Chen 2009.

Methods	Case‐control study to test the association between SARS onset and a range of causative and protective variables in Sun Yan Tzen University hospitals in Guanzhou, Southern China The study collected information on cases and controls retrospectively during the first phase of the SARS epidemic in China (March to May 2003) but there is also a prospective element with antibody confirmation of SARS infection. Analysis plan was similar to that of Liu 2009 with a univariate and multivariate analysis conducted to assess risk factors
Participants	Description of cases. Probable SARS cases were defined using the criteria by the China Health Ministry. Criteria for probable and suspected SARS cases included travel to a SARS epidemic area in the 2 weeks before the onset of symptoms or close contact with a probable SARS patient; fever of ≥ 38°C; chest X‐ray abnormalities; normal or decreased leukocyte count; and no response to treatment by antimicrobial drugs. In this study what appears to have happened is that available Sun Yan Tzen University hospitals HCWs who were willing to be interviewed were bled and those with raised IgG against SARS‐CoV were included as cases. Cases enrolled were 90 out of the possible 112 who had SARS (80%) and 758/846 controls (89%). The choice criterion for interview of cases and controls was availability i.e. being “off duty” during the survey. It is unclear what this means and why such bias was knowingly introduced  Description of controls. Controls were SARS‐CoV negative HCW who had worked in the 2 hospitals attending SARS cases
Interventions	An extensive number of exposure and interventions variables were elicited and quantified with discrete scores. Definitions are absent in most cases  Use of personal protective and control measures Number of gowns worn 0 = single, 1 = double Number of multilayered cotton mask worn 0 = single, 1 = double Number of pairs of gloves worn 0 = single, 1 = double Frequency of wearing shoe cover 0 = never, 1 = sometimes, 2 = often, 3 = every time Frequency of wearing cap 0 = never, 1 = sometimes, 2 = often, 3 = every time Frequency of face shield in SARS ward 0 = never, 1 = sometimes, 2 = often, 3 = every time Frequency of wearing goggles while performing operation for SARS patients 0 = never, 1 = sometimes, 2 = often, 3 = every time Health‐related behaviours Frequency of washing uncovered skin after caring for SARS patients 0 = never, 1 = sometimes, 2 = often, 3 = every time Frequency of washing hands after caring for SARS patients 0 = never, 1 = sometimes, 2 = often, 3 = every time Frequency of washing nasal cavity after caring for SARS patients 0 = never, 1 = sometimes, 2 = often, 3 = every time Frequency of washing oral cavity after caring for SARS patients 0 = never, 1 = sometimes, 2 = often, 3 = every time  SARS patient care Special training for SARS 0 = no, 1 = yes Performing tracheotomy 0 = no, 1 = yes Performing tracheal intubations 0 = no, 1 = yes Caring for "Super Spreading Patient" 0 = no, 1 = Yes Avoiding face to face while caring for patient 0 = never, 1 = sometimes, 2 = often, 3 = every time  Other relevant control measures Method of air ventilation in offices and SARS wards 1 = artificial central ventilation (windows were closed in wards), 2 = natural ventilation (windows were opened in wards), 3 = natural ventilation and additional electronic exhaust fan (windows were opened in wards, at the same time, electronic exhaust fans were used for improving air circulation in wards) Type of equipment for washing hands 1 = automatic tap, 2 = non‐automatic tap, 3 = other
Outcomes	N/A
Notes	The authors conclude that “Some measures, particularly good air ventilation in SARS wards, may be effective in minimising or preventing SARS transmission among HCWs in hospitals”.  The study is biased by the selection of cases and controls (enrolment only of available personnel) and the non‐eligibility (and lack of mention) of HCW who died of SARS (which may be up to 20% of people who were ill during the first wave of SARS). The design and analysis are very similar to those of Ma 2004/Liu 2009 and the design also lacks focus. i.e. it does not test a defined hypothesis, but trawls through large numbers of variables looking for associations. There is no attempt at matching cases with controls and part of the design is prospective (IgG estimation). As a consequence the design distinction between a case‐control and a cohort study is blurred. There is no mention of whether interviewers were blinded to case or control status of interviewees  Data extracted are from the univariate analysis table 3 which is the table reporting both numerators and denominators for cases and controls. Table 4 (multivariate logistic analysis) reports the significant multiple protective associations: caring for super spreading patient and avoiding face to face contact while caring for SARS patient (OR 0.30, 0.15 to 0.60) and wearing gloves coupled with methods of ventilation (various ORs for the various combinations intensity of wearing and ventilation methods, all significant). In the light of so many biases it is difficult to interpret the data but there does seem to be a gradient favouring multiple interventions
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	N/A
Allocation concealment (selection bias)	Unclear risk	N/A
Blinding (performance bias and detection bias) All outcomes	Unclear risk	N/A
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	N/A
Selective reporting (reporting bias)	Unclear risk	N/A