The study cohort consisted of patients admitted to 425 hospitals that adopted EMR during the period 2002 to 2005, and 283 hospitals that did not adopt EMR over the eight years from 2000 to 2007. Table presents patient and disease characteristics in the EMR and non-EMR hospitals. We found that patients were slightly more likely to be male, be Black and have higher DRG weight in EMR hospitals than in non-EMR hospitals. However, age and comorbidity were higher in non-EMR hospitals than in EMR hospitals. Also, because of the large sample size, all other variables were statistically different between EMR and non-EMR hospitals (all p
0.01). Thus, it is more important to focus on the magnitude of any differences rather than their level of statistical significance.
Characteristics of patients and disease two years before and after EMR adoption, based on 425 EMR-adopting hospitals
First, we examined changes in outcomes before (8 quarters) and after (12 quarters) EMR adoption in the 425 hospitals that adopted EMR during the period 2002 to 2005 (Figure ). For each hospital, the year of EMR adoption was designated as year 0. We used piecewise linear regression to assess if there was a change in slope in any of the outcomes over the entire 20 quarters. For 30-day rehospitalization rate and 30-day mortality, there was a significant change in slope over time (p
0.001 for each). In both cases, the cutpoint was in the first quarter of the year of EMR implementation. As shown in Table , the odds of 30-day rehospitalization was stable in the eight quarters prior to EMR adoption and increased at 0.00037 odds per quarter after. Similarly, the odds of 30-day mortality was stable in the eight quarters prior to EMR adoption and decreased at 0.00062 odds per quarter after. There were no significant changes in slope for length of stay or inpatient mortality over the 20 quarters. We used a similar approach with piecewise linear regression to determine the temporal changes in outcomes in the 283 hospitals that did not adopt EMR over the 2000–2007 period. For length of stay, inpatient mortality and rehospitalization rate, there were underlying temporal trends noted in the hospitals that did not adopt EMR within the 2002–2005 window. To address the concern of underline temporal trend, we conducted further analyses focusing on the differences in changes in outcomes over time between EMR adopted and non-EMR adopted hospitals.
Outcomes in EMR adopted group two years before and two years after EMR adoption year.
Slope difference before and after EMR adoption, based on 425 EMR-adopting hospitals
Accordingly, we compared changes over time in the outcomes for hospitals that adopted EMR compared to those that did not. Once again, hospitals that adopted EMR during 2002–2005 were stacked so that the year of EMR adoption was year 0. We randomly assigned hospitals that did not adopt EMR to the same years and stacked them in similar fashion. We calculated the difference in each outcome between the EMR and non-EMR hospitals in the eight quarters before and the eight quarters after the year of EMR adoption (Table ). For example, in Table , the rate of 30-day rehospitalization in EMR hospitals in the eight quarters prior to EMR adoption was 0.46 percent less than in non-EMR hospitals. In the two years after EMR adoption, this difference decreased to −0.26 percent. Thus, relative to non-EMR hospitals, patients in hospitals that adopted EMR had a 0.19 percent higher rate of rehospitalization in the two years after EMR adoption (p
0.01). We also found that EMR was associated with shorter length of stay and lower 30-day mortality, even though the outcome changes associated with EMR adoption were small. Hospitals that adopted the EMR system experienced a 0.11 decrease in length of stay (P
0.05) and 0.18 percent decrease in 30-day mortality (P
However, there were no significant changes in inpatient mortality before and after EMR implementation. For inpatient mortality, length of stay may be an important factor because patients with critical conditions are more likely to both stay longer and die during hospitalization [13
]. Thus, we also compared the inpatient mortality rate after adjusting length of stay with other variables mentioned above, but found no significant difference in outcomes between EMR and non-EMR adopted hospitals.
Moreover, we stratified DRGs into surgical and medical based on the DRG [14
] and compared outcomes between EMR and non-EMR adopting hospitals (Table ). We found that the effect of EMR on outcomes differed by DRG type. EMR reduced the inpatient morality rate in surgical DRGs, but it increased 30-day mortality. In medical DRGs, however, EMR increased length of stay and 30-day rehospitalization but reduced 30-day mortality.
We also compared outcomes of all DRGs across hospital characteristics including number of beds, teaching status and ownership (Table ). Thus, each outcome was analyzed for four stratified sizes (by number of beds), two teaching states and three types of ownership. Across hospital characteristics, we found that the hospitals with large bed size, teaching and profit ownership status are more likely adopt EMR. However, we did not find any association between hospital characteristics and outcomes before and after EMR adoption (all p
0.1), except for ownership and 30-day rehospitalization.