Table 3.

Methodological choices in the area of ‘data analysis’

Topics	Results from studies in inflammatory arthritis (n=58)		Aspects identified by authors of SLRs in other chronic diseases (n=24)
	n/N* (%)	Detailed information	N† (%)	Comments
10. Accounting for skewness of the outcome	10/‡52 (19%)	Sick leave is reported as positively skewed (zero-inflated) in 10/46 (22%) studies. The authors accounted for skewness by: Dichotomising the outcome: 2/10 (20%)20 63 Analysing the outcome as categorical variable: 2/10 (20%)46 50 Using non-parametric bootstrapping: 4/10 (40%)22 45 47 65 Using zero-inflated models‡: 2/10 (20%)58 59 Presenteeism was reported as zero-inflated in 1/40 (3%) study. The authors accounted for skewness by using zero-inflated models‡; 1/1 (100%)58	Not reported	–
11. Accounting for interdependence of outcomes	49/52§ (94%)	Interdependence was accounted for in 49 studies by—n/N*: Assessing sick leave only among employed patients: 43/46 (93%)18–22 24–27 29 31–39 42–47 50–56 58–60 63 65–69 71 73 Assessing presenteeism among employed patients that were not in sick leave: 37/40 (90%)18–22 24–26 31–39 42 43 50–56 58–60 62–64 67–69 71 73	Not reported	–
12. Accounting for contextual factors	25/58 (43%)	Contextual factors were accounted for in 8/22 (36%) RCTs and 17/36 (47%) OBS—n/N*: As covariates/confounders¶ Personal factors: sociodemographics (7/8 RCTs,19 25 36 37 49 67 69 88% and 15/17 OBS,16 20 27 41 44 46 50 56–58 61 65 66 70 72 88%) Work-related factors: workplace support (1/17 OBS50, 6%), nature of work (4/17 OBS,27 41 58 70 24%) Effect modification¶ Personal factors: sociodemographics (1/8 RCT,52 13% and 2/17 OBS,23 63 12%) Work-related factors: nature of work (1/17 OBS,63 6%)	12 (50%)	Adjustment for contextual/confounder factors in the included studies, if any, is performed only for very few factors 101 105 107 109 110 112–117 120

*The denominator may vary according to the topic assessed.

†Number of systematic literature reviews in other chronic diseases in which the authors report on the corresponding topic.

‡Tillet et al (2017) accounted for skewness of both sick leave and presenteeism.

§Studies addressing work status only were excluded from the denominator as interdependence between work outcome domains does not apply to them.

¶Different contextual factors may have been accounted for in the same study.

n/N, number of original studies in which the methodological choice was identified/number of studies in which the topic was possible to assess; OBS, observational longitudinal study; RCT, randomised controlled trial.