Differential costs for the non-adult ADHD population in Catalonia

Health Economics Review

Table 3 Differential annual costs: alternative estimation results

	IV estimation
Matching neighbouring	693.71 (26.25)^c
Propensity score matching	634.26 (10.71)^c
Two-part model	632.27 (8.24)^c
Two-part model (matched units)	609.75 (8.45)^c
FMM 2 latent class (matched units) for positive costs
Class1	757.67 (2.18)^c
Class2	420.04 (0.80)^c

^a, ^b and ^c denote statistical significance levels 10%, 5% and 1%, respectively. The sample size is 967,684; 967,638; 767,099; 84,995; and 124,720, respectively. Standard errors are in parenthesis. Standard errors for FMM estimations are based on mean differences across predictions for both classes. Class 1 refers to regular healthcare users, whereas Class 2 identifies heavy healthcare users after dropping those individuals with high costs. All regressions include as controls: the quarter of birth, age, gender, nationality, and the considered comorbidities (overweight, learning disability, depression, asthma, anxiety, conduct disorder, and defiant disorder), copayment levels plus pharmacy limits and basic health areas fixed effects

ISSN: 2191-1991