Routing for Rural Health: Optimizing Community Health Worker Visit Schedules

Community health worker programs provide healthcare to those living outside the financial and physical reach of the standard health infrastructure. These programs are particularly prevalent in low resource regions. Frequently such programs involve community health workers making household visits across a significant geographical area. We suggest that this problem can be posed as a formal routing and scheduling problem, and to use techniques developed from solving the travelling salesman problem with time windows. In addition, household visits can generate a series of future follow up visits, a feature not often handled in the combinatorial scheduling and routing literature. We present the basic problem and outline potential research directions. Worldwide, community health worker (CHW) programs provide access to health information and interventions for people outside the financial or physical reach of the standard healthcare infrastructure. Such programs include the Lady Health Worker program in Pakistan, the Millennium Village Projects’ CHWs in Africa and Bangladesh Rural Advancement Committee’s (BRAC) CHW programs in multiple countries. Though there is a large amount of variation between different CHW programs, many involve a community health worker making household visits in order to provide health education, dispense common over-the-counter medication, or perform limited medical procedures. There is encouraging evidence that community health workers are associated with positive improvement in community health. For example, recent work by Baqu et al. (2008) found that in a large randomized trial, community health workers were associated with an improvement of over 30% in neonatal survival rates in Bangladesh. Yet in this intervention, each community health worker was responsible for roughly 4000 people, and 20% of children failed to receive a single post-natal visit, let alone the recommended three visits in the week. In addition, Baqu and colleagues report that CHWs “attended less than 5% of all births because of their high workload, travel distances, and difficulties in receiving timely notification of deliveries.” In this and a number of other CHW programs, community health workers need to visit a large number of households, Copyright c © 2010, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. separated by non-trivial distances. There are many household visit types, such as regular check ups, prenatal visits, and visits to dispense pneumonia medication. These visit types come with different time granularity scheduling constraints: regular checkups can occur almost anytime, prenatal visits should be at regular intervals, and a pneumonia patient must be visited frequently over a short time course. In addition, different visit types will take different mean durations of time to complete. Depending on the focus of the CHW program, different visits will have different priorities. In several programs, like the Bangladesh study, CHWs have too many visits to accomplish. In such cases, each day a CHW is faced with a limited amount of time and a large number of potential household visits to do, each of which is associated with different priorities. The CHW must reason about uncertainty over travel and visit duration, the relative proximity of households, and the relative timing constraints of different visits. This is an incredibly complex problem to solve when there is not enough time to do all the visits, and to our knowledge, typically CHW visit scheduling is done in an ad hoc way, either by the community health worker herself, or by program administrators. In these limited resource situations, there is a significant potential for optimization to help use CHWs’ time in the most efficient manner to improve prioritized health outcomes for the community they serve. Given this, we propose posing CHW household visit scheduling as a formal schedule and routing optimization problem. We first frame CHW scheduling and routing as an instance of multiple traveling salesman problems with time windows, and then discuss additional problem features, including one that may lead to novel scheduling research.