[Paper Review] COVID-19: Strategies for Allocation of Test Kits

Arpita Biswas, Shruthi Bannur+2 more|arXiv (Cornell University)|Apr 3, 2020

Machine Learning and Algorithms4 references4 citations

TL;DR

This paper proposes data-driven, adaptive strategies for allocating limited daily test-kits to detect asymptomatic and early-stage COVID-19 cases, focusing on reducing community spread and improving unbiased data collection. It introduces offline and online recommendation frameworks using utility scores—based on risk, uncertainty, or uniform sampling—to prioritize individuals for testing under a daily budget, with rate-limiting to prevent overcrowding at testing centers.

ABSTRACT

With the increasing spread of COVID-19, it is important to systematically test more and more people. The current strategy for test-kit allocation is mostly rule-based, focusing on individuals having (a) symptoms for COVID-19, (b) travel history or (c) contact history with confirmed COVID-19 patients. Such testing strategy may miss out on detecting asymptomatic individuals who got infected via community spread. Thus, it is important to allocate a separate budget of test-kits per day targeted towards preventing community spread and detecting new cases early on. In this report, we consider the problem of allocating test-kits and discuss some solution approaches. We believe that these approaches will be useful to contain community spread and detect new cases early on. Additionally, these approaches would help in collecting unbiased data which can then be used to improve the accuracy of machine learning models trained to predict COVID-19 infections.

Motivation & Objective

• Address the limitations of current rule-based testing policies that miss asymptomatic and community-spread cases.
• Develop systematic test-kit allocation strategies to detect new infections early and collect unbiased data for improving predictive models.
• Optimize limited daily test-kit budgets to enhance containment of community spread while ensuring high-risk individuals are prioritized.
• Design a coordination mechanism between symptom-checker apps and testing facilities to manage testing capacity and prevent overcrowding.
• Provide a framework for utility-based prioritization that supports both public health containment and model improvement objectives.

Proposed method

• Formalize the test allocation problem as selecting K_i individuals per day for testing, based on a utility function U(x) derived from individual risk factors.
• Propose an offline mode where all individuals are scored and ranked using a utility function, then select the top K_i based on their scores.
• Introduce an online mode with time-slot-based rate-limiting: divide the day into six 4-hour slots and allocate α_t * K_i tests per slot, where α_t is derived from prior-day arrival rates.
• Use a threshold-based decision rule in the online mode: recommend testing if an individual's score exceeds the (α_t * K_i)-th highest score from the previous day’s arrivals in the same slot.
• Define utility functions U(x) as either infection probability p_M(y=1|x), entropy H(p_M(y|x)) for uncertainty reduction, or uniform 1 for balanced data collection.
• Coordinate symptom-checker apps with testing centers to estimate daily test-kit availability and allocate dedicated time slots, though time-slot allocation is not modeled in this work.

Experimental results

Research questions

• RQ1 How can test-kits be allocated to detect asymptomatic and early-stage cases more effectively than rule-based approaches?
• RQ2 What utility functions U(x) best balance public health containment with data collection for improving infection prediction models?
• RQ3 How can online test allocation be managed in real time without overwhelming testing facilities, given limited daily test-kit availability?
• RQ4 What role does rate-limiting across time slots play in ensuring equitable and efficient test-kit distribution throughout the day?
• RQ5 In what ways can test-kit utilization be optimized to maximize detection of low-prevalence cases, especially when most results are negative?

Key findings

• The proposed strategies enable systematic testing of asymptomatic individuals who would otherwise be missed by symptom- or contact-based testing policies.
• Using utility functions such as infection probability or entropy allows prioritization of high-risk individuals while still allocating tests to low-risk individuals for data diversity.
• The rate-limiting mechanism in the online mode ensures that test allocations are distributed across the day, preventing large gatherings at testing centers.
• The model supports both public health containment and data collection goals by allowing flexible utility function design based on policy objectives.
• The framework is adaptable to real-world constraints such as limited daily test-kit availability and coordination with existing testing infrastructure.
• While not empirically validated, the approach offers a practical, implementable foundation for improving test allocation in large-population settings like India.