Using longitudinal data from the Health Survey for England to resolve discrepancies in thresholds for haemoglobin in older adults.

Anaemia increases with age and is common among older people. Due to its relationship with morbidity and mortality, accurate diagnosis is important. Thresholds defining the diagnosis of anaemia have been the subject of considerable scientific debate, with both higher and lower cut-offs proposed. High haemoglobin is also a health risk in some but not all studies. Using nationally representative data of 5,329 adults aged 65 + years (Health Survey for England 1998, 2005, 2006), linked to administrative mortality data, this paper describes the relationship between haemoglobin levels and mortality, adjusted for age and other confounders. Among men, a reverse J shaped relationship was observed: relative to the modal group (140-149 g/l), those with 'mild anaemia' of 120-129 g/l haemoglobin had a 56% (95% confidence interval 24-96%) greater mortality hazard, and those with 'severe anaemia', haemoglobin <120 g/l, had an 87% (39-153%) greater hazard. At the other end of the range, those with haemoglobin ≥160 g/l had 32% (2-70%) greater mortality hazard. Haemoglobin levels in women showed a similar but smaller, non-significant pattern: hazard ratio 1·32 (0·91-1·92) for severe anaemia (<110 g/l), and 1·30 (0·95-1·79) for high haemoglobin (≥150 g/l). This research supports the use of the World Health Organization thresholds (130 g/l for men, 120 g/l for women).