论文导读 | Operations Research近期文章精选

推文作者：张曦予

 编者按

本期我们选取了七月后半月来自Operations Research的一篇文章以及来自Management Science的四篇文章以飨读者，内容涉及多个方面，我们选取的文章包含了各个方面对于管理以及运筹进行研究的文章，如研究疲劳对救护车工作表现的影响，对互联医疗的评估，还有如研究企业家和工薪人员的竞争力等较为新颖有趣的文章。

文章1

The Distributional Impact of Fatigue on Performance

疲劳对工作表现的分布影响

基本信息

● 作者：Hessam Bavafa，Jónas Oddur Jónasson

● 发表时间：2023.7.21

● 原文链接：https://doi.org/10.1287/mnsc.2023.4855

● 关键词：

• Fatigue（疲劳）

• Service time distribution（服务时间分布）

• Performance consistency（表现一致性）

• Variability（变异性）

• Operational performance（运营绩效）

• Service operations （服务业务）

• Ambulance operations（救护车业务）

主要内容

摘要：

Little is known about how people-centric factors affect the shape  of service time distributions despite distributional statistics  (variance or quantiles) being key drivers of system performance in many  service industries. We investigate the impact of one people-centric  factor—worker fatigue—on the average, variance, and quantiles of  service times in paramedic operations. Our analysis uses data on the  performance of 368,634 paramedic teams in the London Ambulance Service  over 10 years. We measure fatigue by the number of prior jobs a  paramedic crew has completed during a shift and estimate its impact on  the time it takes the crew to respond to incidents and bring patients to  hospitals. Using a recentered influence function regression approach  with multiple fixed effects, we find that the average time to hospital  increases by 5% throughout the course of an average shift. In addition,  the workers become less consistent with fatigue; service time variance  increases by 39% during a normal shift. Furthermore, we find that in  addition to an upward shift in mean service times, both the upper and  lower tails of the distribution have more weight for fatigued  paramedics. These effects are driven mostly by the performance of  paramedics at the scene, rather than their driving to or from the  incident. The distributional effects of fatigue are only slightly  mitigated by increased experience or reduced system workload. Our work  demonstrates that the impact of people-centric factors can be highly  nonuniform across the service time distribution.

• 文章研究了一个以人为中心的因素--员工疲劳--对辅助医疗操作中服务时间的平均值、方差和定量的影响。

• 方法：文章使用了伦敦救护服务机构 368,634 个辅助医疗团队 10 年来的绩效数据。文章通过护理人员在轮班期间完成的提前的工作数量来衡量疲劳程度，并估算疲劳程度对护理人员应对突发事件和将病人送往医院所需的时间的影响。文章使用具有多重固定效应的重定向影响函数回归方法。

• 结论：文章发现，在平均轮班过程中，送往医院的平均时间会增加 5%。此外，工人的疲劳程度也变得不那么一致；在正常轮班期间，服务时间差异增加了 39%。此外，文章还发现，除了平均服务时间向上移动外，分布的上端和下端对于疲劳的护理人员来说都具有更大的权重。造成这些影响的主要原因是护理人员在现场的表现，而不是他们开车往返事故现场的情况。经验的增加或系统工作量的减少只能稍微缓解疲劳对分布的影响。文章的工作表明，以人为本的因素对整个服务时间分布的影响可能非常不均匀。

文章2

Organizing Data Analytics

组织数据分析

基本信息

● 作者：Ricardo Alonso, Odilon Câmara

● 发表时间：2023.7.18

● 原文链接：https://doi.org/10.1287/mnsc.2023.00207

● 关键词：

• Strategic experimentation（战略实验）

• Bayesian persuasion（贝叶斯劝说）

• tampering（篡改）

• Organizational design（组织设计）

• Information technology（信息技术）

• audit（审计）

主要内容

摘要：

We develop a theory of credible skepticism in organizations to explain  the main tradeoffs in organizing data generation, analysis, and  reporting. In our designer-agent-principal game, the designer selects  the information privately observed by the agent who can misreport it at a  cost, whereas the principal can audit the report. We study three  organizational levers: tampering prevention, tampering detection, and  the allocation of the experimental-design task. We show that motivating  informative experimentation while discouraging misreporting are often  conflicting organizational goals. To incentivize experimentation, the  principal foregoes a flawless tampering detection/prevention system and  separates the tasks of experimental design and analysis.

• 文章提出了组织中的可信怀疑理论，以解释在组织数据生成、分析和报告过程中的主要权衡。

• 在文章的设计者-代理人-委托人博弈中，设计者选择代理人私下观察到的信息，代理人可以有代价地误报信息，而委托人则可以审核报告。文章研究了三个组织杠杆：篡改预防、篡改检测和实验设计任务分配。

• 文章的研究表明，激励信息丰富的实验与阻止误报往往是相互冲突的组织目标。为了激励实验，放弃了完美无瑕的篡改检测/预防系统，并将实验设计和分析任务分开。

文章3

Evaluating the Efficacy of Connected  Healthcare: An Empirical Examination of Patient Engagement Approaches  and Their Impact on Readmission

评估互联医疗的功效：患者参与方法及其对再入院影响的实证研究

基本信息

● 作者：Suparerk Lekwijit, Christian Terwiesch, David A. Asch, Kevin G. Volpp

● 发表时间：2023.7.18

● 原文链接：https://doi.org/10.1287/mnsc.2023.4865

● 关键词：

• Connected healthcare（互联医疗）

• Health information technology（医疗信息技术）

• Medication adherence（用药依从性）

• Readmission analytics（再入院分析）

• Behavioral interventions（行为干预）

主要内容

摘要：

Connected healthcare is a form of health delivery that connects patients  and providers through connected health devices, allowing providers to  monitor patient behavior and proactively intervene before an adverse  event occurs. Unlike the costs, the benefits of connected healthcare in  improving patient behavior and health outcomes are usually difficult to  determine. In this study, we examine the efficacy of a connected health  system that aimed to reduce readmissions through improved medication  adherence. Specifically, we study 975 patients with heart disease who  received electronic pill bottles that tracked medication adherence.  Patients who were nonadherent received active social support that  involved different types of feedback, such as text messages and calls.  By integrating data on adherence, intervention, and readmission, we aim  to (1) investigate the efficacy of connected healthcare in promoting  medication adherence, (2) examine the relationship between medication  adherence and readmission, and (3) develop a dynamic readmission  risk-scoring model that considers medication adherence and use the model  to better target nonadherent patients. Our findings suggest that  patients are more likely to become adherent when they or their partners  receive high levels of intervention that involve personalized feedback  and when the intervention is escalated quickly and consistently. We also  find that long-term adherence to three common heart medications is  strongly associated with reduced readmission risk. Lastly, using  counterfactual simulation, we apply the dynamic readmission risk-scoring  model to our setting and find that, when using an intervention strategy  that prioritizes high-risk patients, we obtain 10% fewer readmissions  while using the same effort level from the patient support team.

• 互联医疗是一种医疗服务形式，它通过互联医疗设备将患者和医疗服务提供者连接起来，使医疗服务提供者能够监控患者的行为，并在不良事件发生前主动干预。与成本不同，互联医疗在改善患者行为和健康结果方面的效益通常难以确定。在本研究中，文章考察了互联医疗系统的功效，该系统旨在通过改善用药依从性来减少再入院率

• 研究对 975 名心脏病患者进行了研究，这些患者接受了可追踪服药依从性的电子药瓶。不坚持用药的患者会得到积极的社会支持，包括不同类型的反馈，如短信和电话。通过整合用药依从性、干预和再入院数据，文章的目标是：（1）研究互联医疗在促进用药依从性方面的功效；（2）研究用药依从性与再入院之间的关系；（3）开发一种考虑到用药依从性的动态再入院风险评分模型，并利用该模型更好地定位不依从的患者。

• 文章的研究结果表明，当患者或其伴侣接受了包括个性化反馈在内的高水平干预，并且干预升级迅速且持续时，患者更有可能坚持用药。文章还发现，长期坚持服用三种常见的心脏病药物与再入院风险的降低密切相关。最后，通过反事实模拟，文章将动态再入院风险评分模型应用到研究的环境中，并发现当使用优先考虑高风险患者的干预策略时，在患者支持团队付出相同努力的情况下，再入院率降低了 10%。

文章4

The Price of Safety: The Evolution of Municipal Bond Insurance Value

Competitiveness of Entrepreneurs and Salaried Workers

企业家和工薪人员的竞争力

基本信息

● 作者：Loukas Balafoutas, Mongoljin Batsaikhan, Matthias Sutter

● 发表时间：2023.7.20

● 原文链接：https://doi.org/10.1287/mnsc.2023.4838

● 关键词：

• competitiveness（竞争力）

• entrepreneurs（企业家）

• Salaried workers（工薪人员）

• profit（利润）

• Field behavior（现场行为）

• experiment（实验）

主要内容

摘要：

We measure the willingness to compete of entrepreneurs and salaried  workers in an experiment. Participants can choose between a piece rate  and a tournament scheme in either private or public. We find that in the  private condition, entrepreneurs are less competitive than salaried  workers, but in the public condition, this ordering is reversed. Survey  data suggest that perceived norms of appropriate behavior, along with  beliefs about the instrumental value of competitiveness for professional  success, can explain why entrepreneurs are more competitive when  decisions are publicly observable. We also find that the latter  condition improves the quality of experimental decisions.

• 文章在一项实验中衡量了企业家和工薪人员的竞争意愿。参与者可以选择计件工资制或锦标赛制。

• 研究发现，在私有条件下，企业家的竞争力低于受薪工人，但在公共条件下，这种排序正好相反。调查数据表明，对适当行为规范的认知，以及关于竞争力对职业成功的工具价值的信念，可以解释为什么在决策可公开观察的情况下，企业家的竞争力更强。我们还发现，后一种情况会提高实验决策的质量。

文章5

Scores for Multivariate Distributions and Level Sets

多元分布和水平集得分

基本信息

● 作者：Xiaochun Meng, James W. Taylor, Souhaib Ben Taieb, Siran Li

● 发表时间：2023.7.24

● 原文链接：https://pubsonline.informs.org/doi/abs/10.1287/mnsc.2023.4819

● 关键词：

• individual treatment effect-个体处理效应

• conditional average treatment effect-条件平均处理效应

• conditional value at risk-条件风险价值

• partial identification-部分识别

• debiased machine learning-去偏移机器学习

主要内容

摘要：

Forecasts of multivariate probability distributions are required for a variety of applications. Scoring rules enable the evaluation of forecast accuracy and comparison between forecasting methods. We propose a theoretical framework for scoring rules for multivariate distributions that encompasses the existing quadratic score and multivariate continuous ranked probability score. We demonstrate how this framework can be used to generate new scoring rules. In some multivariate contexts, it is a forecast of a level set that is needed, such as a density level set for anomaly detection or the level set of the cumulative distribution as a measure of risk. This motivates consideration of scoring functions for such level sets. For univariate distributions, it is well established that the continuous ranked probability score can be expressed as the integral over a quantile score. We show that, in a similar way, scoring rules for multivariate distributions can be decomposed to obtain scoring functions for level sets. Using this, we present scoring functions for different types of level sets, including density level sets and level sets for cumulative distributions. To compute the scores, we propose a simple numerical algorithm. We perform a simulation study to support our proposals, and we use real data to illustrate usefulness for forecast combining and conditional value at risk estimation.

• 通过评分规则可以评估预测的准确性，并对各种预测方法进行比较，文章提出了多元分布评分规则的理论框架，其中包括现有的二次评分和多元连续排序概率评分。

• 文章演示了如何利用该框架生成新的评分规则。在某些多元情况下，需要对水平集进行预测，例如用于异常检测的密度水平集或作为风险度量的累积分布水平集。这就需要考虑此类水平集的评分函数。对于单变量分布来说，连续排列的概率得分可以表示为量化得分的积分，这一点已经得到公认。

• 研究证明，多变量分布的评分规则也可以通过类似的方法进行分解，从而得到水平集的评分函数。由此，文章提出了不同类型水平集的评分函数，包括密度水平集和累积分布水平集。为了计算得分，文章提出了一种简单的数值算法。文章实行了一项模拟研究来支持这个建议，并使用真实数据来说明预测组合和条件风险值估计的实用性。