The impact of paying treatment providers for outcomes: difference-in-differences analysis of the 'payment by results for drugs recovery' pilot

Abstract

Aims: To estimate the effect on drug misuse treatment completion of a pilot scheme to pay service providers according to rates of recovery.

Design: A controlled, quasi-experimental (difference-in-differences) observational study using multi-level random effects logistic regression.

Setting: Drug misuse treatment providers in all 149 commissioning areas in England in the financial years 2011-12 and 2012-13.

Participants: Service users treated in England in 2011-12 and 2012-13.

Intervention and comparators: Linkage of provider payments to performance indicators in eight pilot commissioning areas in England compared with all 141 non-pilot commissioning areas in England.

Measurements: Recovery was measured by successful completion of treatment (free from drugs of dependence) and engagement with services was measured by rates of declining to continue with treatment.

Findings: Following the introduction of the pilot scheme, service users treated in pilot areas were 1.3 percentage points [odds ratio (OR) = 0.859; 95% confidence interval (CI) = 0.788, 0.937] less likely to complete treatment compared with those treated in comparison areas. Service users treated in pilot areas were 0.9 percentage points (OR = 2.934; 95% CI = 2.094, 4.113) more likely to decline to continue with treatment compared with those treated in comparison areas.

Conclusions: In the first year of the pilot 'Payment by Results for Drugs Recovery' scheme in England, linking payments to outcomes reduced the probability of completing drug misuse treatment and increased the proportion service users declining to continue with treatment.

Keywords: Difference-in-differences; drugs policy; multi-level modelling; pay-for-performance; payment-by-results; performance measurement; policy analysis; random effects; unintended consequences.