“Measurement matters, but only when it changes decisions.”
Patient-reported outcome science has spent three decades proving that what patients experience is as important as what clinicians observe. Mark Kosinski, Senior Principal, Patient-Centered Solutions at IQVIA, has been involved in that work since its earliest days. His argument now is that steroid-toxicity measurement belongs in exactly the same framework, not as an add-on, but as a core component of understanding health risk.
“We’re always trying to better understand the impact of disease and its treatment on patient functional health and well-being. Steroid-toxicity is just another component of that. It adds another layer of understanding, and it can be an impediment to improvement from the primary treatment being delivered.”
The clinical stakes are not abstract. Kosinski points to oncology patients receiving steroids as first-line preparation for toxic cancer therapies. Patients and providers may not be aware that mounting steroid-toxicity can actively undermine the effectiveness of the cancer treatment itself.
“If they go blind to the risk associated with toxicity, they’re just going to keep on feeding steroids to patients. That is not going to have a good outcome.”
From scores to decisions: the data infrastructure problem
A patient-reported outcome score, or a steroid-toxicity assessment, has no intrinsic meaning on its own. Kosinski is direct on this point: the scientific value only materializes when the data infrastructure exists to translate scores into interpretable risk signals.
His team at IQVIA builds what he calls reference and benchmark datasets, analogous to the normal ranges printed on a blood glucose report. When a patient’s score falls outside the expected range for their age, gender, and disease, the system flags it. Linked to clinical and claims data, those scores then carry predictive weight: elevated risk of hospitalization, emergency care, or loss of employment.
“We program these interpretations into our assessment system so that when an individual comes into a provider’s office and takes the assessment, there is an immediate feedback report given to both provider and patient - here is what the data are telling you, you are at high risk in certain areas.”
This is the mechanism through which shared decision-making becomes real rather than aspirational. Without actionable interpretation at the point of care, measurement produces numbers. With it, measurement produces conversations.
Population health: payers follow the data
The same logic scales upward. Kosinski describes two landmark examples of data-driven population health management in the US.
Following a landmark JAMA study showing worse four-year outcomes for vulnerable populations in managed care — the chronically ill, the elderly, and the poor — the US government implemented the Medicare Health Outcomes Survey. Managed care plans are now assessed under a “stars” program: those that effectively manage Medicare recipient outcomes receive financial reward.
In California, a major managed care organization layered patient-reported outcomes over clinical and claims data to stratify risk in its dual-eligible population. The combination identified who was genuinely high risk versus those who had simply been heavy past users of services. The highest-risk group was directed to specialist disease management, saving the plan money and improving care quality.
“At a population level, these tools are used to stratify the level of risk so that you can allocate care to where the highest risks are.”
Steroid-toxicity data fit directly into this architecture. As a quantified, validated measure of harm accumulation, the GTI Family refines risk profiles in ways that claims data alone cannot capture.
The patient side: informed consumers of care
Kosinski is emphatic that the measurement conversation cannot stop at the provider and payer level. Patients, particularly in the US, are active consumers of healthcare, and most have never heard of steroid-toxicity.
“They know that they feel great after they take those steroids. They are going to keep using them. But they do not know the harm that it can do 2, 3, 4 years down the road.”
He draws a direct analogy to food labeling. Just as consumers need to understand that a high-sugar product carries long-term dietary risk, patients need to understand that the short-term relief provided by steroids comes with a risk profile that can be measured, tracked, and acted upon. Making that information accessible — in the clinic, and through standardized assessments that give patients a voice in their own care — is, in his view, not optional.
“Such information is vital to understanding treatment practices. If steroids continue to be first-line therapy in certain conditions, you want to monitor and manage the dosing to ensure you are not poisoning people and having bad outcomes years down the road.”
A 35-year perspective
Kosinski started in this field in 1991 at what was then the New England Medical Center in Boston, contributing to the Medical Outcomes Study, the first large-scale application of psychometrics to health-related quality-of-life measurement. He did not expect it to become a field at all.
“I never thought this would blossom into what it has. But it gained momentum through the 90s and into the 2000s when the pharmaceutical industry started to say, ‘we are not only going to show improvements in blood and guts, we are going to show that patients are happy about it and can function better.’”
That same logic now applies to steroid-toxicity. Clinical endpoints are necessary but not sufficient. Demonstrating that a treatment reduces steroid-toxicity burden in a validated, standardized, reproducible way is the next frontier for drug development, population health management, and patient engagement alike.
Mark has more than 30 years of experience in developing and analyzing PRO measures. He was involved with the development and validation of the SF-36®, SF-12®, SF-36®v2, SF-12®v2, and the SF-8 Health Surveys, the Asthma Control Test (ACT), the Headache Impact Test (HIT-6), and other industry-sponsored PRO measures, as well as the development and analysis of item banks for computerized adaptive health assessments.
Mark has co-authored more than 300 peer-reviewed journal articles related to the development and analysis of PRO measures used in clinical research, randomized controlled trials, and population health initiatives. He has also contributed to the development of 10 manuals and interpretation guides for various PRO measures, including the SF-36®, SF-12®, and SF-8.
Over the past 25 years, he has been actively involved in the pharmaceutical industry, analyzing clinical trial data with PRO measures as secondary indicators of treatment efficacy. Mark earned his bachelor’s degree in psychology from the University of Connecticut and his master’s in psychology from Bridgewater State University.