Synopsis of selected studies
Our initial search returned 15,042 unique citations. Screening of titles and abstracts left 1,001 articles that required review. While retrieving the full-text for these studies, we removed 66 non-English publications and 170 citations that were not available online (the review team decided not to pursue them due to time constraint). Of the 765 full-text articles screened, 43 studies were selected for this review (27 controlled, 16 descriptive) [12
]. See Figure .
A flow chart of the study selection process and results.
A synopsis of the 43 studies is shown in Additional file 1
. Twenty-seven of the 43 studies (62.8%) were published between 2005 and 2009. Studies from the United States (11 or 25.6%) and United Kingdom (10 or 23.3%) made up half the publications. The remaining studies were from The Netherlands (5 or 11.6%), Canada (4 or 9.3%), Australia (4 or 9.3%), Norway (2 or 4.6%), New Zealand (2 or 4.6%), and other countries. Fourteen of 43 studies (32.6%) were focused on work practice, 9 (20.9%) on prescribing support, 7 (16.3%) on disease management, 6 (13.9%) on clinical documentation, 4 (7.0%) on preventive care and 4 (9.3%) on patient-physician interaction. For study design, 11 (25.6%) were randomized controlled trial, 7 (16.3%) quasi-experimental, 9 (20.9%) observational and 16 (37.2%) qualitative studies. The observational studies included 6 cross-sectional cohort, 1 time-motion, 1 prospective audit and 1 secondary analysis studies. The qualitative studies included 6 multi-method designs, 6 case series, 3 videotape analyses and 1 interview study.
EMR impacts by study and measure
The 43 studies are summarized by topic area, impacts, influencing factors and our mapping to the CA Framework in Additional file 2
. The differences in impact by country, time period and design are reported in Additional file 3
. When compared between countries with high versus low adoption rates there was no significant difference found in the ratios of positive studies. There was no difference found between the two time periods of 2000-04 and 2005-09. Odds ratio tests showed more positive results (but not significant) from controlled-experimental than controlled-observational studies, and no difference between controlled and descriptive studies.
Table summarizes the impacts by study for the six topic areas. For controlled studies, the number of positive counts ranged from highest in work practice (4/5 studies or 80.0%) to lowest in clinical documentation (1/5 studies or 20.0%). For descriptive studies, the positive count was lower in work practice (5/9 studies or 55.6%), while the remaining areas had too few study counts for meaningful comparison. When combined, 22/43 studies (51.2%) had positive impacts, 13/43 studies (30.2%) had neutral impacts and 8/43 studies (18.6%) had negative impacts. The areas with > 50% positive counts were preventive care (2/3 studies or 66.7%), work practice (9/14 studies or 64.3%) and disease management (4/7 studies or 57.1%). The area with the most negative counts was clinical documentation (3/6 studies or 50.0%).
Number of positive, neutral and negative impacts by study for the six areas
The impacts are grouped according to the micro-level dimensions of the CA Framework in Table . Odds ratio testing showed only minor differences between the controlled and descriptive studies (i.e., neutral impact, OR = 2.5 CI 1.1-5.9). For controlled studies only productivity had > 50% positive count with 10/16 measures (62.5%) positive. For descriptive studies the three measures that had > 50% positive counts were care quality (4/4 measures; 100.0%), information quality (4/6 measures; 66.7%) and productivity (11/17 measures; 64.7%) that were positive. When combined, 50/109 measures (45.9%) showed positive impacts, 39/109 measures (35.8%) showed neutral impacts and 20/109 measures (18.3%) showed negative impacts. Overall, the only measure that had > 50% positive count was productivity where 21/33 measures (63.6%) were positive.
Number of positive, neutral and negative impacts by measure in the CA Framework
Factors influencing EMR adoption and effect
A total of 100 factors that influenced EMR adoption and its effect were identified from the 43 studies (see Additional file 2
). After merging those that were similar we ended up with 48 distinct factors. These factors were mapped to the categories of the CA Framework [refer to [17
]]: 23 of them were micro-level, 16 meso, and 9 macro (see Figure ). The types of influence are elaborated below.
Micro, meso and macro-level factors that influenced EMR success.
At the micro-level
, system quality factors included the availability of templates [25
], interface design [31
] and technical performance (e.g. speed and reliability) [44
]. Information quality factors included the organization, accuracy, completeness and accessibility of the patient record [23
]. Service quality factors included training and technical support [53
], system backup and unexpected downtime [52
]. EMR usage factors included its intent (e.g. quality improvement versus record keeping) [41
], actual strategies for optimal/appropriate use [12
], ease of use [35
], and usage patterns that emerged over time [40
]. Interaction related factors covered patient-physician encounters such as the type of consult (e.g. psychological) [12
], consult room layout [42
] and patients' ability to schedule appointments [27
]. For net benefits, care quality factors covered patient safety [38
], care effectiveness [40
], quality improvement [45
] and guideline compliance [32
]. Productivity factors covered care efficiency [13
], coordination [24
] and net cost including billing, staffing and maintenance costs [25
At the meso-level
, people factors included personal characteristics and expectations such as prior EMR experience of the users [34
], and their personal time investment in exchange for the benefits expected from the system [13
]. Roles/responsibilities covered the need for champions and staff participation [24
], and shift in tasks (e.g. documentation by staff vs. physicians) [13
] that could lead to role ambiguity and conflict [30
]. Organization factors included structure/processes and culture that supported EMR adoption/use [23
], EMR-practice fit (e.g. hybrid EMR/paper systems) [50
], and EMR-supported office and workflow design [30
] such as the placement of computer screens in consult rooms [42
]. Return-on-value focused on demonstrated value at the practice level such as substitution effect from guideline driven test orders and prescribing [51
], and tangible cost-efficiency gain with larger practice size and patient volume [48
]. Implementation factors included the extent that the introduction of an EMR into the practice was planned and carried out as a priority project with dedicated time and resources [52
]. The service support provided during implementation was critical [48
], since they affected the disruptions that physicians and office staff had to overcome while learning to use the EMR and redesign their work routines.
At the macro-level
, factors under healthcare standards included standardized data content [23
], established practice guidelines [32
], and legal documentation requirements [55
] that affected EMR design/performance and user behaviours. They also covered practice standards for clinical guidelines [60
], professional scope of practice [32
] and medico-legal requirements [55
] that governed EMR use. Factors under funding/incentives included remuneration schemes such as pay for performance and fee-for-service that encouraged EMR use [35
] and incentive programs in the form of subsidies to purchase/adopt EMR systems [56
Summary of key findings
Overall, this review found 22/43 EMR studies (51.2%) and 50/109 measures (45.9%) had shown positive impact across the six topic areas examined. When grouped by area, there were modest improvements in preventive care (66.7%), work practice (64.3%) and disease management (57.1%). Clinical documentation showed the least improvement with EMR use (16.7%). Within the dimensions of the CA Framework, EMRs had shown a modest improvement in productivity (63.6%), whereas user satisfaction had the least improvement (18.2%). About one-third of the studies and measures were not able to show an impact. Less than one-fifth of the studies and measures had a negative impact. No significant differences were found based on adoption rates by country, by time period and by study design.
Through this review we found that EMR impact was influenced by many factors. In particular, we were able to extend the CA Framework to EMRs in physician office settings by identifying specific micro, meso and macro level factors that influence EMR adoption and its effect. For instance, at the micro-level, the EMR's technical design, performance and support affected its usage and user satisfaction in the office. At the meso-level, the implementation process and resulting workflow impacted the office's ability to improve productivity and coordination. At the macro-level, incentives such as pay-for-performance were seen as an important driver for EMR adoption since they increased the return on investment made.