Counterfactual Analysis Research Articles

Psychosocial factors mediate social inequalities in health-related quality of life among children and adolescents

BackgroundThe present analysis aimed to assess the mediating role of psychosocial and behavioural factors in socio-economic inequalities in health-related quality of life (HRQoL) among children and adolescents.MethodsCross-sectional data was drawn from the randomly selected SEROCoV-KIDS cohort study in Geneva, Switzerland. Associations of socio-economic conditions (parents’ highest education, household financial situation) with HRQoL, psychosocial (parent–child relationship, school difficulties, friends, extracurricular activities) and behavioural factors (screen time, physical activity, green spaces time, sleep duration), along with associations of psychosocial and behavioural factors with HRQoL, were evaluated with generalized estimating equations. Counterfactual mediation analyses were conducted to test pathways linking socio-economic conditions to HRQoL.ResultsOf 965 children and 816 adolescents, those with disadvantaged financial circumstances were more likely to have a poor HRQoL (adjusted Odds Ratio [aOR]: 3.80; 95% confidence interval [CI]: 1.96–7.36 and aOR: 3.66; 95%CI: 2.06–6.52, respectively). Psychosocial characteristics mediated 25% (95%CI: 5–70%) and 40% (95%CI: 18–63%) of financial disparities in HRQoL among children and adolescents, respectively. Health behaviours were weakly patterned by socio-economic conditions and did not contribute to financial differences in HRQoL.ConclusionsThese findings provide empirical evidence for mechanisms explaining socio-economic disparities in child HRQoL and could inform interventions aimed to tackle health inequalities.

How does family support mitigate the impact of childhood adversity on adolescent mental health?

Abstract Background Childhood adversity is associated with increased risk of developing mental health problems over the life course. We investigated how family emotional support in adolescence may mitigate the impact of childhood adversity on adolescent mental health. Methods We used longitudinal data from the UK Millennium Cohort Study on 9,269 children followed to age 17 years. Individuals were assigned to exposure trajectories characterised using group-based trajectory models that included measures of poverty and family adversities (parental mental ill health, parental alcohol misuse and domestic violence) up to age 14 years. Using counterfactual causal mediation analysis and four-way decomposition modelling approach, we evaluated how perceived family emotional support at age 14 (measured using the three-item Short Social Provisions Scale) influences the association of childhood adversity trajectories on mental health at ages 14 and 17, assessing the relative contributions of mediation and interaction simultaneously. Results Compared with children experiencing low adversity and poverty, those exposed to childhood adversity were almost three times more likely to experience poor mental health (RR 2·99, 95% CI 2 ·41 to 3·57) at age 14 and age 17 (RR 2·58, 95% CI 2·09 to 3·06). Perceived emotional support mitigates up to 18% (95% CI: 9% to 26%) of this effect at age 14, and 13% (95% CI: 5% to 22%) at age 17, mainly due to interactive mechanisms. Conclusions Childhood adversity has a strong effect on mental health, which is partially mitigated by perceived family support in adolescence. Policies that support positive family functioning may be particularly beneficial for children who have experienced adversity. Key messages • Socio-economic – associated mental health problems in adolescence differed based on levels of family support. • Approximately 20% of the social inequality in mental health outcomes in adolescence may be mitigated or reduced by improving levels of family support.

Do poor psychosocial resources mediate health inequalities in type 2 diabetes? The Maastricht Study

Abstract Low socioeconomic position (SEP) has been identified as a risk factor for type 2 diabetes mellitus (T2DM). A potential pathway of this association is psychosocial resources. Disadvantageous environmental conditions result in low control beliefs through adverse resources and inferiority beliefs through social comparison for people with a low SEP. We examined two poor psychosocial resources, low control beliefs and inferiority beliefs, to explain socioeconomic inequalities in T2DM. We analysed mediating roles of low control beliefs and inferiority beliefs by using counterfactual mediation analysis. 8293 participants aged 40-75 living in Limburg, the Netherlands participated in The Maastricht Study starting from September 2010 to October 2020 and were followed up to 10 years with annual questionnaires with a median follow-up time of 7.18 years. SEP (education, income, occupation), control and inferiority beliefs, and (pre)diabetes by oral glucose tolerance test were measured at baseline. Incident T2DM was self-reported per annum. 3.2%(N = 203) of the participants without T2DM at baseline reported incident T2DM. People with low SEP had more often prevalent and incident T2DM (e.g., low education:HR=2.13, 95%CI:1.53-2.97). Lower control beliefs and higher inferiority beliefs were more common among people with low SEP than people with high SEP. Moreover, lower control beliefs and higher inferiority beliefs were risk factors for T2DM (e.g., low control beliefs: HR = 1.50, 95%CI:1.08-2.09). The relationship between SEP and T2DM was partially mediated by control beliefs (7.98-13.56%) and inferiority beliefs (2.16-4.53%). Poor psychosocial resources are important in socioeconomic inequalities in diabetes. Researchers and practitioners should consider the psychosocial profile of people with lower SEP might differ from those with higher SEP and this might interfere with the development, treatment, and prevention of T2DM. We should aim to work with such beliefs, instead of against it. Key messages • Socioeconomic health inequalities in diabetes has many underlying mechanism including psychosocial pathways. • Interventions and research should consider the different environment of low SEP and psychosocial profiles of people. We should aim to work with it, instead of against it.

The effect of patients' socioeconomic status in rehabilitation centers on the efficiency and performance.

Patients' socioeconomic status on hospitals' efficiency in controlling for clinical component characteristics may have a role that has few been studied in rehabilitation centers. Because of the national health insurance system, rehabilitation centers are free of charge. To answer whether a patient's socioeconomic status (SES) is associated with efficiency and performance, we use a counterfactual analysis to get the patient's SES effect "as if" the patient's case was identical to whatever hospital. We restrained the data to patients from public acute care units where the decision on rehabilitation sector admission is based on availability, limiting bias by confounding factors. Besides, an analysis of six pathologies led to the same results. An exhaustive, detailed administrative database on rehabilitation center stays in France. To define the patients' socioeconomic status, we use two sources of data: the information collected at the time of the patient's entry into rehabilitation care and the information collected during the patient's stay in acute care. This double information avoids possible loss of socio-economic details between the two admissions. Patients recruited were exhaustively admitted over the year 2018 for stroke, chronic obstructive pulmonary disease, heart failure, or total hip replacement in France in the acute care unit and then in a rehab center. Mainly the elderly population. Information on patients' demography, comorbidities, and SES are coded due to the reimbursement system. Different dimensions controlling for factors (hospital ownership, patient clinical characteristics, rehabilitation care specificities, medical staff detailed information, and patients' socioeconomic status), were progressively added to control for any differences in baseline data between the two groups. We assess rehabilitation centers' efficiency by combining selected outcome quality indicators (Physical score improvement, Cognitive score improvement, Mortality, Return-to-home). The specific Providers' Activity Index is used to get the performance index. The performance of healthcare institutions is correlated not only to the case mix of their patients but also to the socioeconomic status of the patients admitted. The performance needs to be seen in light of patients' socioeconomic status. The data reveals that patients' socioeconomic status affects rehabilitation care efficiency and performance. In controlling patients' socioeconomic status, for-profit rehabilitation hospitals seemed more efficient than public ones.

Disparities in Salmonellosis Incidence for US Counties with Different Social Determinants of Health Profiles Are Also Mediated by Extreme Weather: A Counterfactual Analysis of Laboratory Enteric Disease Surveillance (LEDS) Data From 1997 to 2019

Understanding disparities in salmonellosis burden is critical for developing effective, equitable prevention programs. Past efforts to characterize disparities were limited in scope and by the analytical methods available when the study was conducted. We aim to address this gap by identifying disparities in salmonellosis incidence between counties with different determinants of health (DOH) profiles. Using national U.S. Laboratory-based Enteric Disease Surveillance (LEDS) data for 1997–2019, age-adjusted county-level salmonellosis incidence/100,000 persons was calculated and linked to publicly available DOH data. We used hurdle counterfactual random forests (CFRFs) to quantify, for each DOH, the risk that (i) ≥1 versus no cases were reported by a county, and (ii) when ≥1 case was reported, whether a high (≥16 cases/100,000 persons) or low incidence (≥1 & <4 cases/100,000 persons) was reported. Risk in both models was significantly associated with demographic DOH, suggesting a disparity between counties with different demographic profiles. Risk was also significantly associated with food, healthcare, physical, and socioeconomic environment. The risk was generally greater for counties with more negative food resources, and for under-resourced counties (e.g., fewer healthcare and social services, fewer grocery stores). Risk was also significantly higher if any extreme weather event occurred. The study also found that underreporting and underascertainment appeared to result in underestimation of salmonellosis incidence in economically marginalized and under-resourced communities. Overall, our analyses indicated that, regardless of other county characteristics, extreme weather was associated with increased salmonellosis incidence, and that certain communities were differentially disadvantaged toward a higher incidence. This information can facilitate the development of community-specific prevention efforts.

The effect of the social security on aggregate effective labor supply: the role of endogenous human capital and of selection effects

Abstract We quantify the importance of endogenous human capital and of selection effects for counterfactual analysis of social security (SS) reforms. The literature typically performs these analyses by using structural models featuring exogenous productivity profiles. However, this approach faces two issues: (i) the estimation of productivity is subject to selection bias, and (ii) productivity is endogenous to the SS reforms. In this paper, we estimate a quantitative overlapping generations model featuring endogenous human capital accumulation using US data. First, we eliminate the SS and find a large positive effect on aggregate effective labor supply ( ${+}10.31\%$ ). Next, we build variants of this model to quantify the two issues (i) and (ii). We find that the endogeneity issue (ii) is quantitatively more important than the selection bias issue (i).

Explaining Recommendation Fairness from a User/Item Perspective

Recommender systems play a crucial role in personalizing user experiences, yet ensuring fairness in their outcomes remains an elusive challenge. This work explores the impact of individual users or items on the fairness of recommender systems, thus addressing a significant knowledge gap in the field. We introduce an innovative approach called A dding-based C ounterfactual F airness R easoning ( ACFR ), designed to elucidate recommendation fairness from the unique perspectives of users and items. Conventional methodologies, like erasing-based counterfactual analysis, pose limitations, particularly in modern recommender systems dealing with a large number of users and items. These traditional methods, by excluding specific users or items, risk disrupting the crucial relational structure central to collaborative filtering recommendations. In contrast, ACFR employs an adding-based counterfactual analysis, a unique strategy allowing us to consider potential, yet-to-happen user-item interactions. This strategy preserves the core user-item relational structure, while predicting future behaviors of users or items. The commonly-used feature-based counterfactual analysis, relying on gradient-based optimization to identify interference on each feature, is not directly applicable in our case. In the recommendation scenario we consider, only interactions between users and items are present during model training—no distinct features are involved. Consequently, the traditional mechanism proves impractical for identifying interference on these existing interactions. Our extensive experiments validate the superiority of ACFR over traditional baseline methods, demonstrating significant improvements in recommendation fairness on benchmark datasets. This work, therefore, provides a fresh perspective and a promising methodology for enhancing fairness in recommender systems.

Environmental and social welfare effects of aviation liberalization: Evidence from Central Asia and China market

We investigate the effects of international aviation liberalization between China and Central Asia on environment and social welfare. Based on monthly international air traffic data between Central Asia and China, we estimate a random-coefficients discrete choice model and perform a counterfactual analysis to quantify the effects of improved air connectivity, mainly considering that the ratio of direct and connecting aviation services between Central Asia and Beijing is the same as those between Central Asia and Russia, between Central Asia and other former Soviet Union countries and between Central Asia and other Asian countries. The main findings are as follows. First, improving air connectivity intensifies competition in the market, resulting in a decrease in the passenger-weighted average airfares between Central Asia and Beijing by 0.86–5.15 US dollars (USD) in 2012 and an increase in the total number of passengers by 2930–20,790. Second, improving air connectivity increases total carbon emissions (CO2) between Central Asia and Beijing by 0.78–6.11 million kilograms (kg) in 2012 but also drives down per capita CO2 emissions in this market by 0.78–2.38 kg. Third, while increasing total environmental externality costs, improving air connectivity significantly enhances consumer welfare and airline profits, resulting in a net social benefit of 14.02–44.12 million USD between Central Asia and Beijing in 2012. Finally, several policy implications are discussed.

Land Access Modes and Agricultural Productivity in Benin

Improving productivity is an important channel for satisfying household food needs through food availability. Increasing the productivity of small-scale farmers is linked to a number of factors, including land access, labor, chemicals, fertilizers and so on. Most farmers resort to expanding their cultivated areas to increase production because of insufficient financial resources available for purchasing inputs during agricultural seasons. This situation, combined with increasing pressure on land, does not favor agricultural production and, by extension, food security. This study therefore assesses the impact of different modes of access to land on agricultural productivity. Regime-switching multinomial regression techniques were used to account for endogeneity bias due to observed and unobserved factors. The results of the study show that access through free loans, tenancy and sharecropping negatively affect agricultural yields. The counterfactual analysis reveals a positive gain estimated at 509.9 kg/ha from land access by landowners. If the lenders were landowners, their gain would be 396.6 kg/ha, whereas the farmers would gain 485.1 kg/ha if they were landowners, and similarly the sharecroppers would gain 389.8 kg/ha if they were landowners. It is clear from these results that improving agricultural yields depends on securing land and requires the establishment of an effective land ownership system. This research impacts land ownership policies, which need to be revised to address customary rights and reduce inequalities in access to secure land. It highlights the way land tenure security drives agricultural advancements and offers actionable recommendations for policy improvements on food security.

Counterfactual-based two-stage trajectory prediction framework for autonomous vehicle navigation systems

Abstract Accurate prediction of a vehicle’s trajectory is essential for autonomous driving and robotic operations in a complex, dynamic environment. However, current methods that rely on historical and environmental cues to predict future trajectories make the inner distance between testing and training environments often overlooked. To address this issue, we propose a counterfactual analysis method specifically designed for vehicle trajectory prediction. We begin by constructing a causal graph that includes historical trajectories, future trajectories, and environmental interactions. This graph captures the causal relationships between these elements. Our approach directly intervenes on the trajectory itself, effectively cutting off the inference from the environment to the trajectory. By doing so, we mitigate the negative impact of environmental bias. Furthermore, by comparing observed and intervened trajectory clues, we highlight the influence of environmental distance. Our goal is to improve the accuracy of trajectory prediction. Our counterfactual analysis module integrates seamlessly into various baseline prediction models. Experimental data demonstrates that our method outperforms others on publicly available Argoverse2 motion forecasting benchmarks.

Asset prices, regime switches, and monetary policy: evidence from China

ABSTRACT This paper develops a regime-switching DSGE model that can capture time-varying volatilities in economic activity and monetary policy. The model is then used to study regime switches in macroeconomic dynamics and monetary policy in China. The results suggest a better fitness of the regime-switching model than conventional constant-parameter models. We also find notable differences in macroeconomic dynamics and monetary policy reactions between high and low volatility regimes. Further counterfactual analysis suggests that, at least in an environment with regime-switching volatilities, the optimal reaction of monetary policy involves responses to asset price variations. The findings of the paper shed new light on the regime-switching properties of the economy and the optimal choice of monetary policy within this context.

Global economic policy uncertainty and oil price uncertainty: Which is more important for global economic activity?

Using monthly data spanning from January 1997 to November 2022, this paper investigates the influence of global economic policy uncertainty (GEPU) and oil price uncertainty (OPU) on global economic activity using a structural vector autoregressive (SVAR) model. The findings from the baseline model reveal the adverse effects of both GEPU and OPU shocks on the global economy. The counterfactual analysis highlights that the negative impact of OPU shocks primarily stems from direct effects, while approximately half of the adverse impact of GEPU shocks is indirectly transmitted through amplifying OPU. Further examination of the indirect effects of these shocks uncovers that lagged effects primarily drive the indirect impact of OPU, whereas the indirect impact of GEPU is mainly driven by common shocks. These findings suggest a more substantial and independent impact of OPU shocks compared to GEPU shocks. Additionally, employing a time-varying vector autoregressive with stochastic volatility (TVP-VAR-SV) model, this paper explores the same dataset. The results of time-varying impulse responses illustrate a diminishing effect of GEPU shocks over the sample period, whereas the impact of OPU shocks initially decreases before subsequently increasing.

Mergers, firm size, and volatility in a granular economy

We study the firm dynamics associated with mergers and acquisitions (M&A) and their implications at the micro and macro levels. Our paper presents three main findings: (i) mergers generate a more fat-tailed firm-size distribution, thereby amplifying granular fluctuations and increasing aggregate volatility; (ii) the impact of mergers depends on strategic market power and endogenous markups; and (iii) under endogenous markups, we provide a novel characterization of the firm size-volatility relationship in which volatility declines disproportionately with size. We build a quantitative model of domestic horizontal mergers and find a sizeable impact of mergers on aggregate volatility using counterfactual analysis.

Do petrol prices affect inflation and inflation expectations? Evidence from New Zealand

Due to their high visibility, petrol prices shape households' inflation expectations. This paper investigates the dynamic effects of petrol price shocks on one- and five-year inflation expectations in New Zealand; effects on headline and core inflation are also analyzed. Estimating partially identified structural vector autoregression models using Bayesian techniques, we show that petrol price shocks had a slightly delayed but persistent effect on one-year inflation expectations, whereas five-year inflation expectations were largely insensitive to these shocks. We also show that headline inflation increased immediately in response to petrol price shocks, while core inflation remained unaffected. Counterfactual analysis reveals that inflation and inflation expectations would have been higher in the absence of petrol price shocks during the second half of 2020. The results underscore the importance of petrol prices in influencing households' short-term inflation expectations, offering valuable insights for policymakers focused on managing the cost of living and inflation expectations in New Zealand.

Public Mass Shootings: Counterfactual Trend Analysis of the Federal Assault Weapons Ban.

Assault weapon and large-capacity magazine bans are potential tools for policy makers to prevent public mass shootings. However, the efficacy of these bans is a continual source of debate. In an earlier study, we estimated the impact of the Federal Assault Weapons Ban (FAWB) on the number of public mass shooting events in the United States. This study provides an updated assessment with 3 additional years of firearm surveillance data to characterize the longer-term effects. This study aims to estimate the impact of the FAWB on trends in public mass shootings from 1966 to 2022. We used linear regression to estimate the impact of the FAWB on the 4-year simple moving average of annual public mass shootings, defined by events with 4 or more deaths in 24 hours, not including the perpetrator. The study period spans 1966 to 2022. The model includes indicator variables for both the FAWB period (1995-2004) and the period after its removal (2005-2022). These indicators were interacted with a linear time trend. Estimates were controlled for the national homicide rate. After estimation, the model provided counterfactual estimates of public mass shootings if the FAWB was never imposed and if the FAWB remained in place. The overall upward trajectory in the number of public mass shootings substantially fell while the FAWB was in place. These trends are specific to events in which the perpetrator used an assault weapon or large-capacity magazine. Point estimates suggest the FAWB prevented up to 5 public mass shootings while the ban was active. A continuation of the FAWB and large-capacity magazine ban would have prevented up to 38 public mass shootings, but the CIs become wider as time moves further away from the period of the FAWB. The FAWB, which included a ban on large-capacity magazines, was associated with fewer public mass shooting events, fatalities, and nonfatal gun injuries. Gun control legislation is an important public health tool in the prevention of public mass shootings.

Variational Bayes machine learning for risk adjustment of general outcome indicators with examples in urology

Risk adjustment is often necessary for outcome quality indicators (QIs) to provide fair and accurate feedback to healthcare professionals. However, traditional risk adjustment models are generally oversimplified and not equipped to disentangle complex factors influencing outcomes that are out of a healthcare professional’s control. We present VIRGO, a novel variational Bayes model trained on routinely collected, large administrative datasets to risk-adjust outcome QIs. VIRGO uses detailed demographics, diagnosis, and procedure codes to provide individualized risk adjustment and explanations on patient factors affecting outcomes. VIRGO achieves state-of-the-art on external datasets and features capabilities of uncertainty expression, explainable features, and counterfactual analysis capabilities. VIRGO facilitates risk adjustment by explaining how patient factors led to adverse outcomes and expresses the uncertainty of each prediction, allowing healthcare professionals to not only explore patient factors with unexplained variance that are associated with worse outcomes but also reflect on the quality of their clinical practice.

Indian Buffet process factor model for counterfactual analysis

This article proposes a factor model-based counterfactual analysis. We explicitly estimate the underlying factor structure of the outcome variables and estimate the counterfactual values of the unit subject to an intervention. With the help of the non-parametric Bayesian Indian Buffet Process prior, our approach is capable of exploring heterogeneous factor exposures, and the number of latent factors is endogenously determined in the estimation process. The flexible Markov Chain Monte Carlo algorithm utilizes the maximal intervention-free information provided by the data, whereas the original synthetic control only uses pre-intervention data. The counterfactual values are estimated by simulating from the posterior predictive distribution so that we can integrate out any parameter estimation uncertainty. We also calculate the posterior predictive upper- and lower-quantile bounds for inference. The two applications, namely California’s Tobacco Control Program and the West German Reunification, demonstrate the usefulness of our approach compared to the synthetic control method and the elastic net model. The Monte Carlo simulation study also shows the robustness of our approach with respect to nonlinear dynamics in the underlying factor process.

Machine learning algorithms to predict mild cognitive impairment in older adults in China: A cross-sectional study

ObjectiveThis study aimed to explore the predictive value of machine learning (ML) in mild cognitive impairment (MCI) among older adults in China and to identify important factors causing MCI. MethodsIn this study, 6434 older adults were selected based on the data of the China Health and Elderly Care Longitudinal Survey (CHARLS) in 2020, and the dataset was subsequently divided into the training set and the test set, with a ratio of 6:4. To construct a prediction model for MCI in older adults, six ML algorithms were used, including logistic regression, KNN, SVM, decision tree (DT), LightGBM, and random forest (RF). The Delong test was used to compare the differences of ROC curves of different models, while decision curve analysis (DCA) was used to evaluate the model performance. The important contributions of the prediction results were then used to explain the model by the SHAP value.The Matthews correlation coefficient (MCC) was calculated to evaluate the performance of the models on imbalanced datasets. Additionally, causal analysis and counterfactual analysis were conducted to understand the feature importance and variable effects. ResultsThe area under the ROC curve of each model range from 0.71 to 0.77, indicating significant difference (P < 0.01). The DCA results show that the net benefits of LightGBM is the largest within various probability thresholds. Among all the models, the LightGBM model demonstrated the highest performance and stability. The five most important characteristics for predicting MCI were educational level, social events, gender, relationship with children, and age. Causal analysis revealed that these variables had a significant impact on MCI, with an average treatment effect of −0.144. Counterfactual analysis further validated these findings by simulating different scenarios, such as improving educational level, increasing age, and increasing social events. ConclusionThe ML algorithm can effectively predict the MCI of older adults in China and identify the important factors causing MCI.

Do poor psychosocial resources mediate health inequalities in type 2 diabetes mellitus? Findings from the Maastricht study.

Low socioeconomic position (SEP) has been identified as a risk factor for type 2 diabetes mellitus (T2DM), and psychosocial resources might be on the pathway in this association. We examined two poor psychosocial resources, low control beliefs and inferiority beliefs, that might link low SEP with T2DM. 8292 participants aged 40-75 living in Southern Netherlands participated in The Maastricht Study starting from September 2010 to October 2020 and were followed up to 10 years with annual questionnaires. SEP (education, income, occupation), low control beliefs, inferiority beliefs, and (pre)diabetes by oral glucose tolerance test were measured at baseline. Incident T2DM was self-reported per annum. We analysed the mediating roles of poor psychosocial resources by using counterfactual mediation analysis. People with low SEP had more often prevalent and incident T2DM (e.g. low education: HR = 2.13, 95%CI: 1.53-2.97). Low control beliefs and high inferiority beliefs were more common among people with low SEP. Moreover, low control beliefs and high inferiority beliefs were risk factors for T2DM (e.g. low control beliefs: HR = 1.50, 95%CI: 1.08-2.09). The relationship between SEP and T2DM was partially mediated by control beliefs (8.0-13.6%) and inferiority beliefs (2.2-4.5%). We conclude that poor psychosocial resources are important in socioeconomic inequalities in diabetes. Researchers and practitioners should consider the psychosocial profile of people with lower SEP, as such a profile might interfere with the development, treatment, and prevention of T2DM. Further research should explore how poor psychosocial resources interact with chronic stress in relation to socioeconomic health inequalities.

Inference of human contact networks based on epidemic data

A key challenge in epidemic modeling is the lack of adequate data on population interactions (such as traffic flow or mobility patterns) which result in the spread of infectious diseases. Knowledge of social contact patterns is crucial for public health professionals to devise effective non-pharmaceutical interventions to control epidemics. This paper focuses on inferring social contact rates from reported infection counts during the spread of an infectious disease, addressing the increased difficulty that arises when dealing with “sparse” contact networks where only a small subset of the edges have non-zero weights. Specifically, a new geographically constrained lasso approach for network reconstruction for non-homogeneous mixing Susceptible-Infected-Removed (SIR) disease spread models is presented. The new network reconstruction method can explicitly account for the spatial proximity of network nodes in estimating the disease transmission rates and predicting the future evolution of the epidemic dynamics. Extensive numerical experiments are presented to show the proposed method outperforms existing approaches in terms of accuracy of contact identification under various graph topologies. A case study based on real data from the COVID-19 pandemic is presented to demonstrate the application of the approach for inferring contact structures and a counterfactual scenario analysis to assess effectiveness of containment strategies.