Extended Data Table 2 Placebo tests with baseline characteristics (Balance tests)

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

  1. *Missingness in baseline variables is shown in Supplementary Table S2.
  2. †The pre-period refers to five years prior to the baseline HbA1c test.
  3. ‡For age, colorectal screening and breast cancer screening, we assumed a quadratic relationship with baseline HbA1c.
  4. BMI = body mass index. GP = General practitioner. P.p. = percentage points. This table shows placebo tests that demonstrate exchangeability of patients above and below the eligibility threshold. We used local linear regression with heteroskedasticity-robust standard errors and triangular kernel weights, within the bandwidth used in our primary analysis. For gender, vaccinations, lifestyle advice, weight loss counselling and statin adherence, effect estimates can be interpreted as the difference in percentage points at the threshold. The results for routine breast screening are restricted to women only. The results for statin adherence are restricted to patients with a record of statin prescription in the pre-period.
Back to article page