Abstract
Global Health Informatics is an emerging field, as demonstrated by several substantial and widely used electronic medical record (EMR) systems along with the emergence of mobile based or“mhealth” systems. We describe here many of the practical lessons we have learned from implementing systems in a wide range of challenging environments over the last decade. Some requirements, like data backups, skilled staff and local leadership are universally important. Others, such as limited power, poor network access and distributed populations, require different designs and strategies in resource poor environments.
Introduction
Over the last decade, a new field of Global Health Informatics or Global eHealth has emerged, supporting the care of patients in resource poor environments in Africa, Latin America, and Asia. Key factors in its recent growth include the availability of more robust, cheaper and lower power hardware, increased internet access, and the emergence of several high profile software projects. These include the DHIS [1], OpenMRS [2], OpenELIS, Baobab Health Trust [3], and Smartcare, all of which are now in use in large numbers of sites in a wide range of countries. Also important is the emergence of mobile health informatics or “mHealth”. Though a new field with limited operational examples, the realization that cell phones can be used in many of the poorest and most remote regions has altered the perception of what is possible and sustainable. Driven by HIV treatment programs initially, substantial resources have been invested in medical information systems in Africa with nationwide expansions planned in several countries. Evidence of successful implementations of these systems is mounting, but effective dissemination of this knowledge is lacking. We describe here some important lessons learned based on a decade of implementing health information systems in developing countries, and examples from other projects.
Approach
Partners In Health (PIH) is a non-governmental organization (NGO) that has worked for 23 years to deliver healthcare to some of the poorest and most vulnerable populations in the world. In 2000, PIH developed a web-based EMR system to support the treatment of multi-drug resistant tuberculosis (MDRTB) in Peru [4]. This system has been in operation for nine years and is now supported and hosted by the Peruvian government. It has been expanded to allow access to TB laboratory data in more than 220 clinics. In 2003, PIH ported the system to support HIV management in Haiti [5]. This system, the HIV-EMR, is still in use, supporting the care of over 18,500 HIV patients in 12 hospitals and clinics. The Mosoriot Medical Record System was created in 2000 by the Regenstrief Institute to allow the collection of data on primary care visits in clinics in Kenya. It allowed better tracking of patient care and more effective and quicker reporting [6].
In 2004, the Regenstrief Institute at the University of Indiana and PIH began a collaboration to create an open source EMR platform designed to support the care of a wide range of diseases [2][7]. This system, OpenMRS, is now developed and supported by a broad collaboration including the South African Medical Research Council, JEMBI (a South African NGO), the Millenium Villages Project, the Rockefeller Foundation, and the Rwandan government, among others. It is in use in at least 24 countries including Kenya, Rwanda, and the US, to support the care of HIV, TB, cardiac disease and primary care. Baobab Health has been developing medical information systems in Malawi for a decade and has created an innovative touch screen approach to clinical data entry and viewing. [3]. This system is in use for the care of over 42,000 HIV patients. A hybrid version linked to OpenMRS is in use by PIH-run clinics in Malawi for primary care. DHIS was developed in South Africa using Microsoft Access and is now widely used there and in several other countries. Users enter aggregate data from health facilities to create reports and overviews of medical care. A web-based version was recently created using open source software. These systems have been used for several years, have demonstrated significant scale in patient numbers and installed sites, and have published descriptions and in some cases evaluations. While many improvements are still needed, a number of lessons have been learned that can help inform other projects. This article is based on the experience of the authors with many of these systems and published reports, and focuses on practical lessons and solutions.
Lessons Learned
The basic task of collecting medical data can be structured in many different ways. The most important distinction for this discussion is whether the data describes individual patients or aggregated numbers. Traditionally, medical information management in developing countries used aggregate data and only recently have complete individual records been collected in health facilities. Individual records are often collected on paper by clinical staff and then entered into an information system by data entry staff either locally or remotely. A few systems, such as Baobab [3], allow direct point of care data entry, cutting out the paper record, at least for core data. Additional data capture strategies, which have had limited impact to date, include scanning forms that are processed with optical character recognition software, and “smart pens” that track the writing on special paper. We describe here important lessons for data collection, analysis and use with the focus on health facilities and patient data.
The importance of local leadership
One of the biggest challenges in implementing successful information systems in resource poor environments is direction and support from local staff. Systems cannot be expected to work unless local staff have a real stake in the process from initial planning to full operation. A local champion who can be taught in more depth on the system and can liaise between clinical staff and developers is a key success factor. Even if there is a competent data manager, there also needs to be a senior member of the clinical staff who can push forward the system. Conversely, this person can communicate issues and areas for improvement to the developers. Aligning the introduction of a system with other organizational improvements can improve its acceptance and use.
The importance of individual patient records
Aggregated data on facilities or patients treated is useful if the patient data is collected and totaled accurately. Unless individual records are kept, it is difficult for staff to keep track of patients seen each day, often as many as 100–200, resulting in poor quality reports [8]. The simplest method is a paper registry with one row per patient, which can be used for daily totals, but is hard to track for longer periods. An alternative strategy is to create a local database with a limited number of variables per patient (5–10) to create reports that are submitted to higher levels in the health system. Though requiring more infrastructure and training, it can lead to major improvements in data quality [8] and local capacity.
The importance of local data use
A related issue is what purposes the data is collected for. The work involved in collecting data and creating required reports can be significant, especially for busy medical and nursing staff. This is compounded by the fact that the reporting systems have traditionally been focused at higher levels in health systems with little attention to the needs of the local staff and the patient care process. These are “systems that just suck,” pulling data centrally, typically with little or no feedback or direct benefit for the reporting facility or its patients. These systems are naturally unpopular, reducing the staff’s motivation to collect good data and denying them the chance to check the resulting statistics. Even a simple local database can create valuable statistics on case mix, outcomes and resource requirements. More complete EMR systems can significantly improve patient care processes and hospital or clinic management [3,6,9]. To be successful, system design must involve local staff, and there must be tools to allow them to access and analyze the data.
Use systems that scale, don’t “reinvent the wheel”
Many clinical data collection projects start small with a simple database for a single disease like HIV. If successful, there is usually a need to scale up numbers and support the management of other diseases. Using existing software validated in developing countries with key functionality already included can improve early results and allow easier growth. The challenge is having the flexibility to adapt and extend the system. OpenMRS was created with these concerns in mind, allowing adaption at multiple levels from form creation, though addition of software modules, to modification of the core code [7].
Design systems with outputs as the primary focus
The purpose of eHealth systems is to create information that can aid tasks such as patient care, clinic management, reporting, supply chain management or disease surveillance. Forms, IT systems and the data entry staff are simply there to serve those goals. There is a common tendency for form design to drive the other steps, frequently resulting in forms that are too large and complex, or lack critical data items needed for reports, or both. As with most tasks, if the larger goals are not clearly defined, then the development process tends to be slow and inefficient, and the product of the process will likely be inadequate.
Defending the core data set
There is a need to create a core data set that allows accurate comparison between clinics, organizations or countries. If such a data set with clear definitions of each variable is agreed upon from the start, it is much more likely that accurate and complete data will be collected. Each site or organization can be given flexibility to add additional variables, but the core data set must be fixed, with only infrequent changes which should be made through a formal update process.
Data management tools and training
Without high quality, timely data, most of the outputs of an information system can be severely compromised. Local staff in developing countries usually have limited exposure to IT systems and data management, making effective training especially important. Data quality can be significantly improved with well-designed systems that provide form field validation and reports of data quality and completeness.
The information system does not have to be onsite
Although EMRs are normally thought of as point-of-care systems on the desk top, sites with no power can be managed with an offsite system. Data is collected on paper forms which are transported to a central site for data entry. Patient summaries and reports can then be printed and sent back with the forms. The AMPATH project in Western Kenya[10] has successfully used this model for 29 clinics and nearly 100,000 HIV patients. The major concern with this method is that paper patient records are offsite for a period of time potentially hampering patient care.
Use the web if possible
Managing servers in remote sites, while successfully achieved in many projects, puts extra burdens on the support team for data backups and software support. For sites with reasonable internet access, maintaining one server in a well-supported site can work effectively [4,5]. In this case local expertise aside from the server site can be limited to basic PC management and internet support. Applications need to minimize page download size, which, in our experience is rarely prioritized in countries with good internet access. Internet bandwidth must be carefully managed in remote sites with controls on the types of data that can be downloaded by staff during the working day.
Benefits of offline data entry and viewing
For sites with unreliable internet, it is important to have alternative ways of working than remote access to the server. If the outages are not frequent or the workload is small, then paper records can be kept until a suitable moment for data entry. A better solution is offline data entry using a client that stores the data locally until it can connect to the server. For entry of follow-up data, it is necessary to download patient lists, which may include brief patient summaries. Such an approach has been deployed in Haiti [5] and Kenya. Perhaps the ideal strategy is use of local servers that automatically synchronize the data with a central site. This allows full use of the EMR on the local network but adds the benefit of access to offsite laboratory date, records of patients seen offsite, and offsite backup. The central site can view and report on all data. The main downside is the complexity of building robust synchronization tools, and possibly large data transfers, but this approach has been successfully employed in Rwanda [11] using OpenMRS, and in Haiti with ISante [12]. Mobile phones and PDAs offer an alternative way of collecting and managing some data offline and can include data synchronization tools.
One successful system beats 10 almost ready
While meeting local requests is important, it is essential to prioritize them by what is feasible and likely to produce good results. Having many applications that are not complete or have partial data leads to frustration and loss of confidence in the system. Conversely, “low hanging fruit“ such as providing clinician access to laboratory data has been popular for decades and recent evaluation studies have strongly confirmed its utility in developing countries. A randomized controlled trial in Peru showed that providing clinicians with access to TB laboratory data reduced delays and errors [13]. A study in Rwanda showed that an EMR system improved clinician knowledge of HIV patients’ most recent CD4 counts [14] and a study in Haiti showed that sending clinicians warning emails about patients with low CD4 counts was associated with a faster start of HIV treatment [15].
Power supply, backups and protection
If a local database or EMR system is used, then stable electrical power is crucial. Standard Uninterrupted Power Supplies provide limited backup time, especially for desktop PC and servers. Low power devices like net-books, fan and driveless PCs, and even servers are now available. Increasingly laptops are being used as servers due to the built in battery. Larger backup systems using several deep cycle batteries have been effectively used in Malawi along with low power hardware to maintain servers functional for over 12 hours at a modest cost [3]. Unstable power and surges due to poor quality power grids, intermittent generator use, or lightning strikes are another major challenge. Good grounding and lightening protection are significant investments but prevent the destruction of more expensive equipment, and protect staff.
Data backups
The increased storage of valuable patient data in remote sites raises the potential for data loss. Using a web-based approach or data synchronization is a major advantage here. In addition, improved availability and cost of portable hard drives and flash memory drives leaves no excuse for not backing up data, but effective protocols and training are essential. A strategy for moving copies of data offsite is also vital if internet access is not available.
Confidentiality and data ownership
Robust software exists for password protecting servers and encrypting data transfers over the web, but this must be implemented carefully along with physical security of servers. The major challenge is training of users in patient confidentiality and password management. Of concern in many countries is that policy and legal frameworks for data ownership are inadequate [16] increasing the risk of abuse by individuals, private organizations or governments.
Auditing user activity and edits
Audit logs of user activity, page viewing and editing complement strategies for data backups and confidentiality. While infrequent with good training, PIH had two major incidents where patient data was destroyed by inadvertent merging of patient records with similar names. As this occurred over many days we could not just reload a backup file. However due to the existence of proper audit logs, complete reconstruction of the data was possible.
Mobile devices for network and power independence
The mHealth approach is rapidly gaining ground in many developing countries, allowing real time data access and management in locations with no infrastructure other than a cell phone tower [17]. The cost of these devices can be much lower than PCs with internet connectivity. An alternative is the use of PDAs offline that can be periodically synchronized with a PC or remote server, which has proved successful in Peru [9]. Currently, low end phones only support SMS as a means of data entry, however, Java-enabled phones that can support forms are falling rapidly in price.
Evaluate your successes and failures
With the ever-growing investment in global eHealth projects it is essential that future projects are guided by evidence of what works and what does not. A recent systematic review of evaluation studies [18] of eHealth systems in developing countries showed early evidence of beneficial impacts, including PDA based tools, laboratory reporting systems, and systems that warn if HIV or TB patients are missing treatment. Evaluation studies have many potential benefits, including justifying a system’s use to funders, governments and local staff. Formative evaluations help address problems in new projects, particularly in the complex and heterogeneous environments of developing countries. In certain cases, evaluation studies are good responses to apparent failures, guiding whether to fix the problems or abandon the approach. Evaluation studies should generally include measures of user satisfaction, and data quality, including comparing electronic data to original documents in clinical sites.
Support and train local developers
Most developing countries are dependent on outside software purchases and development support. We are training local developers of OpenMRS in Rwanda and other sites. This is greatly helped by the use of open source software that local developers can learn on, and indigenous organizations can take ownership of.
Conclusions
eHealth systems are growing rapidly in the developing world and are beginning to show beneficial impacts in clinical and programmatic management. Many practical problems have been solved at this stage and the costs of hardware and software are falling. Sharing of both good and bad experiences is essential if we are to work efficiently and avoid repeating mistakes. GHDonline (www.ghdonline.org) is a forum dedicated to sharing ideas and best practices on eHealth in developing countries. Although this paper is not primarily focused on software design, there are additional issues that will likely impact the success of projects, particularly when scaled up to national level. Open standards for data exchange allow interoperability of different eHealth systems. Formal requirements analysis can help to improve initial designs, complemented by rapid prototyping in the field. Finally, open source software allows sharing of the best software tools which complements the sharing of strategies for implementation and use.
References
- 1).Williamson L, Stoops N, Heywood A. Developing a District Health Information System in South Africa: a social process or technical solution? Stud Health Technol Inform. 2001;84(Pt 1):773–7. [PubMed] [Google Scholar]
- 2).Seebregts CJ, Mamlin BW, Biondich PG, Fraser HSF, Wolfe BA, Jazayeri DG, et al. The OpenMRS Implementers Network. Int J Med Inform. 2009 Nov;78(11):711–20. doi: 10.1016/j.ijmedinf.2008.09.005. [DOI] [PubMed] [Google Scholar]
- 3).Waters E, Rafter J, Douglas GP, Bwanali M, Jazayeri DJ, Fraser HSF. Experience Implementing a Point-of-Care Electronic Medical Record System for Primary Care in Malawi. Proc. Medinfo2010 (in press) [PubMed]
- 4).Fraser HS, Jazayeri D, Mitnick CD, Mukherjee JS, Bayona J. Informatics tools to monitor progress and outcomes of patients with drug resistant tuberculosis in Peru. Proc AMIA Symp. 2002:270–4. [PMC free article] [PubMed] [Google Scholar]
- 5).Fraser HSF, Jazayeri D, Nevil P, Karacaoglu Y, Farmer PE, Lyon E, Smith-Fawzi MK, Leandre F, Choi S, Mukherjee JS. An information system and medical record to support HIV treatment in rural Haiti. BMJ. 2004;329:1142–6. doi: 10.1136/bmj.329.7475.1142. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6).Rotich JK, Hannan TJ, Smith FE, Bii J, Odero WW, Vu N, et al. Installing and implementing a computer-based patient record system in sub-Saharan Africa: the Mosoriot Medical Record System. J Am Med Inform Assoc. 2003 Jul–Aug;10(4):295–303. doi: 10.1197/jamia.M1301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7).Mamlin BW, Biondich PG, Wolfe BA, Fraser HSF, Jazayeri D, Allen C, Miranda J, Tierney WM. Cooking Up An Open Source EMR For Developing Countries: OpenMRS – A Recipe For Successful Collaboration. Proc AMIA Symp. 2006:529–533. [PMC free article] [PubMed] [Google Scholar]
- 8).Mate KS, Bennett B, Mphatswe W, Barker P, Rollins N. Challenges for routine health system data management in a large public programme to prevent mother-to-child HIV transmission in South Africa. PLoS One. 2009;4(5):e5483. doi: 10.1371/journal.pone.0005483. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9).Blaya JA, Cohen T, Rodríguez P, Kim J, Fraser HSF. PDAs to collect tuberculosis bacteriology data in Peru reduce delays, errors, workload, and are acceptable to users: Cluster randomized controlled trial. Int J Infect Dis. 2009 May;13(3):410–8. doi: 10.1016/j.ijid.2008.09.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10).Tierney WM, Rotich JK, Hannan TJ, Siika AM, Biondich PG, Mamlin BW, et al. The AMPATH medical record system: creating, implementing, and sustaining an electronic medical record system to support HIV/AIDS care in western Kenya. Stud Health Technol Inform. 2007;129(Pt 1):372–6. [PubMed] [Google Scholar]
- 11).Allen C, Jazayeri D, Miranda J, Biondich P, Mamlin B, Wolfe B, Seebregts C, Tierney W, Fraser HSF. Experience in implementing the OpenMRS medical record system to support HIV treatment in Rwanda. Stud Health Technol Inform. 2007;129(Pt 1):382–6. [PubMed] [Google Scholar]
- 12).Lober WB, Quiles C, Wagner S, et al. Three years experience with the implementation of a networked electronic medical record in Haiti. AMIA Annu Symp Proc. 2008 Nov 6;:434–8. [PMC free article] [PubMed] [Google Scholar]
- 13).Blaya JA, Shin SS, Yagui M, Yale G, Suarez C, Asencios L, Contreras C, Rodriguez P, Kim J, Cegielski P, Fraser HSF. Electronic lab system reduces errors in National Tuberculosis Program: Cluster randomized controlled trial. IJTLD. 2010;14(8):1009–1015. [PMC free article] [PubMed] [Google Scholar]
- 14).Amoroso C, Akimana B, Wise B, Fraser HSF. Using Electronic Medical Records for HIV Care in Rural Rwanda. Proc. Medinfo 2010 (in press) [PubMed]
- 15).Fraser HS, Allen C, Bailey C, Douglas G, Shin S, Blaya J. Information systems for patient follow-up and chronic management of HIV and TB: a life saving technology in resource poor areas. J Med Internet Res. 2007;9(4):e29. doi: 10.2196/jmir.9.4.e29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16).Willyard C. Electronic records pose dilemma in developing countries. Nature Medicine. 2010;16(3):249. doi: 10.1038/nm0310-249a. [DOI] [PubMed] [Google Scholar]
- 17).Braa Kristin, Purkayastha Saptarshi. Sustainable Mobile Information Infrastructures in Low Resource Settings. Stud Health Technol Inform. 2010;157:127–32. [PubMed] [Google Scholar]
- 18).Blaya J, Fraser HS, Holt B. Evaluations of the Impact of eHealth Technologies in Developing Countries: A Systematic Review. Health Affairs. 2010;29(2):244–251. doi: 10.1377/hlthaff.2009.0894. [DOI] [PubMed] [Google Scholar]