The Wisdom and Justice of Not Paying for “Preventable Complications”

Far too many patients experience preventable harm from medical care in US hospitals. To promote quality and safety, many employers and insurers are linking financial incentives to clinical performance. These programs, often called pay for performance, use a carrot (pay more for better quality) or a stick (pay less for lower quality). To date, most pay-for-performance programs have encouraged physicians to use evidence-based interventions or improve patient satisfaction.¹

The Centers for Medicare & Medicaid Services (CMS) has taken the lead, with many insurers following, in linking pay for performance to reducing harm.² In October 2008, hospitals will no longer derive additional payments they sometimes receive when Medicare patients develop 1 of the following 8 preventable complications: objects (such as surgical instruments or sponges) left in patients after surgery, hospital-acquired urinary tract infections, central line–associated bloodstream infections, administration of incompatible blood products, air embolism, patient falls, mediastinitis after cardiac surgery, and pressure ulcers. In addition, CMS has published that conditions being considered for 2009 expansion of the list include ventilator-associated pneumonia, Staphylococcus aureus septicemia, and deep venous thrombosis or pulmonary embolism.²

The tacit assumption to the “not paid for preventable complications” approach is that an error occurred in a patient's care that, if avoided, would have prevented the harm and ensuing costs. For one complication on the CMS list, foreign objects inadvertently left in patients after surgery, this is undeniably true.

Linking errors to harm for the remaining complications is more complex. For strategies built around the “not paid for preventable complications” concept to be clinically and morally acceptable and to achieve the policy goal of improving quality of care, it must be certain that preventable complications are important and measurable and truly are preventable. In this Commentary, we discuss the CMS initiative in the context of these metrics.

How various stakeholders perceive importance is likely a function of how commonly the event occurs and the amount of preventable harm and marginal costs associated with the event. Although an explicit process is needed to further define importance, all complications listed by CMS have face validity for being important.

The ability to validly measure each complication on the CMS list varies widely. Complications such as central line–associated bloodstream infection can be measured with reasonable accuracy. Accurately diagnosing other complications, such as air embolism, pulmonary embolism, deep venous thrombosis, and ventilator-associated pneumonia, which may be present on admission in some cases, is more difficult. Agreement among experts about these diagnoses is poor or unknown.^{3 - 5} Also, the different methods (medical record vs discharge data) of identifying events can dramatically influence performance,⁵ and aggressive screening strategies can markedly increase case-finding rates without an effect on quality of care. The absence of a standardized surveillance system could lead to a spurious association between more preventable harm and lower-quality care. For example, rates of postoperative deep venous thrombus increased 10-fold at one center when standardized duplex ultrasound screening was implemented for trauma patients for deep venous thrombosis.⁶

The CMS will use discharge data to identify patients who had 1 of the 8 listed preventable complications. Because the sensitivity and specificity of diagnosing many of these complications are imperfect or unknown and because the prevalence for many of these complications is low, measurement error will likely introduce substantial biases. Some patients experiencing the complication will not be diagnosed as having one (a false negative due to low sensitivity), and many more patients who do not have the complication will be diagnosed as having one (false positive due to low specificity).

For example, the most prevalent noninvasive method to diagnose ventilator-associated pneumonia⁴ is the Clinical Pulmonary Infection Score (CPIS), a valid tool based on clinical data.³ The CPIS has 79% sensitivity (percentage of patients with the complication that CPIS identifies as having ventilator-associated pneumonia) and 75% specificity (percentage of patients without the complication that CPIS identifies as not having ventilator-associated pneumonia).⁷

If 1 in 100 hospitalized patients truly develops ventilator-associated pneumonia, in a cohort of 10 000 patients, 100 will develop it and 9900 will not (1% prevalence).⁷ Using the CPIS (with 79% sensitivity), 79 patients will be correctly diagnosed as having ventilator-associated pneumonia (true positives), and 21 patients with it will be labeled as not having it (false negatives). With a 75% specificity, screening all mechanically ventilated patients with the CPIS will falsely label 2475 of the original 10 000 patients as having ventilator-associated pneumonia (9900 × [1−0.75]).

The extent to which hospitals treating these patients will have payment denied under the “not paid for preventable complications” system is uncertain. Withholding payment for patients with false-positive findings is just, but may be unwise because many patients will be falsely identified and subsequently treated for a disorder.^{8 - 9} Clinicians and hospitals will likely lose trust in the whole enterprise if they are falsely identified as having high adverse event rates. In addition, the CMS policy may preferentially target sicker patients, creating disincentives for hospitals to treat these patients. Finally, patients will be misled if hospital rates for these complications are publicly reported, which is likely.

The ability to accurately diagnose most of the CMS complications is poor, if even known, with false positives vastly outnumbering true positives. Thus, more aggressive screening for these complications will lead clinicians to treat many patients without the disease, thereby increasing health care costs and harming quality. Because those who look harder for complications will find more and because most of the diagnoses of complications will be false positives, complication rates will have substantial error, rendering them uninformative. Seen this way, the “not paid for preventable complications” program raises a fundamental question: what false-positive rate is clinically and morally acceptable?

Even when complications can be measured with reasonable accuracy, the degree of preventability also must be estimated accurately, which is no easy task.¹⁰ There is an active debate in the medical literature regarding the extent to which notions of preventability should adhere to traditional scientific standards.¹¹

Experience and face validity may be sufficient evidence of preventability for some rare events, but only if the risk for unintended consequences is small; the outcome easily measured; and the link between exposure and outcome is direct, unambiguous, and not influenced by bias or confounding. For example, for foreign objects retained after surgery, preventability is self-evident and strong scientific evidence not needed.¹²

However, in most health care, the link between an intervention and an outcome is less direct and more subject to bias. Unintended consequences are the norm rather than the exception. As such, it is risky and likely harmful to substantially relax evidentiary standards when making national patient safety policy.

Yet policy makers face a dilemma. Should they chance harming patients by promoting prevention interventions with little to no evidence or hold clinicians and systems accountable for complications that may not be preventable? The CMS program will inevitably be incorrect sometimes, leading it to deny payment for complications that are not preventable and pay for preventable complications. This raises another fundamental question: how much error is acceptable in the estimate of preventability?

One approach to this question is to use a measure of preventability that would evaluate 1 of 2 dimensions. First, a complication would be deemed preventable if a minimally biased effectiveness study with robust measurement from a variety of health care organizations demonstrates that most complications can be prevented. For example, published studies demonstrated that a series of achievable process changes can substantially reduce central line–associated bloodstream infections in a wide variety of hospitals.¹³ In pursuing this strategy, most must be defined and methods must be developed to grade the evidence of effectiveness studies.

Second, a complication would be deemed preventable if substantial variation in complication rates were found among hospitals, assuming those rates were accurately measured.¹⁴ Again, substantial variation must be defined. This method assumes that significant variation in complications rates are not from systematic error or random error, but rather are from differences in the quality of care (preventable complications).

The ability to prevent complications is a matter of degree rather than absolute; the proportion of a complication that must be prevented before a complication is labeled preventable is uncertain. It would be naive to set a threshold of 100% absolute risk reduction since this will rarely if ever be met. For example, although hospitals in Michigan substantially reduced the incidence of central line–associated bloodstream infection, those hospitals did not eliminate these infections.¹³ The few remaining infections were deemed nonpreventable because patients appeared to have received the evidence-based care designed to prevent central line–associated bloodstream infection. Whether these residual infections are preventable is uncertain.

For many complications on the CMS list, the degree of preventability raises concerns. For instance, preventability may differ among patients who developed a pulmonary embolism. One patient who received heparin for an appropriate indication developed heparin-induced thrombocytopenia complicated by deep venous thrombosis and pulmonary embolism. Another patient who had sustained a traumatic spinal cord injury could not receive anticoagulation and had an inferior vena cava filter placed but still developed a pulmonary embolism. These patients received optimal evidence-based care, yet each developed pulmonary embolism. It would be inappropriate to deny payment.

Policies to withhold payment for some complications, such as foreign objects retained after surgery, are likely wise (a reasonable method to decrease preventable complications) and just (true positives vastly exceed false positives). The policy is likely also wise and just for central line–associated bloodstream infections, although some risks remain. Hospitals typically use surveillance definitions that emphasize sensitivity over specificity rather than more accurate clinical definitions when diagnosing central line–associated bloodstream infections; the error introduced by using surveillance definitions is uncertain.¹³

What about the remainder of diagnoses on the initial “not paid for preventable complications” list? The ability to diagnose the remaining complications is error ridden and there is limited to no evidence about their degree of preventability. Although a payment policy is a blunt instrument and the true-positive/false-positive tradeoff may be an unavoidable adverse effect of a generally wise CMS policy (a few penalized for the greater good), diagnoses other than central line–associated bloodstream infection and retained foreign body should be pilot tested before widespread implementation.

There are broader concerns. Nonpayment for complications that are truly not preventable may destroy trust in quality improvement programs, reduce access for patients at-risk for these complications (eg, obese patients at increased risk for decubitus ulcers, deep venous thrombosis, and infections may be shunned), reduce the frequency of diagnosis after admission, and misinform the public when safety and quality results are publicly reported. Also, treating the large number of patients with a false-positive diagnosis identified on admission screening will decrease the quality of care and increase costs of care. The net effect could be greater harm.

The efforts of CMS to align payment incentives with the quality delivered by health care organizations, if done wisely, can stimulate broad improvements in quality and reduce medical costs.¹⁵ To reach the desired effect, complications in the “not paid for preventable complications” category must be important, measurable, and largely preventable. The CMS should start with central line–associated bloodstream infection and retained foreign body, evaluate the benefits and risks of this novel program, and invest resources to better understand how to diagnose and prevent the remaining complications. In the end, CMS must advance the science of quality improvement and measurement for its “not paid for preventable complications” program to realize true improvements in patient outcomes. There is no shortcut.