The Cognitive Screening Gap: What We Found in 12.3 Million Post-Acute Patients

This week we published a piece in McKnight's Long-Term Care News on the cognitive screening gap in post-acute care. The short version: an estimated three quarters of cognitive impairment in this setting never gets documented. This post expands on what we found in the longitudinal data behind that claim, and what changes when screening is embedded directly in the clinical workflow instead of bolted onto an already overloaded shift.

The Scale of the Gap

Roughly three out of four residents with cognitive, neurological, or behavioral health conditions in post-acute care never have those conditions documented. That is not a rough estimate pulled from a single study. It is what shows up when you pull apart a longitudinal record of 12.3 million post-acute patients and 27 billion clinical events and compare what the chart says against what the speech, behavior, and trajectory data says.

Major depressive disorder runs at roughly 48 percent prevalence in skilled nursing. The MDS Section D field is checked on far fewer charts. Mild cognitive impairment runs at roughly 35 percent and gets documented as “dementia, unspecified” months or years after the signal first appeared in speech. Parkinson’s disease gets a diagnosis 2 to 4 years after the first detectable voice tremor. These are not edge cases. They are the median resident.

Why Documentation Fails Despite Good Clinical Intent

The field-based screening tools a Director of Nursing actually has access to, for example the MMSE or MoCA, take 10 to 30 minutes per resident and require a trained clinician in the room. A 120-bed facility running those tools at the cadence clinical practice calls for would spend roughly 240 clinician hours per quarter on the screening step alone. That is two full-time equivalents dedicated to screening, before a single medication review or care conference. No facility has that capacity. None.

So the shift reality wins. The CNA carries 12 to 18 residents. The nurse runs the med pass and the admission paperwork. The Director of Nursing triages the acute problems: a fall, a pressure injury, a family call. Cognitive screening is the step that falls off the list every single day it is not already embedded in the workflow. That is what the McKnight's piece means by a capacity problem. The intent is there. The clock is not.

What Changes With Embedded Screening

A Samsung Health Grade Galaxy device sits on the admission desk. The new resident is handed the tablet during intake, the same moment they would otherwise be answering paperwork questions. GIA® asks six open-ended prompts and listens. Forty seconds of natural speech later, the structured screening result, clinician-ready note, ICD-10 codes, full transcript, and the recorded video are written back to the PointClickCare record in real time. Total resident time at the front end: under 5 minutes.

The clinician review happens on their own cadence, not the resident’s. Only flagged results surface. Average review time per flagged case is under 2 minutes. For an admit cohort of eight residents per week, the total clinician time spent on screening collapses from the quarterly 240-hour figure above to roughly 5 minutes per week. That is the math that lets a facility actually screen every admission, instead of screening the ones they have time for.

The speech biomarker engine behind the screening analyzes 2,500 acoustic and linguistic signals per session, including pause ratio, speech rate, vocal energy, pitch variability, semantic coherence, and word-finding patterns. Those signals are clinically associated with cognitive decline, depression, anxiety, PTSD, Parkinson’s, and further neurological conditions. Accuracy is peer-reviewed and clinically validated: 81.6 percent on depression, 80.0 percent on PTSD, 77.5 percent on anxiety, AUC 0.97 on Parkinson’s. Full clinical research page lists the 19 peer-reviewed studies behind those figures.

What Changes at the Facility P&L

Undetected cognitive impairment carries a direct dollar cost. Untreated depression in skilled nursing is associated with higher readmission rates. Late-stage Parkinson’s diagnosis misses the window for medication titration that reduces fall incidence. Tardive dyskinesia that goes unscreened under CMS F-Tag 605 is a citation risk in the next survey. Each of these is a P&L line: readmission penalties for Enhanced Track ACOs, falls liability, survey citation remediation. The screening step upstream is what changes those downstream lines.

Three CPT codes apply to the clinician review step: 96127 for the brief assessment, 96116 for the neurobehavioral status exam, and 99309 for the evaluation and management visit that results from a flagged screening. For facilities that have never billed at this level because they never had the screening infrastructure to support it, the codes convert from unused to a recurring revenue line in the first month of deployment.

About GIA®

GIA® is the FDA-registered Digital Human® built by Scienza Health on top of digitalhumanOS™. GIA® conducts the screening conversation through video, voice, or landline in 92 languages, analyzes 2,500+ speech biomarkers, and delivers a structured, clinician-ready result back to the EHR in real time. Live in the PointClickCare Marketplace. Deployed on Samsung Health Grade Galaxy devices secured by Samsung Knox. GIA® screens. A licensed clinician reviews and approves every result before it enters the clinical record. GIA® screens for 24 conditions today, with 22 in development.