Skip to main content

Advertisement

Table 1 Assessment of study quality, risk of bias, and confounding

From: Evidence for a causal link between sepsis and long-term mortality: a systematic review of epidemiologic studies

Methodological assessment summary (N=59 studies) %*
1. Exposure definition (sepsis)  
 i) Consensus [7, 8, 2832, 35, 37, 39, 40, 44, 45, 4750, 6374] 49 %
 ii) Claims/ICD codes [11, 2426, 38, 41, 58, 7577] 14 %
 iii) Pathogen-plus (either positive blood culture or pathogen specified) [27, 34, 36, 46, 53, 7881] 20 %
 iv) Pneumonia [33, 40, 42, 43, 52, 8285] 15 %
 v) Peritonitis [21, 51, 86] 5 %
2. Ascertainment of exposure  
 i) Claims/ICD codes data 13 %
 ii) Randomized controlled trial cohorts [44, 45, 63, 66, 69, 70] 10 %
 iii) Non-interventional study cohorts 75 %
3. Selection of the control cohort  
 i) Drawn from ICU patients without sepsis  
  a. Matched [35] 2 %
  b. Not matched [7, 25, 3032, 37, 41] 12 %
 ii) Drawn from hospitalized infected patients (not ICU)  
  a. Matched  
  b. Not matched [30, 41] 3 %
 iii) Drawn from hospitalized non-infected patients (not ICU)  
  a. Matched [36] 2 %
  b. Not matched [30, 34, 3841] 10 %
 iv) Population controls  
  a. Matched [30, 33, 37, 4143] 10 %
  b. Not matched [40, 41] 3 %
 v) No control cohorts/no controls for mortality comparison 73 %
4. Comparability of cohorts on the basis of design or analysis  
 i) Regression models to adjust for confounders  
  a. Regression models in studies reporting control population  
   - Post-acute mortality [7, 25, 30, 3337, 39, 42] 16 %
   - Cumulative mortality  
   - Stratified analysis for post-acute mortality [32, 40, 42, 43] 5 %
   - Non-mortality outcome models [31, 38, 41] 5 %
  b. Regression models using in studies with no controls  
   - Post-acute mortality 22 %
   - Cumulative mortality 20 %
   - No mortality model 31 %
 ii) Studies reporting sepsis dose-response [39, 46, 49, 52, 53] 9 %
5. Assessment of post-acute mortality  
 i) Record linkage with national or regional databases or outcome assessed by contact with patient or relatives 90 %
 ii) No description 10 %
6. Adequacy of follow up  
 i) Complete follow up, all participants accounted for 22 %
 ii) Loss to follow up unlikely to introduce bias (<20 % loss, or >20 % but those lost described and unlikely to be different from those followed) 61 %
 iii) Follow up <80 % and no description of those lost 3 %
 iv) No statement 14 %
7. Report both acute and post-acute mortality (or that information can be determined from the reported data)  
 i) At one year 72.9 %
 ii) Between 2 and 10 years [11, 2932] 8.5 %
 iii) No mortality data/post-acute mortality not estimable/follow-up time unclear [21, 2428] 10.2 %
  1. *Reported proportions (%) corrected to nearest whole number. Study quality and risk of bias was assessed on selection, comparability, and reported outcomes using the Newcastle Ottawa Score checklist (Additional file 1: Table S2). Sepsis case definition evaluates the representativeness of the cohort and ascertainment of exposure by assessing concordance with the sepsis definitions to the definition of Bone et al. [19] or Levy et al. [20]. Studies reporting data from randomized controlled trials (RCTs) are likely to have a lower score as non-sepsis controls are not evaluated