People, Process and Technology: Strategies for Assuring Sustainable Implementation of EMRs at Public-Sector Health Facilities in Kenya

Abstract

The Ministry of Health (MoH) rollout of electronic medical record systems (EMRs) has continuously been embraced across health facilities in Kenya since 2012. This has been driven by a government led process supported by PEPFAR that recommended standardized systems for facilities. Various strategies were deployed to assure meaningful and sustainable EMRs implementation: sensitization of leadership; user training, formation of health facility-level multi-disciplinary teams; formation of county-level Technical Working Groups; data migration; routine data quality assessments; point of care adoption; successive release of software upgrades; and power provision. Successes recorded include goodwill and leadership from the county management (22 counties), growth in the number of EMR trained users (2561 health care workers), collaboration in among other things, data migration(90 health facilities completed) and establishment of county TWGs (13 TWGs). Sustenance of EMRs demand across facilities is possible through; county TWGs oversight, timely resolution of users’ issues and provision of reliable power.

Introduction

Electronic medical records (EMRs) have been embraced across public, private, and faith-based health facilities in Kenya. Perceived benefits driving uptake of EMRs in general include: optimizing documentation of patient encounters; improving communication of information to physicians; improving access to patient medical information; reducing errors; supporting administrative management; forming a data repository for research and healthcare quality improvement; and reducing use of paper^{1, 2, 3, 4}.

The Ministry of Health (MoH) in Kenya rolled out EMRs in facilities beginning November, 2012. KenyaEMR’s implementation was spearheaded by I-TECH with funding from President’s Emergency Plan for AIDS Relief (PEPFAR). The system runs on an open source platform named Open Medical Record System (OpenMRS)⁵. In total, 342 instances of KenyaEMR have been implemented across four regions in Kenya: Central, Nyanza, North Rift Valley, and Western. KenyaEMR initially targeted services within the national HIV/AIDS care and treatment program, as well as related Tuberculosis (TB) and maternal child health (MCH) services. There is presently interest in expanding the system to cover other outpatient primary care services as well as inpatient services. Guidelines from the National AIDS and STIs Control Proramme (NASCOP) defined the scope and workflows within KenyaEMR, and all the implemented instances of the system have the same functionality. Functionality includes patient registration, triage, clinical assessment, documentation of diagnoses, prescriptions and laboratory orders, care alerts and reminders, automated monthly reporting, cohort reporting, and continuous quality indicator reporting. No interoperable provider order entry or laboratory results transfer is presently available and functionality related to billing is also absent, as services are offered at no cost to patients. The vast majority of data is captured in structured formats based upon standardized concepts⁶. Progressive software releases have taken into consideration feedback from end users and stakeholders—including health facility staff, members of County Health Management Teams (CHMTs), and representatives of PEPFAR-supported service delivery implementing partners (SDIPs) tasked with technical assistance to public sector HIV program sites — about desired functionality.

While the anticipated benefits of adopting EMRs may be great, the processes of adopting EMRs can be highly disruptive to conventional workflows⁷. Successfully sustaining EMRs demands a high level of organizational change centered on people, processes and technologies in healthcare, as described in various socio-technical frameworks for information systems success.⁸ This paper focuses on the specific activities undertaken by I-TECH in Kenya in support of the Ministry of Health (MoH) in Kenya focusing on the people, processes and technologies employed in EMRs implementation.

Methods

I-TECH pursued a range of strategies for assuring sustainability of EMRs implementation. We describe each strategy and its results according to the socio-technical “people, process, and technology” framework for organizational capacity to meaningfully use and sustain health information systems. The observations we report are based upon routinely-gathered monitoring and evaluation (M&E) data, as analyzed by the authors (who have been part of project implementation and M&E teams since project inception in 2012). The data has been collected through tools such as form hub, Google sheet on post implementation monitoring tool and quarterly reports shared by I-TECH regional staff, “train smart” application that supports monitoring of EMRs training among other tools. Once collected, I-TECH M&E staff conducts analysis of the data on a quarterly basis for purposes of reporting on project results, successes and challenges to the funder. This study represents a synthesis of these project reports from 2012-June 2016

People:

• Sensitization of leadership

  EMR implementation began with sensitization of local leadership on the site readiness criteria and the steps for preparing public health facilities for EMR implementation, the initial scope of the EMR functionality, the expected role of local leaders in identifying “model sites” for initial implementation (where best practices could be established) as well as subsequent “scale up sites”, and the process of engaging and communicating with target health facilities.. Inadequate communication can lead to divergent understanding among health workers on the role of EMRs, and problems at the organizational level with change management^{9, 10}. Initially, sensitization of local leadership involved outreach to provincial teams in 8 provinces; however, once Kenya adopted a new constitution in 2010, decentralizing governmental functions to 47 newly-formed County governments, sensitization was expanded to leadership in 22 of the 47 counties covering the regions of KenyaEMR implementation. Sensitization was conducted by I-TECH in collaboration with representatives from the national MoH.

• User training

  User training is the process of building skills among target system users to use and maintain the systems. EMRs users were categorized into 3 cadres; end users; health managers; and system administrators. I-TECH supported the MoH in Kenya to develop distinct training curricula targeting each of these groups. The training was conducted in a cascaded mode First, training of trainers (TOTs) were conducted, reaching a small pool of lead trainers in each county. Next, TOT participants helped train KenyaEMR end users as well as “KenyaEMR Champion Mentors” (referred to henceforth as ‘champions’). . Champions were front-line health care workers based at the target health facilities, who were nominated based on high interest and knowledge in the EMRs, willingness to support other users in the EMRs use, and strong post-test scores during the end user trainings. Champions represented a variety of health workers cadres (clinician, data clerk, health records officer, nurse). Champions were prepared for their role through supplemental mentorship trainings. End user trainings also targeted diverse cadres involved in HIV care and treatment programs during training of 3-5 days. The management at the facility was targeted through the health managers’ curriculum which equipped them to provide oversight for the use of the EMRs at the facility level. The system administrators training was designed to prepare IT specialists based within County Health Management Teams (CHMTs) and large health facilities to trouble shoot EMRs hardware and local network problems, data backup and restoration procedures, management of user accounts, and other technical and “help desk” issues.

• Multi-disciplinary teams

  Health facilities have multi-disciplinary teams (MDTs), supported by the PEPFAR SDIPs to address cross-cutting issues in HIV/AIDS service delivery. The MDTs comprise the doctors, clinicians, nurses, peer educators, mentor mothers, nutritionists and counselors working within the Comprehensive Care Center (CCC), the department specializing in HIV/AIDS care Each MDTs is chaired by the doctor or other clinician who directs the CCC and other services (the “facility in-charge”). EMR champions and facility in-charges facilitated the inclusion of the EMR-related topics alongside other agendas discussed by the MDTs. Examples of EMR-related issues addressed by MDTs include: roles and responsibilities in oversight of EMR activities, planning for legacy data migration, routine data quality assessment, and planning to reduce various barriers to EMR use..

• County Technical Working Groups

  As part of the transition of EMRs to sustained local leadership and ownership, I-TECH collaborated with the SDIPs and CHMTs to form eHealth Technical Working Groups (TWGs) to oversee the EMRs agenda within each county. The TWGs deliberately comprised officials with authority and capacity to undertake timely decision making. Each county TWG is typically chaired by the County Director of Health and in his or her absence, by the County Health Records Information Officer (CHRIO). Members include other representatives from the County Health Management Team (CHMT), from EMR technical assistance partners, and from PEPFAR SDIPs.

  Given the access of the TWGs to the county leadership, the groups have facilitated adoption of a number of proposals. Examples include recommendations that senior county leadership take into account EMR skills and experience when managing health worker employee transfers between health facilities (i.e. transferring EMR-experienced health workers to other sites with EMRs), that they plan for EMR-related costs when determining budget allocations during the annual work planning meetings. Allocated budgets should support EMRs maintenance, EMRs review meetings, and routine data quality assessments, among other EMR-related activities.

• OpenMRS community building

  In an effort to increase the local support available for further KenyaEMR software enhancement and support, I-TECH Kenya convened a meeting of Open:MRS implementers in Kenya and organized OpenMRS and KenyaEMR boot camps. The implementers meeting targeted developers and implementers from organizations customizing E:MRs on the OpenMRS platform, in order to share best practices and identify areas of collaboration. The boot camps targeted university students pursuing information teclmology and computer science courses, in their final or penultimate year of study. This was to sensitize and orient them to the basics of the Open:MRS platform and to stimulate interested in health informatics careers.

Process:

• Data migration

  Public sector HIV/AIDS care and treatment programs in Kenya date back to 2005. Facilities offering such care have patient records since the inception of the service by the MoH. The current policy on paper records requires their preservation up to a period of 10 years. As facilities adopted EMRs, there was a need to have the historical patient data captured on the EMRs, to support continuity of care. Effective HIV/AIDS care can only occur when clinicians have access to key historical data for each patient and can track their response to treatment over time.

  To understand the effort needed in legacy data migration, we sampled the largest of the KenyaEMR databases from a large referral hospital in the Nyanza region of Kenya, with 26,971 patients’ records as of January, 2016. The total number of patient visits from the instance was 612,521 and the total number of observations (i.e. individual data elements) was 6,905,588. Therefore, the average number of visits per patient was 22 (612,521/26,971), and the average number of data elements captured per visit was 11 (6,905,588/612,521). With exception of the patients’ demographic information which is mostly captured once, legacy data migration would need to capture 242 data elements per patient. The number of patients per facility ranges between 500 and 26,000 patients. Together, this implies a heavy load of legacy data migration across KenyaEMR sites.

  To support the transfer of paper records to EMRs, data clerks were recruited and orientated to the data capture process to ensure that paper-based records were migrated into the electronic record. Data migration efforts were initiated immediately after the end user training. I-TECH collaborated with CHMTs and SDiPs to conduct intensive legacy data migration initiatives over a period of 3 months, whereby supplemental data clerks were recruited and assigned to specific health facilities.

• Routine Data Quality Assessment (RDQA)

  RDQA refers to the continuous assessment of the quality of data contained in the EMRs. It involves comparing patient data captured on the EMRs against the data contained in the source paper records for consistency of information. KenyaEMR provides an automated RDQA report that draws a random sample of patient records from the EMR, to enable comparison with other data sources, with a sample size intended to estimate the concordance level of data within +/5%. Using the RDQA report, facility personnel or external assessors can compile comparison data values within an Excel-based workbook which is pre-programmed with charts and graphs to show RDQA results. Upon completion of each RDQA, sites identify action plans to address data quality gaps. Sites with weak concordance of data <80%) are encouraged to follow standard operating procedures for data review and cleaning, and to undertake repeated RDQAs every 3-4 months until data quality reaches expected levels.

• Point of Care (PoC) adoption

  EMRs implementation in Kenya occurred in either PoC or Retrospective Data Entry (RDE) mode. PoC refers to a setup where a computer terminal is implemented at every patient service point within CCC, TB and MCH services. RDE refers to a setup where clinicians use paper records to serve patients and later have the paper records captured into the EMR at a single centralized data entry point by themselves or designated health facility workers.

  PoC adoption permits use of clinical decision support functions within KenyaEMR. Available alerts include: patient due for viral load testing; patient eligible for ART; patient lost to follow-up. These features promote timely action towards healthcare quality improvement.

  Upon initial assessment for EMR implementation readiness, facilities that had physical security gaps (e.g. no grills on windows, no bugler-proof doors) were recommended for RDE implementation. County and facility management were informed of the gaps. At sites where the gaps were remediated, subsequent transition from RDE to PoC implementation was possible.

Technology:

• Successive releases

  The initial KenyaEMR software release was based on an HIV/AIDS module for use in CCCs. This was augmented over time in response to the MoH users’ needs and with support from I-TECH to include: maternal child health (MCH) care and Tuberculosis (TB) care modules. Additional features include; Data Quality features; Reports Generation among others. There is presently interest in expanding the system to cover other outpatient primary care services as well as inpatient services

• Power sources

  KenyaEMR implementation involved readiness assessments at each health facility to identify presence or absence of resources needed for EMR implementation. The assessments addressed physical security, electricity, presence of existing EMRs, and other issues. Where facilities were estimated to have electricity up to 75% of the time and met other required factors, the site was approved for implementation. However, power availability was often over-estimated. In reality, power supply is highly dynamic, with frequent power blackouts arising from normal rationing, other faults, or power cuts due to non-payment of bills.

  I-TECH collaborated with the CHMTs and SDIPs to analyze and compare costs of alternative power sources. Kenya has sunlight for more than 90% of the day with exception of the wet season. In the wet season, rationing of power is less common due to sufficient supply of water to the hydroelectric dams that provide electricity⁷. Solar power was found to be cheaper than setting up a generator (US$7,940 vs US$19,250 for a solution supporting a load of up to up to 1.5 Kilowatts, or US$18,940 vs. US$74,690 for a solution supporting a load of up to 6 Kilowatts, each running for 8 hours per day over a period of 10 years (Table 2). Having alternative backup sources of power at health facilities has been identified as a critical factor preventing the piling up of backlogged patient records.

Table 2.

Comparative costs for setting up and running backup source of power (8 hours per day x 10 years)

	Solar (cost in US$)		Generator (Cost in US$)
KW	1.5	6	1.5	6
Fuel Cost per KWh	-	-	0.35	0.35
Fuel Cost per year	-	-	1,533	6,132
Fuel Cost in 10 years	-		15,330	61,320
Number of services (@200hrs)	-	-	146	146
Service Cost	-	-	2,920	10,950
Battery Replacement	1,560	2,340	-	-
Total Running Costs	1,560	2,340	18,250	72,270
Installation Cost	6,380	16,600	1,000	2,420
Total Cost of Ownership	7,940	18,940	19,250	74,690

Results

People

I-TECH supported sensitization of 678 participants from the county leadership. This bore fruits in their facilitation of EMR related activities, including: recommendation of health workers to participate in planed EMR training; supervision of EMR use at the facility level; convening review meeting across the regions where facilities shared their experiences and best practices.

I-TECH further supported the training of over 2561 health care workers (42 doctors, 189 lab technologists, 868 clinical officers, 777 nurses, 370 health records and information officers and 315 other cadres) in 297 trainings across 22 counties in the four regions of Kenya (Table 3).

Table 3.

Distribution of trained health workers for EMRs use across North Rift valley, Nyanza, Western and Central regions of Kenya

Health Care Workers Cadres	Numbers trained
Doctors	42 (2%)
Laboratory Technicians	189 (7%)
Clinical Officers	868 (34%)
Nurses	777 (30%)
HRIO	370(14%)
Other	315 (12%)

Champions conducted one-on-one sessions with users to train them on specific tasks or to bring newly appointed staff members up to speed on the EMR. Altogether, Champions reported mentoring 1 754 facility staff mentored across 180 health facilities. Staff also; interacted with the KenyaEMR “help menu” containing job aids, video tutorials, and two self-paced eLearning modules. In addition, Champions and KenyaEMR end users initiated region and county specific user support groups, via the “Whatsapp” mobile phone application¹¹. The “Whatsapp” groups have been formed across all 22 counties, and serve as a forum for peer-to-peer trouble shooting and bug reporting. On average, about 10 issues are resolved per county per month across the 22 counties implementing KenyaEMR. Facility staffs are further empowered through services such as toll free lines through which they can receive technical support and an online help desk management and bug reporting tool called Redmine¹².

Process

Out of the 342 implemented KenyaEMR instances, 90 health facilities had achieved 100% data migration while 252 were yet to fully migrate the legacy data into the EMR, by January 2016. The persisting need for legacy data migration has been occasioned by factors such as: downtime due to fluctuation of electricity; lost capacity following transfer of EMR trained health workers; increased workload due to double entry of patient records in both paper-based and EMR data systems; lean staff across health facilities; and slow data migration of patient records contained in non-standard MoH tools.

Specific data cleaning efforts have since been initiated to address identified gaps .. In an analysis of baseline vs. follow up ROQAs conducted across 27 health facilities, there were 2,411 patient charts reviewed during baseline ROQAs and 2,381 patient charts reviewed during follow up ROQAs. At baseline, the concordance of data for mandatory data elements was lowest for dates of first and last C04 test (13% and 11%, respectively) and highest for program of enrollment and patient sex (79% and 84% respectively). Follow up ROQA scores improved on average by 1.8 points on a 20-point scale across the 27 health facilities.

Technology

Overall, 208 (61%) initial KenyaEMR implementations were of the PoC model while 134 (39%) were of the RDE model. Additional 27 facilities have been assessed for upgrade to PoC implementation. This has been facilitated by continuous improvement of physical security paving way for deployment of additional computing terminals.

Although power supply is a principal challenge to EMR use, only 50 out of 342 health facilities have an alternative power source that is maintained by the health facility (8 in the Central region, 16 in the North Rift Valley region, 9 in the Nyanza region, and 17 facilities in the Western region). This has greatly facilitated the continuity of EMR use in the absence of power from the national grid.

Overall results

An impact evaluation of the KenyaEMR implementation in improving quality of HIV/AIDS care is planned. Presently, qualitative feedback from EMRs users shared during county-level TWG and EMR review meetings indicate general satisfaction with KenyaEMR. Champions and end users have noted, for example: improved ease of retrieving patient records, reduced data entry errors owing to the presence of validation rules at data entry, reduced duplication of patient records due to easy-to-use patient lookup feature, increased ability to easily detect and resolve duplicate records based upon a feature that identifies potential duplicate records using patient demographic data.

Discussion

Lessons learned for EMR success

Our experience with EMR implementation at 342 health facilities in Kenya underscored many lessons learned for sustainable success of the systems. The process of EMR implementation brings with it numerous expectations, some which are realistic and others which are not, such as: ability to immediately eradicate use of paper records¹³, automatic reduction of errors¹⁴, patient data security and confidentiality, and even loss of jobs for health workers. There is need to manage the expectations and demystify myths well in advance to reduce reluctance to use EMRs, calm users’ frustrations, and set reasonable expectations about the effort required for successful EMR implementation. Further, specific efforts should be made to actively monitor the successes and challenges encountered during the implementation stage. Health workers expectation has been cited as one of the factors that affect users’ satisfaction in EMRs^{15, 16}. Others include practicality of use, impact on delivery of care, organizational support and variation from familiar practice¹⁷.

Without buy-in from the senior leadership it becomes a struggle getting health workers to use EMRs and an EMR project can easily be labeled as a failure^{18, 19. 20, 21} Health facilities have different dynamics in uptake of EMRs and it is important to actively engage health facility leadership before proposing implementation plans²². The process of EMRs implementation is cost intensive in the hardware acquisition, data migration, users training, provision of alternative power sources, among other key activities²³. A costing evaluation covering I-TECH’s direct costs of EMR implementation showed an average site-specific cost of US$9,879 per facility to implement an EMR²⁴. It is paramount that county teams and facilities identify sponsors beforehand to ensure that the funds available can support the different EMRs projects to completion. Recurrent costs including replacement of hardware should further be considered and planned for by the project owners. Besides sourcing funds, the benefits of such interventions should be clear to win the support of the sponsors.

Creating avenues for communication between the system users and the support and development teams ensures that the users never feel stranded and without help. In addition it creates avenues for the EMRs to improve with the growing end-user needs. This further builds confidence for the EMRs leading to indirect influence on the acceptance and readiness to use EMRs. Past experiences show that EMRs have failed where user acceptance needs have not been addressed²⁵. Our experience showed a strong enthusiasm for forums where peers can engage and discuss issues affecting their EMR use. These have reduced the time taken to resolve issues, and have led to knowledge sharing and better capacities across health facilities.

Future Activities

Legacy data migration has proven to be a formidable challenge, but steady progress has been made. As of June 2016, I-TECH, NASCOP, CHMTs and SDIPs were engaged in an intensive push to complete legacy data migration at all sites with remaining backlogs of historical records. Completion of data migration will allow health facilities to engage in data quality improvement activities, and to facilitate transition to Point of Care (PoC) system use which takes advantage of the clinical decision support function within KenyaEMR.

Several SDIPs which have experienced benefits from the EMRs implementation are in advanced stages of planning to support alternative power sources. For example, one partner plans to install solar generated power and voltage stabilizers across 13 health facilities with a high number of HIV/AIDS patients, to maximize system uptime and minimize risks to hardware and data due to power surges. Efforts are underway to move towards paperless operations at sites which have demonstrated strong EMR data quality and with stable power availability. For example, one partner plans to pilot a reduced use of paper tools for parallel data entry in 21 health facilities, an important step in fulfilling the promise that EMRs can enhance rather than hinder the productivity of health care workers.

Users’ needs have further grown leading to extension of the KenyaEMR roadmap to include additional functionality including; support for outpatient and inpatient departments, data exchange between different systems; EMRs to laboratory information systems, EMRs to pharmacy information systems, and incorporation of national unique patient index (NUPI) that would facilitate patient tracking and identification across health facilities.

Conclusion

EMR uptake has grown over time in Kenya and the expectations from users have grown in equal measures. Timelines in EMRs upgrade to include additional features being requested by users will attract confidence and drive data use to greater heights. Availability of reliable power from the main grid and alternative sources will ensure sustained uptime of the EMRs and reduce the temptation to fall back to papers. Data exchange between EMRs will further eliminate duplication of effort in patients’ registration, further strengthening the users’ experience. Overall, the change process in EMR implementation ought to respond to the people, process and technology issues to increase the chances of successful EMR implementation.