1 Data to Analyze

To answer the research question “Does obesity increase the risk of developing diabetes?” in the U.S. context, we do the following:

1.1 Choose a U.S. data source

Data source: National Health and Nutrition Examination Survey (NHANES) (Disease Control and Prevention 2021)
- 2013-2014,
- 2015-2016,
- 2017-2018
Availability: NHANES is a publicly available dataset that can be downloaded for free from the CDC website.
Design: Observational cross-sectional data.

1.2 Confounder identification

Directed acyclic graph (DAG)

(Greenland, Pearl, and Robins 1999)

flowchart TB
  A[Obesity A] --> Y(Diabetes Y)
  L[Confounders C] --> Y
  L --> A

Hypothesized Directed acyclic graph drawn based on analyst’s best understanding of the literature

Exposure: Being obese

Outcome: Developing diabetes

Confounders: Demographic and lab variables

1.3 Identify measured and unmeasured variables in the data

Find variables capturing the following concepts in the data based on a hypothesized DAG.

Role	Data Component	Variables considered based on DAG
Outcome	DIQ	Have diabetes¹
Exposure	BMX	Obese; BMI >= 30
Confounder	(demographic) DEMO	Age, Sex, Education, Race/ethnicity, Marital status, Annual household income, County of birth, Survey cycle year
	(behaviour) SMQ, PAQ, SLQ, DBQ	Smoking², Vigorous work activity, Sleep³, Diet⁴
	(health history / access) DIQ, HUQ	Diabetes family history, Access to care⁵
	(lab) BPX, BPQ, BIOPRO	Blood pressure (systolic, diastolic⁶), Cholesterol, Uric acid, Total Protein, Total Bilirubin, Phosphorus, Sodium, Potassium, Globulin, Total Calcium

14 demographic, behavioral, health history related variables
- Mostly categorical
11 lab variables
- Mostly continuous

1.4 Analytic data

3 cycles of NHANES datasets were merged:

flowchart LR
  A[NHANES] --> C1(2013-2014 cycle) --> ss1(10,175 \nparticipants)
  A --> C2(2015-2016 cycle) --> ss2(9,971 \nparticipants)
  A --> C3(2017-2018 cycle) --> ss3(9,254 \nparticipants)
  ss1 --> ss(7,585 \nafter \nimposing \neligibility \ncriteria)
  ss2 --> ss
  ss3 --> ss
  style A fill:#FFA500;
  style C1 fill:#FFA500;
  style C2 fill:#FFA500;
  style C3 fill:#FFA500;
  style ss1 fill:#FFA500;
  style ss2 fill:#FFA500;
  style ss3 fill:#FFA500;
  style ss fill:#FFA500;

Our study population was restricted to the U.S. population who were

20 years or older and
not pregnant at the time of survey data collection, and
who had available International Classification of Diseases (ICD) codes to ensure we can extract sufficient proxy information for the analysis (discussed next page).

combination of (a) Doctor told you have diabetes, (b) Taking insulin now, (c) Take diabetic pills to lower blood sugar.↩︎
cigarette use (at least 100 cigarettes in life)↩︎
Sleep hours/workdays↩︎
How healthy is the diet↩︎
Routine place to go for healthcare↩︎
average of 4 measurements↩︎