Description of the study population
| Quartiles of maximum C-reactive protein, (range, mg/L) | 1st quartile (0.2–9) | 2nd quartile (10–34) | 3rd quartile (35–101) | 4th quartile (102–754) | p valuea |
|---|---|---|---|---|---|
| Patients, N | 1,200 | 1,459 | 1,641 | 1,652 | |
| Admissions, N | 2,741 | 2,738 | 2,720 | 2,523 | |
| Male, % | 43 | 43 | 50 | 56 | |
| Mean age (years) (95% CI) | 52 (1–102) | 58 (10–106) | 63 (17–108) | 65 (22–107) | |
| Baseline mean Hb (g/L) (95% CI) | 125.6 (104.5–146.7) | 122 (100.2–143.8) | 118.4 (95–141.7) | 116 (92.1–139.8) | <0.0001 |
| Maximum change in Hb (g/L) (95% CI) | −1.4 (–8.6 to 5.8) | −1.6 (–9.8 to 6.6) | −2.8 (–14.1 to 8.4) | −4.9 (–19. 5to 9.6) | <0.0001 |
| Median length of admission, days (IQR) | 0.9 (0.3–2.9) | 1.7 (0.3–5.1) | 3.2 (0.8–8.1) | 6.5 (2.8–12.2) | <0.0001 |
| Median number of blood draws (IQR) | 15 (8–32) | 22 (10–46) | 34 (14–62) | 55 (29–89) | <0.0001 |
CI = confidence interval; Hb = haemoglobin; IQR = Interquartile range.
↵aANOVA for means or Kruskal–Wallis test for medians. Admission model of maximal changes in haemoglobin and C-reactive protein: to test the hypothesis that maximal serum CRP levels were associated with larger decreases in haemoglobin over an admission period, we fitted CRP as a categorical variable because there was a significant departure from a linear trend (p<0.0001, Fig 1).