> ## Content Index
> Fetch the complete content index at: https://www.healthdatacon.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# Laura Gelfman Uses Better Data to Bring Palliative Care to Patients Earlier
- URL: https://www.healthdatacon.io/articles/laura-gelfman-uses-better-data-to-bring-palliative-care-to-patients-earlier/
- Published: 2026-10-16T16:03:44.000Z
- Updated: 2026-10-16T16:03:44.000Z
- Description: At HealthDataCon 2026, Laura Gelfman showed how structured data, clinical triggers and embedded teams can identify unmet palliative care needs earlier and help health systems measure whom they still miss.
- Author: HDM Staff
- Tags: Americas

Two people can share an age and have profoundly different care needs. One 85-year-old may hike regularly, manage her finances and live independently. Another may have advanced dementia, repeated falls and a caregiver responsible for most daily activities.

Laura P. Gelfman, MD, MPH, eFACHDM, used those illustrative patients, Anna and John, to frame a challenge for health data leaders: can their systems distinguish the person who needs prevention from the person who needs symptom relief, caregiver support and help making difficult decisions?

At HealthDataCon 2026, Dr. Gelfman presented “Right Patient, Right Time, Right Team: Data-Driven Access to Palliative Care at Scale.” A professor and vice chair for quality and clinical innovation in geriatrics and palliative medicine at the Icahn School of Medicine at Mount Sinai, she argued that broader access begins with making unmet needs visible.

“Finding the right patient means seeing the need that is often invisible today,” she said.

![Gelfman’s slide contrasts mortality prediction with identifying symptoms, functional decline, caregiver strain and other unmet needs.](https://storage.ghost.io/c/91/db/91db0927-dd9a-4578-9944-921da70f6885/content/images/2026/10/palliative-care-predicting-needs.png)

Gelfman urged health systems to identify unmet needs alongside mortality risk. Slide credit: Laura P. Gelfman / HealthDataCon 2026.

## Identify need before a crisis

Dr. Gelfman began by clarifying what health systems are trying to deliver. Palliative care addresses symptoms, supports patients and families, clarifies priorities and helps coordinate care for people living with serious illness. It can accompany treatment directed at the underlying disease.

The [CAPC definition](https://www.capc.org/about/patient-and-family-resources/?ref=healthdatacon.io) makes the same distinction: access is based on need, and care can begin at any age or stage of serious illness alongside curative treatment. Equating palliative care with hospice can delay referrals until opportunities for earlier support have passed.

That misconception can also become a modeling problem. Mortality is comparatively easy to identify in health records, Dr. Gelfman noted. Symptoms, caregiver strain and functional decline are harder to capture consistently. A system built primarily to predict death may therefore reproduce the late-referral pattern clinicians are trying to change.

“We need to know who’s declining, who’s suffering, who’s having to have conversations, and make decisions about their medical care now,” she said.

Her proposed shift expands the signals used to identify patients: worsening symptoms, loss of function, repeated hospital use, caregiver needs and missing decision-maker information. Mortality risk can contribute, but it cannot explain what assistance a person needs today.

![Six foundational data elements: functional status, medical decision maker, caregiver, serious illness diagnosis, symptoms and goals of care.](https://storage.ghost.io/c/91/db/91db0927-dd9a-4578-9944-921da70f6885/content/images/2026/10/palliative-care-data-foundation.png)

Reliable documentation supports patient identification, resource allocation and measurement. Slide credit: Laura P. Gelfman / HealthDataCon 2026.

## Build the data foundation that care requires

For Dr. Gelfman, better algorithms depend on reliable documentation of six elements: functional status, medical decision maker, caregiver, serious illness diagnosis, symptoms and goals of care. These details often exist in scattered notes, flowsheets or fields that are difficult for clinicians to find.

The distinction between a caregiver and a decision maker matters. Knowing who can make a medical decision does not establish who is available to help at home, what that person can provide or whether the caregiving arrangement is sustainable.

Mount Sinai’s [registry work](https://reports.mountsinai.org/article/gpm2025-using-data-to-improve-the-quality-of-palliative-care-across-all?ref=healthdatacon.io) combines electronic health record and claims information to examine palliative care delivery across settings. Dr. Gelfman described extending that view with patient-reported symptoms and caregiver needs, then using it to identify both people receiving services and those being missed.

Functional status offers a particularly useful signal. A retrospective [cohort study](https://scholars.mssm.edu/en/publications/acute-healthcare-and-palliative-care-utilization-among-hospitaliz/?ref=healthdatacon.io) involving Dr. Gelfman and colleagues examined 18,049 hospitalized adults assessed with the Karnofsky Performance Status scale. Greater impairment was associated with more intensive care use, in-hospital mortality and palliative care consultation.

Yet consultation still occurred several days into admission: a mean of 5.3 days for patients with very severe impairment, compared with 6.7 days for those with mild impairment. The finding points to an opportunity to use information already being collected earlier in the care process. It does not establish that a functional score alone should determine referral.

## Connect every trigger to an action

“Each trigger maps to an action,” Dr. Gelfman said.

In the programs she described, identifying a patient can lead to an automatic consultation, review by an embedded nurse practitioner or discussion during ICU rounds. Examples included advanced cancer with an unplanned admission and specified length of stay, recurrent admissions in advanced heart failure, and an ICU stay of five or more days.

These were locally designed pathways, developed with clinical leaders and tested against available capacity. Worklists help teams triage patients proactively; structured templates capture relevant information; dashboards show volume, reach and quality measures. The rule is useful when an accountable team can respond.

Published evidence illustrates the potential. In Katherine Courtright and colleagues’ [randomized trial](https://pubmed.ncbi.nlm.nih.gov/38227032/?ref=healthdatacon.io), default palliative care orders, which clinicians could cancel, increased consultation from 16.6% to 43.9% among the primary analytic population of 24,065 older hospitalized patients. Consultation occurred an average of 1.2 days earlier.

The trial did not significantly reduce hospital length of stay. Its results demonstrate improved access and timing, while reminding leaders to measure downstream outcomes separately.

Dr. Gelfman also described progress in Mount Sinai’s medical ICU: the share seen during the ICU stay rose from roughly 14% to 46% over the periods she presented. For patients with longer ICU stays, she reported an increase from 27% to 70% after the program’s evolution. These were local program observations, rather than randomized estimates of the intervention’s effect.

## Match the team to the patient

Expanding identification creates another challenge: specialist capacity. Dr. Gelfman argued for matching the intensity and composition of care to the patient’s needs, with generalists providing core skills, supported teams handling additional complexity and specialists focusing on the highest-need patients.

She described a hospital medicine program that began in 2020 with an embedded social worker. Goals-of-care documentation increased from 4% before the model to 89% afterward, she reported. The team later added a nurse practitioner as symptom-management needs became clearer. An embedded advanced heart failure model also reached patients with better function earlier in their illness course.

Patient-reported information can help make those needs actionable. Dr. Gelfman cited Ethan Basch’s symptom-monitoring research, in which severe or worsening symptoms generated alerts for clinical teams. A [single-center analysis](https://jamanetwork.com/journals/jama/fullarticle/2630810?ref=healthdatacon.io) reported median survival of 31.2 months with electronic symptom monitoring versus 26.0 months with usual care; survival was a post hoc analysis.

The larger, 52-practice PRO-TECT trial adds important context. Its [final results](https://pubmed.ncbi.nlm.nih.gov/39920394/?ref=healthdatacon.io) found no survival difference, but electronic monitoring delayed deterioration in physical function, symptoms and health-related quality of life and reduced emergency department use.

Those studies evaluated symptom monitoring during cancer treatment, not the entire palliative care model. Their relevance to Dr. Gelfman’s argument is the connection between patients reporting a need and a care team receiving information it can act on.

## Keep asking who remains unseen

Higher consultation volume is only part of the measurement task. Dr. Gelfman urged teams to define the eligible population, track the share reached and examine time from admission to consultation. Reviewing demographic and clinical characteristics can reveal gaps hidden inside an improving overall average.

“We’re always asking the question of who are we still missing?” she said.

Her team also tracks ICU days, hospital length of stay, readmissions, goals-of-care documentation and discharge services. She described using these measures to explain the program’s contribution to health system priorities, including appropriate resource use and more accurate documentation of clinical severity.

In the discussion, an audience member asked about education and documentation burden. Dr. Gelfman emphasized the daily work of building relationships with frontline teams, demonstrating value in individual cases and approaching colleagues with curiosity. Technology supports that work, but does not replace it.

“You have to have that human interaction and connection, I think, to demonstrate the value,” she said.

Her teams review templates and fields with clinicians and EHR experts, looking for information that can be reused rather than entered again. She also described ongoing professional-society work with major EHR vendors on foundational palliative care data elements, while acknowledging that progress remains slow.

The next stage is to carry this approach across inpatient, outpatient and home-based care. For patients like John, success means that symptoms, declining function and caregiver needs prompt support while there is still time to plan. For the health system, it means making that response a dependable part of care delivery.