Predicting the onset of type 2 diabetes using wide and deep learning with electronic health records

SummaryObjective: To improve the transparency of clinical trial generalizability and to illustrate the method using Type 2 diabetes as an example.Methods: Our data included 1,761 diabetes clinical trials and the electronic health records (EHR) of 26,120 patients with Type 2 diabetes who visited Columbia University Medical Center of New-York Presbyterian Hospital. The two populations were compared using the Generalizability Index for Study Traits (GIST) on the earliest diagnosis age and the mean hemoglobin A1c (HbA1c) values.Results: Greater than 70% of Type 2 diabetes studies allow patients with HbA1c measures between 7 and 10.5, but less than 40% of studies allow HbA1c<7 and fewer than 45% of studies allow HbA1c>10.5. In the real-world population, only 38% of patients had HbA1c between 7 and 10.5, with 12% having values above the range and 52% having HbA1c<7. The GIST for HbA1c was 0.51. Most studies adopted broad age value ranges, with the most common restrictions excluding patients >80 or <18 years. Most of the real-world population fell within this range, but 2% of patients were <18 at time of first diagnosis and 8% were >80. The GIST for age was 0.75. Conclusions: We contribute a scalable method to profile and compare aggregated clinical trial target populations with EHR patient populations. We demonstrate that Type 2 diabetes studies are more generalizable with regard to age than they are with regard to HbA1c. We found that the generalizability of age increased from Phase 1 to Phase 3 while the generalizability of HbA1c decreased during those same phases. This method can generalize to other medical conditions and other continuous or binary variables. We envision the potential use of EHR data for examining the generaliz-ability of clinical trials and for defining population-representative clinical trial eligibility criteria.Citation: Weng C, Li Y, Ryan P, Zhang Y, Liu F, Gao J, Bigger JT, Hripcsak G. A distribution-based method for assessing the differences between clinical trial target populations and patient populations in electronic health records. Appl Clin Inf 2014; 5: 463–479 http://dx.doi.org/10.4338/ACI-2013-12-RA-0105

Download Full-text

Substance use and mental diagnoses among adults with and without type 2 diabetes: Results from electronic health records data

Drug and Alcohol Dependence ◽

10.1016/j.drugalcdep.2015.09.003 ◽

2015 ◽

Vol 156 ◽

pp. 162-169 ◽

Cited By ~ 25

Author(s):

Li-Tzy Wu ◽

Udi E. Ghitza ◽

Bryan C. Batch ◽

Michael J. Pencina ◽

Leoncio Flavio Rojas ◽

...

Keyword(s):

Type 2 Diabetes ◽

Substance Use ◽

Electronic Health Records ◽

Health Records ◽

Electronic Health

Download Full-text

Discovering the Type 2 Diabetes in Electronic Health Records Using the Sparse Balanced Support Vector Machine

IEEE Journal of Biomedical and Health Informatics ◽

10.1109/jbhi.2019.2899218 ◽

2020 ◽

Vol 24 (1) ◽

pp. 235-246 ◽

Cited By ~ 12

Author(s):

Michele Bernardini ◽

Luca Romeo ◽

Paolo Misericordia ◽

Emanuele Frontoni

Keyword(s):

Type 2 Diabetes ◽

Support Vector Machine ◽

Electronic Health Records ◽

Support Vector ◽

Health Records ◽

Electronic Health

Download Full-text

Epidemiology and determinants of non-diabetic hyperglycaemia and its conversion to type 2 diabetes mellitus, 2000–2015: cohort population study using UK electronic health records

BMJ Open ◽

10.1136/bmjopen-2020-040201 ◽

2020 ◽

Vol 10 (9) ◽

pp. e040201 ◽

Cited By ~ 1

Author(s):

Rathi Ravindrarajah ◽

David Reeves ◽

Elizabeth Howarth ◽

Rachel Meacock ◽

Claudia Soiland-Reyes ◽

...

Keyword(s):

Diabetes Mellitus ◽

Type 2 Diabetes ◽

Type 2 Diabetes Mellitus ◽

Electronic Health Records ◽

Secondary Outcome ◽

Health Records ◽

Conversion Rates ◽

Electronic Health ◽

Over Time

ObjectivesTo study the characteristics of UK individuals identified with non-diabetic hyperglycaemia (NDH) and their conversion rates to type 2 diabetes mellitus (T2DM) from 2000 to 2015, using the Clinical Practice Research Datalink.DesignCohort study.SettingsUK primary Care Practices.ParticipantsElectronic health records identified 14 272 participants with NDH, from 2000 to 2015.Primary and secondary outcome measuresBaseline characteristics and conversion trends from NDH to T2DM were explored. Cox proportional hazards models evaluated predictors of conversion.ResultsCrude conversion was 4% within 6 months of NDH diagnosis, 7% annually, 13% within 2 years, 17% within 3 years and 23% within 5 years. However, 1-year conversion fell from 8% in 2000 to 4% in 2014. Individuals aged 45–54 were at the highest risk of developing T2DM (HR 1.20, 95% CI 1.15 to 1.25— compared with those aged 18–44), and the risk reduced with older age. A body mass index (BMI) above 30 kg/m2 was strongly associated with conversion (HR 2.02, 95% CI 1.92 to 2.13—compared with those with a normal BMI). Depression (HR 1.10, 95% CI 1.07 to 1.13), smoking (HR 1.07, 95% CI 1.03 to 1.11—compared with non-smokers) or residing in the most deprived areas (HR 1.17, 95% CI 1.11 to 1.24—compared with residents of the most affluent areas) was modestly associated with conversion.ConclusionAlthough the rate of conversion from NDH to T2DM fell between 2010 and 2015, this is likely due to changes over time in the cut-off points for defining NDH, and more people of lower diabetes risk being diagnosed with NDH over time. People aged 45–54, smokers, depressed, with high BMI and more deprived are at increased risk of conversion to T2DM.

Download Full-text

Assessing occurrence of hypoglycemia and its severity from electronic health records of patients with type 2 diabetes mellitus

Diabetes Research and Clinical Practice ◽

10.1016/j.diabres.2016.09.012 ◽

2016 ◽

Vol 121 ◽

pp. 192-203 ◽

Cited By ~ 28

Author(s):

Anthony P. Nunes ◽

Jing Yang ◽

Larry Radican ◽

Samuel S. Engel ◽

Karen Kurtyka ◽

...

Keyword(s):

Diabetes Mellitus ◽

Type 2 Diabetes ◽

Type 2 Diabetes Mellitus ◽

Electronic Health Records ◽

Health Records ◽

Electronic Health

Download Full-text

Rationale, design and population description of the CREDENCE study: cardiovascular risk equations for diabetes patients from New Zealand and Chinese electronic health records

European Journal of Epidemiology ◽

10.1007/s10654-021-00795-9 ◽

2021 ◽

Author(s):

Jingyuan Liang ◽

Romana Pylypchuk ◽

Xun Tang ◽

Peng Shen ◽

Xiaofei Liu ◽

...

Keyword(s):

Type 2 Diabetes ◽

Cardiovascular Risk ◽

New Zealand ◽

Electronic Health Records ◽

Developing Country ◽

Risk Prediction ◽

Health Records ◽

Cardiovascular Risk Prediction ◽

Electronic Health

AbstractThe cardiovascular risk equations for diabetes patients from New Zealand and Chinese electronic health records (CREDENCE) study is a unique prospectively designed investigation of cardiovascular risk in two large contemporary cohorts of people with type 2 diabetes from New Zealand (NZ) and China. The study was designed to derive equivalent cardiovascular risk prediction equations in a developed and a developing country, using the same epidemiological and statistical methodology. Two similar cohorts of people with type 2 diabetes were identified from large general population studies in China and New Zealand, which had been generated from longitudinal electronic health record systems. The CREDENCE study aims to determine whether cardiovascular risk prediction equations derived in patients with type 2 diabetes in a developed country are applicable in a developing country, and vice versa, by deriving and validating equivalent diabetes-specific cardiovascular risk prediction models from the two countries. Baseline data in CREDENCE was collected from October 2004 in New Zealand and from January 2010 in China. In the first stage of CREDENCE, a total of 93,207 patients (46,649 from NZ and 46,558 from China) were followed until December 31st 2018. Median follow-up was 7.0 years (New Zealand) and 5.7 years (China). There were 5926 (7.7% fatal) CVD events in the New Zealand cohort and 3650 (8.8% fatal) in the Chinese cohort. The research results have implications for policy makers, clinicians and the public and will facilitate personalised management of cardiovascular risk in people with type 2 diabetes worldwide.

Download Full-text