XVIII.6 November + December 2011
Page: 50
Digital Citation

Simplicity and usability

Gerald Douglas, Zach Landis-Lewis, Harry Hochheiser

back to top 

When it comes to electronic medical records (EMRs), complexity is often perceived as being the enemy of usability [1]. Computer information systems that must support the range of medical specialties, disciplinary perspectives, and administrative and regulatory requirements found in many modern healthcare environments can often become unwieldy and cumbersome, turning an opportunity for improving care into a potential cause of errors and confusion [2,3].

These difficulties may not be inevitable. The successful use of simple, straightforward touchscreen EMRs for managing anti-retroviral therapy (ART) and diabetes in Malawi suggests that well-designed interfaces can be used at a national scale by healthcare workers in low-resource environments. The adoption of touchscreen EMRs in Malawi presents intriguing questions: Can these techniques guide development of comparable systems in other low-resource environments and, ideally, offer guidance in handling the complexity of many medical environments?

Malawi is a landlocked country in Southern Africa with a population of 15 million. The government of Malawi provides healthcare at no charge to patients through a network of four central hospitals, 24 district hospitals, and roughly 400 health centers nationwide. Healthcare delivery in Malawi is hindered by high levels of morbidity and mortality resulting from endemic disease (primarily Malaria and HIV/AIDS), and a human-resource crisis in the healthcare sector in which the patient-to-doctor ratio is greater than 50,000:1. By contrast, South Africa has a patient-to-doctor ratio of roughly 1,300:1, and the U.S. ratio is approximately 375:1 [4].

While Malawi spends a greater proportion of its budget on healthcare than most developing countries, medical spending is still very low, with most healthcare being delivered in primary-care outpatient clinics. Systemic challenges associated with the lack of resources include poor clinical-data quality due to reliance on overworked ward clerks with little training in medical documentation, overworked clinicians, and documentation policies that are often poorly explained, poorly understood, or irrelevant to clinical practices.

In 2000, Gerald Douglas founded Baobab Health, a non-governmental organization devoted to the development of effective EMRs for Malawian healthcare. The Baobab anti-retroviral therapy (BART) system is Baobab's largest current deployment, with systems currently deployed at 15 high-burden sites managing clinical data for more than 120,000 patients, representing several million clinical encounters. A companion system for diabetes care has been deployed [5], and efforts in progress include expanding the Baobab model to focus on maternal and child care. Together with companion efforts aimed at providing standardized barcode patient IDs in support of improving continuity of care, Baobab systems have registered more than 1.3 million individuals. A dedication to easy-to-use touchscreen interfaces that require little training for healthcare workers [6] and principled adherence to user interface design guidelines aimed at maintaining this usability provide a shared conceptual framework for these efforts.

back to top  Origins

Baobab's design stems from initial inquiries conducted in 2000 that established a need for a point-of-care system that would reduce both the cognitive and physical burdens of delivering care while collecting complete and accurate data as a transparent byproduct of system use—without imposing additional burdens on clinicians. In contrast with many EMR efforts, a conscious decision was made not to build a generic system [7]. Rather, targeted systems, tailored for disease-specific use cases (initially ART for HIV), would provide specific alerts and content best suited to meeting relevant health challenges. Domain-specific functionality would be supported by cross-cutting modules supporting patient registration, the ordering of lab tests, and the prescribing and dispensing of medications. Simple interfaces would support use by healthcare workers with minimal education (often only high school level) and low computer literacy. Challenges in realizing such a vision included the absence or unreliability of grid power, low staffing, a general lack of supervision, and a severely constrained budget. Ease of use and learnability were identified early on as key requirements for successful deployment and scalability [8].

A touchscreen interface was selected as the most promising approach for providing high levels of learnability and usability, along with the simplicity of an integrated, one-piece solution. A combination of imaginative tinkering, creative procurement—obsolete Internet appliances purchased as surplus on eBay—and infrastructure development for reliable power and connectivity provided a platform for an initial system [9].

The combination of usability requirements and hardware constraints dictated the initial design of the Baobab interface. Given that the mode of interaction would be touchscreen, the physical size of buttons and other controls when rendered on the screen needed to be large enough for reliable finger-touch activation, and sufficiently far apart so as not to activate an adjacent control when making a selection. Additionally, controls could not be placed close to the edge of the screen (approximately 0.3 inch) as finger-touch activation in this region was inhibited by the screen's plastic bezel. This, combined with the already small screen size (10.4 inches) limited the amount of information that could be displayed at any given time. While the small display could be seen as a constraint, it did result in having screens with low information density, presenting an element of simplicity that worked well for users without high levels of computer literacy. The limitations of the touchscreens provided further constraints, as calibration drift made the selection of small targets prohibitively difficult.

A wizard-based approach was adopted to further increase the simplicity of the user interface. Wizard screens were divided into three regions. An upper region occupying roughly 40 percent of the height of the screen was dedicated to displaying information currently being captured. Depending upon the particular task (patient registration, prescribing medication, ordering laboratory tests), this could range from as few as four or five pieces of information to as many as 15 or 20.

Tasks requiring large amounts of data to be collected were broken down into logical groups so that the information density of the screen was maintained without significant loss of global context. The lower portion of the screen was dedicated to displaying context-specific user controls designed to best capture the particular piece of information being collected. For simple questions like "Patient Sex," buttons labeled "Male" and "Female" were displayed. When responses required a broader range of possible responses, such as a drug name or clinical symptom, list boxes were used. Hierarchically organized responses such as the name of a village were displayed using cascading list boxes (see Figure 1).

The third region of the screen was dedicated to user controls required for global navigation (back, next), a button to clear the current entry (reset the page), and a cancel button, located in the upper-right-hand corner of the screen, that would return the user to the previous menu. The simplicity of the interface affords the use of rigorous state modeling in support of error prevention. Controls are enabled only when appropriate, with numeric input via on-screen buttons enabled when task-appropriate, significantly reducing the possibility of entering incorrect data.

The initial Baobab interface was implemented in Microsoft Visual Basic (VB). Early experience with VB, including the realization that native VB controls were too small for touchscreen use, led us to develop a custom set of VB widgets, forming the first generation of the Baobab GUI toolkit. Although these widgets decreased development time and maximized the consistency of the look and feel of the application, limitations of the Visual Basic approach led to a search for alternative platforms. A prototype browser-based system developed using Ruby on Rails was implemented, leveraging Rails' support for the Model-View-Controller design pattern and the interactive capabilities of Javascript to provide a more robust and flexible framework that would be easier to scale and deploy (see Figure 2). The resulting Baobab EMR runs on an open source software stack, using the data model from the OpenMRS open source electronic medical record [10].

Between 2009 and 2011, the majority of the original hardware (customized Internet appliances) were replaced with off-the-shelf hardware. After almost a decade of use, these devices had run with minimal failures, demonstrating far greater reliability and lower cost than a conventional desktop computer. The second-generation Baobab hardware platform builds on this success, using solid-state disks and heat sinks instead of fans for ease of maintainability, and adapted power supplies integrated with Baobab's battery-backed direct current (DC) power architecture [9].

Baobab modules currently in use or development in Malawi include systems for ART, diabetes and hypertension, antenatal care, maternal and pediatric care, pharmaceutical stock management, laboratory specimen management, and radiology, with plans to expand into additional domains based on clinical needs. As of April 2011, Baobab systems in three central hospitals, 12 district hospitals, three HIV testing and counseling centers, and two rural health centers (off the grid) in Malawi have been used to collect patient information on 1.3 million individuals (see Figure 3). A study involving 31 users of the BART system found they were generally satisfied, with 90 percent of users rarely experiencing problems, despite some reported shortcomings with system stability and lack of desired features [11]. Although ART patients represent the lion's share of encounters to date, rapid growth in other systems is anticipated in the near future.

back to top  Lessons Learned

Baobab's commitment to minimal touchscreen interfaces has proven to be a key factor in the success of the initial ART system and subsequent expansion to new domains. User-centered designs supporting underlying clinical processes (as opposed to simply computerizing existing paper forms) provide high levels of usability for healthcare workers with little or no previous computer experience. Software and hardware components have been selected and integrated to meet both functional requirements, such as reliable patient identification, and non-functional needs, such as reliability in the face of intermittent power.

These experiences have led to the development of a set of guiding practices that characterize Baobab's efforts:

  • The use of wizard interfaces to guide users through data collection;
  • Simple displays with low information density;
  • Minimization of free-text data entry;
  • Error prevention rather than recovery; and
  • Targeted systems meeting critical clinical needs.

While deployment at a national scale in Malawi has demonstrated the success of these approaches, expansion into new domains and the possibility of transfer to other low-resource environments present new challenges that we are beginning to address.

Institutionalizing these usability approaches is a key concern. Baobab Health has grown from a small bootstrap effort focusing solely on ART to an established organization with more than 45 employees (including a team of 10 software developers) and a focus on a broader range of clinical problems. This growth has presented challenges in maintaining the vision of the simplicity of the interfaces. The transition from VB to Ruby on Rails caused some additional difficulties, as lessons learned and successful designs were not uniformly transferred to the new system. Although widget toolkits and planned efforts aimed at developing explicit user interface guideline documentation are underway, they are not sufficient. User interface generation toolkits and manual and automated usability inspections may prove helpful in this regard, but resources for the development of these tools are far from abundant.

The complexity associated with added functionality presents another, perhaps more existential, threat to the simplicity of the Baobab interfaces. The guiding philosophy of providing narrowly focused functionality in support of specific clinical domains has helped keep the interfaces from becoming cluttered, but at the cost of forgoing potentially useful clinical data that might be collected by a more complex system. Efforts to augment Baobab systems with clinical tools in support of promoting guideline adherence are underway [12], and successful deployments may lead to calls for additional support for medication management and clinical decision support. Changing workflows and adaptations to specific populations and care environments present further difficulties in understanding needs of more diverse users and communities [13]. Although designing interfaces that respond to these requirements without adding significant complexity may prove challenging, the growing body of work on information and communication technologies for development (ICT4D) will likely provide constructive guidance [14,15,1617].

Broader challenges include fostering sustainability and building upon Baobab's success in Malawi in other environments. As current grant-funded efforts may not be sufficient for both continued expansion into new localities and clinical domains and ongoing operational maintenance, demonstration of both cost and clinical effectiveness will be necessary for sustainability. This evaluation will also support generalization of the model. Through our work at the University of Pittsburgh's Center for Health Informatics for the Underserved, we are interested in bringing this model to other low-resource environments, including those found in underserved areas in developed countries. Can the Baobab model that worked well in hospitals in Malawi add value for inner-city health clinics in the U.S., or other similarly needy clinical domains? Although legal, financial, and regulatory differences may provide obstacles, successful implementation of the Baobab model in the more familiar context of the U.S. healthcare system might provide some insights that could inform the next generations of EMRs and contribute to the quest for meaningful use.

back to top  Acknowledgments

Gary Marsden, Tapan Parikh, and Andy Oram provided insightful comments on early drafts of this article.

back to top  References

1. Pantazi, S.V., Kushniruk, A., and Moehr, J.R. The usability axiom of medical information systems. Int. J. Med. Inform. 75, 12 (2006), 829–839.

2. Ash, J.S., Berg, M., and Coiera, E. Some unintended consequences of information technology in health care: The nature of patient care information system-related errors. J. Amer. Med. Informatics Assoc. 11 (2004), 104–112; http://jamia.bmj.com/content/11/2/104.abstract

3. Koppel, R. et al. Role of computerized physician order entry systems in facilitating medication errors. JAMA 293 (2005), 1197–1203; http://jama.ama-assn.org/content/293/10/1197.abstract

4. World Health Organization. WHO Gobal Health Observatory Data Repository. 2011; http://apps.who.int/ghodata/

5. Allain, T.J. et al. Applying lessons learnt from the 'DOTS' Tuberculosis Model to monitoring and evaluating persons with diabetes mellitus in Blantyre, Malawi. Tropical Medicine and International Health (in press, 2011).

6. Lewis, Z.L. et al. Touchscreen task efficiency and learnability in an electronic medical record at the point-of-care. Stud. Health Technol. Inform. 160 (2010), 101–105.

7. Fraser, H.S. et al. Implementing electronic medical record systems in developing countries. Informatics in Primary Care 13, 2 (2005), 83–95.

8. Blignaut, P.J. and McDonald, T. The user interface for a computerized patient record system for primary health care in a third world environment. J. End User Comput. 11, 2 (1999), 29–33.

9. Douglas, G.P. et al. Using touchscreen electronic medical record systems to support and monitor national scale-up of antiretroviral therapy in Malawi. PLoS Med 7, 8 (2010).

10. Wolfe, B.A. et al. The OpenMRS system: Collaborating toward an open source EMR for developing countries. AMIA Annu. Symp. Proc. (2006), 1146.

11. Msukwa, M.K.B. Evaluation of user perception of the effectiveness, efficiency, satisfaction, challenges, and training of electronic medical record system (EMR) in Malawi. In Department of Community Health. University of Malawi, Blantyre, 2010.

12. Landis-Lewis, Z. et al. The feasibility of automating audit and feedback for ART guideline adherence in Malawi. J. Am. Med. Inform. Assoc. (2011).

13. Marsden, G., Maunder, A., and Parker, M. People are people, but technology is not technology. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 366 (2008), 3795–3804; http://rsta.royalsocietypublishing.org/content/366/1881/3795.abstract

14. Parikh, T., Ghosh, K., and Chava, A. Design studies for a financial management system for micro-credit groups in rural India. Proc. of the 2003 Conf. on Universal Usability (Vancouver, B.C., Canada). ACM, New York, 15–22.

15. Smyth, T. et al. Where there's a will there's a way: Mobile media sharing in urban India. Proc. of the 28th Intern. Conf. on Human Factors in Computing Systems (Atlanta, Georgia). ACM, New York, 2010, 753–762.

16. Medhi, I. et al. Designing mobile interfaces for novice and low-literacy users. ACM Trans. Comput.-Hum. Interact. 18, 1 (2011), 1–28.

17. DeRenzi, B. et al. E-imci: Improving pediatric health care in low-income countries. Proc. of the 26th SIGCHI Conf. on Human Factors in Computing Systems (Florence, Italy). ACM, New York, 2008, 753–762.

back to top  Authors

Gerry Douglas is the founder and board chair of Baobab Health Trust (baobabhealth.org). He is an assistant professor in the Department of Biomedical Informatics at the University of Pittsburgh and the director of the University's Center for Health Informatics for the Underserved, a newly established center to address the challenges of delivering healthcare in low-resource settings through health informatics.

Zach Landis-Lewis is a doctoral student in the Department of Biomedical Informatics at the University of Pittsburgh. Formerly a software developer for Baobab Health, he is currently developing automated methods for clinical practice guideline implementation and clinical audit and feedback with Baobab Health systems in Malawi.

Harry Hochheiser is an assistant professor in the Department of Biomedical Informatics at the University of Pittsburgh. He is a co-author of Research Methods in Human-Computer Interaction (Wiley, 2010).

back to top  Figures

F1Figure 1. The input screen for selecting area of current residence, from the early Visual Basic system.

F2Figure 2. Screenshots from the current Ruby on Rails BART system.

F3Figure 3. The Baobab system in action.

back to top 

©2011 ACM  1072-5220/11/11  $10.00

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.

The Digital Library is published by the Association for Computing Machinery. Copyright © 2011 ACM, Inc.

Post Comment

No Comments Found