Difference-in-differences Estimate Research Articles

Public Periodontal Screening Increases Subsequent Regular Dental Visits: The Life Study.

A public oral health screening, periodontal disease screening, has been implemented in Japan, but it remains unclear whether screening encourages subsequent regular dental visits. This study aimed to examine whether people who underwent periodontal disease screening were more likely to regularly visit dentists after undergoing the screening than before using a difference-in-differences (DID) approach. This study used health care claims data of municipality residents who underwent periodontal disease screening in 2017 or 2018. For each screening recipient, 4 individuals were extracted from those who did not undergo screening as controls. In the DID analysis, we assessed the change in the proportion of dentist visits at least once every 180 d after undergoing screening. A total of 4,050 participants were included in the analysis. The proportion of participants visiting dentists was consistent throughout the study period (approximately 45%) among the participants who did not undergo the screening. However, among the participants who underwent the screening, while the proportion who visited dentists was consistent before screening (approximately 60%), the proportion was higher after screening (1-180 d after, 81.2%). DID analysis indicated that the proportion increased by 12.9% after the screening. In addition, the age-subgroup DID estimates were higher in the younger population (aged 20-35 y, 13.9%; 40-55 y, 12.8%; 60 y, 12.6%). Among people who underwent periodontal disease screening, a higher proportion visited dentists after undergoing the screening than before, suggesting that periodontal disease screening was associated with an increase in subsequent regular dental visits. Our results provide evidence that a public oral health screening could increase regular dental visits, which has the potential to improve and maintain people's oral health, especially in the younger population.

Estimating community-wide indirect effects of influenza vaccination: triangulation using mathematical models and bias analysis.

Understanding whether influenza vaccine promotion strategies produce community-wide indirect effects is important for establishing vaccine coverage targets and optimizing vaccine delivery. Empirical epidemiologic studies and mathematical models have been used to estimate indirect effects of vaccines but rarely for the same estimand in the same dataset. Using these approaches together could be a powerful tool for triangulation in infectious disease epidemiology because each approach is subject to distinct sources of bias. We triangulated evidence about indirect effects from a school-located influenza vaccination program using two approaches: a difference-in-difference (DID) analysis, and an age-structured, deterministic, compartmental model. The estimated indirect effect was substantially lower in the mathematical model than in the DID analysis (2.1% (95% Bayesian credible intervals 0.4 - 4.4%) vs. 22.3% (95% CI 7.6% - 37.1%)). To explore reasons for differing estimates, we used sensitivity analyses and probabilistic bias analyses. When we constrained model parameters such that projections matched the DID analysis, results only aligned with the DID analysis with substantially lower pre-existing immunity among school-age children and older adults. Conversely, DID estimates corrected for potential bias only aligned with mathematical model estimates under differential outcome misclassification. We discuss how triangulation using empirical and mathematical modelling approaches could strengthen future studies.

The impact of fast-food energy posting on college students’ food purchases

BackgroundThe Patient Protection US Affordable Care Act (ACA) energy posting mandate requires restaurant chains to disclose information on the energy content of their food items. Assessments of the effect of menu energy labeling on dietary choices have reported inconsistent findings. ObjectivesThis study examined the impact of menu energy labeling on food items purchased by college students after the mandate was enacted nationally. MethodsStudent food sales data from purchases made at 3 fast-food restaurants during the 2017/2018 and 2018/2019 academic years at a university campus were used for the analysis. The total sample included 1662 students on the university meal plan; these students generated 145,295 food transactions at the restaurants over the study period. We utilized a difference-in-differences (DiD) empirical strategy, comparing changes in transaction-level energy purchases at 2 fast-food restaurants B and C (FFRB and FFRC – treatment groups) that posted energy information in the summer of 2018 with another fast-food restaurant A (FFRA – control group) that began posting energy information before the study period. ResultsWe observed increases in the average energy content per transaction after implementing the menu-labeling policy. The DiD estimates found an increase of 20.6 in the average calories of energy purchased per transaction at the treatment restaurants relative to the control restaurant. In the subgroup analyses, the DiD estimates indicated calories of energy increased: 18.7 for female students, 20.5 for male students, 23.5 for non-Hispanic Black students, 30.2 for students eligible for federal financial aid, and 19.9 for students not eligible for federal financial aid. ConclusionsThe results suggest that the ACA energy menu-labeling policy led to an increase in the energy content per transaction by students at a public university. This paper highlights the need for more research to better understand the determinants of food choice among college students.

An optimal transport approach to estimating causal effects via nonlinear difference-in-differences

Abstract We propose a nonlinear difference-in-differences (DiD) method to estimate multivariate counterfactual distributions in classical treatment and control study designs with observational data. Our approach sheds a new light on existing approaches like the changes-in-changes estimator and the classical semiparametric DiD estimator, and it also generalizes them to settings with multivariate heterogeneity in the outcomes. The main benefit of this extension is that it allows for arbitrary dependence between the coordinates of vector potential outcomes and includes higher-dimensional unobservables, something that existing methods cannot provide in general. We demonstrate its utility on both synthetic and real data. In particular, we revisit the classical Card & Krueger dataset, which reports fast food restaurant employment before and after a minimum wage increase. A reanalysis with our methodology suggests that these restaurants substitute full-time labor with part-time labor on aggregate in response to a minimum wage increase. This treatment effect requires estimation of the multivariate counterfactual distribution, an object beyond the scope of classical causal estimators previously applied to this data.

Specialized nutritious foods and behavior change communication interventions during the first 1000 d of life to prevent stunting: a quasi-experimental study in Afghanistan

BackgroundConsiderable evidence supports the effectiveness of nutritional supplementation with or without nutrition education in preventing stunting in developing countries, but evidence from Afghanistan is scarce. ObjectivesThis project aimed to assess the effectiveness of specialized nutritious food (SNF), social and behavior change communication (SBCC) intervention to prevent stunting among children under 2 y during the first 1000 d of life in Badakhshan, Afghanistan. MethodsWe used a community-based quasi-experimental pre–post study design with a control group. Pregnant and lactating women received a monthly ration of 7.5 kg of super cereal (250 g/d) during pregnancy and the first 6 mo of breastfeeding. Children aged 6–23 mo received 30 sachets of medium-quantity lipid-based nutrient supplement (50 g/sachet/d) monthly. We compared pre- and postintervention assessments of the intervention and control groups to isolate the effect of the intervention on key study outcomes at the endline by difference-in-differences (DID) estimates. ResultsA total of 2928 and 3205 households were surveyed at baseline and endline. DID estimates adjusted for child, maternal, and household characteristics indicated a significant reduction in stunting (DID: −5% (95% confidence interval [CI]: −9.9, −0.2) and underweight (DID: −4.6% (95% CI: −8.6, −0.5) among children <2 y of age. However, DID estimates for wasting among children in the intervention and control groups were not significantly different (DID: −1.7 (95% CI: −5.1, 1.6). Furthermore, exposure to the SBCC messages was associated with improvements in the early initiation of breastfeeding (DID: 19.6% (95% CI: 15.6, 23.6), exclusive breastfeeding under 6 mo (DID: 11.0% (95% CI: 2.3, 19.7), minimum meal frequency (DID: 23% (95% CI: 17.7, 28.2), and minimum acceptable diet (DID: 13% (95% CI: 9.8, 16.3). ConclusionsThe provision of SNF in combination with SBCC during the first 1000 d of life was associated with reduction in stunting and underweight and improvements in infant and young child feeding practices among children under 2 y of age.This trial was registered at clinicaltrials.gov as NCT04581993.

Machine learning-based causal inference for evaluating intervention in travel behaviour research: A difference-in-differences framework

Causal inference with the difference-in-differences (DID) framework is popular in identifying causal effects with observational data and has started to be applied in recent travel behaviour studies. Most relevant transportation research adopts the conventional linear parametric DID model, which is known to be inflexible and restrictive. This study applies non-parametric DID estimators facilitated by machine learning (ML) models for causal inference in a variety of data scenarios. Semi-parametric and doubly robust estimators are established and integrated with the ML-based cross-fitting pipeline. Simulation studies and empirical case studies are conducted to showcase the ability of ML-based DID to detect causal effects from both simulated and real-world datasets. Results suggest that the proposed methods outperform conventional DID models in all data scenarios. Light working models are generally preferred over hyperparameter-dependent ones for their comparable performance, lower computational burden, and higher levels of compatibility to real-world empirical analysis. Empirical case studies also demonstrate how the proposed DID method could be applied to evaluate the impacts of various interventions on travel behaviour in different contexts. The present study adds to the existing travel behaviour literature by leveraging machine learning algorithms and non-parametric estimators to the impact evaluation of external interventions on travel characteristics and expanding the application of causal inference approaches in transportation research.

Advances in Difference-in-differences Methods for Policy Evaluation Research.

Difference-in-differences (DiD) is a powerful, quasi-experimental research design widely used in longitudinal policy evaluations with health outcomes. However, DiD designs face several challenges to ensuring reliable causal inference, such as when policy settings are more complex. Recent economics literature has revealed that DiD estimators may exhibit bias when heterogeneous treatment effects, a common consequence of staggered policy implementation, are present. To deepen our understanding of these advancements in epidemiology, in this methodologic primer, we start by presenting an overview of DiD methods. We then summarize fundamental problems associated with DiD designs with heterogeneous treatment effects and provide guidance on recently proposed heterogeneity-robust DiD estimators, which are increasingly being implemented by epidemiologists. We also extend the discussion to violations of the parallel trends assumption, which has received less attention. Last, we present results from a simulation study that compares the performance of several DiD estimators under different scenarios to enhance understanding and application of these methods.

Effect of Medicaid expansion on cancer treatment and survival among Medicaid beneficiaries and the uninsured.

The Affordable Care Act expanded Medicaid coverage for people with low income in the United States. Expanded insurance coverage could promote more timely access to cancer treatment, which could improve overall survival (OS), yet the long-term effects of Medicaid expansion (ME) remain unknown. We evaluated whether ME was associated with improved timely treatment initiation (TTI) and 3-year OS among patients with breast, cervical, colon, and lung cancers who were affected by the policy. Medicaid-insured or uninsured patients aged 40-64 with stage I-III breast, cervical, colon, or non-small cell lung cancer within the National Cancer Database (NCDB). A difference-in-differences (DID) approach was used to compare changes in TTI (within 60 days) and 3-year OS between patients in ME states versus nonexpansion (NE) states before (2010-2013) and after (2015-2018) ME. Adjusted DID estimates for TTI and 3-year OS were calculated using multivariable linear regression and Cox proportional hazards regression models, respectively. ME was associated with a relative increase in TTI within 60 days for breast (DID = 4.6; p < 0.001), cervical (DID = 5.0 p = 0.013), and colon (DID = 4.0, p = 0.008), but not lung cancer (p = 0.505). In Cox regression analysis, ME was associated with improved 3-year OS for breast (DID hazard ratio [HR] = 0.82, p = 0.009), cervical (DID-HR = 0.81, p = 0.048), and lung (DID-HR = 0.87, p = 0.003). Changes in 3-year OS for colon cancer were not statistically different between ME and NE states (DID-HR, 0.77; p = 0.075). Findings suggest that expanded insurance coverage can improve treatment and survival outcomes among low income and uninsured patients with cancer. As the debate surrounding ME continues nationwide, our findings serve as valuable insights to inform the development of policies aimed at fostering accessible and affordable healthcare for all.

Pay-for-Performance Incentives for Home Dialysis Use and Kidney Transplant

The Centers for Medicare & Medicaid Services' mandatory End-Stage Renal Disease Treatment Choices (ETC) model, launched on January 1, 2021, randomly assigned approximately 30% of US dialysis facilities and managing clinicians to financial incentives to increase the use of home dialysis and kidney transplant. To assess the ETC's association with use of home dialysis and kidney transplant during the model's first 2 years and examine changes in these outcomes by race, ethnicity, and socioeconomic status. This retrospective cross-sectional study used claims and enrollment data for traditional Medicare beneficiaries with kidney failure from 2017 to 2022 linked to same-period transplant data from the United Network for Organ Sharing. The study data span 4 years (2017-2020) before the implementation of the ETC model on January 1, 2021, and 2 years (2021-2022) following the model's implementation. Receiving dialysis treatment in a region randomly assigned to the ETC model. Primary outcomes were use of home dialysis and kidney transplant. A difference-in-differences (DiD) approach was used to estimate changes in outcomes among patients treated in regions randomly selected for ETC participation compared with concurrent changes among patients treated in control regions. The study population included 724 406 persons with kidney failure (mean [IQR] age, 62.2 [53-72] years; 42.5% female). The proportion of patients receiving home dialysis increased from 12.1% to 14.3% in ETC regions and from 12.9% to 15.1% in control regions, yielding an adjusted DiD estimate of -0.2 percentage points (pp; 95% CI, -0.7 to 0.3 pp). Similar analysis for transplant yielded an adjusted DiD estimate of 0.02 pp (95% CI, -0.01 to 0.04 pp). When further stratified by sociodemographic measures, including age, sex, race and ethnicity, dual Medicare and Medicaid enrollment, and poverty quartile, there was not a statistically significant difference in home dialysis use across joint strata of characteristics and ETC participation. In this cross-sectional study, the first 2 years of the ETC model were not associated with increased use of home dialysis or kidney transplant, nor changes in racial, ethnic, and socioeconomic disparities in these outcomes.

Can digital transformation curtail carbon emissions? Evidence from a quasi-natural experiment

Digital infrastructure construction (DIC) is a crucial component of digital transformation, yet its impact on the sustainable development of cities remains uncertain. To address this question, this study employs the Difference-in-Differences (DID) model to evaluate the effects of a typical DIC policy, the Broadband China Pilot (BBCP) policy. The research sample covers data from Chinese cities from 2009 to 2019. The DID estimation demonstrates that the BBCP policy effectively decreased carbon emissions (COE). Robustness tests, such as parallel trend tests, placebo tests, propensity score matching DID, and local projection DID, all confirm the negative impact of the BBCP policy on COE. Heterogeneity analysis reveals that the BBCP policy has a greater mitigation effect on COE in non-resource cities. The BBCP policy has a stronger effect on reducing the COE in cities with a higher proportion of green funds or lower fiscal expenditures. Further analysis indicates that the BBCP policy significantly improves low-carbon ecological efficiency and reduces COE intensity. The results of the mechanism analysis reveal that the BBCP policy lowers COE by enhancing the industrial structure and promoting green technological progress. The study provides policy recommendations in the realm of DIC for developing countries to achieve digital transformation and low COE development concurrently. This research can assist governments in better considering how to implement DIC policies to construct ecologically sustainable cities with low COE.

Medicaid expansion in California and breast cancer incidence across neighborhoods with varying social vulnerabilities.

To investigate changes in breast cancer incidence rates associated with Medicaid expansion in California. We extracted yearly census tract-level population counts and cases of breast cancer diagnosed among women aged between 20 and 64years in California during years 2010-2017. Census tracts were classified into low, medium and high groups according to their social vulnerability index (SVI). Using a difference-in-difference (DID) approach with Poisson regression models, we estimated the incidence rate, incidence rate ratio (IRR) during the pre- (2010-2013) and post-expansion periods (2014-2017), and the relative IRR (DID estimates) across three groups of neighborhoods. Prior to the Medicaid expansion, the overall incidence rate was 93.61, 122.03, and 151.12 cases per 100,000 persons among tracts with high, medium, and low-SVI, respectively; and was 96.49, 122.07, and 151.66 cases per 100,000 persons during the post-expansion period, respectively. The IRR between high and low vulnerability neighborhoods was 0.62 and 0.64 in the pre- and post-expansion period, respectively, and the relative IRR was 1.03 (95% CI 1.00 to 1.06, p = 0.026). In addition, significant DID estimate was only found for localized breast cancer (relative IRR = 1.05; 95% CI, 1.01 to 1.09, p = 0.049) between high and low-SVI neighborhoods, not for regional and distant cancer stage. The Medicaid expansion had differential impact on breast cancer incidence across neighborhoods in California, with the most pronounced increase found for localized cancer stage in high-SVI neighborhoods. Significant pre-post change was only found for localized breast cancer between high and low-SVI neighborhoods.

Medicaid expansion and cancer stage at diagnoses during the COVID-19 pandemic in the United States.

1526 Background: Substantial cancer underdiagnosis, especially for early-stage diseases, occurred during the COVID-19 pandemic in the US in 2020. The Affordable Care Act expanded Medicaid to nonelderly adults with income up to 138% Federal Poverty Level in states that opted in, potentially facilitating residents’ access to care during pandemic-related financial and employment instability, thus aiding in timely detection of cancer. This study examines the association of Medicaid expansion with changes in cancer stage at diagnosis during the COVID-19 pandemic. Methods: Adults aged 18-64 years newly diagnosed with cancer in 2019 and 2021 were identified from the National Cancer Database. Year 2020 was excluded to avoid confounding from distinct geographic patterns of COVID-19 infections. A difference-in-differences (DD) approach was used to compare the changes in proportions of patients diagnosed at stage I/II in 2021 vs 2019 between Medicaid expansion states and non-expansion states. Adjusted DD estimates were calculated with linear probability models and stratified by key sociodemographic factors and cancer type. Results: A total of 427,368 adults newly diagnosed with cancer in 2019 and 393,961 in 2021 were identified. Patients diagnosed with stage I/II decreased nationwide, however, decreases were larger in non-expansion states (from 58.3% in 2019 to 56.9% in 2021) compared to expansion states (61.4% in 2019 to 60.7% in 2021), leading to a DD of 0.92 percentage points in early-stage cancer diagnosis associated with Medicaid expansion (95% CI =0.46-1.38). The association was strongest among individuals aged 18-44 years (DD=1.46, 95% CI=0.42-2.49), men (DD=1.46, 95% CI=0.73-2.19), those living in most socioeconomically deprived areas (DD=1.40, 95% CI=0.41-2.40) and nonmetropolitan areas (DD=1.82, 95% CI=0.56-3.09). When stratified by cancer type, the largest associations were seen for esophageal cancer (DD=3.86, 95% CI=0.26-7.45) and prostate cancer (DD=2.81, 95% CI=1.43-4.18). Conclusions: Substantial cancer underdiagnosis and declines in proportion of early-stage diagnoses continued nationwide in 2021, the second year of the COVID-19 pandemic. However, the pattern was worse in non-expansion states, suggesting a protective effect of Medicaid expansion on cancer early diagnosis during the public health emergency. Findings may inform policy makers and the public in the 10 states that have yet to expand Medicaid eligibility.

From Kyoto to Paris and beyond: A deep dive into the green shift

This study investigates the impact of the Kyoto Protocol and its associated regime, the Paris Agreement, on environmental outcomes and the green transition. We employ a difference-in-differences (DD) specification for country-level panel data from 1995 to 2018. The DD estimates suggest that the Kyoto Protocol of 2005 significantly reduced CO2 emissions and natural resources rents by around 20 percent and decreased the quantity of non-renewable electricity generated by about 36 percent, relative to the pre-Kyoto period. The protocol also led to increases in renewable energy consumption and generation of 18 percent and 60 percent, respectively. Robustness checks using system GMM estimates confirm that the effects of the Kyoto Protocol are likely not biased because of dynamic endogeneity from simultaneity problems or confounding trends. For the Paris Agreement, given documented issues in interpreting the average treatment effects using standard two-way fixed effects (TWFE), we leverage a DD specification with multiple treatment timings. We find no statistically significant effect of the Paris Agreement 2016 on the green transition except that it contributes to reducing CO2 emissions. The absence of a significant impact from the ad hoc Paris Agreement on most outcomes points to the need for greater effort in implementing environmental treaties.

The greenhouse gas reduction effect of critical peak pricing for industrial electricity: Evidence from 285 Chinese cities, 2003–2019

This paper assesses the impact of the Pilot of Critical Peak Pricing for Industrial Electricity (PCPPIE) on industrial Greenhouse Gas (GHG) emissions by employing the Difference-in-Differences (DID) methodology. It is found that the cities participating in PCPPIE underwent a significant 11.3% reduction in the GHGs emitted per industrial enterprise. These findings remain robust across various methodologies, including staggered DID estimators, propensity score matching, instrumental variable methods, and spatial econometric regression. The generalized DID analysis shows that a higher critical peak price corresponds to a more substantial decrease in GHG emissions. PCPPIE achieves weaker GHG reduction in higher-administrative level cities but stronger in those with higher industrial share, and the effectiveness of PCPPIE is compromised by cross-regional electricity imports. PCPPIE reduces GHG emission intensity of the power industry and contributes to additional reductions in GHG emissions from other sources by fostering green technological innovation and facilitating energy substitution. The practical significance of this paper is to reveal a richer dynamic mechanism for China's green and sustainable development, indicating that critical peak pricing is a power pricing strategy worthy of strengthening and optimization in the process of China's carbon peak and carbon neutrality.

Can toll prices changes affect the housing market? The case of the Severn crossing toll removal

PurposeSince 1966, the Severn crossing has been connecting England and Wales. In January 2018, its ownership returned to the UK Government, and this marked the start of a toll-free journey across the two countries and made commuting between the regions more affordable. In this paper, we examine the impact of the toll removal on the property market.Design/methodology/approachWe employ property-level data from the Land Registry and a difference-in-differences (DiD) empirical model for the periods 2016–2018 and 2019–2021 to capture the pre- and post-toll removal dynamics. The DiD estimation allows us to examine the causal relationship between policy changes and property prices.FindingsOur findings suggest that property prices in Newport and Monmouthshire (South East Wales) are positively affected by the policy, which results in a statistically significant increase of 5.8% more than those located in the South West England (Bristol and South Gloucestershire) region in the period 2019–2021. The impact can reach up to 13.1% for properties located in a 10 km radius of the bridge. The results indicate that the toll removal enables the ripple effect across the two markets by reducing commuting costs.Originality/valueThis is the first paper that examines the Severn Crossing case study. Its contribution is significant since we provide empirical evidence on how reduced transportation costs increase property prices in the lowest income region and have the opposite effect on the area with higher incomes and economic activity levels.

Educational and gender heterogeneity of the rural-urban earnings premium: New evidence from Norway

We explore urban earnings premiums for young, native, rural-to-urban movers in Norway. Using an augmented difference-in-differences estimator (DiD-TR) on microdata we challenge previous claims about urban earnings premium's size and sources. Conventional econometric estimators understate the static premium and overstate dynamic premiums. We find that migrants exhibit lower mean but faster pre-move earnings growth than non-migrants. Post-move, the static earnings premium dominates. The observed trajectory is related to frequent pre-move changes of industrial sector, presumably to obtain better job-worker matches. Post-move, these changes occur less frequently. Highly educated females exhibit largest static premiums (34%), less-educated females least (24%), males an intermediate amount. Our findings suggest that cities primarily generate earnings premiums through agglomeration-based efficiencies and superior job-worker matches varying heterogeneously by education and gender.

Does Tuition-Setting Authority Determine Whether Tuition Increases at Community Colleges with New Promise Programs?

ABSTRACT We evaluate whether the Bennett Hypothesis applies to local-level, single-institution promise programs and account for whether colleges have the authority to raise tuition, versus an external entity holding such authority. Using a sample of 29 community colleges affected by promise programs, we analyze changes in tuition across years 2001–02 to 2015–16 using two-way fixed effects difference-in-differences (DD) analyses, robust and weighted DD estimators, and panel event-studies. Among colleges with the sole authority to set their own tuition, the implementation of a promise program produces 14–17% lower tuition rates, suggesting that colleges do not strategically raise rates to capture promise-based financial aid dollars. When a governor or legislature is the sole entity that sets tuition, colleges experience negligible changes to slight declines in tuition levels, demonstrating that elected officials attempt to keep tuition rates stable or lower. When a statewide, systemwide, or local board retains tuition-setting authority, colleges observe no changes to tuition. Overall, our findings suggest a unique scenario for single-institution programs, where the college initiates or partners with a promise program; these colleges are motivated to keep tuition low to avoid counteracting affordability policies.

Designing Difference-in-Difference Studies with Staggered Treatment Adoption: Key Concepts and Practical Guidelines.

Difference-in-difference (DID) estimators are a valuable method for identifying causal effects in the public health researcher's toolkit. A growing methods literature points out potential problems with DID estimators when treatment is staggered in adoption and varies with time. Despite this, no practical guide exists for addressing these new critiques in public health research. We illustrate these new DID concepts with step-by-step examples, code, and a checklist. We draw insights by comparing the simple 2 × 2 DID design (single treatment group, single control group, two time periods) with more complex cases: additional treated groups, additional time periods of treatment, and treatment effects possibly varying over time. We outline newly uncovered threats to causal interpretation of DID estimates and the solutions the literature has proposed, relying on a decomposition that shows how the more complex DIDs are an average of simpler 2 × 2 DID subexperiments.

What is the impact of successive COVID-19 lockdowns on population mental health? Findings from an Australian natural experiment using health service data.

ObjectiveThe causal effect of successive population-wide lockdowns in response to increased COVID-19 cases on mental health has yet to be examined using robust methods. A natural experiment design underpinned by objective data can improve our understanding surrounding the definitive impact of social distancing restrictions.MethodsThe study employed a natural experiment design underpinned by objective data. Health service cost for visits to general practitioners and psychologists and medication dispensing costs served as objective measures of mental health. Difference-in-difference (DID) estimators, which in this study quantify differences in spending changes between groups over time, were produced based on three comparisons: Victoria 2020 lockdown comparison, Victoria 2021 lockdown comparison, and New South Wales (NSW) 2021 lockdown comparison. Specifically, differences in public health service spending during lockdown periods and the same timeframe in 2019 for Victoria and NSW, and control groups (remaining states and territories), were compared.ResultsPositive estimator values indicate that public health service spending for Victoria and NSW increased more during lockdown periods compared to control states and territories. The Victorian lockdowns of 2020 and 2021, but not the NSW lockdown of 2021, resulted in increased public spending for general practitioner mental health consults (2020 DID estimator: $8498.96 [95% CI $4012.84, $12,373.57], 2021 DID estimator: $6630.06 [95% CI $41.27, $13,267.20], all monetary values in AUD$) and short visits to psychologists (2020 DID estimator: $628.82 [95% CI $466.25, $796.00], 2021 DID estimator: $230.11 [95% CI $47.52, $373.98]). The first Victorian lockdown in 2020 and the NSW lockdown in 2021 resulted in greater spending on short visits to clinical psychologists. Spending on long visits to psychologists and clinical psychologists and medication spending did not change.ConclusionsStrict lockdowns can have an adverse impact on population mental health. The impact is particularly evident in those who have a history of previous mental health concerns but does not necessitate extra use of medications, suggesting that psychological care can address the adverse impact of the lockdowns.

THE EFFECT OF PANDEMICS ON TRUST IN DOCTOR IMPROVEMENT AMONG CHINESE OLDER ADULTS: LESSONS FROM COVID-19

Abstract The COVID-19 pandemic has had a major impact on the physician-patient relationship. For older adults, trust in doctors during the pandemic affects their lives and health. This study aimed to estimate the effect of the first wave of COVID-19 transmissibility in China on older adults’ trust in doctors. We obtained individual data from the Chinese Family Panel Studies (CFPS). A total of 2,225 participants aged 65 and older were included in this study. The COVID-19 pandemic was used as a natural experiment. We obtained difference-in-differences (DID) estimates of the pandemic’s effects by exploiting temporal variation in the timing of COVID-19 exposure across participants interviewed from 2018 to 2020 together with the geographical variation in COVID-19 severity at the provincial district level. Overall, the DID estimates showed that the COVID-19 pandemic could significantly increase trust in doctors after controlling for multiple covariates (adjusted β: 0.053, 95% CI: 0.028–0.078). Stratified analyses showed that the effect was only in females and the more educated ones but not in their counterparts. In addition, the COVID-19 pandemic showed a positive effect on trust in doctors among 65-79-year-olds (0.082, 0.056–0.109), but older adults over the age of 80 showed significantly reduced trust in doctors (-0.236, -0.418–-0.055). Our findings confirm the association between the COVID-19 pandemic and trust in doctors among older Chinese people. Hospitals and other medical organizations should pay more attention to improving older adults’ trust in doctors and building a better physician-patient relationship to face the future pandemic.