CareProof Hospital-Grade at Home Index

Indian home-care providers call care “hospital-grade” without a number an auditor can check. A hospital can show monitors, trained staff, equipment dates, an escalation path, and a record. At home the readings are irregular, the team may be one attendant, and the device lives in a bedroom. Copying a ward checklist fails on the first quiet afternoon. The question is whether that promise can be measured without pretending the home is a smaller hospital.

I separated three kinds of statement: what a source establishes, what I take it to mean, and what I propose. The proposed definition is strict. An episode is hospital-grade only when the score is at least 90 and every safety gate passes. The score is six pillars: clinical safety 25%, caregiver competence 20%, equipment 15%, infection and medication 15%, documentation 10%, and patient and family experience 15%. The gates are vetoes, not extra points: deterioration response, equipment eligible for use, critical caregiver skills, and infection or medication process. Hospital NEWS2 bands are only a reference scaffold. Home rules look at the change between consecutive readings, because a single saturation is the wrong question. I did not fit the weights to outcomes. They go to an expert panel before any claim that the standard works.

The engine scored eight fictional episodes on 28 September 2026. The mean is 83. That mean is the wrong headline. One episode meets the proposed standard. Five are not compliant. Arun Patel scores 94 because his observations are stable on prescribed oxygen, and he still fails: concentrator maintenance was due on 12 March 2026. If that date is treated as current, the same engine rescores him at 98 and he meets the standard. The failed record stays on the passport. Suresh Nair deteriorates overnight to an emergency pattern. The attendant does not acknowledge the alert. The episode scores 69. The monitor was eligible. Escalation skill was not. No effectiveness result is claimed. These cases test whether the measure can refuse a flattering number.

The weights, the line at 90, and the home timers are proposals. NEWS2 is not validated here on irregular home data. Eight simulated episodes cannot estimate sensitivity or effect. No second rater has scored the cases, so no agreement statistic is reported. Next I would freeze the indicator list, run the panel of six to eight experts, and only then score 30 episodes with two assessors. A comparative study of 222 episodes is a plan, and it is discarded if the panel says the measure is the wrong measure