We conducted a population-based cohort study of individuals with newly diagnosed type 2 diabetes between January 1, 2000, and December 31, 2019 in the IQVIA Medical Research Data (IMRD). IQVIA, incorporating data from The Health Improvement Network (THIN), is a longitudinal, clinical primary care database of over 18 million patient records in the UK (19). Read codes describing concepts related to health in GP records are used to record diagnoses in the IMRD database (20). Collection for data in IMRD was approved by the NHS South East Multi-centre Research Ethics Committee (MREC) in 2003. We obtained approval to conduct this analysis from the Scientific Review Committee (reference number: 21SRC030).

The Data Extraction for Epidemiological Research (DExtER) tool, an extract transform load-based software framework, was used to extract this dataset (21). This platform enables users to extract high-quality and individual-patient-level data from primary care databases (21). The outcome measures (e.g., prevalence) calculated in IMRD datasets extracted from DExtER have produced comparable results to those from Clinical Practice Research Datalink (CPRD) and other national datasets (21).

Adults ≥18 years with a record of type 2 diabetes diagnosis and registered with an eligible practice for at least 1 year before study entry were eligible for the study. Type 2 diabetes diagnosis was ascertained by the presence of any type 2 diabetes clinical (Read) code in the individual’s medical record. Adults with a recording of type 1 diabetes were excluded.

The main exposure was the risk of DFD based on Read codes that have previously been used in a microvascular complications study (22). Based on NICE guidelines, risk of DFD was categorized into three groups—low, moderate, and high (6). We considered individuals with no evidence of diabetic peripheral neuropathy (DPN), no peripheral vascular disease (PVD), and no presence of foot deformity, impairment, or previous ulcer to be at low risk (6, 22, 23). Individuals presenting with deformity, neuropathy, or non-critical limb ischemia were considered to be at moderate risk (6, 22, 23). Previous ulceration, amputation, and more than 2 of 3 parameters of DPN, PVD, or deformity were defined as high risk (6, 22, 23). The outcome was all-cause mortality among those newly diagnosed with type 2 diabetes with low, moderate, and high risk of DFD.

The index date was defined as 15 months following the date of diagnosis of type 2 diabetes (24), which was chosen because of the requirement to measure foot risk soon after diagnosis of diabetes and reassess the risk annually as per NICE guidelines and Quality Outcomes Framework (QOF) in the UK (6, 25). The QOF indicator is defined as the percentage of patients with diabetes with a record of a foot examination and risk classification within the preceding 15 months (25). Follow-up started at the index date of 15 months post type 2 diabetes diagnosis and ended at exit date defined as the occurrence of one of the following events (whichever came earliest): (a) death, (b) individual left the practice, or (c) study end date (December 31, 2019). In an additional analysis, we took an index date of 30 months following diagnosis of type 2 diabetes, giving additional time for foot risk assessment to take place after a diagnosis of type 2 diabetes.

Baseline characteristics included age, sex, BMI (kg/m²), smoking status, ethnicity, social deprivation status, history of CVD, and HbA1c (mmol/mol). BMI was classified according to NICE BMI classification as follows: underweight (BMI of <18.5 kg/m²), normal weight (BMI of 18.5 to <25 kg/m²), overweight (BMI of 25 to <30 kg/m²), obesity class I (BMI of 30 to <35 kg/m²), obesity class II (BMI of 35 to <40 kg/m²), and obesity class III (BMI of ≥40 kg/m²) (25). Smoking status was categorized as smoker, non-smoker, and ex-smoker. Ethnicity was classified based on UK census ethnic groups (White; Black, African, Caribbean, or Black British; Asian or British Asian; mixed or multiple ethnic groups; and other ethnic groups). The Townsend deprivation index of social deprivation status was based on quintiles with 1 being the least deprived and 5 being the most deprived (26). CVD was defined as atrial fibrillation (AF), heart failure, ischemic heart disease (IHD), and stroke and transient ischemic attacks (TIA). HbA1c was categorized as ≤47.5 mmol/mol, 47.5–58.5 mmol/mol, 58.5–69.4 mmol/mol, and >69.4 mmol/mol (24). Drugs included lipid drugs, metformin, insulin, and other diabetes drugs (glitazones, glinides, acarbose, glucagon-like peptide 1 [GLP-1] receptor agonists, dipeptidyl peptidase-4 [DPP-4] inhibitors, sodium-glucose co-transporter-2 [SGLT-2] inhibitors, and sulfonylureas). Missing data for BMI, Townsend deprivation index, smoking status, ethnicity, and HbA1c were assigned to a separate category and included in the analyses.

In the analysis, means (± SD) were used to summarize continuous variables, and percentages were used to summarize categorical variables. Crude and adjusted HR and 95% CIs were calculated for the occurrence of death in DFD risk groups using a Cox proportional model. The log–log plots were used to check proportional hazards assumption with almost parallel curves indicating that the assumption was not violated. Baseline characteristics including age, sex, Townsend score, ethnicity, smoking status, BMI, CVD, HbA1c, and drug use were included as covariates in the regression model. Kaplan–Meier survival curves were generated for different DFD risk groups, and the log-rank test was performed to test the equality of the survivor function between groups.

Sensitivity analysis I involves setting the index date 30 months after diagnosis with type 2 diabetes, and was performed using the same statistical methods as in the main analysis; sensitivity analysis II concerns the exclusion of individuals who had incomplete data on BMI, smoking status, and Townsend score.

We considered 2-tailed p-value <0.05 to be statistically significant. All statistical analyses were conducted using Stata version 16 software.

In total, 225,787 individuals who had been newly diagnosed with type 2 diabetes were included in the study with 77,346 (34%), 14,929 (7%), and 2,808 (1%) at low, moderate, and high risk of DFD, respectively. There were 1,118 (0.05%) individuals who declined the foot examination and 129,586 (57%) who had no recording of foot risk. Figure 1 described the flow of the study population selection. Baseline characteristics are summarized in Table 1. The population was mostly male (55.7%) and over 50 years old (83.6%). Mean (SD) BMI and HbA1c were observed to be similar in three risk groups (BMI: 31.8 kg/m² [6.6] vs. 31.8 kg/m² [7.1] vs. 31.6 kg/m² [7.4] and HbA1c: 51.5 mmol/mol [12.7] vs. 51.5 mmol/mol [12.5] vs. 52.5 mmol/mol [13.2], respectively); 16.3% of the individuals were active smokers. Individuals who refused a foot examination were more likely to be from more deprived groups. The prevalence of hypertension at baseline was higher in both the moderate-risk (59.3%) and high-risk (59.9%) groups compared with those in the low-risk group (51.6%).

Mortality Among Individuals at Risk of DFD. Among the study population, 34,061 (15.1%) died during the study period. A total of 4,322 (5.6%), 1,904 (12.8%), and 549 (19.6%) deaths occurred in those who were at low, moderate, and high risk of DFD, respectively. Among individuals who declined foot examination and those with no recording, there were 178 (15.9%) and 27,108 (20.9%) deaths, respectively.

Table 2 shows the unadjusted and adjusted HRs from the Cox regression model. Compared with low risk of DFD, the unadjusted hazards of mortality were higher for moderate and high DFD risk groups (HR 2.42, 95% CI [2.29, 2.55], p < 0.001; HR 3.77, 95% CI [3.45, 4.11], p < 0.001, respectively). Figure 2 shows the Kaplan–Meier curve for the mortality rate related to foot risk. The graph lines start to separate from the beginning, representing the significant reduction in mortality in the patient group with low risk compared to moderate and high risk. The difference between the curves was statistically significant (p < 0.001).

After adjusting for age, sex, Townsend score, ethnicity, smoking status, baseline BMI, CVD, baseline HbA1c, and medications, those with a moderate risk of DFD had 1.5 times greater risk of mortality (HR 1.50, 95% CI [1.42, 1.58], p < 0.001), and those with a high risk of DFD had double the risk of mortality (HR, 2.01, 95% CI [1.84, 2.20], p < 0.001) compared to those with a low risk of DFD. In addition, those who declined a foot examination or who had no recording were also 75% and 25%, respectively, more likely to die than those with a low risk of DFD (HR 1.75, 95% CI [1.51, 2.04], p < 0.001; HR 1.25, 95% CI [1.20, 1.30], p < 0.001).

Age, Townsend score, smoking status, CVD events (AF, heart failure, IHD, and stroke), HbA1c, and antidiabetic drug use were associated with increased mortality (Table 3). Individuals who were recorded as being more deprived experienced higher mortality risk. Black, African, Caribbean, or Black British (HR 0.69, 95% CI [0.59, 0.80]); Asian or British Asian (HR 0.61, 95% CI [0.54, 0.68]); and mixed or multiple ethnic groups (HR 0.69, 95% CI [0.54, 0.90]) were less likely to die than those from the White ethnic group population. Former and current smokers had a higher risk of mortality compared to those who never smoked (HR 1.25, 95% CI [1.22, 1.28]; HR 1.77, 95% CI [1.72, 1.83], respectively). Moreover, hazards of mortality were lower in individuals with obesity and significantly higher in those who were underweight compared to those with normal weight (BMI 18.5 to <25 kg/m²). Individuals who were categorized as obesity class II (35 to <40 kg/m²) had approximately 25% reduced risk of mortality compared to those with a healthy weight with 0.75 (95% CI [0.72, 0.78]), while underweight individuals (<18.5 kg/m²) had twice the risk of dying (2.24, 95% CI [2.04, 2.45]). A close association was noted between increasing baseline Hba1c and risk of mortality. Those prescribed lipid-lowering drugs had a 29% lower risk of death (HR 0.71, 95% [0.69, 0.73]), compared to those not prescribed.

In the sensitivity analysis where the index date was set to 30 months after the diagnosis of type 2 diabetes, 190,422 people (among whom 28,065 died) with a diagnosis of type 2 diabetes were included in the analysis. There was a similar trend between DFD risk and mortality with an HR of 1.46 (95% CI [1.39, 1.54]) and an HR of 2.04 (95% CI [1.89, 2.21]) in groups with a moderate and high risk of DFD, respectively, compared to the group with a low risk of DFD (Supplementary Tables 1, 2).

After excluding those with incomplete data for BMI, Townsend score, and smoking status in the main dataset (index date of 15 months post diagnosis), 186,862 individuals were available for analysis (28,015 deaths). Individuals with moderate and high risk of DFD remained at higher risk of death than those at low risk of DFD (Supplementary Table 3, HR 1.47, 95% CI [1.39, 1.57]; HR 2.00, 95% CI [1.81, 2.20], respectively).

This study has a number of strengths ensuring a high-quality study with reliable results including a large sample size. Patients in IMRD are broadly representative of the UK population, and thus, these results should be generalizable. Sensitivity analysis where follow-up started 30 months after the diagnosis of type 2 diabetes allows sufficient time for the exposure to be recorded, and this also showed an increased risk of death for higher DFD risk. Further research is needed to explore the underlying reasons for declining foot examination/absence of recorded data, and the elevated mortality identified in these groups.

Limitations include missing data for some covariates; however, it was a small proportion of the total dataset, and the sensitivity analysis excluding those with missing data showed a consistent association. Renal replacement therapy was not considered as high risk of DFD although it was suggested to be at high risk in NICE guidelines (6). However, there were only a small number of participants (≤0.2%) recorded with this therapy, which did not influence the reliability of the results. We did not have information on the cause of death, and this should be considered in further studies to determine important risk factors that should be targeted to reduce the risk of death among people with newly diagnosed type 2 diabetes who are at risk of DFD.