Diabetes diagnosis based on glucose control levels and time until diagnosis: a regression discontinuity approach to assess the effect on direct healthcare costs

Mora, Toni; Rodríguez-Sánchez, Beatriz

doi:10.1186/s13561-025-00613-y

Table 4 Effects of average glucose level on healthcare costs from the fuzzy regression discontinuity models excluding outliers >99%. Forcing variable: distance from cut-off HbA1c = 6.5%

From: Diabetes diagnosis based on glucose control levels and time until diagnosis: a regression discontinuity approach to assess the effect on direct healthcare costs

	Model 1^a				Model 2^b
	30-day delay	45-day delay	60-day delay	75-day delay	30-day delay	45-day delay	60-day delay	75-day delay
	6-month span
Differences in total healthcare costs for treated (HbA1c ≥ 6.5%) vs. controls (HbA1c < 6.5%)	-3,886.8** (1,910.6)	-3,234.2** (1,573.5)	-2,950.9** (1,453.2)	-2,668.5** (1,314.5)	-3,437.6** (1,661.1)	-2,737.2** (1,357.2)	-2,491.5** (1,238.8)	-2,262.6** (1,121.1)
Number of observations, left of the cut-off	85,889	85,889	85,889	85,889	85,794	85,794	85,794	85,794
Number of observations, right of the cut-off	72,009	72,009	72,009	72,009	71,924	71,924	71,924	71,924
	9-month span
Differences in total healthcare costs for treated (HbA1c ≥ 6.5%) vs. controls (HbA1c < 6.5%)	-5,789.4*** (1,809.4)	-4,733.1*** (1,467.8)	-4,266.2*** (1,314.9)	-3,931*** (1,211.3)	-4,536.5*** (1,663.5)	-3,586.1*** (1,311.6)	-3,236.3*** (1,176.3)	-3,030*** (1,108.4)
Number of observations, left of the cut-off	70,690	70,690	70,690	70,690	70,610	70,610	70,610	70,610
Number of observations, right of the cut-off	56,945	56,945	56,945	56,945	56,876	56,876	56,876	56,876
	12-month span
Differences in total healthcare costs for treated (HbA1c ≥ 6.5%) vs. controls (HbA1c < 6.5%)	-5,064.8*** (1,907.3)	-4,265.3*** (1,559.8)	-3,893.9*** (1,417.9)	-3,656.1*** (1,328.8)	-3,243.5** (1,573	-2,480.2** (1,225.3)	-2,277.7** (1,110.2)	-1,997.2** (1,008.1)
Number of observations, left of the cut-off	54,677	54,677	54,677	54,677	54,618	54,618	54,618	54,618
Number of observations, right of the cut-off	42,850	42,850	42,850	42,850	42,802	42,802	42,802	42,802
	15-month span
Differences in total healthcare costs for treated (HbA1c ≥ 6.5%) vs. controls (HbA1c < 6.5%)	-4,185** (1,705.3)	-3,078.1** (1,353.1)	-2,834.8** (1,218.2)	-2,809.3** (1,161.3)	-3,503.5** (1,538.1)	-2,962.1** (1,291.8)	-2,711.3** (1,179.7)	-2,544.6** (1,113.7)
Number of observations, left of the cut-off	44,576	44,576	44,576	44,576	44,526	44,526	44,526	44,526
Number of observations, right of the cut-off	34,361	34,361	34,361	34,361	34,320	34,320	34,320	34,320
	18-month span
Differences in total healthcare costs for treated (HbA1c ≥ 6.5%) vs. controls (HbA1c < 6.5%)	-4,319.4*** (1,479.3)	-3,242.5*** (1,166.7)	-3,272.6*** (1,122)	-3,041.7*** (1,046.7)	-3,728.8*** (1,373.7)	-3,145.1*** (1,182.1)	-2,842.1** (1,099.8)	-2,640** (1,031)
Number of observations, left of the cut-off	37,035	37,035	37,035	37,035	36,993	36,993	36,993	36,993
Number of observations, right of the cut-off	28,506	28,506	28,506	28,506	28,468	28,468	28,468	28,468
	21-month span
Differences in total healthcare costs for treated (HbA1c ≥ 6.5%) vs. controls (HbA1c < 6.5%)	-4,840.5*** (1,500.6)	-4,227.2*** (1,307.9)	-3,864.9*** (1,190)	-3,646.7*** (1,130.1)	-3,734.7*** (1,347.4)	-3,211.9*** (1,168.7)	-2,899.3*** (1,067.6)	-2,723.5*** (1,013.1)
Number of observations, left of the cut-off	32,179	32,179	32,179	32,179	32,144	32,144	32,144	32,144
Number of observations, right of the cut-off	24,544	24,544	24,544	24,544	24,510	24,510	24,510	24,510
	24-month span
Differences in total healthcare costs for treated (HbA1c ≥ 6.5%) vs. controls (HbA1c < 6.5%)	-5,343.7*** (1,433)	-4,574.5*** (1,217.8)	-4,269*** (1,135.2)	-4,024.7*** (1,074.8)	-3,471.1*** (1,262.3)	-3,005.9*** (1,094)	-2,793.3*** (1,015.4)	-2,656.7*** (968.43)
Number of observations, left of the cut-off	26,834	26,834	26,834	26,834	26,803	26,803	26,803	26,803
Number of observations, right of the cut-off	20,211	20,211	20,211	20,211	20,187	20,187	20,187	20,187

a Model 1 only includes the forcing variable, which refers to the average glucose (HbA1c) level through a laboratory test, with the cut-off value being 6.5
b Model 2 adjusts for age and its square, gender, being Spaniard, drug co-payment level according to the TSI, the adjusted morbidity group, the comorbidities listed in Table 1 and the inverse Mills ratio (IMR), which denotes the probability of having a laboratory test conditional on individual characteristics and the most frequent provider unit
*: p-value < 0.10; **: p-value < 0.05; ***: p-value < 0.01
Clustered standard errors within brackets at the basic health area. A triangular kernel-weighted local polynomial smoothing has been applied, which gives more weight to points close to the threshold

Back to article page

ISSN: 2191-1991

Contact us

General enquiries: journalsubmissions@springernature.com

Health Economics Review

Contact us