Collier 1996.

Methods	RCT comparing 8 different foam mattresses; length of follow‐up not clear but patients assessed weekly. Allocation as follows: mattresses assigned to beds and coded numerically with only the principal investigator and ward link nurse aware of identity of each mattress. Mattresses then allocated to patients "as available".
Participants	Patients on a general medical ward; no further details given.
Interventions	Comparison of 8 foam mattresses: 1. New Standard Hospital Mattress (Relyon) (130 mm) (n = 9). 2. Clinifloat (n = 11). 3. Omnifoam (n = 11). 4. Softform (n = 12). 5. STM5 (n = 10). 6. Therarest (n = 13). 7. Transfoam (n = 10). 8. Vapourlux (n = 14).
Outcomes	Incidence of pressure ulcers. Patients assessed at least weekly throughout hospital stay. No patient developed a pressure ulcer of any grade during whole study.
Notes	9 patients allocated the Cyclone mattress, however, this group was withdrawn from the study at manufacturer's request and data not presented. All mattresses assessed for "grounding", deterioration of cover and contamination of inner foam core, interface pressures. No "grounding" of any mattresses during the evaluation period; softening of the centre of the foam base in Standard and Omnifoam mattresses on completion of study (detected using a "fist test" of unknown reliability). All mattress covers remained intact and inner foam protected.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Only information provided: “Mattresses were randomly allocated to patients on admission as available”.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding (performance bias and detection bias) Pressure ulcer incidence	High risk	“Only the principal investigator and the ward link nurse knew the identification of each mattress”.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	9 patients missing from data in Table 2 as their treatment, Cyclone mattress, was removed during the evaluation process at the request of the manufacturer. No other raw data presented in the paper to evaluate if incomplete outcome data addressed.
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes reported.
Free of other bias ‐ were groups similar at baseline regarding the most important prognostic indicators?	Unclear risk	Not reported.
Free of other bias ‐ was the timing of the outcome assessment similar in all groups?	High risk	"Frequency of assessment was determined by each patient's condition, but in all cases was conducted at least weekly throughout their period in hospital".