Understanding the demographic, clinical biochemistry and health care data associated with malnutrition in the UK Clinical Practice Research Datalink (CPRD), to support development of The EAMIT Tool (East Anglian Malnutrition Identification Tool)

Malnutrition is a common but underrecognized clinical problem associated with sarcopenia, frailty, fragility fractures, and multiple long-term conditions (1) . Malnutrition increases treatment costs, hospital stays, and secondary care recovery. Current methods of identifying malnutrition rely on direct measures of weight loss and BMI which are often less used and so malnutrition is often missed (2) . As the first step in developing a digital application the EAMIT (East Anglian Malnutrition Identification Tool) for identifying malnutrition, using factors recorded in routine clinical healthcare data, including combinations of demographic factors, health care data (e.g. medications) and clinical biochemistry (e.g. CRP), we examined the data for these factors in the CPRD (2) . Individuals screened with the MUST tool (Malnutrition Universal Screening Tool), older than 65 years, and a matched set of people according to age, sex and GP practice were identified within the CPRD Aurum datalink (Feasbility Study reference FS_002595). Four groups were examined for demographic, health care and clinical biochemistry variables according to MUST screening category result: 1) low risk of malnutrition, 2) medium risk 3) high risk (MUST_H), 4) matched unscreened individuals (MUST_0), and either medians (IQR) or percentage of individuals per group were calculated. The percentage of people over 85Y was 47% in in the MUST_H group, versus 43% in MUST_0. Median, IQR BMI ranged from MUST_0 26.3 kg/m 2 (23.1-29.8) to 18.3 kg/m 2 (16.7-20.9) in MUST_H. Median weight loss was -5.9% in MUST_H compared with 0.1% in MUST_ 0. As expected, 50% of people in MUST_H had a BMI <18.5 kg/m 2 compared with 1% in MUST_0, though the percentage of missing values for BMI was 72%-80% in MUST categories, and 50% in MUST_0. The median number of GP visits (year before screening) was 12 (IQR 6-20) per year in MUST_H compared with 8 (IQR 4-16) in MUST_0. Number of people prescribed 5+ medications was 25% higher in MUST_H than MUST_0. Clinical biochemistry data differed by group. The percentage with low albumin (<35g/L) was 6.5% in MUST_0 compared with 22.8% in MUST_H, iron deficiency as low Hb [<120g g/l (women); <130g/l (men)] ranged from 18% in MUST_0 to 36% in MUST_H. CRP, median IQR, was higher in MUST_H (7; 3-25) compared with MUST_0 (5;2-10). Gradients were found across categories of MUST, and between matched individuals, of greater age, lower BMI, greater percentage of weight loss, lower concentrations of haemoglobin and albumin and higher concentrations of CRP. The number of medications and visits to GPs increased according to category of MUST screening. The proportion of missing BMI data indicates the need to develop a new tool but sufficient demographic, health care and clinical biochemistry data exists within the CPRD database to continue developing an algorithm for identifying risk of malnutrition in the community.