Association of Electronic Health Record Use Above Meaningful Use Thresholds With Hospital Quality and Safety Outcomes

This cross-sectional quality improvement study assesses whether electronic health record implementation above meaningful use thresholds is associated with changes in hospital patient satisfaction, efficiency, and safety.

.xlsx CMI = case-mix index, CMS = centers for Medicare and Medicaid, EHR = electronic health record, CHPL = certified health IT product list, NCHS = national center for health statistics.

Discharge Information
Patient-rated quality of information about help needed after discharge and signs to look out for. Reported as percent select best possible response adjusted for patient-mix. Overall Rating of Hospital Patient-rated overall quality of the hospital. Reported as percent select best possible response adjusted for patient-mix. Efficiency (spending) Medicare Spending Per Beneficiary (MSPB) Each hospital's price-standardized risk-adjusted Medicare spending per beneficiary per care episode divided by the national median Medicare spending per beneficiary per care episode. Reported as ratio, interpretable as spending percentage more or less than national median. Safety CLABSI (HAI-1) Central line associated blood stream infections (CLABSIs). Reported as standardized infection rate: ratio of number of infections divided by the number of predicted infections based on national multivariate regression analyses. Interpretable as percent more or less than expected based on national trends. CAUTI  Catheter associated urinary tract infections (CAUTIs). Reported as standardized infection rate: ratio of number of infections divided by the number of predicted infections based on national multivariate regression analyses. Interpretable as percent more or less than expected based on national trends. SSI-Colon (HAI-3) Surgical site infections (SSIs) after colon surgery. Reported as standardized infection rate: ratio of number of infections divided by the number of predicted infections based on national multivariate regression analyses. Interpretable as percent more or less than expected based on national trends. SSI-Abd Hyst  Surgical site infections (SSIs) after abdominal hysterectomy. Reported as standardized infection rate: ratio of number of infections divided by the number of predicted infections based on national multivariate regression analyses. Interpretable as percent more or less than expected based on national trends. MRSA Bacteremia  Methicillin resistant staph aureus bacteremia. Reported as standardized infection rate: ratio of number of infections divided by the number of predicted infections based on national multivariate regression analyses. Interpretable as percent more or less than expected based on national trends. C. diff Infection  Clostridioides difficile infection. Reported as standardized infection rate: ratio of number of infections divided by the number of predicted infections based on national multivariate regression analyses. Interpretable as percent more or less than expected based on national trends.            First, the Hospital Compare datasets for FY2016 and prior include only dimension point scores. These are the number of points awarded for each component, with 0 for being at or below an established baseline, 10 for being at or above an established threshold, and 1-9 points for achieving gradations between the baseline and threshold, which are then summed after also examining improvement since the baseline period and consistency across dimensions to yield the overall domain score. This discrete scale is less suited to analysis, and perhaps less interesting to analyze, than the percent of respondents selecting "top box" scores for each component. These continuous percentages, adjusted for patient-level factors detailed elsewhere, are available only in the FY2017 and FY 2018 datasets, representing CY 2015 and CY 2016 respectively. Thus, the most informative analysis may be performed by looking at the adjusted top-box percentages in 2015 and/or 2016.
Second, the Pain Management component is present in each dataset except for FY2018, when these questions were no longer asked. Conversely, the care transition domain was added in FY2018 and is not present in any other year. If we were to include both FY 2017 and FY 2018 in the analysis, the only way to include both of these components would be to impute their values during the year not asked. Given that this means 50% of the data for each of these components would be imputed, this presents a significant barrier to the validity of the model resulting from this outcome and thus we favor one year or the other and include the respective dimension for that year.
The Efficiency domain, consisting of the sole MSPB measure, is unchanged from its inclusion in the HVBP in FY2015. It has been reported by calendar year in the Hospital Compare datasets for much longer. This, then, gives us little direction in selecting years to analyze. However, in looking at the components of the domain scores themselves rather than the points awarded, we have believe that the MSPB may be a better outcome to analyze than the domain score. The Efficiency domain score represents points awarded for performing at, below, or between pre-established baseline and threshold scores. The MSPB measure itself represents each hospital's average price-standardized, risk adjusted spending per care episode divided by the national median of spending per episode. Using this measure avoids the problem of a large percent of the hospitals having 0 points (since scoring below 50th percentile of the baseline period earns 0 points). Using the MSPB directly will allow us to model across the entire spectrum of hospitals.
The Safety domain is composed of PSI-90 (itself a combination of several measures of accidents, injuries, and other adverse events), HAI 1-6 (rates of various kinds of hospital acquired infections), and PC-01 (rate of elective delivery before 39 weeks). As seen in supplement 4, the performance periods for PSI-90 are less regular than those for the Engagement and Efficiency domains and overlap significantly. While FY2015 has PSI-90 performance periods are contemporaneous only with the pre-2014 MU measures, FY2016 and on have performance periods that include both pre-2014 and post-2014 MU measures. Thus, the PSI-90 outcome is not conducive to analysis with post-2014 MU measures. While PSI-90 for FY2015 may be analyzed with pre-2014 MU measures this is a departure from the above considerations favoring the analysis of CY2015 and/or CY2016 data, and thus may be suited to separate analysis in the interest of keeping the current project cohesive. HAI 1-4 as reported in FY2016 to FY2018 have performance periods in CY2014 to CY2016, and similarly HAI 5-6 and PC-01 have performance periods in CY2015 to CY2016 for FY 2017 and FY 2018 respectively. While HAI 1-6 are reported as standardized infection ratios, PC-01 is reported as an unadjusted percentage, making it less amenable to