In-hospital screening for diabetes mellitus with HbA1c in an internal medicine department was not useful; a prospective pilot study

Abstract

Background. Screening inpatients for diabetes mellitus may be a good opportunity to detect undiagnosed cases and several studies have demonstrated the feasibility and usefulness of this practice. HbA1c has been suggested as the method of choice due to the effects of acute illness on glucose. The aim of this study was to evaluate a screening protocol based on HbA1c to identify inpatients with undiagnosed diabetes mellitus in an internal medicine department. Methods. We conducted a prospective study of all admissions in the internal medicine department of a 412-bed community hospital in Greece during a 6-month period. Candidates for screening based on the American Diabetes Association’s recommendations were screened with HbA1c. Patients with very poor health status and patients with conditions that may interfere with HbA1c measurement or interpretation were excluded. Results. Of 463 patients (median age 74) only a small proportion (14.9%) were candidates for screening with HbA1c. Known diabetes mellitus, a low admission glucose, severe anemia or blood loss and poor health status were the most common reasons of exclusion. Among the 55 screened patients, 7 had diabetes (based on HbA1c ≥ 6.5%). However, in only 1 of them HbA1c was above target considering the patients’ health status. Categorical agreement (no diabetes, prediabetes, diabetes) between morning glucose and HbA1c was low. However, the concordance between a morning glucose < 125 mg/dl and HbA1c < 6.5% was > 90%. Conclusions. In settings similar to ours (very elderly patients, high rate of conditions that confound the use of HbA1c and high rate of patients with poor health status), untargeted screening of inpatients with HbA1c is unlikely to be cost-effective. A morning glucose during hospitalization may be a better first step for screening.

Assessment of glycaemic status in adult hospital patients for the detection of undiagnosed diabetes mellitus: A systematic review

“…Two studies 26,35 reported similar findings when a blood glucose threshold of >7.0 mmol/L was applied. Karakonstantis et al 35 report that 21% of participants with a blood glucose >7.0 mmol/L had an HbA1c ≥48 mmol/mol (≥6.5%), compared with 8% of participants with a blood glucose <7.0 mmol/L.…”

“…Four studies took place in the United States, 25‐28 four in Australia 29‐32 and one each in South Africa, 33 Ireland, 34 Greece 35 and Guyana 36 . Ethnicity was not reported for seven out of 12 studies 25,29‐31,33‐35 . Out of five studies reporting participants’ mean age, 26,27,29,32,33 the highest was 67 years, 27 and the lowest was 50.6 years 33 .…”

“…All but one 33 carried out HbA1c testing during admission rather than outpatient follow‐up, and three of these 26,31,32 tested HbA1c on blood taken at the same time as the sample used for random glucose testing. Out of the 12 studies, seven used HbA1c ≥48 mmol/mol (≥6.5%) to diagnose diabetes, 29,30,32‐36 one used HbA1c >48 mmol/mol (>6.5%), 25 and four used values that were lower than 48 mmol/mol (6.5%) 26‐28,31 29 where 7.5% of hyperglycaemic participants were followed up with HbA1c, HbA1c reference testing was consistently high; in 10 studies 90% or more hyperglycaemic participants were assessed for diabetes using HbA1c 25‐28,31‐36 …”

“…These included study populations consisting only of elective surgical patients, 33 or those selected using diabetes risk factors. 25,35 A proportion of the participants studied by Wexler et al 27 were admitted to ICU, so the study was considered to have a high degree of concern regarding applicability. Hng et al 30 was also considered to have limited applicability regarding patient selection as the inclusion criteria also included people aged between 16 and 18 years.…”

“…29,34 Six studies were considered at high risk of bias for patient selection. One used extensive exclusion criteria that may limit generalisability of the results to the realworld setting, 35 whereas the other five recruited study populations by convenience sampling. 25,26,31,33,36 One of these 25 was a case control study, and for the purposes of data extraction, data from the study and control populations were assessed; however, there is a high risk of bias with this method of interpretation because the selective sampling method may have influenced pre-test probability for the studied population.…”