Skip to main content

Advertisement

Log in

How does the quality of care for type 2 diabetic patients benefit from GPs-nurses’ teamwork? A staggered difference-in-differences design based on a French pilot program

  • Research article
  • Published:
International Journal of Health Economics and Management Aims and scope Submit manuscript

Abstract

In many countries, policies have explicitly encouraged primary care teams and inter-professional cooperation and skill mix, as a way to improve both productive efficiency gains and quality improvement. France faces barriers to developing team working as well as new and more advanced roles for health care professionals including nurses. We aim to estimate the impact of a national pilot experiment of teamwork between general practitioners (GPs) and advance practitioners nurses (APN)–who substitute and complement GPs–on yearly quality of care process indicators for type two diabetes patients (T2DP). Implemented by a not-for-profit meso-tier organisation and supported by the Ministry of Health, the pilot relied on the voluntary enrolment of newly GPs from 2012 to 2015; the staffing and training of APNs; skill mixing and new remuneration schemes. We use latent-response formulation models, control for endogeneity and selection bias by using controlled before-after and quasi-experimental design combining coarsened exact matching–prior to the treatment, at both GPs (435 treated vs 973 control) and T2DP levels –, with intention to treat (ITT; 18,310 in each group) and per protocol (PP, 2943 in each group) perspectives, as well as difference-in-differences estimates on balanced panel claims data from the National Health Insurance Fund linked to clinical data over the period 2010–2017. We show evidence of a positive and significant positive impact for T2DP followed-up by newly enrolled GPs in the pilot compared to the pretreatment period and the control group. The effect magnitudes were larger for PP than for ITT subsamples.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7

Similar content being viewed by others

Notes

  1. See for further details: https://geodes.santepubliquefrance.fr/#c=home and https://www.santepubliquefrance.fr/les-actualites/2020/journee-mondiale-du-diabete-14-novembre-2020.

  2. See Federico Rios-Avila website, https://friosavila.github.io/playingwithstata/main_csdid.html.

References

Download references

Acknowledgements

We would like to acknowledge the Directorate of Strategy, Studies and Statistics (DSES) of the National Health Insurance (NHI) and the not‐for‐profit organization Asalée for having facilitated accessibility to the dataset and especially the following people: Claude Gissot, Frédéric Bousquet, Pierre Bergman, Mehdi Gabbas, and Brice Dufresne (DSES) and Jean Gautier and Amaury Derville (Asalée). We are also grateful to Denis Raynaud for his support and Fabien Daniel and Charlie Menard for early involvement on data collection and management. This research was undertaken within the framework program of the Institute for Research and Information in Health Economics (IRDES). Finally, we warmly thank all the staff and health care professionals involved for their support.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Julien Mousquès.

Ethics declarations

Ethical approval

Decree 2014‐1181 of 13 October 2014 from the French Council of State.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix 1

Appendix 1

Callaway and Sant’Anna event study DiD estimator

CS estimator is a two-step estimation process. The first step consists in estimating the unit cohort and time-specific treatment effects and the second step aggregating them to produce measures of overall treatment effects.

The first step is based on the following general formula of the ATT, with G the time period when a unit first becomes treated, for a unit g (group) at time t, the so-called group-time ATT:

  • \(ATT\left(g,t\right)={\mathbb{E}} \left[{Y}_{t}\left(g\right)-{Y}_{t}\left(0\right)| {G}_{ g}=1\right]\)

Where \(ATT\left(g,t\right)\) is the expected difference between the observed outcome for the treated and for the untreated at time t. It allows for heterogeneity in ATT across cohorts or groups or over time. Following the notation suggests by Rios-Avila,Footnote 2 with never treated group as counterfactual as it is the case here, the ATT, for the group (g) at time t, could be specified as follows:

  • \(ATT\left(g,t\right)=\left[{EY\left(g\right)}_{t}-{EY(NT)}_{t}\right]-\left[{EY\left(g\right)}_{g-1}-{EY(NT)}_{g-1}\right]\)where g is the index for treated cohorts or groups (the year when g is treated) and t is the year index (t = 2010,…, 2017), EY the expected value of the outcome, at time t or g-1 (one year before the treatment), for the treated (g) or never treated (NT) groups. The first part calculates the differences in outcomes at time t, the second the difference at time g-1. When t < g, ATT(g,t) can be used to test the robustness of the parallel trend assumption.

  • \({m}_{g,t}^{nt}\left(X\right)={\mathbb{E}} \left[{Y}_{t}\left(0\right)-{Y}_{t-1}\left(0\right)| X, {G}_{ g}=1\right]= {\mathbb{E}} \left[{Y}_{t}\left(0\right)-{Y}_{t-1}\left(0\right)| X, NT=1\right]\)

The group-time ATT is identified following a conditional parallel trends assumption based on the never treated (nt) group (3), estimate through the doubly robust estimator (dr) (Sant’Anna & Zhao, 2020), here with a vector of pre-treatment covariates X measuring patient characteristics (age quartile class, gender, NHI scheme, complementary public health insurance for the deprived population, and groups of diabetes drug therapy) and GP or patient list characteristics (age, gender, practice location 2SFA group, the number of registered patients encountered at least once during the year and their proportions: female, age 60–69 years old, age over 70, beneficiaries of the NHI health insurance for the deprived population, and beneficiaries of the salaried worker NHI schemes).

Finally, under these above-mentioned assumptions, group time ATTs are semi-parametrically point-identified:

  • \({ATT}_{dr}^{nt}\left(g,t\right)={\mathbb{E}} \left[\left(\frac{{G}_{g}}{{\mathbb{E}}\left[{G}_{g}\right]}-{\frac{\frac{{p}_{g}\left(X\right) NT}{{1-p}_{g}\left(X\right)}}{{\mathbb{E}}\left[\frac{{p}_{g}\left(X\right) NT}{{1-p}_{g}\left(X\right)}\right]}}\right)\left({Y}_{t}-{Y}_{g-1}-{m}_{g,t}^{nt}\left(X\right)\right)\right]\)

This is a weighted (in the first part of the equation) average of the difference in the outcome. This re-weighting process aims to balanced covariates for both treated and control groups. To give the intuition of how it works let says that you up-weight or down-weight observations depending on how similar or not they are between treated and control groups, and you did this for each group g and time t.

Considering \({\beta }^{g,t}\)=ATT (g,t) under limited anticipation assumptions and homogenous treatment effects, it can be obtained by running the following population linear regression:

  • \({Y}^{g,t}={\alpha }_{1}^{g,t}+{\alpha }_{2}^{g,t}\cdot {G}_{g}+{\alpha }_{3}^{g,t}\cdot 1\left\{T=t\right\}+ {\beta }^{g,t}\cdot ({G}_{g}\times 1\left\{T=t\right\})+\gamma \cdot {X}+{\varepsilon }^{g,t}\)

Also, it is worth noting that standard errors are clustered at the GP level.

Since Callaway and Sant’Anna DiD specification for staggered treatment proposed numerous ATT’s, they also propose useful aggregated ATT (AGGTT) that follows this general formula:

  • \(AGGTT\left(g,t\right)=\frac{\sum \left({w}_{g,t}*ATT\left(g,t\right)\right)}{\sum \left({w}_{g,t}\right)}\)

where w(g,t)_is a weight of how much information was used to estimate ATT(g,t) (Fig. 

Fig. 8
figure 8

Per protocol subsample, ATT estimates for fundoscopy/ophthalmologist visit, depending on the group or cohort and time, over the period 2010–2017

8).

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Gilles de la Londe, J., Afrite, A. & Mousquès, J. How does the quality of care for type 2 diabetic patients benefit from GPs-nurses’ teamwork? A staggered difference-in-differences design based on a French pilot program. Int J Health Econ Manag. 23, 433–466 (2023). https://doi.org/10.1007/s10754-023-09354-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10754-023-09354-z

Keywords

JEL Classification

Navigation