For most Americans, third-party insurance companies reimburse health care providers eased on the diagnosis-related group codes (DRGs) originally established by Medicare. When Medicare changes policies, it is likely that private health insurance companies will follow. Recently, the Centers for Medicare & Medicaid Services announced a policy change on reimbursements for “hospital-acquired conditions” — a bureaucratic term for medical errors.
Until now, a hospital could be reimbursed for the costs of treating Medicare patients will ailments resulting from hospital mistakes. Under the new regulation, if a surgeon amputates the wrong limb or leaves an object behind after surgery, or hospital staff administer an incompatible blood type or permit an air embolism, CMS will not reimburse the added cost of subsequent treatment. The policy change is intended to give health care providers a financial incentive to avoid committing these errors.
In our discussion below, we analyze one specific medical error of special interest: mediastinitis.
Third-party health insurance is riddled with perverse incentives that distort both the demand and the supply for heath care. Consumers overutilize medical services because they do not pay the full price. Health care providers deliver more (and more expensive) services than consumers otherwise would buy because someone else — a third-party insurer — is paying almost the entire bill. Insurers (both private and public) try to rein in both. For consumers, they increase co-payments, require pre-approval, or deny coverage. For health care providers, they have adopted various prospective payment regimes in which they provide a fixed dollar amount for each DRG, and it’s up to providers to work within that cap. These systems lead both consumers and providers to engage in new rounds of strategic behavior to maximize coverage and third-party reimbursements.
One problem with third-party reimbursement in which incentives can be improved without leading to adaptive, strategic behavior concerns “hospital-acquired conditions.” A “hospital-acquired condition” is bureaucratic language for medical error. By reimbursing hospitals for the costs of medical error, insurers implicitly reward them for committing error.
On August 1, 2007, the Centers for Medicare and Medicaid Services announced a final regulation that stops reimbursing hospitals for certain medical errors. The relevant section of the CMS announcement summarizes why it took any action to stop reimbursing for medical errors (Congress required it to do so) and hints at why it chose which specific errors to target. Unsurprisingly, many health care providers submitted public comments objecting to items on the list. However, several hospital associations suggested that CMS include mediastinitis as a non-reimbursable medical error, and CMS agreed.
WHAT IS MEDIASTINITIS?
Mediastinitis is a life-threatening infection with a mortality rate estimated at 19-47%. It is considered a preventable medical error when it occurs subsequent to coronary artery bypass graft surgery (CABG), in which the sternum is bisected to permit surgical entry to the thoracic cavity. By denying reimbursement, CMS intends to create a financial incentive for hospitals to improve quality.
The American College of Cardiology has published a scoring system for assessing the risk of mediastinitis (first link, PDF Table 2). Based on the patient’s score, the risk of mediastinitis ranges from 0.3% (score = 0) to > 7.3% (score greater than or equal to 12). Factors that increase a patient’s score are advanced age, obesity, chronic obstructive pulmonary disease, diabetes and dialysis, high left ventricular ejection fraction, and surgical urgency (a proxy for other aspects of coronary condition). Pre-operative infection risk is less than 1% until the score exceeds 5 points. In addition, the risk of contracting mediastinitis is greater when both internal mammary arteries are used for grafts (p. 12).
Scores are derived from an evidence-based statistical model (p. 10):
The regression models that generated the scores for these prediction rules were based on 14,971 patients receiving isolated CABG surgery between 1999 and 2002. The dependent variables and observed event rates are as follows: in hospital mortality (2.5%); cerebrovascular accident, defined as a new focal neurologic event persisting at least 24 hours (1.6%); and mediastinitis during the index admission defined by positive deep culture and/or gram stain and/or radiographic findings indicating infection and requiring re-operation (1.0%).
ACC reports that the mortality rate from mediastinitis is about 25% (p. 12). It can be treated by immediate surgical debridement (removal of the portion of the sternum and nearby tissue that is infected) or sternectomy (removal of the sternum). Surgical debridement is a debilitating procedure with a months-long recovery period, but if successful, it permits the patient to resume a generally normal life. Sternectomy is permanently disabling.
ACC lists numerous ways to reduce the risk of post-operative infection; some are hospital practices but others speak to the quality of the thoracic surgeon, but all are within the control of the community of health care providers. It is important to remember that half of all surgeons who perform CABG surgery and half of all hospitals at which CABG is performed are below average in quality. Hospitals have excellent data on surgeons’ performance, and surgeons are similarly well-informed concerning practices at hospitals where they perform surgery. Thus, both hospitals and surgeons have considerable ability to influence each others’ conduct.
MARKET FAILURE: ASYMMETRIC INFORMATION
This is not true for patients, however. Patients generally do not have access to full and accurate information about both hospital and surgeon quality, and thus they are unable to make informed choices. It is infeasible for some patients to exercise informed choice because of the urgency of some CABG procedures.
The CABG market thus displays the classical symptoms of market failure due to asymmetric information. Hospitals and surgeons have excellent information about their own risks and can readily assess the pre-operative risk of every patient they serve. Patients, however, have little or no information about hospital and surgeon quality. Increasingly, hospitals in major urban areas are advertising to attract customers because, at least in theory, patients have many choices. The typical advertising pitch attempts to steer prospective patients to captive surgical teams, which then automatically secure the business for their hospitals. Both surgeons and hospitals make quality claims that are unvalidated, valid on average but not necessarily valid for them, or inherently impossible to validate. Washington Hospital Center’s heart unit is typical:
- Unvalidated: “We have an unrivaled research capability.”
- Valid on average but not necessarily valid for WHC: “Studies in the New England Journal of Medicine confirm that hospitals with experience – high volumes of given procedures – have better patient outcomes.”
- Inherently impossible to validate: “a national reputation for leadership, excellence, superior quality and compassionate care.”
- Outputs, not outcomes: “We perform more than 1,750 coronary artery bypass operations per year, with more than 50 percent performed on the beating heart, a significantly higher percentage than other hospitals in the region.”
WHC publishes on-line a 2004 report titled “Cardiac Surgery Outcomes” (last link, PDF), but despite the title most of these data concern inputs and outputs and not outcomes. The report says the overall mortality rate in heart surgery was 5%, and 0.7% for elective CABG procedures. The report is silent concerning the incidence of mediastinitis, or any other kind of medical error.
CAN THIS MARKET FAILURE BY REMEDIED?
In principle, it could be easily remedied by requiring full disclosure of medical errors and other attributes of hospital and surgeon performance prior to securing patient consent. These data could be disclosed for all to see on the hospitals’ and surgeons’ websites. Hospitals and surgeons treat a higher-risk mix of patients would be expected to have a higher incidence of operative mortality and post-operative infection, but data on pre-operative risk also can be disclosed.
Because insurance reimbursements are the same regardless of quality, non-emergency patients would gravitate toward hospitals and surgeons that have better outcomes and not just better marketing. This would put enormous pressure on lower-quality hospitals and surgeons to either improve or exit the market. As lower-quality providers exit the market, even patients requiring emergency CABG surgery would benefit.
Patients can rely on advice from cardiologists, and this may be better than having no information at all, but cardiologists also lack access to validated information about hospital and surgeon performance. Thus, what cardiologists possess is largely anecdotal and limited to their own experience at hospitals where they have admitting privileges.
THE LIMITS OF WHAT REGULATION CAN ACCOMPLISH
An important caveat is that hospitals and surgeons must disclose valid information. Surgeons and hospitals alike have incentives to misclassify medical errors to improve how their performance appears. By denying reimbursement for medical errors such as mediastinitis, CMS may have created an incentive for surgeons and hospitals to misclassify patients with this condition into reimbursable DRGs. Thus, if the reported incidence of mediastinitis declines, it could be attributable to either improved quality or increased data misclassification.
Patients with mediastinitis are expensive to treat, and treatment must be undertaken immediately to be successful. Thus, the new regulation creates an even more perverse incentive for surgeons and hospitals: delay treatment so that more patients die.
As it implements the new regulation, CMS should pay careful attention to these (and probably other more subtle) adaptive responses.
Neutral Source managing editor Richard Belzer had 2-vessel CABG surgery at Washington Hospital Center on October 4, 1999. performed by heart surgeon Dr. Ammar S. Bafi. He was discharged on October 8. On October 15, he experienced a sudden high fever, which was immediately reported to the WHC heart surgery team. On October 16, he was advised by heart surgeon Dr. Anjum G. Qazi to treat the fever with ibuprofen. On October 17, he was readmitted to WHC. On October 19 and 21, he underwent sternal debridement surgery for mediastinitis performed by Dr. Susan Otero, and was maintained in a coma until approximately October 28. He was discharged on November 2. The infection recurred in February 2000, at which time Dr. Otero performed two additional sternal debridement surgeries.
Upon readmission to WHC on October 17, he was diagnosed by Dr. Qazi with an unspecified infection, not mediastinitis subsequent to CABG surgery performed at WHC by his colleague, Dr. Bafi. Misclassification reduced the apparent incidence of mediastinitis attributable to WHC, Dr. Bafi, and the heart surgery group employed at WHC
If misclassification is random, it has no effect on the incidence of mediastinitis of a surgeon, a group of affiliated surgeons, or a hospital. However, if misclassification is systematic — that is, if admitting physicians have a propensity to misclassify certain patients with post-operative mediastinitis, such as those patients whose infections were caused by the admitting physician or a colleague — then the reported rate of infection is biased downward. Surgeons in the same group are especially vulnerable to the temptation to cover up each others’ errors.
This problem could be significantly attenuated if hospitals did not permit physicians potentially responsible for medical errors to make subsequent diagnoses that might implicate them or their colleagues. In any case where a medical error is plausible, hospitals should require that these diagnoses to be made by the most disinterested qualified physician rather than the attending physician, who is almost always conflicted.
As CMS begins to use denial of Medicare reimbursement as a financial incentive to improve quality. it should take seriously the risk that hospitals and attending physicians, who have inherent conflicts of interest in the diagnosis of medical error, will adapt to the new rules by misclassifying patients or even allowing them to die.
Section 5001(c) of the [Deficit Reduction Act of 2005] requires hospitals to begin reporting on October 1, 2007 the secondary diagnoses that are present on the admission (POA) of patients. By October 1, 2007, the Secretary must select at least two conditions that are: (1) high cost or high volume or both; (2) assigned to a higher paying DRG when present as a secondary diagnosis; and (3) reasonably preventable through application of evidence-based guidelines. Beginning October 1, 2008, cases with these conditions would not be assigned to a higher paying DRG unless they were present on admission.
In the proposed rule, CMS considered 13 and proposed to select 6 conditions that would be subject to the provision. While there were some public comments that said CMS should only select serious preventable events (Object left in surgery, blood incompatibility and air embolism), CMS believes there is a significant public health interest in selecting more. Based on public comments, CMS decided not to select 1 of the original six conditions (septicemia) but added an additional 2 (falls and mediastinitis—a preventable surgical site infection that follows heart surgery). In addition, CMS selected catheter associated urinary tract infections, pressure ulcers and vascular catheter associated infections. CMS also indicated that it will work to create a code to identify ventilator associated pneumonia as well as determine when septicemia and deep vein thrombosis are not present on admission and preventable in the hospital so that these conditions may be included in the future.
Section V. Management Strategies: Reduction of Perioperative Mortality and Morbidity
E. Reducing the Risk of Perioperative Infections
Several methods exist to reduce the risk of wound infections in patients undergoing CABG. These include the following:
- Interval reporting to individual surgeons regarding their respective wound-infection rates
- Adherence to sterile operative technique
- Skin preparation with topical antiseptics
- Clipping rather than shaving the skin
- Avoidance of hair removal
- Reduction of operating room traffic
- Laminar-flow ventilation
- Shorter operations
- Minimization of electrocautery
- Avoidance of bone wax
- Use of double-glove barrier techniques for the operating room team
- Routine use of a pleural pericardial flap
- Aggressive perioperative glucose control in patients with diabetes through the use of continuous intravenous insulin infusion (reduces perioperative hyperglycemia and its associated infection risk)
- Avoidance of homologous blood transfusions after CABG may reduce the risk of both viral and bacterial infections (this is due to an immunosuppressive effect of transfusion). Leukodepletion of transfused blood also reduces this effect. This can be accomplished by regional blood blanks at the time of donation or at the bedside by use of a transfusion filter.
Preoperative antibiotic administration reduces by 5-fold the risk of postoperative infection. Efficacy depends on adequate drug tissue levels before microbial exposure. Cephalosporins are the agents of choice. Table 7 identifies appropriate choices, doses, and routes of therapy. A 1-day course of intravenous antimicrobials is as effective as 48 hours or more. Therapy should be administered within 30 minutes of incision and again in the operating room if the operation exceeds 3 hours.
If deep sternal wound infection does occur, aggressive surgical debridement and early vascularized muscle flap coverage are the most effective methods for treatment, along with long-term systemic antibiotics.