Source-linked AI summary
Non-pharmaceutical interventions during the COVID-19 pandemic: a rapid review
Nicola Perra
TL;DR
The paper addresses the poorly developed empirical understanding of feedbacks between infectious diseases and human behavior. It reviews evidence on NPIs, models, and behavioral patterns, finding heterogeneous transmission and unequal capacity to reduce mobility and disease burden across socioeconomic groups.
Problem
Human behavior affects disease spread and is affected by illness, but epidemiology lacks a well-developed, standardized, empirically grounded account of this feedback loop.
Method
The review synthesizes epidemic modeling approaches, including compartmental and Bayesian statistical models, to examine transmission, behavioral change, and NPI effects.
Results
SARS-CoV-2 transmission is heterogeneous, with a minority of cases responsible for many infections and restaurants, hotels, and workplaces identified as riskier locations; NPI adoption varies across socioeconomic groups.
Takeaways & Limitations
Financial support and food programs may help NPI adoption, while spontaneous behavioral changes can precede official containment orders.
Takeaways & Limitations
The review is constrained by the rapidly expanding and highly diversified COVID-19 literature, making exhaustive manual coverage difficult.
Abstract
from arXiv · showhide
Infectious diseases and human behavior are intertwined. On one side, our movements and interactions are the engines of transmission. On the other, the unfolding of viruses might induce changes to our daily activities. While intuitive, our understanding of such feedback loop is still limited. Before COVID-19 the literature on the subject was mainly theoretical and largely missed validation. The main issue was the lack of empirical data capturing behavioral change induced by diseases. Things have dramatically changed in 2020. Non-pharmaceutical interventions (NPIs) have been the key weapon against the SARS-CoV-2 virus and affected virtually any societal process. Travels bans, events cancellation, social distancing, curfews, and lockdowns have become unfortunately very familiar. The scale of the emergency, the ease of survey as well as crowdsourcing deployment guaranteed by the latest technology, several Data for Good programs developed by tech giants, major mobile phone providers, and other companies have allowed unprecedented access to data describing behavioral changes induced by the pandemic. Here, I aim to review some of the vast literature written on the subject of NPIs during the COVID-19 pandemic. In doing so, I analyze 347 articles written by more than 2518 of authors in the last $12$ months. While the large majority of the sample was obtained by querying PubMed, it includes also a hand-curated list. Considering the focus, and methodology I have classified the sample into seven main categories: epidemic models, surveys, comments/perspectives, papers aiming to quantify the effects of NPIs, reviews, articles using data proxies to measure NPIs, and publicly available datasets describing NPIs. I summarize the methodology, data used, findings of the articles in each category and provide an outlook highlighting future challenges as well as opportunities
1 Introduction
Human behavior both shapes infectious-disease transmission and changes in response to disease, yet epidemiology still lacks a well-developed theory or standardized approach for capturing this feedback loop.
- Only 15% of pre-COVID papers on behavior and infectious diseases used empirical data, while most models lacked representative data and validation.
- Researchers used games, surveys, and behavioral datasets to study social distancing, vaccination attitudes, altruism, self-interest, and disease-related risk perceptions.
2 Disclaimer
The COVID-19 literature expanded so rapidly that manually reviewing all relevant papers became impractical, while this review’s PubMed-centered search introduces a disciplinary bias.
- PubMed searches for “COVID” returned more than 71,000 results, compared with more than 135,000 Google Scholar matches.
- The review focuses mainly on epidemiological implications of NPIs, including epidemic models and surveys of awareness and adoption.
- Using PubMed for much of the sample inevitably biases coverage toward interdisciplinary, biomedical, and life-science journals.
3 Data collection, inclusion principles, annotation, and classification
The review combines a hand-curated list with PubMed-based collection, classifies papers into seven methodology- or aim-based categories, and supplements the sample with author and citation data.
- Data collection: The review began with 57 influential papers and expanded the sample through a PubMed query submitted on November 10th.
- Classification: The sample was classified into seven categories: epidemic models, surveys, comments or perspectives, NPI effects, reviews, proxy-data studies, and datasets.
- Classification: The seven categories represent each paper’s most representative methodology and/or aim, with epidemic models comprising 29% and surveys 28%.
- Annotation: Each paper was assigned to one and only one main category, despite overlap and unclear boundaries among cases.
- Bibliometric data: Semantic Scholar supplied author and citation information, covering more than 2518 authors and over 9300 citations as of December 19th, 2020.
- Bibliometric data: Epidemic models were the most cited category by both total and median citations, whereas surveys had the lowest median citations.
4 Epidemic models
Epidemic-model studies use mechanistic, statistical, metapopulation, and agent-based approaches to represent disease spread and assess NPIs. Across these models, intervention timing, targeting, mobility, and behavioral adaptation substantially shape projected transmission and outcomes.
- Compartmental models: Compartmental models commonly follow SEIR-like structures and use epidemic data to estimate transmission, intervention effects, and behavioral changes.These models represent infection states and may add compartments for mask use, awareness, isolation, or other behaviors.
- Compartmental models: Timing and combined implementation of NPIs were critical: only compound interventions reduced the reproductive number below one, while delayed timing weakened suppression.Models also found targeted interventions cutting the right tail of transmission rates more efficient than population-level social distancing in some settings.
- Compartmental models: 80% mask adoption with 50% efficacy could reduce projected COVID-19 deaths by 17%−45% over two months in one model.Another model estimated that 70% adoption in New York and 80% in the USA, with efficacy above 70%, could lead to elimination under its assumptions.
- Metapopulation models: Metapopulation models couple compartmental disease dynamics across mobile sub-populations, enabling geographic analyses of reopening, mobility, and superspreading.One individual-based model estimated that 2% of cases caused 20% of infections, while another found that reducing time spent at points of interest was more efficient than reducing mobility broadly.
- Statistical models: Statistical models link current infections to prior infections and estimate intervention effects from observed data, including cross-country policy portfolios.One hierarchical analysis found lockdowns produced the largest variation in Rt among the examined NPIs, while patchy adoption was associated with worse health outcomes.
5 Surveys
Surveys provided direct observations of how NPIs affected health, behavior, medical practice, activities, and social contacts across diverse countries and populations. Across these studies, findings document substantial behavioral changes, uneven NPI adoption, disruptions to care, and reduced social contact.
- Survey scope: Surveys collected direct evidence on NPI effects across health and wellbeing, adoption and awareness, medical practice, human activities, and contacts.The review organized surveys into five categories covering these domains.
- Survey scope: 85 countries were represented among survey participants, while survey authors were affiliated with institutions in 41 countries and published across 65 journals.The United States, Italy, and China were the most represented countries among participants and affiliations.
- Health and wellbeing indicators: Mental-health surveys covered loneliness, domestic abuse, psychological state, and groups including frontline workers, students, parents, and children.Loneliness varied by age and gender: younger people, especially females, reported larger increases, while both very young and older groups appeared more affected.
- Adoption of NPIs: Female and highly educated respondents generally reported greater COVID-19 risk perception and NPI compliance, while age effects varied across studies and urban-rural barriers affected adoption.Peri-urban and rural areas in India showed barriers involving knowledge and opportunities to adopt hygienic behaviors.
- Changes on medical practice: 74% of caregivers across 29 countries reported losing access to at least one therapy, while 56% retained services and 36% lost a healthcare provider.Telemedicine was often useful, but socioeconomic, linguistic, health-related, revenue, and communication challenges complicated remote care.
- Effects of NPIs and contacts: 85% of bird watchers across 25 countries reduced travel and targeted local sites, while NPIs also altered leisure, education, mobility, consumption, and research activity.The review also reports a factor 7–8 reduction in social contacts in Wuhan and Shanghai, a factor 4 reduction in the UK, and post-lockdown increases of 5–17% in China that remained 3–7 times below pre-pandemic levels.
6 Comments and/or perspectives
Comments and perspectives examine how NPIs affected medical practice, modeling, activities, and strategies for lifting measures, while also drawing lessons from successful suppression efforts.
- Change to medical practice: Comments and perspectives discuss changes to medical procedures and workflows across specialties including mental health, emergency, substance-abuse, oncological, pediatric, and other care.NPIs altered routines, social support, and standard procedures, prompting recommendations such as new routines, digital interventions, community outreach, and family support.
- Modeling: The category also addresses epidemic modeling, transmission dynamics, novel data streams, and opportunities to use mobile-phone data for surveillance, mobility monitoring, and contact tracing.These applications span early case detection, population-level mobility measurement, epidemic-model inputs, and hotspot-risk identification after epidemic peaks.
- Effect of NPIs: Other perspectives reflect on NPI effects on children, education, gambling, food, other diseases, the environment, corporations, and porn consumption.
- Lifting NPIs: Four articles consider relaxing NPIs, including second-wave risks and children’s return to school after prolonged isolation.One perspective argues that suppression approaches may provide time to build testing and contact-tracing infrastructure, while warning that missing infrastructure could permit resurgence.
- Success stories: Perspectives on South Korea and Taiwan identify early action, rapid testing, aggressive contact tracing, social distancing, and widespread mask adoption as central elements of suppression.
7 Quantifying the effects on NPIs
The reviewed studies classify NPI effects on disease spread, human behavior, and medical practice, using country-specific, multicountry, and proxy-based evidence. Findings include reduced incidence and transmission, behavioral shifts, and substantial disruptions to healthcare delivery.
- Classification: NPIs are grouped into voluntary bottom-up responses and authority-imposed top-down measures, although the distinction can be concurrent and unclear.Examples include face covering, hygiene, social distancing, border closures, school closures, curfews, and lockdowns.
- Effects on spreading: Studies of disease spread examine country-specific, multicountry, and other-disease effects, including travel bans, mobility restrictions, surveillance, testing, and contact tracing.The reviewed evidence uses epidemic, mobility, case, travel-history, and contact-tracing data.
- Effects on spreading: 13% lower incidence was associated with implementation of any NPI in the reviewed multicountry evidence.Stronger reductions were associated with higher GDP per capita, a higher proportion of people aged about 65, and country security index.
- Effects on behaviors: NPI studies also characterize effects on online activity, health, mobility, spending, and financial-market performance.Examples include shifts in psycholinguistic features and increased interest in nature during strict NPIs.
- Effects on medical practice: Medical-practice studies report projected avoidable cancer-death increases of 8–10% for breast cancer and 15–17% for lung cancer after diagnostic and treatment disruptions.Other studies describe missed HIV visits and reduced therapy dispensing, ICU-capacity effects, and reduced incidents in a long-term-care institution.
8 Reviews
The review articles are organized into four themes covering NPIs, medical-practice changes, NPI adoption, and digital technologies. They synthesize effectiveness, adaptation, implementation challenges, and opportunities alongside privacy, legal, ethical, and data-bias concerns.
- Review sample: The review sample contains 21 articles, included because they provide broad overviews of particular topics or contexts.Some could also be classified as perspectives or comments.
- Review themes: Reviews are grouped into four themes: NPIs, change in medical practice, adoption of NPIs, and impact of digital technologies.The taxonomy covers effectiveness and modeling, medical adaptation, adoption challenges, and technology-enabled behavioral data.
- Review themes: NPI reviews summarize effectiveness, effects, and modeling, with some evidence indicating that portfolios of measures appear more effective than single interventions.The reviewed modeling work includes an Indian context, while other reviews synthesize evidence and approaches across the globe.
- Review themes: Medical-practice reviews cover surgery, ophthalmology, routine vaccination campaigns, and dentistry.They summarize approaches, results, and reported adaptations to NPIs during the pandemic.
- Review themes: Adoption reviews link health-behavior uptake to socioeconomic and psychological factors and describe practical barriers such as limited access to water and soap.The adoption evidence includes several African-country contexts.
- Review themes: Digital-technology reviews describe opportunities for capturing behavior and informing modeling and policy while highlighting privacy, legal, ethical, and data-bias concerns.Mobile phones are cited as an example of the technologies considered.
9 Measuring NPIs via proxy data
Proxy-data studies measured NPI adoption and mobility changes using digital traces from phones, platforms, apps, and media. Findings link social ties, political and socioeconomic characteristics, and voluntary behavior to NPI adherence and mobility reductions.
- Articles classified proxy-data studies into NPI adoption and mobility changes.
- Adoption of NPIs: Proxy measures drew on mobile phones, Twitter, ad-hoc apps, Wikipedia, Reddit, Google, and television viewing.
- Adoption of NPIs: Counties with stronger Facebook ties to China and Italy were more likely to comply with NPIs.The authors suggest that information from highly affected areas might influence risk perception and behavior change.
- Adoption of NPIs: Mobility and NPI adherence differed by political affiliation, education, climate-change views, and neighborhood wealth.The reviewed findings indicate heterogeneous behavioral responses across social groups.
- Mobility changes: About half of proxy-data papers studied mobility changes associated with travel bans, school closures, remote working, and lockdowns.Common sources included mobile-phone platforms, Facebook, Google, and other mobile applications.
- Mobility changes: Mobility reductions began before stay-at-home mandates, and variation in mobility correlated above 0.7 with disease growth in 20 of 25 counties.Disease changes became visible 9–12 days later and remained associated for up to three weeks.
10 Datasets
The review identifies publicly shared NPI datasets covering intervention implementation, mobility, and physical contacts. These resources span governmental responses, mobile-phone mobility measures, and contact matrices across many countries.
- Classification: NPI datasets were divided into implementation, mobility, and contacts categories.
- Implementation of NPIs: Implementation datasets primarily described governmental responses, including detailed timelines and intervention typologies.The Oxford COVID-19 Government Response Tracker provides 17 indicators of governmental responses.
- Implementation of NPIs: The Oxford tracker provides overall response, stringency, containment and health, and economic support indexes for modeling use.
- Mobility: Mobility datasets captured NPI-associated changes using measures such as origin-destination matrices and mobile-phone observations.One Italian dataset covered more than 80,000 users from January 18 to April 17.
- Contacts: The Socrates project assembled pre- and post-COVID contact matrices across multiple countries and provided an interactive extraction tool.
11 Conclusions and outlook
The reviewed literature characterizes SARS-CoV-2 transmission, disease-risk heterogeneity, and unequal effects of NPIs across socioeconomic groups. Despite unprecedented behavioral datasets and data-driven modeling, a validated theory of the behavior–disease feedback loop remains absent.
- Most models estimated R0 in the 3–4 range, while transmission remained partly undetected because of limited testing and asymptomatic infections.
- Transmission was heterogeneous, with a minority of cases responsible for a large fraction of infections and several high-risk settings identified.Restaurants, hotels, workplaces, households, and extended-family interactions were highlighted in the review.
- High-income individuals reduced mobility more than low-income individuals, while disadvantaged groups experienced higher infection rates and disease burden.The review connects these disparities with healthcare access and preexisting comorbidities.
- COVID-19 shifted NPI research from mainly theoretical approaches toward data-driven modeling using unprecedented high-resolution behavioral datasets.The review notes that much theoretical work on behavioral change was neglected during the pandemic.
- A validated theory describing the feedback loop between behavior and infectious diseases remains an open challenge.