What Is the Karnofsky Score and How Does It Work?

The Karnofsky Performance Status score, usually called the KPS, is a clinician-assigned rating from 0 to 100 that measures how well a person can carry out everyday activities while living with a serious illness. Developed in the late 1940s during early chemotherapy trials for lung cancer, it has become one of the most widely used tools in oncology and palliative care for gauging a patient’s functional ability, guiding treatment decisions, and estimating prognosis.¹ The score looks deceptively simple, but the way it is assigned, interpreted, and acted on involves real complexity that affects patients directly.

How the Scale Works

The KPS runs in increments of 10 from 100 (no complaints, no evidence of disease) down to 0 (dead). Each 10-point step maps to a short description of what the patient can and cannot do. At 80 or 90, you can carry out normal activities with some effort or minor symptoms. At 70, you can care for yourself but cannot carry on normal activity or do active work. At 50 or 60, you need varying amounts of help with personal care. Below 40, you are severely disabled and likely need hospital-level support. The descriptions are intentionally broad, which makes the tool quick to use but also introduces room for disagreement about where a given patient falls.

The scale groups loosely into three tiers. Scores of 80 to 100 generally mean you can function independently. Scores of 50 to 70 indicate you need some assistance but are not bedridden. Scores below 50 suggest you are largely unable to care for yourself. These thresholds come up constantly in clinical decision-making, particularly around whether someone is a candidate for aggressive treatment.

Why the Score Matters for Treatment Decisions

In oncology, the KPS often acts as a gatekeeper. A review of 96 oncology clinical trials found that roughly 40% set restrictive thresholds, enrolling only patients with a KPS of 70 or higher (or the equivalent on the ECOG scale). Just 20% of those studies included patients with lower functional status.² That means a large fraction of people living with cancer are excluded from the very trials that might offer them new options, simply because their score falls below an arbitrary cutoff.

The rationale is understandable: patients with poor functional status tend to tolerate aggressive treatments less well, and including them can muddy a trial’s results. But the consequence is that the evidence base is built almost entirely around higher-functioning patients, leaving clinicians with less data when treating people who score lower. In brain metastases, for example, standard prognostic tools treat a KPS below 70 as a marker of poor prognosis with expected survival under three months. Yet recent research suggests that surgical resection can still improve functional status and facilitate further treatment even in patients with low KPS, challenging the assumption that a low score should automatically rule out intervention.³

How Well the Score Predicts Outcomes

The KPS does carry real prognostic weight. Early research established that a patient’s score at the time of evaluation accurately predicted early death, though a high initial score did not necessarily guarantee long survival.⁴ In other words, the scale is better at identifying who is doing badly right now than at forecasting the future for someone who currently appears well. A rapidly dropping score, however, was found to predict deterioration and death within months.

In more specialized contexts, the picture gets nuanced. In allogeneic stem cell transplants, for instance, the KPS is an independent predictor of overall mortality, but a comorbidity index designed specifically for transplant patients showed significantly greater predictive power for complications and non-relapse death.⁵ The KPS captures something real about how a patient is doing, but it is a blunt instrument that does not account for the specific medical baggage a person carries. A patient with a KPS of 80 who has diabetes, kidney problems, and a history of lung disease is in a very different position from a KPS 80 patient without those conditions, even though the number is the same.

The Problem of Who Assigns the Score

One of the most underappreciated issues with the KPS is that the person doing the scoring changes the result. In a study of over 450 palliative care outpatients, weighted agreement between nurses and physicians was moderate to good, with a weighted kappa of 0.74 for the KPS. But physicians consistently rated patients as functioning better than nurses did on related scales.⁶ That optimism bias among physicians has turned up repeatedly in the literature.

A separate study comparing multiple observers found that the exact agreement rate for the KPS was only about 49%, meaning that in roughly half of cases, two clinicians looking at the same patient picked different scores. When the analysis accounted for how close the disagreements were (using weighted statistics), agreement improved to “substantial,” but the raw disagreement rate underscores how subjective the process can be.⁷ The ECOG performance status scale, which uses a simpler 0-to-4 range, showed somewhat better raw agreement in that same study, likely because its fewer categories leave less room for splitting hairs.

When Patients and Doctors Disagree

The gap between clinician ratings and patient self-ratings is an even bigger issue, and it has real consequences. In a study of patients with advanced cancer, physicians and patients disagreed on the KPS about two-thirds of the time. Physicians tended to rate patients as functioning better than patients rated themselves. Critically, this disagreement was not just academic: when physician and patient scores diverged, patients had a higher risk of death, with an adjusted hazard ratio of 1.16 for KPS disagreement.⁸

Why would disagreement itself predict worse outcomes? The most likely explanation is that the disagreement flags something the clinician’s quick assessment misses. Patients who report worse function than their doctor perceives may be experiencing symptoms, fatigue, or declines that are not yet visible in a clinic visit. The physician’s rating is based on observation and conversation during a brief encounter; the patient’s rating reflects what life actually feels like day to day.

Research in brain metastasis patients reinforced this finding from a different angle, showing only modest correlation between physician-reported KPS scores and patient-reported functional scores, with a correlation coefficient of just 0.46.⁹ That level of agreement is essentially saying the two ratings share less than a quarter of their variance. If a patient tells you they are struggling more than their KPS suggests, the patient’s perspective deserves serious weight.

KPS Versus the ECOG Scale

The other major performance status tool you will encounter is the ECOG (Eastern Cooperative Oncology Group) scale, which runs from 0 (fully active) to 4 (completely bedridden). Because both scales attempt to measure the same thing, clinicians and researchers sometimes try to convert between them. The two scales are highly correlated at the population level, but translating an individual patient’s KPS into an ECOG grade introduces substantial error, especially at lower functional levels where the spread between possible conversions widens.¹⁰

In practice, most oncology trials specify which scale to use. The ECOG scale dominates cooperative group trials in the United States, while the KPS remains the standard in neuro-oncology and many European protocols. For older cancer patients, both scales showed similar ability to screen for patients who might benefit from a comprehensive geriatric assessment, with comparable discriminatory power.¹¹ The choice between them often comes down to institutional tradition and the specific clinical context more than any inherent superiority of one over the other.

Uses Beyond Cancer

Although the KPS was born in oncology, its utility has spread to other fields where functional status matters for prognosis and decision-making. In cirrhosis, a study of nearly a thousand patients found that KPS measured shortly after hospital discharge was a strong predictor of three-month mortality. Patients in the lowest KPS category (scores of 10 to 40) had a mortality rate of 23%, compared to 11% for intermediate scores and 5% for scores of 80 to 100. A model combining KPS with age and the standard liver disease severity score outperformed the liver score alone.¹² The finding makes intuitive sense: how well a person functions at home captures something about their reserve that lab values and imaging cannot fully express.

In kidney transplantation, lower KPS is independently associated with a higher risk of death or delisting while on the waiting list. The effect is dramatic at the low end of the scale: patients with severely impaired performance status had nearly five times the risk of death or delisting compared to those with no impairment. This association held across all age groups, though it was slightly less pronounced in patients 65 and older.¹³

In geriatrics more broadly, the KPS has shown value as a proxy for overall health status in outpatient settings. One study of 134 geriatric patients found that the KPS correlated strongly with standard geriatric instruments measuring activities of daily living, and it actually performed as well or better at predicting hospitalizations, survival, and institutionalization.¹⁴ For a single number assigned in a few minutes, that is notable predictive power.

Derived Scales and Adaptations

The limitations of the KPS in palliative care led to the development of the Palliative Performance Scale (PPS), which modifies the KPS framework to better suit patients approaching end of life. The PPS adds explicit attention to oral intake, level of consciousness, and the specific type of activity a patient can manage, making it more descriptive for clinicians working in hospice and palliative settings.¹⁵ It retains the 0-to-100 structure and 10-point increments, so it feels familiar to anyone used to the KPS, but its criteria are more specific at the lower end of the scale where palliative decisions concentrate.

Other research groups have developed flowchart-based tools designed to improve agreement between raters. One such tool, the Functionality Assessment Flowchart, achieved substantially higher inter-observer agreement than either the KPS or the ECOG scale by walking clinicians through a structured series of yes-or-no questions rather than leaving them to choose a category based on a paragraph description.¹⁶ These efforts recognize that the KPS’s biggest weakness is not the concept behind it but the looseness of how scores are assigned.

What a Changing Score Means

A single KPS snapshot tells you something useful, but the trajectory of the score over time can tell you more. Early work showed that a rapidly declining score predicted death within a few months, even when the starting score was reasonably high.¹⁷ This makes the KPS more valuable as a longitudinal tool than as a one-time measurement. If your score drops by 20 points between visits, that signals a shift in your disease course that deserves attention regardless of where the score started.

Tracking this trajectory in busy clinics has its own challenges. Longitudinal patient-reported performance assessment has been shown to be feasible in cancer clinics and to correlate with survival, with changes in patient-reported impaired performance carrying significant prognostic weight.¹⁸ The practical takeaway: if your care team is not tracking your functional status over time and only recording a single number at enrollment, they may be missing meaningful changes.

Automating the Process

Extracting KPS data from medical records for research purposes is notoriously tedious. Performance status is often buried in clinic notes rather than stored as a discrete, searchable data point. Recent work has explored using optical character recognition combined with natural language processing to pull KPS values from scanned documents automatically. One study found that this approach achieved excellent accuracy with less than 1% incorrect results, while cutting the median time needed to review documents by a third for records that contained performance status data. For records that turned out not to contain performance status information at all, a situation researchers frequently encounter, the time savings were even more dramatic: a median of 18 minutes versus 108 minutes for manual review.¹⁹

These tools are primarily useful on the research and administrative side, not for assigning scores in the first place. But they point to a broader trend of making performance status data easier to collect, track, and analyze at scale, which could eventually improve how the score is used in clinical decision-making.

What Patients and Families Should Know

If you or a family member is living with cancer or another serious illness, the KPS is likely influencing your care in ways that are not always made explicit. It may determine whether you are eligible for a clinical trial, whether your oncologist recommends aggressive treatment or transitions to palliative approaches, and how your prognosis is discussed. A few things are worth keeping in mind.

First, the score is subjective. Two clinicians can look at the same patient and arrive at different numbers, and your own sense of how you are functioning may differ from what your doctor records. If you feel your score does not reflect your reality, say so. The research on patient-physician disagreement suggests that your self-assessment carries independent prognostic information that the clinician’s rating alone may not capture.²⁰

Second, a low score is not necessarily a sentence. The cutoff of 70 is used widely as a dividing line between “eligible for treatment” and “not eligible,” but evidence from surgical cohorts and other settings suggests that some patients below that threshold can still benefit from intervention. If a treatment option is being withheld solely because of your KPS, it is reasonable to ask whether that cutoff is based on evidence specific to your situation or on a convention carried over from clinical trial design.

Third, trends matter more than single numbers. A stable KPS of 60 means something very different from a KPS of 60 that was 80 three months ago. Ask your care team how your score has changed over time, not just what it is today. That trajectory often tells a more honest story about where things are heading than any individual rating can.