (current) data realm.
So, to recap the data offerings that would satiate my current appetite for price transparency:
- Claims data, aggregated by DRG and patient characteristics to obtain expected procedures
- Claims data, aggregated by hospital and procedure charge
- Combine these two alongside insurance reimbursement rates to give a patient’s total estimated personal expense at every hospital
- Hospital procedural outcomes data to evaluate cost savings and determine at which hospital(s) to have the necessary procedures performed
This is the type of data that would be useful to me as a consumer.
Data Nerd,
Interesting approach, but I still have a very big problem with charge data, whether sourced off claims forms or elsewhere.
The media is clamoring for price transparency, what we really need is payment transparency. Why should Medicare pay $5,000 for a Hip Replacement, Atena pay $7,500, Medicaid $5,200, Assurance pay $6,700? The usual argument is they deliver volume, but that’s not true. None of these payors guarantee a provider any volume at all. The only one that might is a managed care plan with a defined population.
Your analysis would be even far more enlightening if done on actual payment, not charges. But that would be REAL Transparency which no insurance company or provider would ever voluntary agree too.
FLPoggio, Change:Healthcare (Nashville, TN) is one such company that is tackling price transparency at the payment level.
Excellent post. I also agree with FLPoggio that paid claim data is far more useful than charge data for people with insurance. He’s wrong, though, to say that no insurer would ever voluntarily agree to provide it. I directed the project to build a similar system for a New York based carrier. Unless they removed it after I left, it’s still on their member site. I know of several other commercial insurers that do this as well. Sadly, few people aware of it. If you want to research what it looks like, CIGNA may be the best example.
Also, if you have an HMO or a rich-benefit PPO, this information actually isn’t very useful personally because you pay very little out of pocket, and what you do pay is largely copays that are independent of the adjudicated claim amount.
The people who need this data the most are the uninsured (where charge data from Medicare may be useful) and those in high-deductible, limited benefit plans.
Is it that hard to understand this? The commercials pay more because providers lose money on Medicaid. It’s cost shifting and focusing on payment isn’t going to help anything; especially as we enter into a post exchange world where there will be alot more plans that start to look more like Medicaid from a payment perspective.
Healthcare providers do not understand their own unit or activity based costs. What they understand is procedure costs as a component of what they can charge and, yes, that is variable. How can you take a systems engineering approach to improving the system when we can’t even tell you accurately what it costs (and map that cost to procedure cost) to have a nurse on hand for 6 hours? Is it her loaded salary? Is it 1/52 of the procedure cost? Does her cost change when she work a shift in L&D versus a shift in the NICU? Even the ACO structure won’t completely solve this problem as shared gains only matter if the initial size of the pie is large enough to share.
We do have a cost problem in healthcare and transparency is a solution to it. Only, before we worry about being transparent with the outside, we need to truly be transpareny internally and actually understand healthcare costs at a unit and activity level.
Thank you both for your insightful comments. My background is strictly data, so I only have my own personal health insurance experience on which to base assumptions about what data would be needed to forecast my personal cost savings analysis.
My original assumption was that knowing what hospitals charge for a procedure can be coupled with knowing the rate at which an insurance provider pays the hospital, i.e. if hospital charges $10,000 for procedure X, and I know that my insurance pays 75% or charges for procedure X, I know they’ll pay the hospital $7500 and since my deductible/out-of-pocket maximum is reached at $5000, I can assume that I’m only only the hook for $2500.
But, as you both point out, we could just collect what my insurance pays each hospital for that procedure directly, rather than the rate of payment since this is a hypothetical ideal dataset. Forgive my lack of vision here. I’m used to skirting holes in the data and trying to come up with best-case results, rather than dreaming up ideal data sources. Having payment data as you suggest would also possibly offer more insight into what these negotiated rates are based upon, if not volume, and level the playing field. interesting to note which insurers are or would be willing to share the data and which would hold back should a magic wand grant our data wish.
However I would argue, JD, that the market trend has been going in a direction of higher out-of-pocket amounts for both HMO and most affordable PPO’s that, while they may be insignificant to much of the HIStalk readership, comprise a large proportion of the average American’s take-home pay and will be deal-breakers when it comes to actually receiving the care promised under the ACA. Great that we can get more preventive care, but may be insignificant if price transparency and other market pressures do not force these “charges” and “payments” into the realm of ‘affordable’ for both payers and patients. I for one would like to see the ACA live up to its name in this regard, regardless of my personal politics.
RE: Really?
Providers presumably have access to all of this internal data should they choose to take a nosce te ipsum approach to understand how their own prices are formulated (and I do think they should, if they haven’t already). Consumers, however, have fragmented access to data that is rarely representative of the full market picture. They are expected to make cost-saving financial and lifestyle decisions with access to a fraction of the medical and financial data that providers already have. Why should public resources be made available to help providers make better choices? Would that result in lower cost for the consumer? The secretive nature of the chargemaster to date seems to suggest not. The focus going forward needs to be on the consumer, and then allow the competitive marketplace to affect how providers respond in their price setting.
Data Nerd,
If the focus has to be to be on the consumer, which I do not disagree with, that will take a complete overhaul of the current healthcare system. Keep in mind the convoluted, complex system we have today has evolved over fifty years. It’s that system that created all the data elements you and the Big Data folks want to use to analyze and change things. And as I have said before all that price data is severely ‘ warped’.
In summary you cannot take data points generated by a government /third party system and give them to a consumer and expect them to understand its weaknesses and make valid decisions.
Lastly, our new government systems (ARRA, PPACA, P4P, VBP, etc) do nothing to correct the warped data, in fact, I believe they will only warp things even more.
Data Nerd – I admire your courage in taking up the challenge :).
All the comments on this thread bring home a great point – data modeling for healthcare is hard and CMS should be applauded for taking baby steps in making more data available. Last I checked, no private insurer or provide has made downloadable spreadsheets of relevant data available on their sites.
Coming to the data set that you have proposed: I think aggregating DRG data by appropriate cohorts and layering it with outcomes and quality is a great idea. I still think that including procedures does not add any additional insight, especially in acute care settings. As long as a hospital has great outcomes for a condition and I can look at their average price for my cohort for treating that condition, it is not important to know the expected procedures.
The other important outcome of a data set like this would be from policy perspective.
Hopefully, CMS is reading this blog and are working on the next layer of data for us that can answer some of these questions for us.