Abstract
Objectives
Unwarranted variation refers to differences in treatments that are not driven by patients’ medical needs. It reflects inefficiencies and inequities in the provision of health care. At present, a method for attribution of practice variation towards health care providers is lacking; more specifically, how much of the practice variation is caused by actual patient needs, the rate of General Practitioner (GP) referrals, or differences in specialists’ treatments remains unclear. This omission has sparked debates by health care providers, and hampers the design and implementation of effective policy. This paper presents a novel method for the decomposition of practice variation based on different stages of the patient journey. Our application aims (1) to decompose regional practice variation in the treatment of varicose veins in the Netherlands and (2) to investigate where the variation originates: at the specialist’s stage (i.e. decision whether or not to treat the patient), or at a prior stage (i.e. patients’ preference or GP’s decision to refer).
Methods
Claims data for 7 million health insurance plan holders were used to calculate and compare the two probabilities of interest: (a) the probability of a GP referral to a specialist and (b) the probability of treatment conditional on being referred for varicose veins in different municipalities. Relevant need and preference factors were identified, and the calculated probabilities were adjusted for the former using logistic regression. Bootstrapped coefficients of variation were computed for the adjusted probabilities.
Results
The logistic model explained 16% of the regional variation in the probability of a GP referral and 53% of the regional variation in the probability of treatment. The unweighted coefficient of variation of the needs-adjusted probability of referral is consistently higher than that of the need-adjusted probability of referral: higher coefficients were observed in 98.2% of the bootstrapped samples.
Discussion
Patient preferences and the GP’s referral habit explain more of the practice variation than the specialist’s propensity to provide treatment for varicose veins, even after adjusting for patients’ need factors. The choice made by the specialist only partially compensates for the impact of the patient’s propensity to seek medical help and for the referral habit of the GP. Our method can be applied to other treatments in order to obtain in-depth insights into regional practice variation.
Unwarranted variation refers to differences in treatments that are not driven by patients’ medical needs. It reflects inefficiencies and inequities in the provision of health care. At present, a method for attribution of practice variation towards health care providers is lacking; more specifically, how much of the practice variation is caused by actual patient needs, the rate of General Practitioner (GP) referrals, or differences in specialists’ treatments remains unclear. This omission has sparked debates by health care providers, and hampers the design and implementation of effective policy. This paper presents a novel method for the decomposition of practice variation based on different stages of the patient journey. Our application aims (1) to decompose regional practice variation in the treatment of varicose veins in the Netherlands and (2) to investigate where the variation originates: at the specialist’s stage (i.e. decision whether or not to treat the patient), or at a prior stage (i.e. patients’ preference or GP’s decision to refer).
Methods
Claims data for 7 million health insurance plan holders were used to calculate and compare the two probabilities of interest: (a) the probability of a GP referral to a specialist and (b) the probability of treatment conditional on being referred for varicose veins in different municipalities. Relevant need and preference factors were identified, and the calculated probabilities were adjusted for the former using logistic regression. Bootstrapped coefficients of variation were computed for the adjusted probabilities.
Results
The logistic model explained 16% of the regional variation in the probability of a GP referral and 53% of the regional variation in the probability of treatment. The unweighted coefficient of variation of the needs-adjusted probability of referral is consistently higher than that of the need-adjusted probability of referral: higher coefficients were observed in 98.2% of the bootstrapped samples.
Discussion
Patient preferences and the GP’s referral habit explain more of the practice variation than the specialist’s propensity to provide treatment for varicose veins, even after adjusting for patients’ need factors. The choice made by the specialist only partially compensates for the impact of the patient’s propensity to seek medical help and for the referral habit of the GP. Our method can be applied to other treatments in order to obtain in-depth insights into regional practice variation.
| Original language | English |
|---|---|
| Publication status | Accepted/In press - 2 Jul 2020 |
Keywords
- Health care
- Hospital
- Primary care
- Unwarranted practice variation
