- Email: [email protected]
Physician productivity and the ambulatory EHR in a large academic multi-specialty physician group
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
journal homepage: www.intl.elsevierhealth.com/journals/ijmi
Physician productivity and the ambulatory EHR in a large academic multi-specialty physician group Adam D. Cheriff ∗ , Akshay G. Kapur, Maggie Qiu, Curtis L. Cole Weill Cornell Medical College, New York, NY, United States
a r t i c l e
i n f o
a b s t r a c t
Article history:
Purpose: The impact of the ambulatory electronic health record (EHR) on physician productiv-
Received 14 September 2009
ity is poorly understood. Fear of productivity loss remains a major concern for practitioners
Received in revised form
and health care delivery organizations and inhibits system adoption. This study describes
23 April 2010
the changes in physician productivity after the implementation of a commercially available
Accepted 23 April 2010
ambulatory EHR system in a large academic multi-specialty physician group. Methods: Weill Cornell faculty members implemented on the EpicCare (Epic Systems) EHR between 2001 and 2007 were identified as potential study participants. Monthly visit volume,
Keywords:
charges, and work relative value units (wRVUs) were compared pre and post each provider’s
Electronic health record
EHR implementation go-live date. Practitioners who lacked at least 6 months of pre- and
Ambulatory care
post-implementation visit volume and charge data were excluded. Practitioners who did not
Physician productivity
meet pre-determined system proficiency metrics were additionally identified and became
Information system adoption
the basis of a non-adopter comparison group. Results: 203 physicians met criteria for the analysis. The eligible providers were divided into an adopter and non-adopter cohort based on system proficiency benchmarks. Those practitioners who adopted the EHR had a statistically significant increase in average monthly patient visit volume of 9 visits per provider per month. The non-adopter cohort’s visit volume was statistically unchanged. Both the EHR adopters and non-adopters had statistically significant increases (22% and 16% respectively) in average monthly charges in the post-implementation period. Average monthly wRVUs were statistically unchanged in the non-adopter cohort, but showed a positive and statistically significant increase of 12 wRVUs per provider per month for the adopter group. The EHR adoption group showed an incremental increase in productivity once practitioners achieved 6 or more months experience with the EHR, consistent with a “ramp-up” period. A multivariable regression model did not reveal any association between the post-EHR implementation change in wRVUs and several potential confounding variables, including baseline provider average monthly visit volume and wRVUs, date of system adoption, and specialty categorization. Conclusion: Provider productivity, as measured by patient visit volume, charges, and wRVUs modestly increased for a cohort of multi-specialty providers that adopted a commercially available ambulatory EHR. The productivity gain appeared to become even more pronounced after several months of system experience. This objective data may help persuade apprehensive practitioners that EHR adoption need not harm productivity. The baseline differences in productivity metrics for the adopters and non-adopters in our study suggest that there
∗ Corresponding author at: 575 Lexington Avenue, Third Floor, New York, NY 10022, United States. Tel.: +1 212 746 0471; fax: +1 212 746 7402. E-mail address: [email protected] (A.D. Cheriff). 1386-5056/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ijmedinf.2010.04.006
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
493
are fundamental differences in these groups. Further characterizing these differences may help predict EHR adoption success and guide future implementation strategies. © 2010 Elsevier Ireland Ltd. All rights reserved.
1.
Introduction
1.1.
Background
As electronic health record (EHR) adoption in the ambulatory setting accelerates, larger efforts are being made to study the information technology’s effect on various aspects of care delivery. The scientific evaluation of these systems helps us to more precisely understand the benefits and limitations of EHRs in order to enhance their design and overcome barriers to adoption. Significant barriers to EHR adoption remain in the United States. The U.S. considerably lags other parts of the world in terms of penetrance of the EHR [1]. As of 2008, only 4% of surveyed ambulatory practitioners used a fully functional EHR. Perhaps more surprising was the finding that even among very large physician groups with more than 50 providers, the adoption rate of a full-featured EHR was only about 17% [2]. There are many factors that have historically limited the rate of adoption of heath information technology, reflecting the complex misalignment of incentives that confront providers and healthcare delivery systems alike. Wildly divergent perceptions abound with regards to the potential benefits and pitfalls of an EHR implementation. HIT evangelists promote nearly unlimited potential of the technology to transform patient care and improve delivery efficiency, while rank-and-file providers resist, citing concerns about the expense, unreliability, and work-flow limitations of the systems [3]. A diverse literature does exist that demonstrates a variety of clinical benefits of the EHR. Many studies have demonstrated improved outcomes and quality via computerized order-entry systems and/or targeted decision support initiatives [4,5]. Further, survey data of those providers who have adopted partial or full functioned EHR systems indicate that most adopters consider the EHR to have had positive effects on care delivery, with enhancement of clinical decision quality, provider communication, and medical record accessibility [2]. The economics of EHR implementation are exceedingly complex. Health care delivery systems, both large and small, face significant financial decisions in the acquisition, implementation, and maintenance of electronic health records. Because of the substantial investments required, many have sought to demonstrate some form of economic return to justify the expenditures. There is a persistent attempt to demonstrate that EHR implementations result in net positive financial benefit. The return on investment (ROI) literature for EHR implementations is multifaceted, but does highlight the potentially positive effects of these systems [6]. Many of the first and most successful efforts to demonstrate investment returns focused on the use of clinician order-entry systems within hospital settings. In 2006, The Brigham and Women’s hospital estimated
a 10-year net savings of $16.7 million net savings attributable to its internally developed CPOE system [7]. Literature describing the EHR ROI experience within the ambulatory setting is considerably more sparse. Wang et al. performed a seminal cost-benefit analysis of an internally developed EHR in the primary care setting within the Partners HealthCare System [8]. After tallying the per-provider cost of system implementation over a five-year period, they concluded a per-provider net benefit of $86,400. They arrived at this return via a series of assumptions of savings that would be achieved by the EHR in terms of chart pulls, transcription costs, drug and ancillary service utilization, and accuracy and efficiency of charge capture. Grieger et al. were also able to demonstrate a positive return on investment in a pilot study describing the effects of the implementation of a commercially available EHR within five small ambulatory care practices within the University of Rochester Medical Center [9]. The authors concluded that the practices were able to achieve a net annual cost savings of $9,983 per provider and recoup initial costs within 16 months. Savings were attributed to reduced chart preparation and filing costs, a reduction in support staff, transcription costs, patient cycle time, and a shifting towards higher intensity evaluation and management billing codes.
1.2.
Rationale
Ongoing efforts to demonstrate ROI from ambulatory EHR implementations will likely continue, but it is unclear if the assumptions used in each setting will be broadly generalizable. A critical analysis of such data is unlikely to be a persuasive argument for providers who resist adoption. Attempts to demonstrate a net financial return from an EHR implementation underestimate the complexity of the expense and benefit allocation. Particularly in the case of large academic medical centers, there are a large set of stakeholders who make varying amounts of capital and time investments. The financial returns are diffuse and rarely directly passed on to those who bear the costs of implementation [3]. What has been less described in the EHR literature to date is the actual effect of implementation of the ambulatory EHR on physician productivity. This “top line” analysis is likely to have more valence with doctors who are skeptical of the ROI data and how it applies to them personally. Survey data has consistently demonstrated that one of the main barriers to system adoption is provider perception that the system will compromise productivity [2,10]. There is some evidence from time-motion studies to suggest that EHRs do not significantly slow down patient visits [11]. Anecdotal evidence suggests that an EHR implementation can transiently compromise physician productivity for some short adoption period, but that productivity levels then return to or exceed preimplementation levels. To date, there are few rigorous studies that have attempted to quantify these effects.
494
1.3.
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Objective
This study describes the effect on physician productivity of the implementation of a commercially available ambulatory EHR in a large multi-specialty academic physician group. Via retrospective analysis, the study compares individual practitioner visit volumes, charges, and work relative value units (wRVUs) before and after implementation of the EHR. It attempts to address the question of whether EHR implementation is associated with significant changes in physician productivity, as represented by the number of monthly patient visits and the associated billable units.
2.
Methods
The Weill Cornell Physician Organization is made up of more than 750 faculty providers in a wide variety of subspecialties of medicine. Weill Cornell Physicians see close to 1 million annual ambulatory patient visits via approximately 100 distinct practice groups. The vast majority of the practices are located on the main campus on the Upper East Side of Manhattan. As a quaternary care center, the practice is highly specialized with primary care making up roughly 15% of the total visit volume. Slightly more than 50% of patients have commercial managed care insurance, 25% have Medicare, and 7% have Medicaid. The Weill Cornell Physician Organization has a mature implementation of the GE Centricity Business Solutions (formerly IDX) practice management software which it began installing in 1984. All practitioner visit and billing data reside within the GE system. The implementation of the clinical EHR (EpicCare, Epic Systems) began in mid-2001 and has been a slow rolling implementation by practice unit, primarily focusing on the subset of Weill Cornell Physicians who provide outpatient evaluation and management services. The subspecialty sequence of the EHR roll-out was determined centrally by Physician Organization administration. Faculty members could not receive training or “go-live” on the EHR system outside of the pre-determined implementation period for his/her respective practice unit. The total number of Weill Cornell attending faculty providers eligible for full ambulatory EHR implementation was approximately 500 providers. About 250 of Weill Cornell’s faculty providers practice within subspecialties that primarily used niche health information systems (e.g. radiology, pathology, anesthesiology, etc.) The majority of Weill Cornell EHR adopters use nearly full system functionality, including electronic documentation, CPOE, results review, and charge capture.
2.1.
Eligibility criteria
Attending faculty physicians using the EHR at Weill Cornell Medical College were potential participants in the study. Attending provider productivity data from services rendered in a supervisory capacity within house-officer clinics were excluded. Initial inclusion required a provider having been trained on the EpicCare Ambulatory EHR sometime between the years 2001 and 2007. 343 providers met this initial criterion. To be eligible for further analysis, the physicians implemented
on the EHR needed to have a minimum of 6 months of scheduling and billing data within the GE Centricity Business (IDX) practice management system before and after his/her EHR golive date. The requirement for at least 6 months of practice data both before and after system adoption was necessary to eliminate any data skew. Given the nature of Weill Cornell’s academic practices, there is a great deal of month to month variation in productivity markers. Faculty clinicians often have rotating teaching or “on-service” responsibilities that impact ambulatory practice productivity. A total of 203 physicians met the requirement for 6 months of practice data pre- and post-system implementation. Most of the providers who were excluded on this basis either had little or no practicing career at Weill Cornell before the EHR implementation or left the institution within 6 months of having made the transition. The 203 eligible physicians were then rated with an EHR proficiency index to identify those practitioners whose ultimate EHR use after go-live was either so poor or non-existent to indicate that they never completed the implementation. The proficiency score was comprised of a weighted index1 which factored the percentage of patient encounters that were closed within the EHR and the closure lag time. This system proficiency index resulted in a per-provider numeric rating between 0 and 100. A rating of zero represented no encounter closure activity in the post-implementation period. A rating of 100 was achieved if a provider closed 100% of his/her EHR encounters with a median lag of 0 days (encounter closed on the date of service). Those practitioners whose proficiency rating was below 40 were considered system non-adopters and analyzed separately as a comparison group. The adoption threshold of 40 corresponded to an encounter closure rate of less than 20% with a median closure lag of greater than 2 days after the date of service. This was a level of proficiency at Weill Cornell that would necessitate a provider’s system reimplementation. A calculated proficiency score of 40 also was approximately one standard deviation below the mean EHR system proficiency score for the entire 203-physician cohort (mean = 70, SD = 28).
2.2.
Data acquisition
IRB approval was obtained to extract the pertinent study data. To measure physician productivity, monthly visit volume, charges, and wRVU data was extracted from the practice management system. Only ambulatory visit volume was included. Work RVUs in the pre- and post-implementation periods were based on published 2006 relative values [12]. For those physicians meeting inclusion criteria, data was extracted for up to 12 months pre-EHR go-live and up to 18 months post-EHR golive date. Extending the pre-and post-data collection periods helped to avoid skew from any possible seasonal variations of productivity and to account for the possibility of any potential post go-live ramp-up or post-implementation adjustment period.
1 Per-provider EHR proficiency index was derived using the following calculation: [(# of closed office visits/total # arrived appointments) × 70] + [30 − median number of closure lag days].
495
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
2.3.
data, the regression model required a log transformation of these data types.
Data analysis
In order to compare the productivity data for the provider population, monthly per-provider metrics (relative to go-live date) were computed for each of the three productivity variable categories; visit volume, charges, and wRVUs. The monthly productivity metrics were averaged across each study period for each provider. Months in which a provider did not have any ambulatory visit volume were excluded from the analysis. Work RVU data was further normalized relative to visit volume for both the pre-and post-implementation periods. In an effort to detect effects on productivity in the immediate post-implementation period, the productivity metrics in the post-implementation period were further subdivided into a “ramp-up” period (up to the first 5 months post-EHR go-live) and “post-ramp-up” period (6 months and beyond EHR go-live date). Because the provider productivity metrics were not normally distributed, a Wilcoxon-signed rank test was conducted to measure the significance of the difference between each of the pre- and post-productivity metrics for the adopter and non-adopter cohorts. Productivity changes were further analyzed using a multivariate regression model. This analysis attempted to identify any associations between changes in post-EHR-implementation average monthly wRVUs and the potentially confounding variables of date of EHR implementation, pre-implementation provider average monthly visit volume, pre-implementation provider average wRVUs, and provider specialty category (medical vs. surgical). Because of the non-normal distribution of the visit volume and wRVU
3.
Results
3.1.
Demographics
Application of inclusion and exclusion criteria to eligible providers resulted in an EHR adopter group of 147 providers and a non-adopter group of 56 providers. The mean computed system proficiency score, demographics, and clinical specialty distribution for each cohort is summarized in Table 1.
3.2.
Study findings
Average monthly patient visit volume, charges, wRVUs, and wRVUs per visit were compared in the pre- and postimplementation period for both the adopter and non-adopter cohorts. The findings are summarized in Table 2. The non-adopter group’s visit volume was statistically unchanged. The EHR-adopter group demonstrated a statistically significant increase in monthly average visit volume of 9 visits per provider per month. The average change in visit volume increased to 10 visits per provider per month beyond the ramp-up period. Both groups had a statistically significant increase in average monthly charges per provider. The non-adopter average monthly charges increased by 16% ($9.7K), while the EHR adopter group showed a larger 22% ($10.2K) increase.
Table 1 – Provider and practice characteristics. Adopters (N = 147)
Non-adopters (N = 56)
Provider proficiency score Mean (SD)
84 (14)
28 (10)
Provider demographics Gender Male Female Age—mean (SD) Years in practice—mean (SD)
71% 29% 51 (10) 21 (10)
80% 20% 56 (11) 26 (11)
Specialty distribution Medicine subspecialties Cardiovascular medicine Digestive GI/hepatology Hematology–oncology Internal medicine/primary care Pulmonary medicine Obstetrics/gynecology Anesthesia pain management Dermatology Neurology Ophthalmology Otorhinolaryngology Pediatric subspecialties Neurological surgery Rehabilitation medicine General surgical subspecialties Urology
9% 3% 12% 5% 4% 11% 0% 3% 7% 3% 3% 16% 5% 2% 14% 2%
20% 11% 4% 0% 4% 4% 4% 0% 9% 11% 2% 5% 4% 0% 11% 14%
496
EHR adopter group Pre
Visit volume Mean Standard deviation Change pre to post P value† Charges Mean Standard deviation Change pre to post P value wRVUs Mean Standard deviation Change pre to post P value wRVU/visit Mean Standard deviation Change pre to post P value †
EHR non-adopter group
Post
Pre ≥6 months
Total
Months 1–5
119 108
128 111 9 (8%) <.0001
125 112 6 (5%) <0.05
129 111 10 (8%) <.0001
$45,886 $49,024
$56,099 $64,633 $10,213 (22%) <.0001
$51,586 $60,278 $5,700 (12%) <.05
$57,836 $67,311 $11,950 (26%) <.0001
$62,027 $61,711
$71,728 $73,176 $9,701 (16%) 0.06
$66,703 $65,716 $4,676 (8%) 0.24
$73,527 $77,028 $11,500 (19%) 0.05
158.54 136.43
170.58 138.47 12.04 (8%) 0.05
164.13 135.51 5.60 (4%) 0.34
172.91 142.16 14.37 (9%) <.05
203.89 185.67
195.57 169.81 −8.32 (−4%) 0.90
189.48 160.19 −14.41 (−7%) 0.48
196.70 175.41 −7.19 (−4%) 0.88
5.84 12.59
5.80 13.38 −0.04 (−1%) 0.45
5.66 12.19 −0.18 (−3%) 0.06
2.17 4.37 −0.22 (−9%) <.05
P values were calculated with the use of a Wilcoxon-signed rank test.
2.44 6.99 0.05 (2%) <.05
2.13 4.09 −0.26 (−11%) <.05
81 74
Total 84 74 3 (4%) 0.21
Months 1–5
≥6 months
Total
2.39 5.56
Total
Post
83 74 2 (2%) 0.50
83 73 2 (2%) 0.30
6.05 14.73 0.2 (3%) 0.39
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Table 2 – Average monthly per-provider productivity metrics in the pre- and post-EHR implementation periods.
0.63 0.04 0.66 0.47 0.76 11.68 (24.49) 24.33 (11.69) −2.77 (6.37) −4.49 (6.14) 0.003 (0.01) 17.07 (19.25) 18.97 (9.19) −10.28 (5.01) 3.08 (4.82) −0.0002 (0.008) 0.04 0.38 0.71 0.8 0.11
0.38 0.04 0.04 0.52 0.97
0.05 21.56 (11.12) 0.02 20.00 (8.81) 0.04
Log transformation was applied to non-normally distributed data.
21.89 (10.37) −4.97 (5.65) −2.06 (5.45) 0.002 (0.009) 15.75 (9.91) Multivariable model EHR adoption status (adopter vs. non-adopter) Log of pre-implementation wRVU† Log of pre-implementation visit volume† Time from first EHR go-live (days) Specialty category (surgical vs. medicine)
†
20.38 (9.88)
P value Beta coefficient (SE) P value Beta coefficient (SE) P value Beta coefficient (SE)
497
Billable units with wRVU values were analyzed in the preand post-implementation periods for the adoption and intervention group. Overall, 82% of the unique CPT codes captured in the analysis period had associated wRVU values which accounted for 89% of the captured charges. The non-adopter group’s per-provider average monthly wRVU value was statistically unchanged. The adopter group’s average monthly wRVU total increased by a statistically significant 12 wRVUs per provider per month. The post-ramp-up period showed an even larger change of 14.4 wRVUs per provider per month. The wRVU/visit ratio was not significantly different in the post-implementation period for the non-adopter group. For EHR adopters, wRVU/visit fell by a statistically significant 0.22 wRVUs per visit in the entire post-implementation period, despite a statistically significant increase of 0.05 wRVUs per visit in the ramp-up period. Table 3 displays the results of a regression analysis on the entire study cohort to detect possible associations with several independent variables. The regression model tests whether the change in post-EHR-implementation wRVUs is significantly associated with EHR adoption status, date of EHR implementation, pre-implementation provider visit volume, pre-implementation wRVU level, and provider specialty category (medical vs. surgical). Only the provider EHR adoption status was significantly associated with a post-implementation change in wRVUs. None of the other potentially confounding characteristics of the providers were found to be significantly associated with a change in post-implementation wRVUs.
4.
Univariable model EHR adoption status (adopter vs. non-adopter)
Total post-wRVU
Table 3 – Univariable and multivariable linear regression for change in wrvu (N = 203).
Months 1–5 post-wRVU
≥6 months post-wRVU
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Discussion
The results of this study provide previously unavailable insight into provider productivity in a large multi-specialty academic physician group during a prolonged ambulatory EHR implementation. Weill Cornell’s experience with its ambulatory EHR implementation suggests that fears about the EHR adversely affecting long-term physician productivity may be unfounded. In fact, it would appear that those providers who adopted the ambulatory EHR, on average, demonstrated increased productivity. This analysis showed that those faculty members who adopted the EHR had higher recorded average visit volumes and had higher average monthly charges and wRVUs than prior to EHR usage. A comparison subgroup of providers who failed to adopt the EHR did not show improved visit volume or wRVUs. It is not surprising that both the adoption and comparison groups saw increases in charges per unit visit in the post-implementation period. This likely reflects rising fee schedules over time. However, the wRVU analysis, which represented the majority of billable units, suggests that when controlling for changing fee schedules, productivity still increases. While this subgroup analysis is not a true control, the finding is provocative. The relatively simplistic regression model built to further assess the post-adoption change in wRVUs did not suggest that the changes in productivity were associated with several other measurable independent variables.
498
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Though there appears to be an association with EHR adoption and increased average monthly patient visits and wRVUs, this does not necessarily suggest that the technology allowed physicians to complete more work or higher intensity output per unit time A plausible explanation is that the EHR is simply more efficient at capturing and recording effort that was already being exerted. This might help explain the unexpected finding that the ratio of wRVU/visit actually seems to have fallen in the post-implementation period for EHR adopters. Our expectation was that wRVU per visit would have significantly increased upon adoption of the EHR based on the potential for better data capture and perhaps higher intensity billing at each visit. In fact, we detected a statistically significant decrease in wRVU/visit. One possible explanation for this is that much of the incremental volume captured by the EHR may have been for low intensity services like counseling visits, lab visits, or vaccination visits that previously went undocumented and unbilled. This would have had the effect of increasing the overall visit volumes and wRVUs, but lowering the ratio of wRVU to visit. Further exploration of post-EHR changes in visit type distribution, CPT frequency, and E/M coding intensity might be helpful in future analyses to help understand this phenomenon. We can extrapolate the financial value of the productivity increase for EHR adopters in the Weill Cornell Physician Organization. This study revealed an average increase of approximately 12 wRVUs per month per provider in the post-EHR implementation period for system adopters. Our recent global institutional experience suggests a ratio of approximately $285 worth of charges for every wRVU. Considering approximately 500 practitioners who are eligible for the ambulatory EHR and a gross collection ratio of 44%, the “bottom line” monetary value of EHR adoption could be approximately $9 million annually. It is worth noting that this equates to a net benefit of approximately $18,000 annually per physician, which closely approximates the adjusted figure of $17,000 of annual net benefit in the Wang et al. analysis [8]. It is also within the range of the federal incentive payments for physician adoption of EHRs and should help justify ongoing maintenance of the investment by physicians themselves, after the government subsidy has expired. This analysis also reinforces the widely held notion that productivity gains attributable to the EHR may not be realized in the immediate post-implementation or “ramp-up” period. This is not at all surprising given Weill Cornell’s implementation methodology which explicitly recommends a decrease in visit volume in the “go-live” period to acquire system proficiency. Our data suggests that the investment to master the EHR does pay off based on higher productivity indices beyond the ramp-up period.
4.1.
Study limitations
This study has some obvious limitations. The findings should be interpreted cautiously as our methodology amounts to a retrospective “before and after” study, rather than a prospective randomized control trial. One could have presented the productivity changes for the entire cohort of faculty physi-
cians, including adopters and non-adopters. This would have approximated an “intention to treat” methodology. Because EHR system adoption was not actually mandated, nonadoption cannot really be deemed a “treatment failure.” Since inclusion of non-users could mask or dilute possible positive and/or negative effects of the EHR on productivity, we opted to analyze the data by considering each cohort independently. Because of different baseline characteristics of the cohorts, we labeled the non-adopters as a “comparison” group as it cannot strictly be considered a “control” group. While the non-adopter group does not constitute a rigorous control, it does represent a real world comparison group of users who have resisted EHR implementation and use. The univariate analysis did show significant positive changes in the post-implementation period for the system adopters. Clearly there were many potential confounding variables that may have influenced provider productivity in addition to adoption of the EHR, though our staggered implementation times statistically mitigates this effect somewhat. Our basic regression analysis did indicate that even when controlling for date of implementation, baseline provider visit volume, and subspecialty category, EHR adopters still had improved average monthly productivity after system adoption. While these findings are encouraging and strengthen the notion that EHR adoption was associated with higher productivity, there are many other important confounding variables that were difficult for us to quantify in any meaningful way. Our EHR implementation occurred over many years. During that time, our Physician Organization was very aggressive in promoting best practices that would presumably have positive impacts on provider proficiency. For example, the Physician Organization has commissioned several internal and external consulting engagements to identify opportunities for improved practice operations within the clinical departments. A great deal of effort has been made to optimize the use of our practice management system to achieve better utilization of provider patient-care schedules and higher quality patient registration data. During the study period, major investments were made in reporting and business intelligence tools that allowed practice managers to more efficiently manage all aspects of the revenue cycle. Lastly, during the same period as our EHR implementation, the institution initiated a formal office of billing compliance. The education provided to providers with regards to accurate E/M coding almost certainly had implications for the provider productivity markers considered in this study.
4.2.
New questions
The dissimilar nature of the EHR adopters and non-adopters is potentially telling in terms of the characteristics that are most likely to predict implementation failure. In our small cohort, those providers who failed to adopt the EHR tended to have lower patient visit volume, but higher baseline average monthly charges and wRVUs. One could argue that the higher baseline productivity markers in the adoption group reflected a more optimized subset of practitioners. This raises
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Summary points What was already known on the topic? • There is a large body of evidence that demonstrates the clinical benefits of the electronic health record, particularly with regards to CPOE and decision support tools. Fewer studies have addressed the financial implications of EHR adoption in the ambulatory care setting. • Previous investigations have revealed meaningful return on investment for electronic health record implementations, but these analyses have focused on aggregate institutional costs and savings, rather than physician productivity. • Electronic health record adoption in the ambulatory setting has been impeded by provider fears of reduction in productivity. What this study added to our knowledge? • Adoption of a commercially available ambulatory EHR has the potential to increase physician productivity as measured by average monthly patient visit volume and provider work relative value units. • This analysis provided objective support of anecdotal observations that a “ramp up” period of approximately 6 months post-EHR-implementation exists. Providers achieve additional productivity gains beyond this adjustment period. • The findings provide evidence that physician productivity need not be harmed by the transition to the electronic health record. This will help overcome provider resistance to EHR adoption.
the question as to whether the EHR would only increase productivity for practitioner with a low or non-optimized baseline. It is worth noting the data trend that average monthly wRVUs decreased for non-adopters in the immediate post-implementation period. It is conceivable that this group of providers may have had their productivity disproportionately disrupted during the ramp-up period, leading ultimately to system abandonment. There is certainly a “chicken and egg” question that requires further exploration. Does the EHR dramatically influence provider productivity or do productivity characteristics of providers predict system adoption?
4.3.
Meaning and impact
With the U.S. Government’s enactment of the Health Information Technology for Economic and Clinical Health Act (HITECH) [13], there will be renewed interest in provider proficiency with the EHR as it relates to “meaningful use.” There have been previous attempts in the literature to describe a spectrum of EHR usage sophistication [14]. As objective models of user proficiency mature, it will present an opportunity to identify correlations between the quality of EHR usage and the magnitude of any resulting productivity changes.
499
As HIT adoption accelerates nationally, understanding the effects of the EHR on physician productivity may be more relevant than attempts to demonstrate recoupment of sunk costs. The very real benefits of the EHR in terms of structured data capture, decision support, and provider communication are difficult to quantify in financial terms. Given the overwhelming regulatory pressures and intra-operability demands facing the U.S. health care system, ongoing attempts to demonstrate the technology’s ability to pay for itself may be misplaced. Further limiting the actual value of these exercises is the cost of these analyses and the difficulty quantifying the “soft” ROI of improved clinical care. A robust heath information technology infrastructure is now part of the cost of doing business in modern medicine. However, our experience in a large academic multi-specialty physician group suggests the potential to implement the EHR and still derive a modest increase in overall physician productivity. The fact that the providers can exceed preimplementation productivity levels within 6 months of an EHR transition should hopefully alleviate some barriers to system adoption and help to further align the incentives of the individual practitioners and their associated care delivery organizations.
Acknowledgements The authors wish to acknowledge the outstanding contributions of Alison M. Edwards for her assistance with the statistical analyses in this manuscript. The authors also wish to acknowledge the efforts of the EHR implementation staff of the Weill Cornell Physician Organization. We would like to additionally thank the administrative leadership of the Weill Cornell Physician Organization for its unwavering support of health information technology and our efforts to evaluate its impact.
references
[1] D.W. Bates, Physicians and ambulatory electronic health records, Health Aff. 24 (September/October (5)) (2005) 1180–1189. [2] C.M. DesRoches, E.G. Campbell, R.R. Sowmya, et al., Electronic health records in ambulatory care—a national survey of physicians, N. Engl. J. Med. 359 (July (1)) (2008) 50–60. [3] D. Blumenthal, J.P. Glaser, Information technology comes to medicine, N. Engl. J. Med. 356 (June (24)) (2007) 2527–2534. [4] B. Chaudhry, J. Wang, S. Wu, et al., Systematic review: impact of health information technology on quality, efficiency, and costs of medical care, Ann. Intern. Med. 144 (2006), E-12–E-22. [5] R. Amarasingham, L. Plantinga, M. Diener-West, et al., Clinical information technologies and inpatient outcomes, Arch. Intern. Med. 169 (2) (2009) 108–114. [6] N. Menachemi, R.G. Brooks, Reviewing the benefits and costs of electronic health records and associated patient safety technologies, J. Med. Syst. 30 (2006) 159–168. [7] R. Kaushal, A.K. Jha, C. Franz, et al., Return on investment for a computerized physician order entry system, J. Am. Med. Inform. Assoc. 13 (May/June (3)) (2006) 261–266.
500
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
[8] S.J. Wang, B. Middleton, L.A. Prosser, et al., A cost-benefit analysis of electronic medical records in primary care, Am. J. Med. 114 (April) (2003) 397–403. [9] D.L. Grieger, S.H. Cohen, D.A. Krusch, A pilot study to document the return on investment for implementing an ambulatory electronic health record at an academic medical center, J. Am. Coll. Surg. 205 (July (1)) (2007) 89–96. [10] N. Menachemi, Barriers to ambulatory EHR: who are “imminent adopters” and how do they differ from other physicians? Inform. Prim. Care 14 (2006) 101–108.
[11] H.G. Lo, L.P. Newmark, C. Yoon, et al., Electronic health records in specialty care: a time-motion study, J. Am. Med. Inform. Assoc. September/October (14) (2007) 609–615. [12] Centers for Medicare & Medicaid Services, PFS Relative Value Files, 2006, http://www.cms.hhs.gov/ PhysicianFeeSched/PFSRVF/list.asp?listpage=2. [13] American Recovery and Reinvestment Act of 2009, H.R. 1, 111th Cong., 1st sess., 2009. [14] R.H. Miller, I. Sim, J. Newman, Electronic medical records in solo/small groups: a qualitative study of physician user types, Stud. Health Technol. Inform. 107 (1) (2004) 658–662.
journal homepage: www.intl.elsevierhealth.com/journals/ijmi
Physician productivity and the ambulatory EHR in a large academic multi-specialty physician group Adam D. Cheriff ∗ , Akshay G. Kapur, Maggie Qiu, Curtis L. Cole Weill Cornell Medical College, New York, NY, United States
a r t i c l e
i n f o
a b s t r a c t
Article history:
Purpose: The impact of the ambulatory electronic health record (EHR) on physician productiv-
Received 14 September 2009
ity is poorly understood. Fear of productivity loss remains a major concern for practitioners
Received in revised form
and health care delivery organizations and inhibits system adoption. This study describes
23 April 2010
the changes in physician productivity after the implementation of a commercially available
Accepted 23 April 2010
ambulatory EHR system in a large academic multi-specialty physician group. Methods: Weill Cornell faculty members implemented on the EpicCare (Epic Systems) EHR between 2001 and 2007 were identified as potential study participants. Monthly visit volume,
Keywords:
charges, and work relative value units (wRVUs) were compared pre and post each provider’s
Electronic health record
EHR implementation go-live date. Practitioners who lacked at least 6 months of pre- and
Ambulatory care
post-implementation visit volume and charge data were excluded. Practitioners who did not
Physician productivity
meet pre-determined system proficiency metrics were additionally identified and became
Information system adoption
the basis of a non-adopter comparison group. Results: 203 physicians met criteria for the analysis. The eligible providers were divided into an adopter and non-adopter cohort based on system proficiency benchmarks. Those practitioners who adopted the EHR had a statistically significant increase in average monthly patient visit volume of 9 visits per provider per month. The non-adopter cohort’s visit volume was statistically unchanged. Both the EHR adopters and non-adopters had statistically significant increases (22% and 16% respectively) in average monthly charges in the post-implementation period. Average monthly wRVUs were statistically unchanged in the non-adopter cohort, but showed a positive and statistically significant increase of 12 wRVUs per provider per month for the adopter group. The EHR adoption group showed an incremental increase in productivity once practitioners achieved 6 or more months experience with the EHR, consistent with a “ramp-up” period. A multivariable regression model did not reveal any association between the post-EHR implementation change in wRVUs and several potential confounding variables, including baseline provider average monthly visit volume and wRVUs, date of system adoption, and specialty categorization. Conclusion: Provider productivity, as measured by patient visit volume, charges, and wRVUs modestly increased for a cohort of multi-specialty providers that adopted a commercially available ambulatory EHR. The productivity gain appeared to become even more pronounced after several months of system experience. This objective data may help persuade apprehensive practitioners that EHR adoption need not harm productivity. The baseline differences in productivity metrics for the adopters and non-adopters in our study suggest that there
∗ Corresponding author at: 575 Lexington Avenue, Third Floor, New York, NY 10022, United States. Tel.: +1 212 746 0471; fax: +1 212 746 7402. E-mail address: [email protected] (A.D. Cheriff). 1386-5056/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ijmedinf.2010.04.006
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
493
are fundamental differences in these groups. Further characterizing these differences may help predict EHR adoption success and guide future implementation strategies. © 2010 Elsevier Ireland Ltd. All rights reserved.
1.
Introduction
1.1.
Background
As electronic health record (EHR) adoption in the ambulatory setting accelerates, larger efforts are being made to study the information technology’s effect on various aspects of care delivery. The scientific evaluation of these systems helps us to more precisely understand the benefits and limitations of EHRs in order to enhance their design and overcome barriers to adoption. Significant barriers to EHR adoption remain in the United States. The U.S. considerably lags other parts of the world in terms of penetrance of the EHR [1]. As of 2008, only 4% of surveyed ambulatory practitioners used a fully functional EHR. Perhaps more surprising was the finding that even among very large physician groups with more than 50 providers, the adoption rate of a full-featured EHR was only about 17% [2]. There are many factors that have historically limited the rate of adoption of heath information technology, reflecting the complex misalignment of incentives that confront providers and healthcare delivery systems alike. Wildly divergent perceptions abound with regards to the potential benefits and pitfalls of an EHR implementation. HIT evangelists promote nearly unlimited potential of the technology to transform patient care and improve delivery efficiency, while rank-and-file providers resist, citing concerns about the expense, unreliability, and work-flow limitations of the systems [3]. A diverse literature does exist that demonstrates a variety of clinical benefits of the EHR. Many studies have demonstrated improved outcomes and quality via computerized order-entry systems and/or targeted decision support initiatives [4,5]. Further, survey data of those providers who have adopted partial or full functioned EHR systems indicate that most adopters consider the EHR to have had positive effects on care delivery, with enhancement of clinical decision quality, provider communication, and medical record accessibility [2]. The economics of EHR implementation are exceedingly complex. Health care delivery systems, both large and small, face significant financial decisions in the acquisition, implementation, and maintenance of electronic health records. Because of the substantial investments required, many have sought to demonstrate some form of economic return to justify the expenditures. There is a persistent attempt to demonstrate that EHR implementations result in net positive financial benefit. The return on investment (ROI) literature for EHR implementations is multifaceted, but does highlight the potentially positive effects of these systems [6]. Many of the first and most successful efforts to demonstrate investment returns focused on the use of clinician order-entry systems within hospital settings. In 2006, The Brigham and Women’s hospital estimated
a 10-year net savings of $16.7 million net savings attributable to its internally developed CPOE system [7]. Literature describing the EHR ROI experience within the ambulatory setting is considerably more sparse. Wang et al. performed a seminal cost-benefit analysis of an internally developed EHR in the primary care setting within the Partners HealthCare System [8]. After tallying the per-provider cost of system implementation over a five-year period, they concluded a per-provider net benefit of $86,400. They arrived at this return via a series of assumptions of savings that would be achieved by the EHR in terms of chart pulls, transcription costs, drug and ancillary service utilization, and accuracy and efficiency of charge capture. Grieger et al. were also able to demonstrate a positive return on investment in a pilot study describing the effects of the implementation of a commercially available EHR within five small ambulatory care practices within the University of Rochester Medical Center [9]. The authors concluded that the practices were able to achieve a net annual cost savings of $9,983 per provider and recoup initial costs within 16 months. Savings were attributed to reduced chart preparation and filing costs, a reduction in support staff, transcription costs, patient cycle time, and a shifting towards higher intensity evaluation and management billing codes.
1.2.
Rationale
Ongoing efforts to demonstrate ROI from ambulatory EHR implementations will likely continue, but it is unclear if the assumptions used in each setting will be broadly generalizable. A critical analysis of such data is unlikely to be a persuasive argument for providers who resist adoption. Attempts to demonstrate a net financial return from an EHR implementation underestimate the complexity of the expense and benefit allocation. Particularly in the case of large academic medical centers, there are a large set of stakeholders who make varying amounts of capital and time investments. The financial returns are diffuse and rarely directly passed on to those who bear the costs of implementation [3]. What has been less described in the EHR literature to date is the actual effect of implementation of the ambulatory EHR on physician productivity. This “top line” analysis is likely to have more valence with doctors who are skeptical of the ROI data and how it applies to them personally. Survey data has consistently demonstrated that one of the main barriers to system adoption is provider perception that the system will compromise productivity [2,10]. There is some evidence from time-motion studies to suggest that EHRs do not significantly slow down patient visits [11]. Anecdotal evidence suggests that an EHR implementation can transiently compromise physician productivity for some short adoption period, but that productivity levels then return to or exceed preimplementation levels. To date, there are few rigorous studies that have attempted to quantify these effects.
494
1.3.
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Objective
This study describes the effect on physician productivity of the implementation of a commercially available ambulatory EHR in a large multi-specialty academic physician group. Via retrospective analysis, the study compares individual practitioner visit volumes, charges, and work relative value units (wRVUs) before and after implementation of the EHR. It attempts to address the question of whether EHR implementation is associated with significant changes in physician productivity, as represented by the number of monthly patient visits and the associated billable units.
2.
Methods
The Weill Cornell Physician Organization is made up of more than 750 faculty providers in a wide variety of subspecialties of medicine. Weill Cornell Physicians see close to 1 million annual ambulatory patient visits via approximately 100 distinct practice groups. The vast majority of the practices are located on the main campus on the Upper East Side of Manhattan. As a quaternary care center, the practice is highly specialized with primary care making up roughly 15% of the total visit volume. Slightly more than 50% of patients have commercial managed care insurance, 25% have Medicare, and 7% have Medicaid. The Weill Cornell Physician Organization has a mature implementation of the GE Centricity Business Solutions (formerly IDX) practice management software which it began installing in 1984. All practitioner visit and billing data reside within the GE system. The implementation of the clinical EHR (EpicCare, Epic Systems) began in mid-2001 and has been a slow rolling implementation by practice unit, primarily focusing on the subset of Weill Cornell Physicians who provide outpatient evaluation and management services. The subspecialty sequence of the EHR roll-out was determined centrally by Physician Organization administration. Faculty members could not receive training or “go-live” on the EHR system outside of the pre-determined implementation period for his/her respective practice unit. The total number of Weill Cornell attending faculty providers eligible for full ambulatory EHR implementation was approximately 500 providers. About 250 of Weill Cornell’s faculty providers practice within subspecialties that primarily used niche health information systems (e.g. radiology, pathology, anesthesiology, etc.) The majority of Weill Cornell EHR adopters use nearly full system functionality, including electronic documentation, CPOE, results review, and charge capture.
2.1.
Eligibility criteria
Attending faculty physicians using the EHR at Weill Cornell Medical College were potential participants in the study. Attending provider productivity data from services rendered in a supervisory capacity within house-officer clinics were excluded. Initial inclusion required a provider having been trained on the EpicCare Ambulatory EHR sometime between the years 2001 and 2007. 343 providers met this initial criterion. To be eligible for further analysis, the physicians implemented
on the EHR needed to have a minimum of 6 months of scheduling and billing data within the GE Centricity Business (IDX) practice management system before and after his/her EHR golive date. The requirement for at least 6 months of practice data both before and after system adoption was necessary to eliminate any data skew. Given the nature of Weill Cornell’s academic practices, there is a great deal of month to month variation in productivity markers. Faculty clinicians often have rotating teaching or “on-service” responsibilities that impact ambulatory practice productivity. A total of 203 physicians met the requirement for 6 months of practice data pre- and post-system implementation. Most of the providers who were excluded on this basis either had little or no practicing career at Weill Cornell before the EHR implementation or left the institution within 6 months of having made the transition. The 203 eligible physicians were then rated with an EHR proficiency index to identify those practitioners whose ultimate EHR use after go-live was either so poor or non-existent to indicate that they never completed the implementation. The proficiency score was comprised of a weighted index1 which factored the percentage of patient encounters that were closed within the EHR and the closure lag time. This system proficiency index resulted in a per-provider numeric rating between 0 and 100. A rating of zero represented no encounter closure activity in the post-implementation period. A rating of 100 was achieved if a provider closed 100% of his/her EHR encounters with a median lag of 0 days (encounter closed on the date of service). Those practitioners whose proficiency rating was below 40 were considered system non-adopters and analyzed separately as a comparison group. The adoption threshold of 40 corresponded to an encounter closure rate of less than 20% with a median closure lag of greater than 2 days after the date of service. This was a level of proficiency at Weill Cornell that would necessitate a provider’s system reimplementation. A calculated proficiency score of 40 also was approximately one standard deviation below the mean EHR system proficiency score for the entire 203-physician cohort (mean = 70, SD = 28).
2.2.
Data acquisition
IRB approval was obtained to extract the pertinent study data. To measure physician productivity, monthly visit volume, charges, and wRVU data was extracted from the practice management system. Only ambulatory visit volume was included. Work RVUs in the pre- and post-implementation periods were based on published 2006 relative values [12]. For those physicians meeting inclusion criteria, data was extracted for up to 12 months pre-EHR go-live and up to 18 months post-EHR golive date. Extending the pre-and post-data collection periods helped to avoid skew from any possible seasonal variations of productivity and to account for the possibility of any potential post go-live ramp-up or post-implementation adjustment period.
1 Per-provider EHR proficiency index was derived using the following calculation: [(# of closed office visits/total # arrived appointments) × 70] + [30 − median number of closure lag days].
495
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
2.3.
data, the regression model required a log transformation of these data types.
Data analysis
In order to compare the productivity data for the provider population, monthly per-provider metrics (relative to go-live date) were computed for each of the three productivity variable categories; visit volume, charges, and wRVUs. The monthly productivity metrics were averaged across each study period for each provider. Months in which a provider did not have any ambulatory visit volume were excluded from the analysis. Work RVU data was further normalized relative to visit volume for both the pre-and post-implementation periods. In an effort to detect effects on productivity in the immediate post-implementation period, the productivity metrics in the post-implementation period were further subdivided into a “ramp-up” period (up to the first 5 months post-EHR go-live) and “post-ramp-up” period (6 months and beyond EHR go-live date). Because the provider productivity metrics were not normally distributed, a Wilcoxon-signed rank test was conducted to measure the significance of the difference between each of the pre- and post-productivity metrics for the adopter and non-adopter cohorts. Productivity changes were further analyzed using a multivariate regression model. This analysis attempted to identify any associations between changes in post-EHR-implementation average monthly wRVUs and the potentially confounding variables of date of EHR implementation, pre-implementation provider average monthly visit volume, pre-implementation provider average wRVUs, and provider specialty category (medical vs. surgical). Because of the non-normal distribution of the visit volume and wRVU
3.
Results
3.1.
Demographics
Application of inclusion and exclusion criteria to eligible providers resulted in an EHR adopter group of 147 providers and a non-adopter group of 56 providers. The mean computed system proficiency score, demographics, and clinical specialty distribution for each cohort is summarized in Table 1.
3.2.
Study findings
Average monthly patient visit volume, charges, wRVUs, and wRVUs per visit were compared in the pre- and postimplementation period for both the adopter and non-adopter cohorts. The findings are summarized in Table 2. The non-adopter group’s visit volume was statistically unchanged. The EHR-adopter group demonstrated a statistically significant increase in monthly average visit volume of 9 visits per provider per month. The average change in visit volume increased to 10 visits per provider per month beyond the ramp-up period. Both groups had a statistically significant increase in average monthly charges per provider. The non-adopter average monthly charges increased by 16% ($9.7K), while the EHR adopter group showed a larger 22% ($10.2K) increase.
Table 1 – Provider and practice characteristics. Adopters (N = 147)
Non-adopters (N = 56)
Provider proficiency score Mean (SD)
84 (14)
28 (10)
Provider demographics Gender Male Female Age—mean (SD) Years in practice—mean (SD)
71% 29% 51 (10) 21 (10)
80% 20% 56 (11) 26 (11)
Specialty distribution Medicine subspecialties Cardiovascular medicine Digestive GI/hepatology Hematology–oncology Internal medicine/primary care Pulmonary medicine Obstetrics/gynecology Anesthesia pain management Dermatology Neurology Ophthalmology Otorhinolaryngology Pediatric subspecialties Neurological surgery Rehabilitation medicine General surgical subspecialties Urology
9% 3% 12% 5% 4% 11% 0% 3% 7% 3% 3% 16% 5% 2% 14% 2%
20% 11% 4% 0% 4% 4% 4% 0% 9% 11% 2% 5% 4% 0% 11% 14%
496
EHR adopter group Pre
Visit volume Mean Standard deviation Change pre to post P value† Charges Mean Standard deviation Change pre to post P value wRVUs Mean Standard deviation Change pre to post P value wRVU/visit Mean Standard deviation Change pre to post P value †
EHR non-adopter group
Post
Pre ≥6 months
Total
Months 1–5
119 108
128 111 9 (8%) <.0001
125 112 6 (5%) <0.05
129 111 10 (8%) <.0001
$45,886 $49,024
$56,099 $64,633 $10,213 (22%) <.0001
$51,586 $60,278 $5,700 (12%) <.05
$57,836 $67,311 $11,950 (26%) <.0001
$62,027 $61,711
$71,728 $73,176 $9,701 (16%) 0.06
$66,703 $65,716 $4,676 (8%) 0.24
$73,527 $77,028 $11,500 (19%) 0.05
158.54 136.43
170.58 138.47 12.04 (8%) 0.05
164.13 135.51 5.60 (4%) 0.34
172.91 142.16 14.37 (9%) <.05
203.89 185.67
195.57 169.81 −8.32 (−4%) 0.90
189.48 160.19 −14.41 (−7%) 0.48
196.70 175.41 −7.19 (−4%) 0.88
5.84 12.59
5.80 13.38 −0.04 (−1%) 0.45
5.66 12.19 −0.18 (−3%) 0.06
2.17 4.37 −0.22 (−9%) <.05
P values were calculated with the use of a Wilcoxon-signed rank test.
2.44 6.99 0.05 (2%) <.05
2.13 4.09 −0.26 (−11%) <.05
81 74
Total 84 74 3 (4%) 0.21
Months 1–5
≥6 months
Total
2.39 5.56
Total
Post
83 74 2 (2%) 0.50
83 73 2 (2%) 0.30
6.05 14.73 0.2 (3%) 0.39
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Table 2 – Average monthly per-provider productivity metrics in the pre- and post-EHR implementation periods.
0.63 0.04 0.66 0.47 0.76 11.68 (24.49) 24.33 (11.69) −2.77 (6.37) −4.49 (6.14) 0.003 (0.01) 17.07 (19.25) 18.97 (9.19) −10.28 (5.01) 3.08 (4.82) −0.0002 (0.008) 0.04 0.38 0.71 0.8 0.11
0.38 0.04 0.04 0.52 0.97
0.05 21.56 (11.12) 0.02 20.00 (8.81) 0.04
Log transformation was applied to non-normally distributed data.
21.89 (10.37) −4.97 (5.65) −2.06 (5.45) 0.002 (0.009) 15.75 (9.91) Multivariable model EHR adoption status (adopter vs. non-adopter) Log of pre-implementation wRVU† Log of pre-implementation visit volume† Time from first EHR go-live (days) Specialty category (surgical vs. medicine)
†
20.38 (9.88)
P value Beta coefficient (SE) P value Beta coefficient (SE) P value Beta coefficient (SE)
497
Billable units with wRVU values were analyzed in the preand post-implementation periods for the adoption and intervention group. Overall, 82% of the unique CPT codes captured in the analysis period had associated wRVU values which accounted for 89% of the captured charges. The non-adopter group’s per-provider average monthly wRVU value was statistically unchanged. The adopter group’s average monthly wRVU total increased by a statistically significant 12 wRVUs per provider per month. The post-ramp-up period showed an even larger change of 14.4 wRVUs per provider per month. The wRVU/visit ratio was not significantly different in the post-implementation period for the non-adopter group. For EHR adopters, wRVU/visit fell by a statistically significant 0.22 wRVUs per visit in the entire post-implementation period, despite a statistically significant increase of 0.05 wRVUs per visit in the ramp-up period. Table 3 displays the results of a regression analysis on the entire study cohort to detect possible associations with several independent variables. The regression model tests whether the change in post-EHR-implementation wRVUs is significantly associated with EHR adoption status, date of EHR implementation, pre-implementation provider visit volume, pre-implementation wRVU level, and provider specialty category (medical vs. surgical). Only the provider EHR adoption status was significantly associated with a post-implementation change in wRVUs. None of the other potentially confounding characteristics of the providers were found to be significantly associated with a change in post-implementation wRVUs.
4.
Univariable model EHR adoption status (adopter vs. non-adopter)
Total post-wRVU
Table 3 – Univariable and multivariable linear regression for change in wrvu (N = 203).
Months 1–5 post-wRVU
≥6 months post-wRVU
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Discussion
The results of this study provide previously unavailable insight into provider productivity in a large multi-specialty academic physician group during a prolonged ambulatory EHR implementation. Weill Cornell’s experience with its ambulatory EHR implementation suggests that fears about the EHR adversely affecting long-term physician productivity may be unfounded. In fact, it would appear that those providers who adopted the ambulatory EHR, on average, demonstrated increased productivity. This analysis showed that those faculty members who adopted the EHR had higher recorded average visit volumes and had higher average monthly charges and wRVUs than prior to EHR usage. A comparison subgroup of providers who failed to adopt the EHR did not show improved visit volume or wRVUs. It is not surprising that both the adoption and comparison groups saw increases in charges per unit visit in the post-implementation period. This likely reflects rising fee schedules over time. However, the wRVU analysis, which represented the majority of billable units, suggests that when controlling for changing fee schedules, productivity still increases. While this subgroup analysis is not a true control, the finding is provocative. The relatively simplistic regression model built to further assess the post-adoption change in wRVUs did not suggest that the changes in productivity were associated with several other measurable independent variables.
498
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Though there appears to be an association with EHR adoption and increased average monthly patient visits and wRVUs, this does not necessarily suggest that the technology allowed physicians to complete more work or higher intensity output per unit time A plausible explanation is that the EHR is simply more efficient at capturing and recording effort that was already being exerted. This might help explain the unexpected finding that the ratio of wRVU/visit actually seems to have fallen in the post-implementation period for EHR adopters. Our expectation was that wRVU per visit would have significantly increased upon adoption of the EHR based on the potential for better data capture and perhaps higher intensity billing at each visit. In fact, we detected a statistically significant decrease in wRVU/visit. One possible explanation for this is that much of the incremental volume captured by the EHR may have been for low intensity services like counseling visits, lab visits, or vaccination visits that previously went undocumented and unbilled. This would have had the effect of increasing the overall visit volumes and wRVUs, but lowering the ratio of wRVU to visit. Further exploration of post-EHR changes in visit type distribution, CPT frequency, and E/M coding intensity might be helpful in future analyses to help understand this phenomenon. We can extrapolate the financial value of the productivity increase for EHR adopters in the Weill Cornell Physician Organization. This study revealed an average increase of approximately 12 wRVUs per month per provider in the post-EHR implementation period for system adopters. Our recent global institutional experience suggests a ratio of approximately $285 worth of charges for every wRVU. Considering approximately 500 practitioners who are eligible for the ambulatory EHR and a gross collection ratio of 44%, the “bottom line” monetary value of EHR adoption could be approximately $9 million annually. It is worth noting that this equates to a net benefit of approximately $18,000 annually per physician, which closely approximates the adjusted figure of $17,000 of annual net benefit in the Wang et al. analysis [8]. It is also within the range of the federal incentive payments for physician adoption of EHRs and should help justify ongoing maintenance of the investment by physicians themselves, after the government subsidy has expired. This analysis also reinforces the widely held notion that productivity gains attributable to the EHR may not be realized in the immediate post-implementation or “ramp-up” period. This is not at all surprising given Weill Cornell’s implementation methodology which explicitly recommends a decrease in visit volume in the “go-live” period to acquire system proficiency. Our data suggests that the investment to master the EHR does pay off based on higher productivity indices beyond the ramp-up period.
4.1.
Study limitations
This study has some obvious limitations. The findings should be interpreted cautiously as our methodology amounts to a retrospective “before and after” study, rather than a prospective randomized control trial. One could have presented the productivity changes for the entire cohort of faculty physi-
cians, including adopters and non-adopters. This would have approximated an “intention to treat” methodology. Because EHR system adoption was not actually mandated, nonadoption cannot really be deemed a “treatment failure.” Since inclusion of non-users could mask or dilute possible positive and/or negative effects of the EHR on productivity, we opted to analyze the data by considering each cohort independently. Because of different baseline characteristics of the cohorts, we labeled the non-adopters as a “comparison” group as it cannot strictly be considered a “control” group. While the non-adopter group does not constitute a rigorous control, it does represent a real world comparison group of users who have resisted EHR implementation and use. The univariate analysis did show significant positive changes in the post-implementation period for the system adopters. Clearly there were many potential confounding variables that may have influenced provider productivity in addition to adoption of the EHR, though our staggered implementation times statistically mitigates this effect somewhat. Our basic regression analysis did indicate that even when controlling for date of implementation, baseline provider visit volume, and subspecialty category, EHR adopters still had improved average monthly productivity after system adoption. While these findings are encouraging and strengthen the notion that EHR adoption was associated with higher productivity, there are many other important confounding variables that were difficult for us to quantify in any meaningful way. Our EHR implementation occurred over many years. During that time, our Physician Organization was very aggressive in promoting best practices that would presumably have positive impacts on provider proficiency. For example, the Physician Organization has commissioned several internal and external consulting engagements to identify opportunities for improved practice operations within the clinical departments. A great deal of effort has been made to optimize the use of our practice management system to achieve better utilization of provider patient-care schedules and higher quality patient registration data. During the study period, major investments were made in reporting and business intelligence tools that allowed practice managers to more efficiently manage all aspects of the revenue cycle. Lastly, during the same period as our EHR implementation, the institution initiated a formal office of billing compliance. The education provided to providers with regards to accurate E/M coding almost certainly had implications for the provider productivity markers considered in this study.
4.2.
New questions
The dissimilar nature of the EHR adopters and non-adopters is potentially telling in terms of the characteristics that are most likely to predict implementation failure. In our small cohort, those providers who failed to adopt the EHR tended to have lower patient visit volume, but higher baseline average monthly charges and wRVUs. One could argue that the higher baseline productivity markers in the adoption group reflected a more optimized subset of practitioners. This raises
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
Summary points What was already known on the topic? • There is a large body of evidence that demonstrates the clinical benefits of the electronic health record, particularly with regards to CPOE and decision support tools. Fewer studies have addressed the financial implications of EHR adoption in the ambulatory care setting. • Previous investigations have revealed meaningful return on investment for electronic health record implementations, but these analyses have focused on aggregate institutional costs and savings, rather than physician productivity. • Electronic health record adoption in the ambulatory setting has been impeded by provider fears of reduction in productivity. What this study added to our knowledge? • Adoption of a commercially available ambulatory EHR has the potential to increase physician productivity as measured by average monthly patient visit volume and provider work relative value units. • This analysis provided objective support of anecdotal observations that a “ramp up” period of approximately 6 months post-EHR-implementation exists. Providers achieve additional productivity gains beyond this adjustment period. • The findings provide evidence that physician productivity need not be harmed by the transition to the electronic health record. This will help overcome provider resistance to EHR adoption.
the question as to whether the EHR would only increase productivity for practitioner with a low or non-optimized baseline. It is worth noting the data trend that average monthly wRVUs decreased for non-adopters in the immediate post-implementation period. It is conceivable that this group of providers may have had their productivity disproportionately disrupted during the ramp-up period, leading ultimately to system abandonment. There is certainly a “chicken and egg” question that requires further exploration. Does the EHR dramatically influence provider productivity or do productivity characteristics of providers predict system adoption?
4.3.
Meaning and impact
With the U.S. Government’s enactment of the Health Information Technology for Economic and Clinical Health Act (HITECH) [13], there will be renewed interest in provider proficiency with the EHR as it relates to “meaningful use.” There have been previous attempts in the literature to describe a spectrum of EHR usage sophistication [14]. As objective models of user proficiency mature, it will present an opportunity to identify correlations between the quality of EHR usage and the magnitude of any resulting productivity changes.
499
As HIT adoption accelerates nationally, understanding the effects of the EHR on physician productivity may be more relevant than attempts to demonstrate recoupment of sunk costs. The very real benefits of the EHR in terms of structured data capture, decision support, and provider communication are difficult to quantify in financial terms. Given the overwhelming regulatory pressures and intra-operability demands facing the U.S. health care system, ongoing attempts to demonstrate the technology’s ability to pay for itself may be misplaced. Further limiting the actual value of these exercises is the cost of these analyses and the difficulty quantifying the “soft” ROI of improved clinical care. A robust heath information technology infrastructure is now part of the cost of doing business in modern medicine. However, our experience in a large academic multi-specialty physician group suggests the potential to implement the EHR and still derive a modest increase in overall physician productivity. The fact that the providers can exceed preimplementation productivity levels within 6 months of an EHR transition should hopefully alleviate some barriers to system adoption and help to further align the incentives of the individual practitioners and their associated care delivery organizations.
Acknowledgements The authors wish to acknowledge the outstanding contributions of Alison M. Edwards for her assistance with the statistical analyses in this manuscript. The authors also wish to acknowledge the efforts of the EHR implementation staff of the Weill Cornell Physician Organization. We would like to additionally thank the administrative leadership of the Weill Cornell Physician Organization for its unwavering support of health information technology and our efforts to evaluate its impact.
references
[1] D.W. Bates, Physicians and ambulatory electronic health records, Health Aff. 24 (September/October (5)) (2005) 1180–1189. [2] C.M. DesRoches, E.G. Campbell, R.R. Sowmya, et al., Electronic health records in ambulatory care—a national survey of physicians, N. Engl. J. Med. 359 (July (1)) (2008) 50–60. [3] D. Blumenthal, J.P. Glaser, Information technology comes to medicine, N. Engl. J. Med. 356 (June (24)) (2007) 2527–2534. [4] B. Chaudhry, J. Wang, S. Wu, et al., Systematic review: impact of health information technology on quality, efficiency, and costs of medical care, Ann. Intern. Med. 144 (2006), E-12–E-22. [5] R. Amarasingham, L. Plantinga, M. Diener-West, et al., Clinical information technologies and inpatient outcomes, Arch. Intern. Med. 169 (2) (2009) 108–114. [6] N. Menachemi, R.G. Brooks, Reviewing the benefits and costs of electronic health records and associated patient safety technologies, J. Med. Syst. 30 (2006) 159–168. [7] R. Kaushal, A.K. Jha, C. Franz, et al., Return on investment for a computerized physician order entry system, J. Am. Med. Inform. Assoc. 13 (May/June (3)) (2006) 261–266.
500
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 9 ( 2 0 1 0 ) 492–500
[8] S.J. Wang, B. Middleton, L.A. Prosser, et al., A cost-benefit analysis of electronic medical records in primary care, Am. J. Med. 114 (April) (2003) 397–403. [9] D.L. Grieger, S.H. Cohen, D.A. Krusch, A pilot study to document the return on investment for implementing an ambulatory electronic health record at an academic medical center, J. Am. Coll. Surg. 205 (July (1)) (2007) 89–96. [10] N. Menachemi, Barriers to ambulatory EHR: who are “imminent adopters” and how do they differ from other physicians? Inform. Prim. Care 14 (2006) 101–108.
[11] H.G. Lo, L.P. Newmark, C. Yoon, et al., Electronic health records in specialty care: a time-motion study, J. Am. Med. Inform. Assoc. September/October (14) (2007) 609–615. [12] Centers for Medicare & Medicaid Services, PFS Relative Value Files, 2006, http://www.cms.hhs.gov/ PhysicianFeeSched/PFSRVF/list.asp?listpage=2. [13] American Recovery and Reinvestment Act of 2009, H.R. 1, 111th Cong., 1st sess., 2009. [14] R.H. Miller, I. Sim, J. Newman, Electronic medical records in solo/small groups: a qualitative study of physician user types, Stud. Health Technol. Inform. 107 (1) (2004) 658–662.