IPM Take
Cardiology is entering an era of measurement overload. Every service line now has dashboards, quality indicators, productivity targets, readmission rates, length-of-stay goals, registry benchmarks and soon AI-generated operational alerts. Measurement is necessary. But measurement is not care. The policy danger is that health systems may reward what is easy to count, not what matters most: safer decisions, better recovery, equitable access, continuity, patient experience and long-term outcomes. AI will not solve this if it simply makes bad metrics faster.
Executive Summary
A Cardiovascular Business interview with Joel Sauer, executive vice president of consulting at MedAxiom, highlights a growing problem in cardiovascular care: quality metrics can become distorted when organisations treat them as targets rather than tools.
Sauer discussed Goodhart’s Law, often summarised as the idea that when a measure becomes a target, it can stop being a good measure. In cardiology, this can happen when hospitals focus too aggressively on reducing length of stay or readmissions without enough clinical context. A shorter hospital stay may look efficient, but premature discharge can increase risk. A lower readmission rate may look successful, but only if patients are not being held in emergency departments or observation status simply to avoid a formal readmission.
The article also points to physician compensation. Many compensation models rely heavily on work relative value units, or wRVUs, which reward clinical volume. If volume is the only thing that is paid and recognised, physicians may be discouraged from participating in quality improvement, care coordination, multidisciplinary meetings or administrative work that improves long-term performance but does not directly generate reimbursement.
AI and enterprise analytics could help cardiology teams move from raw data to actionable insight. But the same tools could also amplify the wrong incentives if they optimise for poorly chosen targets. The next challenge for cardiovascular leadership is not collecting more data. It is deciding which metrics deserve attention, which ones distort behaviour, and how to keep clinical judgment at the centre.
Why it matters
- Policymakers and payers: Value-based payment and quality reporting depend on metrics. Poorly designed measures can encourage gaming, misclassification or narrow care priorities.
- Hospital and cardiology leaders: Dashboards should support decisions, not replace them. Leaders need fewer, sharper and more clinically meaningful metrics.
- Clinicians: Quality measurement can help improve care, but only if it recognises clinical complexity, patient mix and work that does not generate immediate billing.
- Data and AI leaders: AI should not simply rank performance faster. It should help explain context, identify risk, reduce burden and support human decision-making.
- Patients and advocates: A hospital can look better on paper while patients experience poorer transitions, rushed discharge or fragmented follow-up. Patient outcomes must remain the endpoint.
Cardiology has a data problem. Not too little data. Too much of it.
Every modern cardiovascular service line now lives inside dashboards: length of stay, readmissions, wRVUs, imaging volumes, cath lab throughput, heart failure admissions, clinic access, referral leakage, registry performance, complications, patient satisfaction and revenue. Soon, AI will sit on top of all of it, promising to filter the noise.
That sounds like progress. It can be…but it can also become dangerous.
A Cardiovascular Business interview with Joel Sauer of MedAxiom puts the problem bluntly. Cardiology leaders need to be careful about the quality metrics they focus on, because once a measure becomes the target, it can stop serving the patient.
This is Goodhart’s Law in a hospital coat.
Length of stay is the obvious example. Reducing unnecessary hospital days is good. Nobody should remain admitted because the system is inefficient. But if leadership treats shorter length of stay as the primary goal, the incentive can shift from “discharge safely” to “discharge quickly.” The number improves. The patient may not.
Readmissions create the same trap. Reducing avoidable readmissions is a legitimate policy objective. It can reflect better discharge planning, medication reconciliation, follow-up and patient education. But if the target becomes avoiding the readmission label at all costs, organisations may be tempted to hold patients in emergency departments or observation status rather than admit them, even when admission may be clinically appropriate.
That is the dark side of metric-driven care: the number looks clean because the reality has been moved somewhere else.
The United States has made readmissions a major policy lever through the Medicare Hospital Readmissions Reduction Program. CMS describes the programme as a value-based purchasing model that links payment to quality and encourages hospitals to reduce avoidable readmissions through better communication, care coordination and discharge planning. The programme includes conditions and procedures highly relevant to cardiology, including acute myocardial infarction, heart failure and coronary artery bypass graft surgery.
The goal is reasonable. The implementation risk is real.
This does not mean metrics are bad. Cardiovascular medicine needs measurement. Without data, quality improvement becomes anecdote. Performance measures can expose variation, identify gaps and drive better care. ACC/AHA performance and quality measures for chronic coronary disease, for example, include practical measures around lipid management, high-intensity statin use, blood pressure management, tobacco intervention, cardiac rehabilitation referral and appropriate imaging.
Those are not bureaucratic details. They are core components of better cardiovascular care.
The problem starts when measurement becomes detached from meaning.
A dashboard may show that one cardiac imaging technologist completes fewer studies per day than colleagues. Without context, that looks like underperformance. With context, it may show that the technologist is handling the most complex cases, frailest patients or technically difficult scans. A crude productivity metric can punish expertise.
Physician compensation raises the stakes further. If cardiologists are rewarded mainly through wRVUs, the system is telling them what it values: billable volume. Yet cardiovascular quality also depends on work that is often invisible in productivity models: case conferences, protocol design, quality committees, peer review, patient communication, registry improvement, care coordination and mentoring.
If nobody pays for that work, leaders should not be surprised when it disappears. This is where AI enters the story.
Cardiovascular information systems, enterprise analytics platforms and AI tools could help organisations turn data into insight. AI might identify which metrics are drifting, where delays occur, which patients are at risk of decompensation, or where documentation and outcomes diverge. Used well, it could reduce dashboard fatigue and help clinicians focus on actionable signals.
Used badly, it could industrialise Goodhart’s Law.
An AI system trained to optimise length of stay may help discharge patients faster, but not necessarily better. A model designed to predict readmission risk may become a tool for coding, routing or avoidance rather than care improvement. An analytics platform built around wRVUs may reinforce volume-based medicine while pretending to support value-based care.
AI does not make a bad metric wise. It makes it scalable. The policy lesson is simple: every metric should be forced to answer three questions.
First, does this measure reflect something patients actually value?
Second, can clinicians act on it without harming something else?
Third, could this metric be gamed, misread or over-optimised?
If the answer to the third question is yes, the metric needs safeguards. It needs clinical review, balancing measures, equity checks and patient-centred outcomes.
For length of stay, the balancing measures might include readmissions, emergency visits, patient-reported recovery and post-discharge complications. For readmissions, they might include observation stays, emergency department boarding, mortality, access to follow-up and patient experience. For productivity, they might include quality participation, teamwork, teaching, equity work and care coordination.
Metrics should not be single levers. They should be part of a balanced system.
This is also a global issue. Health systems everywhere are moving toward value, quality reporting, benchmarking and AI-enabled management. The risk is that governments and payers will reward what is measurable because it is administratively convenient, not because it is clinically sufficient.
The National Academies warned years ago that performance measurement can create unintended consequences, including burden, misclassification, gaming and adverse selection. AHRQ’s patient safety work has also highlighted risks such as measure fixation, tunnel vision, misinterpretation and gaming when quality indicators are poorly implemented.
Cardiology should take that warning seriously.
The future of cardiovascular care will be data-rich. That is unavoidable. The question is whether it will also be clinically intelligent.
The best cardiology leaders will not be those with the biggest dashboards. They will be those who know which metrics to ignore, which to challenge, and which to connect back to the patient sitting in front of them.
The goal is not a prettier dashboard. The goal is better care.
