Formal social control — the regulation of behavior through explicit, codified rules enforced by state institutions with the authority to impose sanctions — constitutes the most visible and most extensively studied dimension of the crime control enterprise. Its institutions — the police, the courts, the correctional system, and the regulatory agencies that enforce civil and administrative law — represent the state’s organized response to the behavioral norm violations that criminal law defines, and their operation raises fundamental sociological questions about legitimacy, effectiveness, fairness, and the relationship between formal institutional control and the informal social processes that criminological research consistently identifies as the primary determinants of behavioral compliance. Sociology and Criminology engages formal social control not merely as the subject of policy evaluation but as a sociological institution whose structure, logic, and consequences reflect the broader organization of power, inequality, and normative consensus in the societies that deploy it.
The sociological analysis of formal social control begins with the recognition that criminal law is not a neutral codification of universal moral prohibitions but a selective system that defines some harmful behaviors as criminal while leaving others unregulated, that enforces those definitions in ways that reflect the distribution of power across the social structure, and that produces consequences — individual and community level — that extend far beyond the specific deterrence and incapacitation effects that the formal justifications for criminal punishment invoke. Understanding formal social control sociologically requires examining both what it does (its effects on crime and criminal careers) and what it is (a social institution embedded in structures of power and inequality that shapes and is shaped by those structures in ways that crime-focused evaluation alone cannot reveal).
Introduction
The sociological study of formal social control encompasses multiple analytical levels: the macro-level analysis of how societies differ in their criminal law systems, incarceration rates, and institutional arrangements for crime control; the meso-level analysis of how specific criminal justice institutions — police departments, prosecutorial offices, courts, correctional systems — operate as organizations with their own cultures, incentive structures, and institutional logics; and the micro-level analysis of how individual encounters between citizens and criminal justice agents — police stops, arraignments, sentencing hearings — are shaped by the organizational and structural contexts within which they occur.
Each analytical level reveals different dimensions of formal social control’s sociological character. Macro-level comparative analysis documents that formal social control is not a natural or inevitable response to crime but a historically and culturally specific institutional arrangement that varies enormously across societies — from the extraordinarily punitive American carceral state to the rehabilitative Scandinavian systems that produce comparable or better crime control outcomes with dramatically lower incarceration rates. Meso-level organizational analysis reveals the gap between the formal rules that supposedly govern criminal justice decision-making and the informal organizational cultures, discretionary practices, and structural pressures that shape actual decisions. And micro-level interactional analysis reveals the social mechanisms — racial profiling, prosecutorial charging discretion, judicial sentencing variation — through which the structural inequalities of the broader society are reproduced within and amplified by the formal control apparatus.
Deterrence Theory and Formal Sanctions
The Deterrence Doctrine
Deterrence theory — the proposition that the threat of formal punishment reduces criminal behavior by raising its expected costs — provides the primary theoretical justification for the formal criminal justice system and the primary theoretical framework for evaluating its effectiveness. The classical deterrence framework, developed by Beccaria and Bentham in the eighteenth century and formalized in contemporary criminological research, identifies three dimensions of formal sanctions as the primary determinants of deterrent effect: certainty (the probability that a criminal act will result in detection, arrest, and punishment), severity (the magnitude of the punishment imposed), and celerity (the speed with which punishment follows the criminal act).
Contemporary deterrence research — synthesized in systematic reviews by Nagin and colleagues — has consistently found that certainty of punishment is a substantially more powerful deterrent than severity, while celerity shows relatively weak independent effects after controlling for certainty. Research exploiting natural experiments in sanction certainty — including studies of police hiring, clearance rate variation, and changes in prosecution practices — has found consistent crime reductions associated with increased certainty, while research examining the crime deterrence effects of increasing sentence severity — including mandatory minimum sentencing expansions, three-strikes laws, and sentence enhancement policies — has found consistently weak or null effects on crime rates. This finding — that certainty matters but severity has minimal incremental deterrent effect beyond moderate levels — has profound implications for the political economy of American criminal justice, which has consistently invested in severity (longer sentences, mandatory minimums, mass incarceration) rather than certainty.
Specific and General Deterrence
The deterrence literature distinguishes between specific deterrence (the crime-preventing effect of punishment on the punished individual) and general deterrence (the crime-preventing effect on the general population through the example of observed punishment). The empirical evidence on specific deterrence is substantially more pessimistic than the theoretical expectation: research on the crime consequences of incarceration consistently finds that incarceration does not reduce subsequent offending relative to alternative sanctions, and in many studies finds that incarceration increases subsequent offending through the social bond-weakening, labeling, and criminal network-embeddedness mechanisms that the sociological criminology literature documents.
Research by Nagin, Cullen, and colleagues synthesizing the specific deterrence evidence found that incarceration does not deter future offending relative to community-based sanctions for similar offenders, and that the severity of the prison experience — which classical deterrence theory predicts should amplify deterrent effect — is either uncorrelated with or positively associated with subsequent recidivism. These findings directly contradict the deterrence theory justification for incarceration as the primary formal sanction and support the reform arguments for community-based alternatives that the evidence-based criminology movement has advanced through systematic review and meta-analysis.
Table 1. Formal Social Control Mechanisms: Sociological Analysis
| Mechanism | Theoretical Justification | Empirical Evidence | Sociological Critique |
|---|---|---|---|
| Deterrence (certainty) | Rational choice: raising criminal costs prevents crime | Moderate-strong: police presence, clearance rates reduce crime | Assumes rational calculation; ignores informal social control primacy |
| Deterrence (severity) | Greater punishment → greater deterrence | Weak-null: sentence severity has minimal incremental deterrent effect | Marginal utility of punishment; most crime involves limited future calculation |
| Incapacitation | Incarcerated individuals cannot commit crime in community | Moderate but declining: high incarceration rates → diminishing returns | Replacement effects; community social organization costs |
| Rehabilitation | Treatment reduces criminal propensity | Moderate-strong: RNR-based programming reduces recidivism | Requires sustained investment; undermined by punitive prison culture |
| Labeling (iatrogenic) | Formal sanctioning worsens outcomes through stigma | Moderate: formal contact elevates subsequent offending for some | Supports diversion, restorative justice alternatives |
| Specific deterrence | Punishment deters the sanctioned individual | Weak-negative: incarceration often elevates subsequent offending | Social bond weakening; criminal network embeddedness; stigma |
Incapacitation and Its Limits
Incapacitation — the crime-preventing effect of physically removing offenders from the community through incarceration — provides a theoretically straightforward mechanism for formal crime control that does not depend on the contested empirical assumptions of deterrence: even if incarceration does not deter, it prevents incarcerated individuals from committing crimes in the community during the period of confinement. Research on the incapacitation effect has found positive estimates — incarceration does reduce crime through this mechanism — but with significantly diminishing returns as incarceration rates increase: research by Spelman found that a 10% increase in incarceration rates at American operating levels produces only about a 2% reduction in crime, a benefit-cost ratio that compares unfavorably to prevention and treatment alternatives at the margin of current investment.
The diminishing returns to incapacitation reflect two countervailing mechanisms. First, as incarceration rates increase, the marginal offenders being incarcerated have lower individual crime rates — the most prolific and violent offenders were already being incarcerated, so additional incarceration captures progressively less dangerous individuals with lower per-person incapacitation benefits. Second, incarceration produces community-level costs that offset some of its direct incapacitation benefits: the concentrated incarceration that Clear documents disrupts community social organization, weakens informal social control, and creates the reentry conditions that elevate recidivism, reducing the net crime-prevention benefit of high incarceration rates below what individual-level incapacitation studies predict.
The Sociology of Police
Police as Formal Control Institution
The police constitute the most visible component of the formal social control apparatus and the institution whose daily operation most directly shapes the relationship between citizens and the state’s crime control function. Sociological research on policing has documented that police exercise their formal authority through a combination of legal rules (whose formal scope substantially understates the discretion that operational policing actually involves) and informal organizational cultures that shape how formal authority is deployed in ways that official policy documents do not acknowledge or control.
Research on police discretion — the degree to which officers exercise independent judgment in deciding whether to arrest, cite, warn, or take no action in encounters with potential law violators — has established that most police encounters do not result in arrest, and that the factors that determine arrest are not limited to the legal elements of offenses but include the demeanor of the suspect, the victim’s preference, the officer’s perception of the suspect’s social status, and the organizational pressures that shape officers’ risk-benefit calculations. Research by Black on the sociology of law documented that legal decisions — including police arrest decisions — systematically vary with the social characteristics of the parties involved, with higher-status complainants and lower-status suspects producing the highest arrest rates, regardless of the legal merits of the case.
Racial Disparities and Structural Bias
The racial disparities in police stops, searches, use of force, and arrests — among the most consistently documented findings in the sociology of police — reflect both the structural concentration of poverty and disadvantage in minority neighborhoods that determines patrol deployment and the specific racial biases, both implicit and explicit, that operate in individual officer decision-making. Research by Gelman, Fagan, and Kiss using hierarchical modeling of New York stop-and-frisk data found that Black and Latino civilians were stopped at substantially higher rates than white civilians in comparable situations after controlling for neighborhood crime rates — establishing that racial disparities in policing are not entirely attributable to differential criminal behavior but reflect differential enforcement of comparable behavior.
The sociological significance of these disparities extends beyond their immediate injustice to their systemic crime consequences through the legitimacy mechanism. Research by Desmond, Papachristos, and Kirk found that police killings of Black Americans substantially reduced subsequent crime reporting in Black communities — establishing that the legitimacy costs of racially disparate enforcement translate into concrete crime control deficits that undermine the formal control function even as they serve other institutional purposes. This finding connects the justice and the effectiveness arguments for racially equitable policing in a way that instrumental arguments alone cannot achieve.
Courts and Prosecutorial Discretion
The criminal courts — the institutional context within which guilt is determined and sentences are imposed — exercise formal social control through processes that are simultaneously governed by formal legal rules and substantially shaped by the informal organizational cultures, professional norms, and structural pressures that produce the “courtroom workgroup” described in Eisenstein and Jacob’s foundational research. The courtroom workgroup — the ongoing working relationship among prosecutors, defense attorneys, and judges who interact repeatedly in the same court — develops informal norms for how cases should be processed and what outcomes are appropriate for different offense and offender types, producing substantial consistency within jurisdictions alongside dramatic variation across them.
Prosecutorial discretion — the authority of prosecutors to decide whether and how to charge arrestees, what charges to file, and what plea agreements to offer — is the most consequential and least formally constrained decision-making power in the criminal justice system. Research on prosecutorial charging decisions has documented racial and socioeconomic disparities that persist after controlling for legally relevant case characteristics, establishing that the formal authority of prosecutorial discretion is exercised in ways that reproduce structural inequality through the case-processing decisions that formally neutral charging criteria do not adequately constrain. Research by Rehavi and Starr using federal sentencing data found that prosecutors charge Black men with more serious offenses than comparable white defendants, and that this charging disparity explains a substantial portion of the racial sentencing gap that sentencing research has documented.
The Sociology of Punishment
Incarceration Rates and Social Structure
The United States incarcerates approximately 2 million individuals on any given day — a rate of approximately 630 per 100,000 population that is the highest of any democracy and approximately five times the rate of comparable European nations. This extraordinary incarceration rate is not a natural response to higher crime rates (American crime rates are not five times higher than European rates) but a product of specific political and policy choices — mandatory minimum sentencing, truth-in-sentencing requirements, the war on drugs, and the political economy of crime that rewards punitiveness and punishes perceived leniency — that produce a carceral state whose sociological consequences the research literature has extensively documented.
Research by Western, Pettit, and colleagues has established that incarceration in the United States has become a specific life-course event that is nearly universal among young Black men with limited education in disadvantaged communities — with estimates suggesting that over 60% of Black male high school dropouts will be incarcerated at some point in their lives. This concentration of incarceration in specific populations and communities represents a structural feature of American social organization whose consequences — for families, communities, and the social bonds that crime prevention requires — cannot be adequately captured by individual-level recidivism studies that ignore the community-level costs of concentrated incarceration.
Collateral Consequences and Social Reintegration
The formal consequences of conviction — imprisonment, fines, probation — are accompanied by a web of collateral consequences that extend the formal social control apparatus far beyond sentence completion: employment restrictions that bar former offenders from licensed occupations; housing restrictions that deny public housing to individuals with criminal records; educational loan restrictions that limit access to higher education; civic restrictions including voting disenfranchisement; and family-law consequences including restrictions on child custody. Research by Mauer and Chesney-Lind has documented over 40,000 specific collateral consequences in state and federal law that collectively impose a lifelong formal exclusion from social participation whose crime consequences — through the collateral consequence spiral of reduced legitimate opportunity, elevated economic strain, and blocked social mobility — compound the direct effects of incarceration on criminal career persistence.
The Carceral State and Social Stratification
Sociologist Michelle Alexander’s analysis in The New Jim Crow (2010) — which argued that mass incarceration functions as a system of racialized social control that maintains racial hierarchy through the legal mechanisms of criminal justice rather than the discredited mechanisms of Jim Crow — generated both widespread public attention and scholarly debate about the relationship between formal social control and racial stratification. The sociological case for understanding mass incarceration as a system of social stratification rather than simply a crime control mechanism rests on multiple empirical pillars: the racial concentration of incarceration that makes the prison system’s effects on Black communities comparable in some respects to those of legally codified racial exclusion; the collateral consequences that permanently reduce the social membership and economic participation of formerly incarcerated individuals; and the political economy that has sustained mass incarceration despite its limited crime control effectiveness through the racial politics of crime that both parties have exploited for electoral advantage.
Research by Wacquant, Western, and others has extended this analysis beyond Alexander’s specific claims to examine the broader sociology of the carceral state as an institution of governance that simultaneously manages surplus populations, reproduces racial stratification, and responds to the political demands of constituencies for whom punitiveness signals moral seriousness about crime. This structural analysis of formal social control’s political economy does not deny the crime control functions of the criminal justice system but insists that those functions do not fully explain its structure or operation — that the carceral state is as much a social institution with its own political and economic logic as it is a crime control apparatus.
Labeling Theory and the Iatrogenic Effects of Formal Control
Secondary Deviance and Career Criminality
Labeling theory — developed by Lemert, Becker, and subsequently by Tannenbaum, Schur, and others — proposes that formal social control produces secondary deviance: criminal behavior that develops specifically in response to the label of criminal imposed by official processing. Lemert’s distinction between primary deviance (rule violation before official labeling) and secondary deviance (rule violation that develops in response to the reorganization of identity and social relationships around the deviant label) remains one of the most sociologically important contributions to criminological theory, establishing that the formal control apparatus designed to reduce crime can itself generate the stable criminal careers it is supposed to prevent.
Research testing labeling theory using quasi-experimental designs has found consistent evidence of labeling effects on subsequent offending, with the strongest effects for early and formal criminal justice contact. Research by Bernburg and Krohn using propensity score matching found that official labeling — police arrest relative to comparable incidents without arrest — significantly elevated subsequent criminal offending, with the effect operating through the social bond-weakening and criminal network-embeddedness mechanisms that labeling theory predicts. Research on the effects of juvenile record access, employment screening, and housing screening has documented the practical mechanisms through which the criminal label produces the social exclusion that elevates recidivism.
Restorative Justice and Therapeutic Jurisprudence
Restorative justice approaches — which replace or supplement the adversarial punishment model with processes that bring together offenders, victims, and communities to repair harm and reintegrate offenders — represent the most theoretically coherent alternative to conventional formal social control available in contemporary criminal justice. Research on restorative justice programs across multiple crime types and multiple countries has found generally positive effects on victim satisfaction, offender recidivism, and community involvement outcomes, with meta-analyses finding effect sizes comparable to those of evidence-based rehabilitative programming. The theoretical mechanism — that restorative processes restore the social bonds and normative commitments that conventional punishment severs — is directly consistent with the social bond and labeling theory frameworks that identify conventional punishment’s limitations.
Drug courts — which divert substance-involved offenders from conventional criminal prosecution into judicially supervised treatment programs — represent the most extensively evaluated therapeutic jurisprudence alternative, with a large literature of rigorous evaluations finding consistent reductions in recidivism among participants relative to comparable offenders processed conventionally. Research syntheses by Lowenkamp, Holsinger, and Latessa, and by the National Drug Court Institute, consistently find drug court participation associated with 10–20% reductions in recidivism, with the strongest effects for programs that follow RNR principles in targeting the highest-risk offenders for the most intensive services.
Criminal Justice Reform and the Evidence Base
What Works in Formal Social Control
The evidence-based criminology movement’s application to formal social control has produced a clearer picture of what formal interventions work — reducing crime and recidivism — and which are criminologically ineffective or counterproductive. Hot spots policing and problem-oriented policing show consistent positive effects on crime at targeted locations. Focused deterrence strategies produce meaningful reductions in gang-involved violence. Risk-Needs-Responsivity-based rehabilitative programming reduces recidivism among high-risk offenders by 10–20% in meta-analyses. Drug courts and other therapeutic jurisprudence alternatives to incarceration reduce subsequent criminal involvement among participants while reducing criminal justice costs.
What the evidence does not support — but what American criminal justice has most extensively deployed — is severity-based deterrence through incarceration expansion. The research consensus that additional incarceration beyond moderate levels produces minimal deterrent effects while imposing substantial social organization costs through concentrated disadvantage and community-level disruption provides the empirical foundation for the decarceration argument that justice reinvestment policy has begun to translate into legislative reform. The pace of that reform, however, remains far slower than the evidence supports. The political and institutional barriers to evidence-based formal social control reform — the career incentives of criminal justice professionals, the electoral incentives of politicians, the organizational inertia of institutions designed around punitive practices, and the racial politics of crime that have consistently rewarded punitiveness in American political culture — are as important as any evidential uncertainty in explaining the gap between what criminological research establishes about effective formal control and what American criminal justice practice actually deploys.
Community Supervision: Probation and Parole
Probation and parole — the formal supervision of individuals in the community rather than in prison — constitute the most numerically significant components of formal social control: approximately 4.5 million Americans are under community supervision at any given time, compared to approximately 2 million in prison and jail. Research on the crime prevention effectiveness of community supervision has found generally positive effects when supervision is organized around the risk-needs-responsivity (RNR) principles that the correctional rehabilitation literature has established: targeting supervision intensity to offender risk level, focusing supervision contacts on criminogenic needs, and delivering treatment services matched to individual learning styles.
Research by Pew Charitable Trusts and by the Urban Institute on probation and parole reform has documented that the primary source of returns to incarceration in American corrections is not new crime but technical violations — the failure to comply with supervision conditions (missing appointments, failing drug tests, traveling without permission) — that send large numbers of individuals back to incarceration without the criminal justice processing that new crimes require. Research finding that reducing technical revocations produces minimal increases in new criminal activity has provided the empirical foundation for supervision reform policies — including graduated sanctions for technical violations and the elimination of supervision conditions not directly related to criminogenic needs — that reduce incarceration through administrative reform while maintaining meaningful supervision for public safety purposes.
Wrongful Conviction and the Limits of Formal Truth-Finding
The formal social control apparatus’s claim to legitimacy rests substantially on its accuracy — its ability to correctly distinguish the guilty from the innocent and impose sanctions only on those who have committed the crimes for which they are prosecuted. Research on wrongful conviction — the formal process of identifying and exonerating individuals convicted of crimes they did not commit — has documented that the formal truth-finding process is substantially less accurate than its procedural safeguards are designed to ensure. Research by Gross and colleagues using the National Registry of Exonerations data has found that murder, rape, and drug conviction cases show particularly high rates of documented wrongful conviction, with estimates suggesting that 4–5% of individuals convicted of capital murder in the United States may have been innocent.
The sociological significance of wrongful conviction research extends beyond its implications for individual cases to its implications for the systemic accuracy of formal social control. If the most carefully processed cases — serious felony prosecutions that receive the most investigative and prosecutorial attention — show meaningful wrongful conviction rates, the accuracy of the full case volume that moves through the system with far less careful attention is likely substantially lower. Research on the specific mechanisms of wrongful conviction — eyewitness misidentification, false confession, informant testimony, forensic evidence error, and prosecutorial misconduct — documents the specific institutional failures through which formal social control misidentifies its targets, with implications for reform of each specific mechanism that the wrongful conviction literature has generated.
Conclusion
Formal social control is simultaneously the most studied and the most contested institution in criminology — studied because its operations are consequential, documented, and subject to rigorous evaluation; contested because its purposes (crime control, retribution, incapacitation, rehabilitation, restorative justice) are genuinely in tension with each other, its effects on crime are more limited and more conditioned than its formal justifications claim, and its distribution of enforcement across the social structure reflects inequalities of power that the sociology of law and the critical criminology traditions have documented extensively.
Formal social control is a sociological institution as much as a crime control mechanism — embedded in structures of power and inequality, shaped by organizational cultures and political economies, and producing consequences that extend far beyond the individual deterrence and incapacitation effects that its formal justifications invoke. The sociological analysis of formal social control simultaneously establishes what it can and cannot achieve (certainty deters more than severity; rehabilitation works when evidence-based; incarceration’s social bond costs undermine its deterrence benefits), explains why American formal control is structured as it is despite the evidence (the political economy of punitiveness, racial politics of crime, and institutional interests of the carceral apparatus), and points toward the reforms that evidence-based criminology consistently supports but political processes consistently resist.
The gap between what the sociology of formal social control establishes about effectiveness and the structure of American formal control remains among the most consequential and most frustrating in the policy landscape of criminological research. Closing it requires not only the continued production of rigorous evidence but the political conditions that make evidence-based reform politically feasible — conditions that the justice reinvestment movement, bipartisan reform coalitions, and the growing public recognition of mass incarceration’s social costs have begun to create, even as the pace of change remains far below what the evidence justifies.
References
- Beccaria, C. (1764/1995). On crimes and punishments (R. Davies, Trans.). Cambridge University Press.
- Becker, H. S. (1963). Outsiders: Studies in the sociology of deviance. Free Press.
- Bernburg, J. G., & Krohn, M. D. (2003). Labeling, life chances, and adult crime: The direct and indirect effects of official intervention in adolescence on crime in early adulthood. Criminology, 41(4), 1287–1318. https://doi.org/10.1111/j.1745-9125.2003.tb01020.x
- Black, D. (1976). The behavior of law. Academic Press.
- Braga, A. A., Weisburd, D., & Turchan, B. (2018). Focused deterrence strategies and crime control: An updated systematic review and meta-analysis of the empirical evidence. Criminology & Public Policy, 17(1), 205–250. https://doi.org/10.1111/1745-9133.12353
- Clear, T. R. (2007). Imprisoning communities: How mass incarceration makes disadvantaged neighborhoods worse. Oxford University Press.
- Desmond, M., Papachristos, A. V., & Kirk, D. S. (2016). Police violence and citizen crime reporting in the Black community. American Sociological Review, 81(5), 857–876. https://doi.org/10.1177/0003122416663494
- Dobbie, W., Goldin, J., & Yang, C. S. (2018). The effects of pre-trial detention on conviction, future crime, and employment. American Economic Review, 108(2), 201–240. https://doi.org/10.1257/aer.20161503
- Gelman, A., Fagan, J., & Kiss, A. (2007). An analysis of the New York City police department’s “stop-and-frisk” policy. Journal of the American Statistical Association, 102(479), 813–823. https://doi.org/10.1198/016214506000001040
- Lemert, E. M. (1951). Social pathology. McGraw-Hill.
- Nagin, D. S. (2013). Deterrence in the twenty-first century. Crime and Justice, 42(1), 199–263. https://doi.org/10.1086/670398
- Nagin, D. S., Cullen, F. T., & Jonson, C. L. (2009). Imprisonment and reoffending. Crime and Justice, 38(1), 115–200. https://doi.org/10.1086/599202
- Pager, D. (2007). Marked: Race, crime, and finding work in an era of mass incarceration. University of Chicago Press.
- Petersilia, J. (2003). When prisoners come home: Parole and prisoner reentry. Oxford University Press.
- Rehavi, M. M., & Starr, S. B. (2014). Racial disparity in federal criminal sentences. Journal of Political Economy, 122(6), 1320–1354. https://doi.org/10.1086/677255
- Rushin, S., & Edwards, G. S. (2017). De-policing. Cornell Law Review, 102(3), 721–782.
- Sherman, L. W., & Weisburd, D. (1995). General deterrent effects of police patrol in crime “hot spots”: A randomized, controlled trial. Justice Quarterly, 12(4), 625–648. https://doi.org/10.1080/07418829500093221
- Travis, J., Western, B., & Redburn, F. S. (Eds.). (2014). The growth of incarceration in the United States. National Academies Press. https://doi.org/10.17226/18613
- Tyler, T. R. (1990). Why people obey the law. Yale University Press.
- Western, B. (2006). Punishment and inequality in America. Russell Sage Foundation.
- Wildeman, C., & Wang, E. A. (2017). Mass incarceration, public health, and widening inequality in the USA. The Lancet, 389(10077), 1464–1474. https://doi.org/10.1016/S0140-6736(17)30259-3