Mobility Atlas research: tilt-in-space-wheelchair-guide
Data and method for the article at mobilitatlas.com/blog/tilt-in-space-wheelchair-guide/
License: CC BY 4.0 (credit "Mobility Atlas")
Prepared: 22 October 2026

WHAT WAS READ
FDA MAUDE medical device reports filed under product code IOR (Wheelchair, Mechanical; 21 CFR 890.3850) or ITI (Wheelchair, Powered; 21 CFR 890.3860) whose text contains "tilt in space", "tilt-in-space", "power tilt", "tilt system", "tilt function", "tilt actuator", "tilt module", "tilt mechanism", "space tilt", "recline", "reclining" or "power recline" (openFDA phrase search), read on 22 October 2026: (a) every such report received before 1 January 2011 (45); (b) every such report received from 1 January 2016 (243); (c) a systematic sample of 200 of the 1,118 received from 2011 to 2015, taken as 50 consecutive reports at positions 0, 279, 559 and 838 of the openFDA result list sorted by date received. The four runs fell between January 2011 and September 2014, so no 2015 report is in the sample. Pool: 488 reports, all coded: 309 about power wheelchairs, 175 about manual wheelchairs, 4 about other devices. The stratum of each report is in strata.csv; the full date-sorted list of the 1,118 reports, with each report's position and whether it was sampled, is in frame_2011_2015.csv (read again on 23 October 2026, openFDA data last updated 13 October 2026).
FDA recall records for product codes IOR and ITI whose text names tilt, recline or an actuator, from the local copy of openFDA recalls (19 records, 16 about tilt or recline systems).
Makers: grouped from the manufacturer and brand names in each report (Invacare, Medline, Permobil, Pride/Quantum, Sunrise Medical, other).

HOW IT WAS CODED
1. A codebook (codebook.txt) was written before any report was coded. Its SHA-256 fingerprint, recorded at 2026-10-22T05:17:01Z, is b640d3ef5157ec3d1a54afdb1866a0371e42d617d3bd62f5143cd949fe0f77c7. Any later change to the codebook would change this fingerprint.
2. The reports were split into chunks. Two coders (AI language models working separately, each in a private scratch folder, each without access to the other's answers, told to read every report and not to use keyword scripts) coded every report in every chunk against the codebook.
3. Agreement between the two coders: device 99.6% (kappa 0.991); seating 99.4% (kappa 0.973); problem 95.7% (kappa 0.929); injury 99.6% (kappa 0.986); fall 99.8% (kappa 0.991); user 99.0% (kappa 0.946); what was found 98.4% (kappa 0.97).
4. Every report where the coders differed was decided by the lead analyst under the written rules below (40 reports marked adjudicated). A consistency pass then re-applied the rules across the whole table and changed codes in 36 more reports (marked consistency).
5. Counts in the article were produced by script from the final table, not by hand. Some sub-counts (parts, places or words named in reports) are keyword counts over the report text; the exact word patterns are listed under KEYWORD COUNTS. Different wording would give somewhat different counts.

ADJUDICATION RULES
T1: PROBLEM (dropped back or seat vs structure): whenever the backrest, or a tilted, reclined or raised seat, dropped, fell, gave way, collapsed, came off, lost support or flipped back (with or without the user falling), code A, whatever part is said to have broken (hinge, bracket, weld, back cane, bolt, screw, gas spring). The codebook names a dropped seat or back under A, and A comes before S in its list. Code S only when a structural part (seat or back frame, back cane, tilt or recline frame, bracket, bolt, weld, hardware) broke, cracked, bent, loosened, separated or wobbled and the report does not say the back or seat dropped (it still held, or the break was found with no drop described). A seat that drops because the folding crossbar of a non-tilting seat broke stays S.
T2: PROBLEM (trapped in a position): P when the report says the user, not only the seat, was stuck, locked, stranded or trapped in a tilted, reclined, raised or standing position, could not get out or return upright, or had to be lifted out or rescued (fire department, EMTs, caregiver), as the codebook says ("the user trapped in a position") and because P comes first in its list. This holds whatever caused it (error message, failed actuator, a death with the user unable to come back up). A when the seat or mechanism is stuck and no user is said to be held in it, when the user is only said to be in the chair, when the function merely stopped working for days or weeks, or when the user is stuck upright (the ordinary sitting position).
T3: PROBLEM (mechanism vs seating electrics): C when the report names a seating electrical part as the fault (seating switch or toggle, tilt or recline module, TRAM or control box, wiring, harness, connector, potentiometer, programming), or reports smoke, fire, sparks, burning or melting in the seating system, actuator included, even when the tilt or recline then stopped, ran on or worked intermittently. A when the report names a mechanical failure (actuator, motor, gearbox, gas spring, cable, release handle, lock) or describes only the behaviour (stuck, would not tilt, moved by itself) with no electrical part named, or only suspects an electrical cause; also A when an uncommanded movement or a mechanical break came first and smoke followed.
T4: PROBLEM (allegation vs maker finding): code what the reporter alleges, even when the maker found a different part. A reporter who names the brakes or a "brake cable" is coded G even if the maker found the recline handle fractured. Use the part an inspection or photo names only when the reporter names no part ("frame issue", "the back is broken", "gave way"), then code that part (seat tubes bent: S; push handle weld: G; bolt or bracket under a back that gave way: T1). A tip or fall that the report describes with no complaint about the device stays T even when the maker later finds a contributing fault; a fall that the reporter blames on the chair's own movement (it threw her out on return from recline) is A.
T5: DEVICE (power vs manual when the type is not stated): the chair's brand and model decide over the generic brand_name field ("POWERED WHEELCHAIR", "POWER CHAIR 20 WIDE BY 18 DEEP FOR PMI" are Invacare form defaults). Manual (M): Medline reclining wheelchairs, Invacare Solara 2G/3G and Tracer/Spree transport and recliners, Quickie Iris, Zippie (manual pediatric tilt), Freedom Designs Gizmo, Drive Viper GT Recline, Bentley. Power (P): Invacare TDX, FDX, 3G Storm/Torque/Arrow, Pronto M51/M91; Permobil; Pride and Quantum Q6 Edge; Quickie P-series, QM/Q-series power, S646. When the text itself states the type, the text decides.
T6: SEATING (y/n): push handles, armrests, arm assemblies, leg rests, footrests, wheel locks and brakes are not seating, so seating is n when the event is about them, unless the part that failed is the recline release handle or lever (on Medline recliners a push handle that moves without the recline handle is the recline not locking: y). Seating is y when the event happened while the seat was being tilted, reclined or raised, or when the chair tipped or the user fell while tilted or reclined and no part outside the seating system is said to have failed. Damage to a seating part that happened in a fall caused by something else (brakes, casters) does not make it y. For device O, seating is n: the flag describes a wheelchair's seating system.
T7: INJURY: y only when harm to the body or medical care is stated (bruise, cut, swelling, soreness or pain, fracture, pressure injury, hospital or ER, stitches) or the event type is Death. The event type Injury alone, being thrown out, striking the head with no harm stated, discomfort or a fit complaint, or a harm only feared ("could have been crushed") is n.
T8: FINDING (N vs X): D when the maker, dealer or a service inspection confirms the fault or finds a broken, bent, missing or defective part on the device or in photos or video, or a dealer technician diagnoses a part that must be replaced, whatever cause the maker then names (shipping damage, abuse, unknown). N when the device or returned parts were inspected and no fault was found, or the maker or dealer concludes the cause lay outside the device (modification, misuse, accident, user positioning) without confirming a fault. X when nothing was inspected, when the maker says only that the malfunction was "not confirmed" without stating an inspection, and when a photo-only review ends "Not Confirmed" because the issue could not be identified from photos and a sample was required, as in knee rule K10.
T9: DEVICE O and PROBLEM: hydrobaths, clinical and treatment recliners, lift recliners and strollers (Zippie Voyage) are O. Their PROBLEM is X unless the report alleges that the other device itself failed (a hydrobath tilt catch sticking, a treatment recliner missing bolts and needing a recline mechanism, a stroller seat detaching from its base); then it is Q, as in knee rule K1. The wheelchair letters A, C, S, P, G name wheelchair parts and are not used for another device. Seating is n for O (T6).
T10: USER: y when the report attributes the event to what the user or a third party did: misuse, modification, unapproved or user-replaced hardware, a caregiver breaking a part while loading the chair, a patient who "managed to tip over" the chair, or continued use of a chair known to need repair, against the instructions, that the report says caused the harm or damage. n when the user is only present, when an IFU warning is quoted without saying it was ignored, or when a third party acted only after the event (a neighbor trying to repair a chair already stuck).
T11: FALL: y when the user left the seat and went to the floor, ground or another surface, was thrown or ejected, or the chair tipped over with the user. n when the back dropped or the user slumped but stayed in the seat, or someone caught the user.
T12: PROBLEM (first failure and loosely named parts): code the first failure in the chain; damage done in or after a fall (a recline actuator casing cracked when the chair rolled off a lift because the brakes were slow) does not count. An unspecified "frame" broken on a chair with nothing placing it in the seat, back or tilt frame is G. A hazard of a named part (a sharp wheel edge that cut toes) is coded to that part (G). A fire whose location is not stated is Q.

KEYWORD COUNTS (Python regular expressions, matched against the upper-case report text without its header line)
KW = {'fracture': 'FRACTUR|BROKEN (LEG|ANKLE|FEMUR|TIBIA|HIP|ARM|WRIST|BONE)', 'head': '\\bHEAD\\b', 'trapped_hours': 'HOURS|OVERNIGHT|DAYS', 'stuck': 'STUCK|TRAPPED|COULD NOT (GET|RETURN|COME)|UNABLE TO (GET|RETURN|COME)', 'pressure': 'PRESSURE (SORE|INJURY|ULCER|WOUND)|BED ?SORE|DECUBITUS', 'fire': 'FIRE|SMOKE|MELT|BURN', 'actuator': 'ACTUATOR', 'gas_spring': 'GAS (SPRING|CYLINDER|STRUT)|PISTON', 'back_cane': 'BACK ?CANE|BACKREST|BACK REST', 'uncommanded': 'UNCOMMAND|BY ITSELF|ON ITS OWN|WITHOUT (INPUT|COMMAND)', 'van_ramp': '\\bVAN\\b|RAMP|LIFT', 'caregiver': 'CAREGIVER|CARE ?GIVER|NURSE|STAFF|ATTENDANT|AIDE'}

QUALITY NOTES (disclosed in full)
- The article went through fact-check passes by checkers who had not written it (a full pass with a reader review, then a control pass over the changes), each recomputing numbers from the files here and reading quotes against their sources; every error found was corrected before publication.

AMOUNTS
Every report text was searched for money figures; none gives one. Prices in the article are prices shown on makers', dealers' and retailers' pages and in the Mobility Atlas catalog, listed in tilt-in-space-wheelchair-guide-documents.csv and the article sources.

LIMITS
- Reports are not rates. Nobody publishes how many devices of each model are in use.
- A report is an allegation or a description; the FDA says the submission of a report "itself does not necessarily demonstrate that the device caused or contributed to the adverse outcome or event."
- Counts from 2011 to 2015 are counts within the sample, not totals for those years.
- The 2011-2014 sample is almost entirely short Invacare reports with no inspection, and many 2026 reports are Medline retrospective-review filings; both are filing patterns, not verdicts on those makers.
- The phrase search does not find reports that say only "reclined" or "recliner", or name tilt only in the product name; such reports are outside the pool.
- Some events were filed twice (a user's MedWatch report and the maker's report); likely duplicates are flagged in the article and counted as filed.

FILES
codebook.txt        the codebook as fingerprinted
coded_reports.csv   one row per coded report: both coders' answers, the final code, and how a disagreement was resolved (resolution, rule_applied)
report_pool.csv     the reports in the pool with product code, brand, date received and event type
The full report texts are public: search MAUDE or openFDA by report key.
tilt-in-space-wheelchair-guide-documents.csv  the documents and pages read, with links