Skip to main content

Abdulaziz Alkattan

From a DBMI Fireside Chat. Summary by ChatGPT 5.5.


His path into clinical informatics came from seeing system failures firsthand. During residency and especially during the first COVID surge at NYP Queens, he saw how fragmented systems, resource shortages, documentation requirements, and poor interoperability directly affected care. That experience shifted him from thinking about individual clinical practice to thinking about the healthcare system itself.

COVID exposed how much organizational context matters. His burnout study found that clinicians who were redeployed to unfamiliar units had higher burnout, and that excess burnout persisted even after redeployment ended. Comments suggested that loss of familiar colleagues, workflows, and local knowledge were major contributors.

Healthcare data problems are often workflow problems disguised as data problems. A timestamp or field in an EHR does not necessarily tell you what actually happened clinically. His team recently went into EDs at Cornell and Columbia and interviewed physicians, consultants, and clerks simply to understand what particular events in a dataset meant in real life.

That “bridge” role is a major value of biomedical/clinical informatics. He argues that you need people who understand enough clinical workflow + data + informatics to translate between clinicians and statisticians/engineers. Purely technical people may not know what the data represents; clinicians may not know how the data infrastructure works.

Operational work and academic research have different standards of evidence. His key point: ask what obstacle you are trying to overcome. For a local operational change, relatively basic data plus firsthand experience may be enough. For a publishable research claim or a large/generalizable change, much greater rigor is required.

His staffing example makes that concrete. He analyzed when surgical ED consults actually arrived and showed that the existing 6–6 resident schedule mismatched demand. They changed the schedule and added swing coverage. The evidence was good enough to improve operations even though it would not necessarily constitute a strong research study.

For operational leaders, simple analysis can have enormous leverage. Many hospitals collect only the data required for CMS, registries, insurance, and reimbursement. Something as straightforward as showing patient volume, burnout, or workload patterns can reveal something leadership has literally never seen before.

A critical professional skill is knowing how to simplify rigorous work. His advice is essentially: learn to do the rigorous analysis, but then be able to reduce it to the headline a busy clinician or administrator needs to make a decision. More analysis is not always more useful operationally; sometimes it obscures the actionable finding.

NYP has two somewhat separate informatics worlds:

  1. EHR/operational informatics — order sets, notes, workflows, clinical decision support, Epic governance.
  2. Data/analytics infrastructure — DSS, Snowflake, registries, research datasets, analytics/AI.
    His role increasingly sits at the interface between them.

He sees the EHR’s fundamental design objective as part of the problem. His argument is that EHRs are fundamentally optimized around billing and administrative requirements, not patient care, data science, or clinician usability. Clinicians therefore spend huge amounts of time documenting rather than caring for patients.

Because of that, AI layered on top of the current system does not automatically fix healthcare. If the underlying incentives and workflows are bad, AI may simply automate extraction, billing, denials, or other parts of the same dysfunctional structure. He argues that the system ultimately needs to be reoriented toward care rather than financial optimization.

His broader view is that meaningful healthcare-system change ultimately requires policy, because many of the incentives producing documentation burden, administrative expansion, and financial optimization are structural rather than merely technical. He is also careful to note that universal or centralized systems elsewhere still have substantial problems, so interoperability alone is not the solution.