OXYGEN, CONSUMABLES, AND EQUIPMENT ============================================================================ Ideas 76 and 78. Demand prediction is a solved problem applied to an unsolved one, and predictive maintenance on machines that cannot be replaced the same day.
WHY THESE TWO SIT TOGETHER -------------------------- Both are arithmetic over run-hours. Oxygen consumption is a function of what every ventilator and every mask is set to; equipment failure risk is a function of how long each machine has run since it was last serviced. The same data collection serves both, and separating them would duplicate it.
WHY THIS IS NOT A FORECASTING MODEL ----------------------------------- Oxygen demand is computable, not predictable. A ventilator at FiO2 0.6 and 12 litres per minute consumes a known quantity. Adding a learned model over arithmetic that is already exact would add error, not accuracy.
What is genuinely uncertain is the ADMISSION rate, and this engine does not pretend to forecast it. It computes current draw, states hours of supply remaining at that draw, and shows what happens under stated scenarios the unit chooses. The uncertainty is named and handed to a human.
THE INDIAN CONTEXT IS NOT DECORATION ------------------------------------ The 2021 oxygen crisis was not a failure of prediction. Units knew their consumption. What they lacked was the lead time between "we will run out" and "the tanker must leave now", which is a different quantity and the one this engine reports.
So the headline number is not days of supply. It is HOURS UNTIL THE ORDER MUST BE PLACED, which is days of supply minus delivery lead time, and that subtraction is the entire value.
Sources: standard oxygen consumption arithmetic; manufacturer service intervals. Every figure must be replaced with the unit's own before use.
meta
version0.1.0
ownerUNASSIGNED — set to the biomedical engineering lead
verified bynot set
boundary
Computes current draw and states hours remaining at that draw. Does not
forecast admissions, and says so rather than implying a prediction.
headline
The reported figure is hours until the order must be placed, not hours of
supply. The difference is the delivery lead time, and it is the number
that was missing in 2021.
oxygen
sources
- idnasal_cannula
labelNasal cannula
lpm formulaflow_lpm
noteFlow is the consumption. Nothing further to compute.
- idface_mask
labelSimple face mask
lpm formulaflow_lpm
- idnrbm
labelNon-rebreather mask
lpm formulaflow_lpm
typical lpm15
- idhfnc
labelHigh-flow nasal cannula
lpm formulaflow_lpm * (fio2 - 0.21) / 0.79
note
The heaviest consumer per patient in the unit. Air entrainment applies
exactly as it does for a ventilatoronly the fraction above room air
comes from the supply. 60 L/min at FiO2 1.0 draws the full 60; the
same flow at FiO2 0.4 draws 14.4, not 24 — multiplying flow by FiO2
counts the oxygen already present in the entrained air as if it came
from the tank, and over-estimates by about two-thirds at that setting.
device caveat
This is the arithmetic for a device entraining room air. A blender-fed
device supplied from both an oxygen and a medical-air line divides its
draw differently, and a turbine device with an oxygen inlet differs
again. Confirm against the specific device before relying on the
figure for procurement.
- idventilator_invasive
labelInvasive ventilation
lpm formulaminute_volume * (fio2 - 0.21) / 0.79
note
Only the fraction above room air is drawn from the supply. A
ventilator at FiO2 0.21 consumes almost nothing beyond drive gas.
- idniv
labelNon-invasive ventilation
lpm formulaminute_volume * (fio2 - 0.21) / 0.79 * leak_factor
leak factor1.4
noteMask leak is real consumption and is routinely left out.
reserve
critical hours6
warning hours24
note
Hours of supply at current draw. Compared against delivery lead time,
not against a fixed target.
consumables
- idvent_circuit
labelVentilator circuit
per patient day0.15
unitcircuit
stops caretrue
noteChanged on soiling rather than on schedule; usage is variable.
- idhme_filter
labelHME filter
per patient day1.0
unitfilter
stops caretrue
- idsuction_catheter
labelClosed suction catheter
per patient day1.0
unitcatheter
stops caretrue
- idcrrt_set
labelCRRT circuit set
per patient day1.5
unitset
stops caretrue
note
Filters clot unpredictably. Usage runs above the nominal figure and the
shortfall appears at 3 AM.
- idart_line_set
labelArterial line transducer set
per patient day0.2
unitset
stops carefalse
- idsyringe_pump_line
labelSyringe pump extension set
per patient day3.0
unitset
stops carefalse
equipment
- idventilator
labelVentilator
service interval hours5000
critical at percent90
single point of failuretrue
what degrades first
- partOxygen sensor
interval hours8760
symptomFiO2 reading drifts before it fails outright
why it matters
A drifting sensor is more dangerous than a dead one. The delivered
oxygen is wrong and nothing alarms.
wrong denominator
Run-hours is the wrong measure and this figure is a placeholder.
Galvanic cells deplete with cumulative OXYGEN EXPOSURE — FiO2
multiplied by time — so a ventilator run at FiO2 1.0 exhausts a
sensor far faster than one run at 0.3 for the same hours.
Manufacturers rate them in the region of one to two years of
continuous exposure, and replacement is normally driven by
calibration failure rather than by any elapsed count. Use the
manufacturer's figure and the unit's calibration log.
- partInspiratory valve
interval hours8000
symptomPressure delivery becomes inconsistent
- partInternal battery
interval hours17520
2 years
symptomTransport time falls without warning
why it matters
Discovered during a transfer, which is the worst possible moment
and the reason battery age is worth tracking separately.
- idinfusion_pump
labelInfusion pump
service interval hours8760
critical at percent95
single point of failurefalse
what degrades first
- partOcclusion sensor
interval hours8760
symptomLate or missed occlusion alarms
why it matters
A pump that does not alarm on occlusion silently stops a
vasopressor.
- iddialysis_machine
labelCRRT machine
service interval hours4000
critical at percent90
single point of failuretrue
- idmonitor
labelBedside monitor
service interval hours17520
critical at percent95
single point of failurefalse
- iddefibrillator
labelDefibrillator
interval basiscalendar
NOT run-hours
service interval days365
daily check requiredtrue
critical at percent80
single point of failuretrue
note
The only machine here governed by the calendar rather than by run-hours,
and the reason is that it sits unused for months and must work on the
first attempt.
why run hours fail
A standby defibrillator accrues almost no run-hours, so a run-hour rule
reports it as "in service" indefinitely — including for a machine nobody
has checked in years, which is precisely the failure this entry exists to
catch. Readiness is governed by the daily user check and annual
electrical safety testing.
not yet modelled
The data model carries no calendar field, so the daily check cannot be
tracked here yet. This engine reports the defibrillator as NOT ASSESSABLE
rather than as in service, because a reassuring answer from the wrong
measurement is worse than no answer.