EMR use is not associated with better diabetes care Patrick J. O’Connor, MD, MPH, A. Lauren Crain, PhD, Leif I. Solberg, MD, Stephen E. Asche, MA, William.

EMR use is not associated with better diabetes care

Patrick J. O’Connor, MD, MPH, A. Lauren Crain, PhD, Leif I. Solberg, MD, Stephen E. Asche, MA, William A. Rush, PhD, Robin R. Whitebird, PhD, MSW

Electronic medical record (EMR)

$10+ billion spent on EMR in last 5 years 300 EMR vendors (EMR Institute) Office EMRs now used by > 35% of

physicians Typical features of an EMR: High expectations that EMRs will improve

care quality since 1980; IOM reports 1992

Research Question Do patients receiving care at clinics using

EMRs have better quality of diabetes care, compared to patients receiving care at clinics not using EMRs?

Project Quest Multi-site 3 year study involving 19

medical groups, 85 clinics, 700 providers and 7865 adult DM or CHD patients

Designed to identify patient, physician, and clinic factors related to quality of care for adults with diabetes or heart disease

Funded by Agency for Healthcare Research and Quality (AHRQ)

Project Quest Diabetes Sample

Diabetes patients in 1998 (based on ICD-9 and pharmacy codes)

HealthPartners insurance in 1998 19+ yrs old in 1998 Returned patient survey Self-report confirmed having diabetes Consented to chart audit Linked to a clinic in which a clinic medical director

completed a survey N=1,491 DM patients from N=60 clinics

Data Sources Administrative data

Diabetes determination (based on diagnosis & pharmacy codes), limited demographic information

Patient survey (2000) Socio-demographic information

Clinic medical director survey (2000) Report on use of EMR Other clinic variables

Chart audit (1999, 2000, 2001) HbA1c, LDL, SBP (last in each year)

EMR item “Does your clinic use computerized

medical record systems that include provider entry of data” Asked of 60 clinic medical directors 14 / 60 (23.3%) replied “yes”

Diabetes patients at clinics with and without an EMR

EMR (n=441) Non-EMR (n=1050)

Age (mean)* 64.2 60.7

Female (%)* 51.5 43.8

Duration DM (mean)*

11.5 10.3

Charlson (mean)

1.6 1.4

* p < .05

Diabetes patients at clinics with and without an EMR

EMR Non-EMR

A1c (mean, sd)

7.3 (1.21)(n=359)

7.3 (1.34)(n=877)

LDL (mean, sd)

101.4 (30.1)(n=246)

101.8 (30.0)(n=680)

SBP(mean, sd)

132.5 (17.6)(n=397)

130.8 (17.3)(n=934)

Year 2001 clinical values. Bivariate analysis. * p < .05

Multilevel analysis Uses clinical values in all 3 years Models clinical value pooled across all 3 years,

and change in clinical values over time Models time within person within provider

within clinic (“clean” hierarchy) Used MLWin Patient covariates: age, sex, education, duration

of DM, Charlson score, CHD disease, BMI Provider covariate: physician specialty

Multilevel analysis: HbA1c and change in HbA1c

Coeff SE p

Intercept 7.31 - -

EMR present

-0.07 .11 .56

Patient and provider covariates included

Change over time analysis: LR test p=.14

Multilevel analysis: LDL and change in LDL

Coeff SE p

Intercept 106.4 - -

EMR present

0.1 1.7 .95

Patient and provider covariates included

Change over time analysis: LR test p=.37

Multilevel analysis: SBP and change in SBP

Coeff SE p

Intercept 128.8 - -

EMR present

1.18 .82 .15

Patient and provider covariates included

Change over time analysis: LR test p=.90

Strengths of Study

Large number of patients with diabetes Multiple data sources (patient, provider,

clinic medical director) Use of hierarchical analytic models to

accommodate nested data Uniform data collection procedures and

standards at all clinics

Potential Limitations Study only involved 60 clinics in one state, generalizability to other

regions or patient populations is uncertain Observational study precludes causal inference Clinic systems already in place Didn’t examine process measures as dependent variables (e.g.,

test rates) Clinic EMR use examined in isolation (no other clinic variables

considered in same analysis) We don’t have information on 1) features / functionality of the EMR,

2) extent to which EMR is used, 3) extent to which practitioners are trained to use the EMR

Some patients may link to multiple doctors, who link to multiple clinics, but we have simplified the hierarchy

Conclusions EMR use not associated with better

glucose, BP, or lipid control in adults with diabetes

Compare to Other Studies Meigs ’02 at Mass General Clinics—EMR

increased A1c tests but did not improve A1c level

Montori ’02 at Mayo—EMR improved number of A1c tests but did not improve A1c or LDL level

O’Connor ’01 at HPMG—EMR use led to more A1c tests, but worse A1c levels

Crabtree ’06 at NJ clinics—EMR using clinics no better than non-EMR for DM care

Implications

Anticipated benefits of very expensive EMRs for improving diabetes (and other chronic disease) care have yet to be realized

Office systems not yet redesigned to take advantage of EMR potential

Physician training to use EMRs not standardized or optimized

More research needed if the potential of very expensive EMRs to support better care is to be realized

Questions or Comments

[email protected]

Appendix slides

Diabetes identificationDiabetes identified using a method

with estimated sensitively of 0.91 and positive predictive value of 0.94. Data on A1c and CHD were obtained from a medial record review.

See paper draft for detail

Recruitment rates, sampling

QUEST successfully recruited: 19 of 22 eligible medical groups 85 of 86 eligible clinics within those

medical groups See paper draft for details on

sampling: 19 MG, all clinics in these MG, minimum of 10 pts per clinic (DM and CHD sample)

Survey response rates

Survey response rates of medical groups (100%), clinics (98%), providers (55%) and chart audit consent rate of patients responding to surveys (about 80%) exceeded levels needed to power the analysis.

Patient Factors Analyzed

*Age *Educational Level*Duration of Diabetes*Comorbidity*Gender*BMI

Physician Factors Analyzed

Years Experience (Post-Residency)

GenderSpecialty (FP, IM)

Measures of Clinic Systems in Clinic Surveys Expanded Roles for Nurses/Teams Registries Electronic Medical Records Monitoring of Clinical Status Prioritization based on Risk, RTC Active Interventions:

Visit Planning Active Outreach Patient Activation

Where is the Variance?

--80-90% of variance at Patient/Time level

--5% of variance at Physician level

--5% of variance at Clinic level--2-4% of variance at Medical

Group level