Despite electronic health records' (EHRs') perceived potential to improve care, the 2011 National Ambulatory Medical Care Survey revealed that only 55 percent of U.S. physicians had adopted an EHR (Jamoom et al. 2012). Of these physicians, 74 percent believe it enhances patient care (Jamoom et al. 2012). However, physicians frequently cite financial barriers to adoption: the lack of reliable information about return on investment, decreased productivity, and that costs are borne by the practice but most potential savings accrue to payers (Boonstra and Broekhuis 2010; Police, Foster, and Wong 2010; Yan, Gardner, and Baier 2012). The context for EHR adoption has changed substantially with the Medicare and Medicaid Meaningful Use incentives, and it is now frequently viewed as inevitable (Song et al. 2011). Nevertheless, the experience of practices that adopted EHRs prior to these incentives remains relevant, providing information on the impact on productivity, volume, staffing, and income (Song et al. 2011).
We previously reported the hardware, software, and time-and-effort costs of implementing a commercially developed EHR in 26 primary care practices (Fleming et al. 2011). We now report the impact on productivity, staffing, and financial measures in those same practices.
- Top of page
- Supporting Information
While expenses do increase and productivity does decrease following EHR implementation, the effects are not as large or persistent as might be expected. We saw increased practice expenses of approximately $1,650 per physician FTE per month, which corresponds closely to the monthly $1,425 per physician cost of EHR maintenance ($1,225 for software licensing, networking, and hosting and technical support through a third-party vendor, and $200 of additional maintenance personnel support) reported in the evaluation of the costs associated with HTPN's EHR implementation (Fleming et al. 2011). The other contributing expense was the 3 percent increase in staffing.
We also observed decreased productivity and patient volume postimplementation, but by 12 months, performance was only approximately 4 percent below preimplementation, suggesting full recovery could be achieved. Net income per physician also decreased but rebounded, suggesting little long-term detriment. As intensity of services did not change significantly, the initial drop in net income was likely related to increased expenses and decreased productivity, rather than changes in case mix.
Decreased payment per work RVU could cause decreased net income per physician. However, this decrease was offset by an increasing secular trend, suggesting that annual contracting rates play a determinative role. The overall lack of change in payment per work RVU during our study suggests a flattening of reimbursement rates (best seen in Table 2 for years 2007–2009), such that these should have little impact on changes in the net income measures.
A decrease in net income could result from lower net collection—for example, if integrating the EHR with the practice management system proved problematic, affecting practices' ability to collect the charges billed. This is unlikely here: HTPN continued using its preexisting practice management system, which was not integrated with the EHR; the codes/charges for services rendered continued to be entered manually. Furthermore, collection typically only affects income from self-pay patients, which constitute only 3 percent of HTPN patients.
One potential issue in interpreting our results is the impact of the 2008 recession. Its effect was mitigated by (but also confounded with) the early versus late adopter variable in our models. The 1-year postimplementation activity for early adopters occurred before the recession, and a subgroup analysis found their productivity and net income recovered completely at >12 months. The late adopters did not recover during the study period, but we cannot tell how much of this difference is attributable to differences between early and late adopters versus the recession. In addition, since the impact of the recession was not uniform across the United States (Semega 2009), the effects may not be broadly generalizable.
Our results show both consistencies and inconsistencies with similar evaluations. Like the 2010 MGMA Electronic Health Record Impacts on Revenue, Costs, and Staffing report, we saw increased costs for equipment and software maintenance and support staff (Gans 2010). We did not observe that the revenue increases MGMA reports, which could be attributable to our shorter postimplementation time (<2 years vs. >5 years). One academic pediatric primary care center reported increased revenue at 2 years postimplementation, but this was explained by a 13 percent increase in coding for detailed level visits and a 7 percent decrease for problem-focused visits (Samaan et al. 2009). While we did not look at visit coding, visit intensity did not change significantly, suggesting no similar change. This is interesting given the concern that EHRs result in visit “upcoding” (Abelson and Creswell 2013). Our results suggest this does not necessarily follow when the practice management system remains separate and consistent with that used prior to EHR implementation. However, lack of integration between the EHR and practice management system may also represent an opportunity missed to achieve greater productivity and efficiency through more efficient charge capture, and reduced time and effort for manual charge entry.
Differences in preimplementation practice management systems and coding practices may also explain why we did not see the significantly higher average monthly patient visits and work RVUs per physician with EHR adoption reported by the Weill Cornell physician group (Cheriff et al. 2010). They also observed a significant decrease in work RVU per visit, leading to speculation that the EHR captured and recorded low-intensity services that were previously undocumented and unbilled, making it appear that volume and productivity had increased (Cheriff et al. 2010). Finally, EHR implementation in five University of Rochester Medical Center ambulatory care practices had a neutral impact on efficiency and billing (similar to the small effects we saw on visits per physician FTE and payments received per work RVU), but total savings of $14,055 per provider, ongoing annual savings of $9,983 per provider, and a reduction of 1 staff FTE per physician FTE (Grieger, Cohen, and Krusch 2007). The savings were primarily due to reductions in chart pulls and support staff (Grieger, Cohen, and Krusch 2007). Since we did not separate chart pulls from the activities accounted for under “staff per physician FTE,” it is hard to compare our results. However, the opposite effect on support staff needs suggests that they play different roles in the practices, and detail on their activities is needed to truly determine the impact of the EHR.
Our observational study carries the inherent limitation of possible imbalance in unobserved differences that confound the outcomes. However, all practices ultimately implemented the EHR, and ≥12 months of data pre- and postimplementation were included for each practice, minimizing any imbalance. In addition, the interrupted time series design reduces the threat of historical events and selection bias by looking at an intervention occurring at different times across all included practices (Mercer et al. 2007).
The other substantial limitation is that we evaluated a product from a single vendor in a single network, meaning our results may not be generalizable to dissimilar settings—for example, managed care—or products. When HTPN chose its EHR in 2004–2005, GE Centricity Office typically ranked in the top three products, as reported through such mechanisms as the Towards the Electronic Patient Record annual conference. In 2006, when implementation began, it was the most widely used ambulatory care EHR in the United States and ranked fifth by KLAS (Enrado 2006). As such, HTPN's experience should have been similar or superior to that of other practices implementing an EHR. Like other high-ranked products, GE Centricity Office incorporated CPOE, e-prescribing, and electronic documenting and charting, and had the ability to interface with disparate hospital systems (Enrado 2006); it differed in that it did not integrate the billing and scheduling applications. GE Centricity Office's successor was rated as “average” across all categories at the end of our study period (KLAS).
Having studied only primary care practices also limits the generalizability. While subspecialties that function similarly to primary care might have a similar experience, those that do not (e.g., surgical subspecialties with high patient volumes, which would magnify any per visit time losses or gains; or cardiology, which relies heavily on ancillary testing services) may see very different effects. Furthermore, we evaluated an EHR geared toward general ambulatory care rather than any of the specialty-specific EHRs that have been developed.
Neither our results nor other recent reports show persistent substantial decreases in productivity or financial performance following EHR implementation. Nevertheless, the physicians whose bottom lines are affected may find the changes practically significant and must weigh them against the evidence of EHRs' impact on quality of care (Crosson et al. 2012; Reed et al. 2012; Walsh et al. 2012). Future research should examine staff roles and visit capture by paper-based records: if practices considering EHR implementation can identify which models their practices follow, they will be better able to predict the EHR's impact. Research is also needed on long-term effects: additional experience may enable practices to realize gains in productivity and net income.