AJMC: In your first study, you used publishedMedicare costs and outcomes fromrecent stroke prevention trials to estimatethe costs of managing AF patients in variousways. Why did you do this study?Don't we already have good estimates ofthe national costs for stroke?
Dr Caro: Three reasons. The first reason wasthat cost studies have a shelf life of only 3years or so. Unlike randomized trials with scientificinformation, economic studies have nostaying power. The costs change with timebecause prices change and so does clinicalpractice, so they need constant updating. Thesecond reason was that past studies of strokehave not been done from the practical perspectiveof Medicare. Instead, they have beendone from the more theoretical societal perspective;so our studies were an attempt tofocus on items that are relevant to Medicare.A third reason for our studies was that neweconomic and clinical information on strokeprevention and treatment has become availablefrom newly published trials.
AJMC: How is figuring out the nationalexpenditure for stroke relevant to everydaymanagement of patients?
Dr Caro: Generally, cost studies are notaimed at practicing physicians worriedabout the patients in front of them.However, many physicians wear 2 hats andalso have population-level concerns andadministrative interests. The results of ourstudies are clearly most relevant for decisionmakers involved in Medicare and at thenational level. They should also have applicabilityto local decision makers in healthplans who might share some of thatMedicare perspective.
AJMC: You estimated that the annualdirect costs of AF-related stroke toMedicare was nearly $8 billion [Study 1Highlights]. Given that the AmericanHeart Association had previously estimatedall stroke costs [not just AF-related] at$33.1 billion, was that a surprisingresult?
Dr Caro: It may surprise many people. Adirect or "real" cost of this magnitude forwhat many consider to be a relatively smallmedical problem is impressive.
AJMC: Are these all hospital expenses? Ordo the costs include medications, physicians,nursing homes, or expenses borneby patients and families?
Dr Caro: These are the direct Medicare-billablecosts only. The $7.9 billion does notinclude out-of-pocket costs, outpatient medicationcosts, or costs to patients and theirfamilies. This is not because these costs areinsignificant, but because they're not part ofMedicare's remit. Also, these are costs toMedicare, not the charges that providersmay make.
AJMC: Is the whole $7.9 billion spent in1 year?
Dr Caro: No. We account for all the eventcosts, including hospitalizations, follow-up,and the initial rehabilitation and nursinghome care to the extent permitted by CMS[Centers for Medicare & Medicaid Services].That first-year cost is about $2.6 billion. Thecosts in subsequent years are then includedin terms of their "net present value." Inother words, the model takes the view thatonce a Medicare-eligible person has a stroke,Medicare has become liable for the care ofthat patient in the future, whether theyactually set aside that money today or not.The net present value accounts for thosefuture costs at a reduced rate constant withtheir occurrence in later years, which wecalculated to be about $5.4 billion. Ofcourse, those downstream costs are considerablyless than the real total costs becauseMedicare does not pay for long-term nursinghome care, or modifications to the home,or long-term rehabilitation, or the manyimplications to the family and caregivers. So,again, we've only measured a fraction of thetotal costs, the fraction that applies toMedicare.
AJMC: Since your study is based onMedicare data, how is it relevant to amanaged care commercial population?
Dr Caro: Generally speaking, Medicare datamay not be directly applicable to managedcare plans. In this case, our study is germanebecause AF is very much a disease of theolder Medicare-eligible population. To evaluatethese same economic questions from acommercial plan's point of view, however,you would need to account for the coveragethey provide these patients. You'd need toask: Are there carveouts? Are they in a managedMedicare plan? And so on.
AJMC: So, if you were a managed caremedical director, how would you interpretthese cost numbers?
Dr Caro: I would look at these numbers fortheir relative magnitudes rather than theirabsolute value. The bottom lines will differby plan, by composition of the population,and so on. The key idea is that stroke costs,no matter how you look at them, are a significantproportion of the total system costs.The related conclusion, based on the secondpart of our study, is that we can reduce thesecosts by optimizing management.
AJMC: Are you saying that stroke costs arehigh? Relative to what? Other diseases?
Dr Caro: Yes, the costs for treatment andfor prevention should be compared to thosefor other diseases. Unfortunately, most coststudies done today are isolated and do notdirectly compare the costs of different diseases.We really should be asking, "If I treatdisease X, what do I have to give up in treatmentof disease Y?" We don't do that now.We treat these studies as independent problems,when they are actually not independentat all.
AJMC: Do your cost calculations takeinto account the bleeding side effects ofwarfarin?
Dr Caro: Yes. In fact, in the treated populationsof this model, the costs of managingthe bleeds—both major and not so major—amount to a little over a third of the totalcosts.
AJMC: How did you decide on the model'sstroke rates for patients in anticoagulationclinics and those in routine medicalcare?
Dr Caro: The model assumes, somewhatgenerously, that anticoagulation clinics willachieve the same results as shown for warfarinin the recent large SPORTIF [StrokePrevention using Oral Thrombin Inhibitor inAtrial Fibrillation] clinical trials. I say theseresults are generous because clinical trialsgenerally provide more intensive managementthan you would see even in most anticoagulationclinic settings. For thewarfarin-treated population in routine care,we derive the rates of strokes by adjustingthe anticoagulation clinics' rates based onpublished analyses comparing clinics versusroutine practice.
AJMC: It's just a calculated rate? Not anactual rate measured in practice?
Dr Caro: Yes, it's calculated. Stroke rates inroutine practice are actually quite difficultto study because of what is known as theHawthorne effect. Once you study aprocess, it tends to change—the truly routineaspect is lost. People behave differentlywhen they know they are being studied.When doctors or nurses agree to participatein studies, they tend to bone up on theguidelines and see their patients moreoften. Patients are also more compliant. Inaddition, the people who agree to participatein studies tend to be different from theaverage patient. These effects are very hardto eliminate. The best we can do is look atdata collected for other purposes, where youcan assume that they reflect people behavingmore naturally. With these kinds ofdata, the treatment and event rates in routinepractice are usually quite poor. This isnot just an indictment of the routine practiceof anticoagulation for AF patients. It'salso true that the patients being seen inroutine practice tend to be sicker and morecomplicated or "messier" than those whoare eligible for clinical trials.
AJMC: Getting back to your model, howdo the stroke rates compare in these 2treated groups?
Dr Caro: The stroke rates in usual practicewere assumed to be about 2 1/2 times higherthan in the clinics. The rates were 1.8% inthe clinic patients and 4.5% in the usual caregroup.1 Again, the clinics were assumed tobe as effective as the SPORTIF warfarin trialsin preventing stroke, and usual practicewas assumed to be 2 1/2 times worse. That relativedifference is still considerably less thanhas been shown in some literature where itwas 10 times worse. In other words, this wasa somewhat conservative estimate of whatwould happen in actual practice.
AJMC: In your second study, you makethe statement that only 11% of the benefitof anticoagulation is currently beingobtained in stroke prevention. How didyou arrive at such a low percentage?
Dr Caro: We started with a theoretical maximumbenefit that could be derived fromanticoagulation based on the efficacy ratesseen in the SPORTIF trials. We asked ourselves,"What would happen if the total efficacyreported in the trials could manifestitself in real-world clinical practice?" Thenwe compared that ideal with how we are currentlyoperating, and we found that onlyabout one tenth of that theoretical maximumbenefit is currently being achieved.
AJMC: Doesn't that percentage vary fromhealth system to health system?
Dr Caro: Yes. A top-notch clinic in a carefullymanaged and selected population isprobably operating at 60% or 70% of the theoreticalmaximum benefit. This also impliesthat some clinics are delivering essentiallynone of the potential benefit.
AJMC: In your models, you estimated thehealth and cost benefits of shifting AFpatient populations from no preventivetreatment to warfarin treatment or fromwarfarin usual care to warfarin in theanticoagulation clinic setting. In the second"What if…" scenario, you estimatesavings of $1.3 billion by shifting half ofall usual or routine care patients intooptimal anticoagulation therapy [Study 2Highlights]. Is it really practical to assumethat healthcare providers can shift 50% oftheir patients into optimal treatment?
Dr Caro: In general, no. Under conditionsexisting today, with our current treatments,an assumption that 50% of patients can beshifted to optimized treatment is unrealistic.We've tried for 20 years to put more peopleon warfarin after the major clinical trialsproved beyond a doubt that this drug worksand yet still have half the population or morenot receiving a benefit. It's also overly optimisticbecause we know many people simplycannot be treated because of contraindications,living in nursing homes where nobodyfollows them, or a variety of other reasons.So, on a countrywide basis, the "what if" of50% is very optimistic. However, in individualmanaged care plans, or in settings startingfrom a very low level of care, it mayactually be feasible. And, as new treatmentscome on line, if we can avoid some of theissues with warfarin, we should be able toachieve that level of optimized anticoagulationfor stroke prevention.
AJMC: In your second current scenario,why are the total costs of stroke for the38 000 patients who received anticoagulationactually higher than the totalcosts for the 58 000 stroke patients incurrent scenario number 1 who did notreceive any anticoagulation?
Dr Caro: Because those costs shown in thestudy highlights are for the scenario of carenot just for managing the strokes. Theyinclude costs for other emboli (renal andsystemic), for the visits and monitoringrequired with warfarin, and also for managingthe treatment-related bleeds. In thetreated population, the costs are higherbecause you need to pay for all 3 components.For example, the costs for just theemboli themselves are $2.6 billion in thetreated population and $4 billion in theuntreated group. This is a reminder thattreatments are not free, especially a treatmentwith fairly severe side effects.
AJMC: How much were the bleedingcosts?
Dr Caro: The annual bleed costs per yearwith warfarin are about $2 billion, a significantpart of the total. This is a reflectionof how expensive it is to treat a majorbleed. Acutely, it's more expensive thanstroke treatment, which basically involvessome initial imaging followed by supportivecare. A gastrointestinal bleed can requireintensive care units, surgery, and bloodtransfusions.
AJMC: You estimate $1.3 billion in savingsby shifting half of patients currentlytaking warfarin into a setting of anticoagulationclinics. Where do these savingscome from?
Dr Caro: The economics here are driven bythe fact that the clinic shifts 2 things in theproper direction by keeping the anticoagulationlevels within a tighter range: first, itdecreases the number of bleeds, and second,it decreases the rate of strokes. By avoidingthe high and low INRs [international normalizedratios], the bleed and stroke costsdrop. These avoided costs exceed the costsof the management itself if you ignore theoperational and capital costs [ie, personnel,buildings, equipment, furniture, etc] thatwould be required to set up and run the newclinics.
AJMC: Why didn't you account for thecosts of optimization in your study?
Dr Caro: Because Medicare would not paythe capital costs needed to establish newclinics. They might pay for the monitoringwithin an established clinic, but they wouldnot invest to build the clinic. If providersthemselves set up the new clinic, they wouldhave to absorb those costs. In fact, that iswhat happens in most health systems today;physicians simply set up the clinic as anextra service, essentially a loss leader, inorder to manage more patients.
AJMC: How will the availability of a newanticoagulant such as an oral directthrombin inhibitor impact your economicmodel?
Dr Caro: It would be very easy to include thisdrug in a model and we've actually done somepreliminary modeling based on SPORTIF IIIand V trials. However, we still don't know theprice of the product and we don't know howit will behave in actual practice. A lot willdepend on which patients get the new drugfirst. For example, if the best patients—thatis, the low-bleed risk patients and those whoare already in clinics or are meticulous abouttaking their drugs regularly—are put on thedrug first, then we would not expect to seemuch savings because they already do well onwarfarin. If these "good" patients are theearly adopters, the benefit needle won't shiftmuch. But if people who are currently poorlyoptimized begin to adopt it, then we'll see bigshifts in the benefit.
AJMC: Won't reducing or eliminating theneed for anticoagulation clinics also shiftthe overall cost benefit?
Dr Caro: That depends very much on yourpoint of view. For a health system or medicalgroup that currently pays for anticoagulationclinics, cutting back on these willlikely allow redeployment of theseresources and pharmacists into other areasof prevention or education. For Medicare,however, dropping clinics probably won'taffect overall costs either way since they'renot paying capital costs to the clinics. Themain advantage to Medicare of an effectivenonwarfarin treatment is improved qualityof AF care. Since anticoagulation clinic supplyis limited and since they tend to be concentratedin large urban centers that missmuch of the population, a new treatmentmight optimize management in a more feasibleway than by expanding clinics. Thisoptimization will increase the overall qualityof care.
The American Journal of Managed Care
Research conducted by J. Jaime Caro, MDCM, focuses on the economic costs of stroke in patients withatrial fibrillation (AF), and how much can be saved by optimizing anticoagulation in these patients. Hisresearch will have practical implications not only for researchers but also for the clinicians and decisionmakers within managed care settings who must allocate resources for competing prevention programs. Apanelist at the roundtable organized by , Dr Caro recently reportedpreliminary results from his economic analyses at the 2004 World Stroke Congress in Vancouver, BC.He spoke with an editor from about those studies.
Arch Intern Med.
1. Chiquette E, Amato MG, Bussey HI. Comparison ofan anticoagulation clinic with usual medical care anticoagulationcontrol, patient outcomes, and health carecost. 1998;158:1641-1647.