Extended Data Table 3 Regression discontinuity results of being eligible for and of being referred to, intensive lifestyle counselling

From: Quasi-experimental evaluation of a nationwide diabetes prevention programme

  1. *Effects for HbA1c and diabetes complication were adjusted for diabetes medication prescription; effects for lipid levels were adjusted for lipid-lowering medication prescription; effects for diastolic and systolic blood pressure were adjusted for blood pressure-lowering medication; and effects for mortality and MACE hospitalization were adjusted for all three medication groups. All relevant medications are listed in our Open Science Framework project (see code availability statement).
  2. †Sample size within MSE-optimal bandwidth.
  3. ‡Sample restricted to those without prior lipid-lowering medication prescription. At baseline, 411 288 (20.0%) people had already received at least one prescription for a lipid-lowering medication.
  4. §Sample restricted to those without prior blood pressure-lowering medication prescription. At baseline, 749 884 (36.5%) people had already received at least one prescription for a blood pressure-lowering medication.
  5. BMI = Body mass index. BP = Blood pressure. MACE = Major adverse cardiovascular event. RD = Risk difference (i.e., difference in the probability of the outcome in percentage points). This table shows the main regression discontinuity results for the primary cohort including effect size estimate, 95% confidence intervals (95% CI), P values, mean-squared error (MSE) optimal bandwidth (BW) and sample size. The effects were estimated in local linear regressions with heteroskedasticity-robust standard errors and triangular kernel weight and evaluated using two-sided t-tests (p < 0.05). The definition of all outcomes is detailed in the Supplementary Information 1.2. Additional details are available in Methods (‘Main Analysis’). We compared statistical significance to results using robust bias-corrected confidence intervals, which yielded the same statistical inferences except for diabetes medication, which was no longer significant (Extended Data Table 4). Results for the secondary cohort are available in the Supplementary Information S4.
Back to article page