Everyone loves prevention. It may seem strange then, to learn that one of the biggest barriers keeping prevention from reaching its full potential is the current set of performance measures that, ironically, were created to promote them. The reason is that current measures are promoting activities that are inaccurate and inefficient. It is as though explorers who are trying to reach the North Pole have been given a compass that is sending them to Greenland.
This problem is being addressed by a new project conducted by NCQA and funded by the Robert Wood Johnson Foundation. The objective is to evaluate a new type of measure of healthcare quality called GCVR (Global Cardiovascular Risk). The new measure will have an important effect on the prevention of cardiovascular conditions.
To understand how, we need first to understand the limitations of current measures. For reasons that were appropriate when they were initially introduced – about 20 years ago — current performance measures were designed to be simple: simple to implement (e.g. collect the necessary data, do the calculations), and simple to remember and explain. This was accomplished in three main ways. One was to create separate performance measures for different risk factors. Thus there are separate measures for blood pressure control, cholesterol control, glucose control, tobacco use, and so forth.
While a performance measure for any one risk factor might take into account a few other risk factors to some extent, none of them incorporate all the relevant risk factors in a physiologically accurate way. A second simplification is that current measures are based on care processes and treatment goals for biomarkers, rather than on health outcomes. Thus a blood pressure measure asks if a patient with hypertension is controlled to a systolic pressure below 140 mmHG. A third simplification is the use of sharp cut points to determine the need for and success of treatment. For example, patients with hypertension are counted as properly treated if their systolic pressures are below 140 mmHG, otherwise not.
The simplifications may seem reasonable and innocent, but they are not. The focus on care processes and biomarker goals leaves providers, patients, and everyone else ignorant about the effects of different actions on the health outcomes people actually care about. There is no natural way to determine how “good” an improvement in quality is or make decisions based on the magnitude of the benefits. A related problem is that it is virtually impossible for providers, patients, and policymakers to determine how the different measures should be compared and prioritized. Lacking any better information, they usually consider the different measures to be equally important.
For example, the measures developed for Accountable Care Organizations are equally weighted. But in fact they are not equally important in terms of their effects on health outcomes. An analysis we did in collaboration with the American Diabetes Association and American Heart Association found more than tenfold differences in the effects of improvements in performance measures for cardio metabolic conditions[i]. Current measures also suffer from false positives, false negatives, and blind spots. For a false positive, take a person with systolic pressure equal to 142 mmHG, but otherwise no risk factors. The blood pressure performance measure will give full credit to reducing that person’s blood pressure to 138 mmHG (i.e., below the 140 mmHG cut point), even though the benefit of such treatment is very low. For a false negative, take a person who has a systolic pressure of 138 mmHG but lots of other risk factors. There is great value to reducing that person’s pressure, but the current performance measure will not recognize any value at all because the patient is already below the designated cut point of 140 mmHG. For a blind spot, take a person with a systolic pressure of 200 mmHG whose pressure can be safely brought to only 145 mmHG. Because the 140 mmHG cut point cannot be reached, the current performance measure will give no credit, even though the clinical value would be very high. Because of the simplifications current performance measures give providers quite inaccurate signals about how to prevent cardiovascular events in their patients. This both misses opportunities and wastes resources.
The Global Cardiovascular Risk score, which is based on the Archimedes GO Score[ii], was designed to address each of the limitations of current measures. First, it is based on health outcomes, not care processes or goals for treating biomarkers. Second, it encompasses all risk factors, outcomes, and treatments into a single measure. It avoids the false assumption that all risk factors and interventions are equally important, and instead gives each intervention credit (in calculating the GCVR score) in proportion to its actual effect on health outcomes. By being a single integrated measure, rather than directing providers to spend roughly equal time on dozens of specific care processes and treatment goals, GCVR recognizes that different interventions have different effects and leaves it to providers to determine the best way to allocate their resources to prevent bad outcomes.
GCVR also accurately registers the continuous nature of risk factors, giving credit to the treatment of each person in proportion to the actual reduction in risk achieved, not whether some artificial cut point was crossed. A final benefit is that the GCVR score has a very natural interpretation; it is the proportion of potentially preventable cardiovascular outcomes that are being prevented with current care. This greatly enhances its ability to help everyone get where they want to go.
Why should these changes in quality measurement make a difference? To switch analogies, imagine two chefs directed to prepare Julia Child’s recipe for Coq au Vin[iii]. One is told she will be judged on whether she used between 2½ to 3 pounds of frying chicken (score one point if she does), ¼ cup of cognac (one point), a bay leaf (one point), and had ≤ 35 minutes of cooking time (one point). Four points = top score. The other chef is told she will be judged on how good her Coq au Vin tastes to a panel of foodie diners. Which chef will prepare the better meal? Which method of measuring quality will detect the superior chef? Which method would you choose as the basis for rewarding your chef? This analogy has obvious flaws, but it captures the main differences between current performance measures and GCVR – a list of ingredients with artificial thresholds weighted equally, versus direct measurement of the quality of the actual outcome you care about, with the chefs providers left free to determine the best ways to accomplish it.
How much difference will GCVR make? That will depend on the populations and current levels of performance in different settings, but in one analysis we did that looked at only blood pressure and cholesterol treatments use of the GO score on which the GCVR is based would deliver more than 60% greater clinical benefit – that is, 60% more adverse events prevented – at the same level of resources as current performance measures[iv]. Not a bad return from merely giving providers a more accurate target to shoot for.
David M. Eddy, MD is the Founder and Chief Medical Officer Emeritus of Archimedes, San Francisco.