OBJECTIVE

To investigate the association between ambulatory medication reconciliation and health care utilization in patients with diabetes.

RESEARCH DESIGN AND METHODS

In this retrospective cohort analysis, we studied adults taking at least one diabetes medication treated in primary care practices affiliated with two academic medical centers between 2000 and 2014. We assessed the relationship between the fraction of outpatient diabetes medications reconciled over a 6-month period and the composite primary outcome of combined frequency of emergency department (ED) visits and hospitalizations over the subsequent 6 months.

RESULTS

Among 261,765 reconciliation assessment periods contributed by 31,689 patients, 176,274 (67.3%), 27,775 (10.6%), and 57,716 (22.1%) had all, some, or none of the diabetes medications reconciled, respectively. Patients with all, some, or no diabetes medications reconciled had 0.354, 0.377, and 0.384 primary outcome events per 6 months, respectively (P < 0.0001). In a multivariable analysis adjusted for demographics and comorbidities, having some or all versus no diabetes medications reconciled was associated with a lower risk of the primary outcome (rate ratio 0.94 [95% CI 0.90–0.98; P = 0.0046] vs. 0.92 [0.89–0.95; P < 0.0001], respectively). Introduction of feedback to individual providers was associated with a significant increase in the odds of all diabetes medications being reconciled (2.634 [2.524–2.749]; P < 0.0001).

CONCLUSIONS

A higher fraction of reconciled outpatient diabetes medications was associated with a lower frequency of ED visits and hospitalizations. Individual performance feedback could help to achieve more comprehensive medication reconciliation.

Medication discrepancies—unintended differences in medication information for a given patient between different sites of care—are common (16). Many medication discrepancies have the potential to lead to patient harm (3,5,7,8). Medication reconciliation, a standardized process of comparing a patient’s medication record to the medications the patient is actually taking, can reduce medication discrepancies (911). The World Health Organization has named medication reconciliation as one of its top five high-risk areas for standardization in patient safety (12), and medication reconciliation has been one of The Joint Commission’s National Patient Safety Goals since 2005 (13).

Although most studies of medication reconciliation have focused on the inpatient environment (10), inaccurate medication records remain a challenge in ambulatory settings as well (1416). Some studies have indicated that ambulatory medication reconciliation can improve accuracy of outpatient medication records (1719). However, whether ambulatory medication reconciliation can decrease health care utilization, particularly emergency department (ED) visits and hospitalizations (20), remains uncertain.

Patients with diabetes are at a particularly high risk for medication errors (21,22). Adverse drug events stemming from medications used to treat diabetes commonly lead to ED visits and hospitalizations (2325). We therefore studied whether ambulatory medication reconciliation in patients with diabetes is associated with a lower risk of ED visits and hospitalizations.

### Study Design

We performed a retrospective cohort study among patients with diabetes to determine the association between the fraction of outpatient diabetes medications reconciled over a specified period (see below) and a subsequent risk of ED visits and hospitalizations.

### Study Cohort

Adult patients with diabetes followed by primary care physicians affiliated with Brigham and Women’s Hospital (BWH) and Massachusetts General Hospital (MGH) between 1 January 2000 and 30 June 2014 were studied. Patients were included in the analysis if they fulfilled all the following criteria: 1) were at least 18 years old, 2) had documented evidence of diabetes (diabetes on the electronic medical record [EMR] problem list, HbA1c of at least 7.0% (53 mmol/mol), or at least three ICD-9 codes for diabetes), 3) were followed for at least 12 months during the study period, and 4) had taken at least one diabetes medication while they were followed. Patients with missing zip codes were excluded to enable adjustment for median household income by zip code. During the study period, BWH and MGH had programs that actively encouraged outpatient medication reconciliation since 2012, including individual provider performance feedback (26). These programs were not limited to patients with diabetes and/or diabetes medications. This study was approved by the Partners HealthCare System institutional review board (Somerville, MA), and the requirement for written informed consent was waived.

### Study Measurements

A patient was entered into the study on the last of the following dates: 1) first primary care note, 2) 18th birthday, 3) first documented evidence of diabetes, 4) first record of a diabetes medication, or 5) 1 January 2000. The patient was then followed to assess outpatient diabetes medication reconciliation until the first of the following: 1) last primary care note or 2) 30 June 2014. To account for the fact that effects of medication reconciliation were not expected to be observed over a prolonged period, the study follow-up period was divided into 6-month-long reconciliation assessment periods that started on 1 January and 1 July of each calendar year while the patient was in the study. A unique reconciliation assessment period served as the unit of analysis. Study outcomes were evaluated during the subsequent 6 months (outcome ascertainment period). For example, for a reconciliation assessment period between 1 January 2003 and 30 June 2003, the outcome ascertainment period was between 1 July 2003 and 31 December 2003. A single patient could contribute multiple reconciliation assessment and outcome ascertainment periods to the analysis.

The categorized fraction of diabetes medications reconciled during the reconciliation assessment period served as the predictor variable. A medication was considered reconciled if any of the following actions were performed in the EMR (in accordance with BWH and MGH policies): 1) medication record was edited, 2) prescription was generated, or 3) medication was confirmed on a special EMR reconciliation screen. Because the possible numeric fractions of medications reconciled highly depend on the number of medications in the denominator, the fraction of medications reconciled was represented in the analysis as a categorical variable with three possible values: none, some (which included all values not equal to 0 or 1) and all. Information on the dosage of medications prescribed or taken was available to the reconciling providers but was not used in these categorizations.

The combined number of hospitalizations or ED visits at BWH or MGH during the subsequent outcome assessment period served as the composite primary outcome. In addition to the composite primary outcome, 1) hospitalizations for any cause and 2) ED visits for any cause were analyzed as secondary outcomes. ED visits for any cause followed by a hospitalization were excluded to avoid double-counting; the combination of the two events was counted as a single hospitalization. All data for the study were obtained from the EMR at Partners HealthCare, an integrated health care delivery network founded by BWH and MGH.

### Statistical Analysis

Summary statistics were calculated by using frequencies and proportions for categorical data and means, SDs, and medians for continuous variables. In univariable analysis, the relationship between the fraction of diabetes medications reconciled and the study outcomes was evaluated using the Kruskal-Wallis test.

In multivariable analysis, a negative binomial model was used to account for overdispersion. This model was used to determine the association between the categorized fraction of diabetes medications reconciled and study outcomes while accounting for longitudinal repeated measurements within individual patients and providers’ random effects and adjusting for the patient’s demographics (age, sex, race/ethnicity, primary language, median household income by zip code, marital status, and health insurance status), study year, Charlson comorbidity index (CCI), treatment with insulin, categorized fraction of reconciled nondiabetes medications, smoking history, systolic and diastolic blood pressure, BMI, HbA1c level, logarithm of estimated glomerular filtration rate (eGFR), number of hospital admissions in the 12 months preceding the reconciliation assessment period, number of primary care visits during the reconciliation assessment period, and the logarithm of the number of diabetes and nondiabetes medications the patient was taking.

In addition, we analyzed the factors associated with the higher fraction of reconciled diabetes medications. We used a multinomial logistic regression model to determine the association between predictor variables (patient demographics, CCI, smoking history, blood pressure, BMI, HbA1c level, eGFR, number of diabetes medications, treatment with insulin, number of hospital admissions in the 12 months preceding the reconciliation assessment period, number of primary care visits during the reconciliation assessment period, and whether the reconciliation assessment period took place after reconciliation feedback was instituted at BWH and MGH) and the categorical outcome variable (none, some, or all) representing the fraction of diabetes medications reconciled while accounting for clustering within individual patients. Because patients who were only taking one diabetes medication could not have some (but only all or none) of their medications reconciled, their data were analyzed separately.

Multiple imputation procedure was used to account for missing data (blood pressure, BMI, HbA1c, and eGFR) both in the primary analysis and in the evaluation of the factors associated with medication reconciliation. Interpretation of P values was adjusted for multiple hypothesis testing using the Simes-Hochberg method (27,28). All analyses were performed with SAS 9.4 statistical software (SAS Institute, Cary, NC).

We identified 32,880 adult patients with diabetes who were followed in primary care settings for at least 1 year and were taking diabetes medications during the study period. We excluded 191 patients who did not have income by zip code information available. The remaining 31,689 patients represented by 261,765 reconciliation assessment periods were included in the analysis. Among these, 176,274 (67.3%), 27,775 (10.6%), and 57,716 (22.1%) reconciliation assessment periods had all, some, and no diabetes medications reconciled, respectively. Most (21,015 [66.3%]) of the study patients had a baseline HbA1c ≥7.0% (53 mmol/mol); they were followed on average for 57 months (Table 1). During the reconciliation assessment periods, patients had a median of four primary care visits and took a median of one diabetes medication (Supplementary Table 1).

Table 1

Baseline characteristics of study patients

VariableValue
Participants, n 31,689
Age (years), mean (SD) 59.4 (13.7)
Female sex 15,914 (50.2)
White race 18,724 (59.1)
Median household income by zip code ($1,000), mean (SD) 67.2 (25.5) English as the primary language 25,483 (80.8) Government insurance 16,457 (51.9) Married 15,408 (48.6) History of smoking 17,159 (54.2) Treated with insulin 12,305 (38.8) Follow-up (days), mean (SD) 1,757 (1,458) BMI (kg/m2), mean (SD) 32.0 (6.2) HbA1c, mean (SD) % 8.0 (1.9) mmol/mol 64 (11.8) SBP (mmHg), mean (SD) 130.7 (16.0) DBP (mmHg), mean (SD) 75.7 (9.8) eGFR (mL/min/1.73 m2), mean (SD) 76.0 (25.0) CCI, mean (SD) 6.2 (4.2) VariableValue Participants, n 31,689 Age (years), mean (SD) 59.4 (13.7) Female sex 15,914 (50.2) White race 18,724 (59.1) Median household income by zip code ($1,000), mean (SD) 67.2 (25.5)
English as the primary language 25,483 (80.8)
Government insurance 16,457 (51.9)
Married 15,408 (48.6)
History of smoking 17,159 (54.2)
Treated with insulin 12,305 (38.8)
Follow-up (days), mean (SD) 1,757 (1,458)
BMI (kg/m2), mean (SD) 32.0 (6.2)
HbA1c, mean (SD)
% 8.0 (1.9)
mmol/mol 64 (11.8)
SBP (mmHg), mean (SD) 130.7 (16.0)
DBP (mmHg), mean (SD) 75.7 (9.8)
eGFR (mL/min/1.73 m2), mean (SD) 76.0 (25.0)
CCI, mean (SD) 6.2 (4.2)

Data are n (%) unless otherwise indicated. DBP, diastolic blood pressure; SBP, systolic blood pressure.

### Diabetes Medication Reconciliation and ED Visits and Hospitalizations

Over the subsequent 6-month outcome assessment period, patients with all, some, or no diabetes medications reconciled had a mean of 0.354, 0.377, and 0.384 composite primary outcome events, respectively (P < 0.0001). Over the same period (Fig. 1), patients with all, some, or no diabetes medications reconciled had 0.223, 0.232, and 0.230 ED visits (P = 0.0296) and 0.132, 0.145, and 0.154 hospitalizations (P < 0.0001), respectively. No consistent relationship was observed for reconciliation of nondiabetes medications. Among 258,526 reconciliation assessment periods with at least one nondiabetes medication, periods with no, some, and all nondiabetes medications reconciled were followed by a mean of 0.278, 0.397, and 0.335 primary outcome events, respectively.

Figure 1

Diabetes medication reconciliation and rate of hospitalizations and ED visits. A: Primary composite outcome. B: Hospitalizations. C: ED visits. Error bars indicate SE.

Figure 1

Diabetes medication reconciliation and rate of hospitalizations and ED visits. A: Primary composite outcome. B: Hospitalizations. C: ED visits. Error bars indicate SE.

Close modal

In multivariable analysis, a larger fraction of diabetes medication reconciled remained associated with lower frequency of the composite primary outcome (Table 2). Compared with patients who had no diabetes medications reconciled, those who had some of their diabetes medications reconciled had a rate ratio (RR) of the primary outcome events of 0.94 (0.90–0.98; P = 0.0046), and patients who had all diabetes medications reconciled had an RR of 0.92 (0.89–0.95; P < 0.0001). Findings for secondary outcomes were similar (Table 2). Other patient characteristics associated with lower frequency of the composite primary outcome events were female sex, lower HbA1c, better kidney function, lower CCI, no insulin therapy, lower frequency of primary care visits, and absence of hospitalizations in the preceding 12 months.

Table 2

Effect of patient characteristics and medication reconciliation on frequency of ED visits and hospitalization

Combined ED visits and hospitalizations
ED visits
Hospitalizations
VariableRR (95% CI)P valueRR (95% CI)P valueRR (95% CI)P value
Agea (years)
<50 2.027 (1.935–2.124) <0.0001 2.064 (1.970–2.162) <0.0001 1.690 (1.597–1.789) <0.0001
50–64 1.242 (1.198–1.288) <0.0001 1.230 (1.186–1.276) <0.0001 1.194 (1.144–1.246) <0.0001
65–74 0.978 (0.947–1.011) 0.1859 0.944 (0.913–0.976) 0.0008 1.046 (1.007–1.086) 0.0212
Female 0.955 (0.932–0.979) 0.0002 0.974 (0.950–0.999) 0.4170 0.888 (0.862–0.915) <0.0001
White 0.903 (0.878–0.929) <0.0001 0.835 (0.811–0.860) <0.0001 1.035 (1.000–1.072) 0.0499
Government insurance 1.169 (1.141–1.199) <0.0001 1.177 (1.147–1.207) <0.0001 1.130 (1.096–1.165) <0.0001
Englishb 1.087 (1.052–1.123) 0.005 1.066 (1.032–1.102) 0.0001 1.108 (1.064–1.155) <0.0001
Married 0.865 (0.845–0.886) <0.0001 0.829 (0.809–0.849) <0.0001 0.931 (0.904–0.958) <0.0001
Incomec 0.945 (0.940–0.950) <0.0001 0.937 (0.932–0.943) <0.0001 0.963 (0.957–0.970) <0.0001
History of smoking 1.066 (1.041–1.091) <0.0001 1.070 (1.044–1.096) <0.0001 1.036 (1.006–1.067) 0.0168
SBPd 0.999 (0.998–1.00) 0.0268 0.999 (0.998–1.000) 0.0049 1.000 (1.000–1.001) 0.6072
DBPe 1.00 (0.998–1.001) 0.5556 1.001 (0.999–1.002) 0.2943 1.000 (1.000–1.000) 0.0089
BMIf 1.004 (1.002–1.006) <0.0001 1.002 (0.999–1.004) 0.0056 1.000 (0.998–1.002) 0.9172
HbA1c 1.036 (1.029–1.044) <0.0001 1.034 (1.026–1.042) <0.0001 1.047 (1.037–1.056) <0.0001
eGFRg 0.854 (0.833–0.875) <0.0001 0.882 (0.860–0.903) <0.0001 0.732 (0.712–0.753) <0.0001
CCI 1.172 (1.168–1.176) <0.0001 1.144 (1.139–1.148) <0.0001 1.171 (1.166–1.76) <0.0001
Treated with insulin 1.271 (1.238–1.305) <0.0001 1.259 (1.225–1.294) <0.0001 1.286 (1.245–1.329) <0.0001
Primary care visits 1.038 (1.035–1.040) <0.0001 1.038 (1.036–1.041) <0.0001 1.027 (1.024–1.030) <0.0001
Number of nondiabetes medicationsg 1.446 (1.414–1.478) <0.0001 1.450 (1.417–1.484) <0.0001 1.455 (1.415–1.497) <0.0001
Number of diabetes medicationsg 0.864 (0.838–0.890) <0.0001 0.861 (0.834–0.888) <0.0001 0.879 (0.847–0.913) 0.0283
Hospitalizations in the preceding 12 months 1.382 (1.365–1.399) <0.0001 1.318 (1.304–1.333) <0.0001 1.354 (1.338–1.370) <0.0001
Nondiabetes medications reconciledh
Some 1.011 (0.968–1.055) 0.6330 1.012 (0.968–1.059) 0.5950 1.094 (0.956–1.65) 0.7340
All 1.136 (1.081–1.194) <0.0001 1.127 (1.070–1.888) <0.0001 1.177 (1.105–1.254) <0.0001
Diabetes medications reconciledi
Some 0.938 (0.898–0.980) 0.0046 0.927 (0.886–0.970) 0.0010 0.942 (0.893–0.994) 0.0283
All 0.920 (0.892–0.950) <0.0001 0.917 (0.888–0.947) <0.0001 0.917 (0.883–0.952) <0.0001
Combined ED visits and hospitalizations
ED visits
Hospitalizations
VariableRR (95% CI)P valueRR (95% CI)P valueRR (95% CI)P value
Agea (years)
<50 2.027 (1.935–2.124) <0.0001 2.064 (1.970–2.162) <0.0001 1.690 (1.597–1.789) <0.0001
50–64 1.242 (1.198–1.288) <0.0001 1.230 (1.186–1.276) <0.0001 1.194 (1.144–1.246) <0.0001
65–74 0.978 (0.947–1.011) 0.1859 0.944 (0.913–0.976) 0.0008 1.046 (1.007–1.086) 0.0212
Female 0.955 (0.932–0.979) 0.0002 0.974 (0.950–0.999) 0.4170 0.888 (0.862–0.915) <0.0001
White 0.903 (0.878–0.929) <0.0001 0.835 (0.811–0.860) <0.0001 1.035 (1.000–1.072) 0.0499
Government insurance 1.169 (1.141–1.199) <0.0001 1.177 (1.147–1.207) <0.0001 1.130 (1.096–1.165) <0.0001
Englishb 1.087 (1.052–1.123) 0.005 1.066 (1.032–1.102) 0.0001 1.108 (1.064–1.155) <0.0001
Married 0.865 (0.845–0.886) <0.0001 0.829 (0.809–0.849) <0.0001 0.931 (0.904–0.958) <0.0001
Incomec 0.945 (0.940–0.950) <0.0001 0.937 (0.932–0.943) <0.0001 0.963 (0.957–0.970) <0.0001
History of smoking 1.066 (1.041–1.091) <0.0001 1.070 (1.044–1.096) <0.0001 1.036 (1.006–1.067) 0.0168
SBPd 0.999 (0.998–1.00) 0.0268 0.999 (0.998–1.000) 0.0049 1.000 (1.000–1.001) 0.6072
DBPe 1.00 (0.998–1.001) 0.5556 1.001 (0.999–1.002) 0.2943 1.000 (1.000–1.000) 0.0089
BMIf 1.004 (1.002–1.006) <0.0001 1.002 (0.999–1.004) 0.0056 1.000 (0.998–1.002) 0.9172
HbA1c 1.036 (1.029–1.044) <0.0001 1.034 (1.026–1.042) <0.0001 1.047 (1.037–1.056) <0.0001
eGFRg 0.854 (0.833–0.875) <0.0001 0.882 (0.860–0.903) <0.0001 0.732 (0.712–0.753) <0.0001
CCI 1.172 (1.168–1.176) <0.0001 1.144 (1.139–1.148) <0.0001 1.171 (1.166–1.76) <0.0001
Treated with insulin 1.271 (1.238–1.305) <0.0001 1.259 (1.225–1.294) <0.0001 1.286 (1.245–1.329) <0.0001
Primary care visits 1.038 (1.035–1.040) <0.0001 1.038 (1.036–1.041) <0.0001 1.027 (1.024–1.030) <0.0001
Number of nondiabetes medicationsg 1.446 (1.414–1.478) <0.0001 1.450 (1.417–1.484) <0.0001 1.455 (1.415–1.497) <0.0001
Number of diabetes medicationsg 0.864 (0.838–0.890) <0.0001 0.861 (0.834–0.888) <0.0001 0.879 (0.847–0.913) 0.0283
Hospitalizations in the preceding 12 months 1.382 (1.365–1.399) <0.0001 1.318 (1.304–1.333) <0.0001 1.354 (1.338–1.370) <0.0001
Nondiabetes medications reconciledh
Some 1.011 (0.968–1.055) 0.6330 1.012 (0.968–1.059) 0.5950 1.094 (0.956–1.65) 0.7340
All 1.136 (1.081–1.194) <0.0001 1.127 (1.070–1.888) <0.0001 1.177 (1.105–1.254) <0.0001
Diabetes medications reconciledi
Some 0.938 (0.898–0.980) 0.0046 0.927 (0.886–0.970) 0.0010 0.942 (0.893–0.994) 0.0283
All 0.920 (0.892–0.950) <0.0001 0.917 (0.888–0.947) <0.0001 0.917 (0.883–0.952) <0.0001

DBP, diastolic blood pressure; SBP, systolic blood pressure.

aAge >74 years served as the comparison group.

bRecorded as the primary language.

cMedian household income by zip code by $10,000. dSBP by 10 mmHg. eDBP by 10 mmHg. fBMI by 1 kg/m2. gParameterized as a natural logarithm. hNo nondiabetes medications reconciled served as the comparison group. iNo diabetes medications reconciled served as the comparison group. In a secondary analysis, the association between the composite primary outcome and reconciliation of insulin and sulfonylureas only was weaker than the association with reconciliation of all diabetes medications; the RR for reconciliation of all insulins and sulfonylureas was 0.95 (95% CI 0.92–0.98; P = 0.002) and for reconciliation of some insulins and sulfonylureas 0.97 (0.95–1.00; P = 0.34). An analysis of the difference of the effect of reconciliation of diabetes medications before versus after 2012 (when individual feedback to clinicians was introduced) showed no statistically significant difference (0.98; 95% CI 0.90–1.06; P = 0.58). ### Predictors of Diabetes Medication Reconciliation In multivariable analysis of factors associated with a higher fraction of reconciled diabetes medications among 124,408 reconciliation assessment periods with multiple diabetes medications, feedback to individual providers was associated with a significant increase in the probability of all diabetes medications being reconciled at the expense of the lower probability of only some diabetes medications being reconciled (Table 3). Other factors associated with higher probability of all diabetes medications being reconciled included a greater number of primary care visits, nonwhite race, primary language other than English, and higher HbA1c level. The results were similar among the 133,357 reconciliation assessment periods with only one diabetes medication (Table 3). Table 3 Effect of patient characteristics on diabetes medication reconciliation Reconciliation assessment periods with more than one diabetes medication Reconciliation assessment periods with one diabetes medication All diabetes medications reconciled Some diabetes medications reconciled All medications reconciled VariableOR (95% CI)P valueOR (95% CI)P valueEstimate (95% CI)P value Agea (years) <50 0.966 (0.932–1.002) 0.0659 0.936 (0.897–0.977) 0.0024 0.935 (0.906–0.965) <0.0001 50–64 1.071 (1.044–1.098) <0.0001 1.06 (1.029–1.091) 0.0001 1.023 (1.001–1.045) 0.039 65–74 1.067 (1.037–1.098) <0.0001 1.004 (0.971–1.037) 0.8303 1.034 (1.009–1.058) 0.006 Female sex 0.978 (0.946–1.011) 0.1848 1.053 (1.014–1.094) 0.0078 1.031 (1.003–1.06) 0.029 White 0.778 (0.749–0.807) <0.0001 0.851 (0.815–0.889) <0.0001 0.872 (0.846–0.899) <0.0001 Government insurance 1.045 (1.009–1.081) 0.0129 1.05 (1.009–1.092) 0.0167 0.97 (0.942–0.998) 0.039 Englishb 0.716 (0.684–0.75) <0.0001 0.899 (0.852–0.948) <0.0001 0.804 (0.774–0.834) <0.0001 Married 0.921 (0.892–0.951) <0.0001 0.99 (0.953–1.027) 0.5864 0.993 (0.966–1.02) 0.596 Incomec 0.983 (0.976–0.989) <0.0001 0.989 (0.982–0.997) 0.0067 0.972 (0.967–0.977) <0.0001 History of smoking 1.088 (1.054–1.124) <0.0001 1.073 (1.034–1.114) 0.0002 1.08 (1.052–1.11) <0.0001 SBPd 1.005 (0.993–1.017) 0.3895 0.999 (0.986–1.013) 0.8984 0.985 (0.975–0.995) 0.003 DBPe 0.964 (0.944–0.983) 0.0003 0.982 (0.959–1.005) 0.1219 0.998 (0.981–1.015) 0.789 BMIf 1.015 (1.012–1.017) <0.0001 1.01 (1.007–1.013) <0.0001 1.008 (1.006–1.01) <0.0001 HbA1c 1.047 (1.036–1.059) <0.0001 1.021 (1.009–1.034) 0.0009 1.043 (1.032–1.053) <0.0001 eGFRg 1.274 (1.223–1.327) <0.0001 1.223 (1.166–1.283) <0.0001 1.079 (1.042–1.117) <0.0001 CCI 0.988 (0.983–0.992) <0.0001 1.001 (0.995–1.006) 0.7883 0.983 (0.979–0.987) <0.0001 Treated with insulin 0.861 (0.832–0.89) <0.0001 1.005 (0.966–1.046) 0.8147 0.911 (0.88–0.943) <0.0001 Number of diabetes medications 0.933 (0.906–0.961) <0.0001 1.731 (1.679–1.786) <0.0001 NA NA Primary care visits 1.091 (1.086–1.095) <0.0001 1.041 (1.036–1.046) <0.0001 1.052 (1.049–1.056) <0.0001 With provider feedback 2.634 (2.524–2.749) <0.0001 0.56 (0.529–0.593) <0.0001 2.304 (2.225–2.385) <0.0001 Hospitalizations in the preceding 12 months 0.935 (0.917–0.954) <0.0001 0.926 (0.905–0.947) <0.0001 0.972 (0.955–0.989) 0.002 Reconciliation assessment periods with more than one diabetes medication Reconciliation assessment periods with one diabetes medication All diabetes medications reconciled Some diabetes medications reconciled All medications reconciled VariableOR (95% CI)P valueOR (95% CI)P valueEstimate (95% CI)P value Agea (years) <50 0.966 (0.932–1.002) 0.0659 0.936 (0.897–0.977) 0.0024 0.935 (0.906–0.965) <0.0001 50–64 1.071 (1.044–1.098) <0.0001 1.06 (1.029–1.091) 0.0001 1.023 (1.001–1.045) 0.039 65–74 1.067 (1.037–1.098) <0.0001 1.004 (0.971–1.037) 0.8303 1.034 (1.009–1.058) 0.006 Female sex 0.978 (0.946–1.011) 0.1848 1.053 (1.014–1.094) 0.0078 1.031 (1.003–1.06) 0.029 White 0.778 (0.749–0.807) <0.0001 0.851 (0.815–0.889) <0.0001 0.872 (0.846–0.899) <0.0001 Government insurance 1.045 (1.009–1.081) 0.0129 1.05 (1.009–1.092) 0.0167 0.97 (0.942–0.998) 0.039 Englishb 0.716 (0.684–0.75) <0.0001 0.899 (0.852–0.948) <0.0001 0.804 (0.774–0.834) <0.0001 Married 0.921 (0.892–0.951) <0.0001 0.99 (0.953–1.027) 0.5864 0.993 (0.966–1.02) 0.596 Incomec 0.983 (0.976–0.989) <0.0001 0.989 (0.982–0.997) 0.0067 0.972 (0.967–0.977) <0.0001 History of smoking 1.088 (1.054–1.124) <0.0001 1.073 (1.034–1.114) 0.0002 1.08 (1.052–1.11) <0.0001 SBPd 1.005 (0.993–1.017) 0.3895 0.999 (0.986–1.013) 0.8984 0.985 (0.975–0.995) 0.003 DBPe 0.964 (0.944–0.983) 0.0003 0.982 (0.959–1.005) 0.1219 0.998 (0.981–1.015) 0.789 BMIf 1.015 (1.012–1.017) <0.0001 1.01 (1.007–1.013) <0.0001 1.008 (1.006–1.01) <0.0001 HbA1c 1.047 (1.036–1.059) <0.0001 1.021 (1.009–1.034) 0.0009 1.043 (1.032–1.053) <0.0001 eGFRg 1.274 (1.223–1.327) <0.0001 1.223 (1.166–1.283) <0.0001 1.079 (1.042–1.117) <0.0001 CCI 0.988 (0.983–0.992) <0.0001 1.001 (0.995–1.006) 0.7883 0.983 (0.979–0.987) <0.0001 Treated with insulin 0.861 (0.832–0.89) <0.0001 1.005 (0.966–1.046) 0.8147 0.911 (0.88–0.943) <0.0001 Number of diabetes medications 0.933 (0.906–0.961) <0.0001 1.731 (1.679–1.786) <0.0001 NA NA Primary care visits 1.091 (1.086–1.095) <0.0001 1.041 (1.036–1.046) <0.0001 1.052 (1.049–1.056) <0.0001 With provider feedback 2.634 (2.524–2.749) <0.0001 0.56 (0.529–0.593) <0.0001 2.304 (2.225–2.385) <0.0001 Hospitalizations in the preceding 12 months 0.935 (0.917–0.954) <0.0001 0.926 (0.905–0.947) <0.0001 0.972 (0.955–0.989) 0.002 DBP, diastolic blood pressure; NA, not applicable; OR, odds ratio; SBP, systolic blood pressure. aAge >74 years served as the comparison group. bRecorded as the primary language. cMedian household income by zip code by$10,000.

dSBP by 10 mmHg.

eDBP by 10 mmHg.

fBMI by 1 kg/m2.

gParameterized as a natural logarithm.

In this large retrospective cohort study, we found that higher rates of reconciliation of diabetes medications were associated with lower frequency of hospitalizations and ED visits and that higher rates of medication reconciliation were observed after introduction of performance feedback at the individual provider level. These findings have important implications for both individual clinician practice of medicine and health policy.

Medication reconciliation is an important patient safety measure. It can reduce medication discrepancies, potential adverse drug events, and actual adverse drug events (9,10,19). However, the effect of inpatient medication reconciliation on health care utilization remains uncertain (9,10), and little information is available on the relationship between health care utilization and ambulatory medication reconciliation. To our knowledge, the current study is the first to show a strong association between higher fraction of reconciled outpatient medications and lower rates of ED visits and hospitalization, an important finding in this era of intensifying efforts for controlling health care costs. Although the analysis was retrospective in nature and, therefore, cannot provide definitive evidence of a causal link, the finding of a dose-response relationship between diabetes medication reconciliation and both the composite primary outcome and hospitalizations offers an additional degree of reassurance.

Medication reconciliation is not without its own costs. The medication reconciliation process typically takes 15–20 min and sometimes as long as 30–60 min (2932). The resulting high cost of nurse/pharmacist time spent on verifying the patient’s medication information is a known barrier to medication reconciliation (33,34). This barrier can be even higher in the ambulatory setting where 15–20 min required for medication reconciliation can consume the entire time slot allotted for the patient’s encounter with the clinician. Consequently, both individual providers and health system administrators must be aware of the balance between the costs and downstream benefits of the medication reconciliation process. The current study provides some of the information necessary to illuminate this balance.

Medication reconciliation can take different forms. The system used at the institutions included in this analysis during the study period required reconciliation at the individual medication level. However, many EMR systems allow wholesale reconciliation through a click of a button, which attests that the provider reviewed all the patient’s medications. Our findings, therefore, may not be generalizable to institutions that use a significantly different process for medication reconciliation.

The current analysis only found health care utilization benefits with reconciliation of diabetes medications. This could be related to some diabetes medications having low therapeutic indices, and incorrect dosing can easily lead to adverse events (e.g., hypoglycemia) that require emergency health care services. In a national study, insulin was the second most common medication implicated in adverse drug events leading to an ED visit by older patients (24). Similarly, insulins were the second most common and oral diabetes medications the fourth responsible for adverse drug events leading to hospitalizations among older patients (25). Reconciliation of diabetes medications could therefore decrease the risk of the patient taking an incorrect dose and prevent an adverse drug event that otherwise would have led to a hospitalization or an ED visit, explaining our findings. Although we could not find a consistent relationship between ED visits/hospitalizations and reconciliation of nondiabetes medications as a whole, reconciliation of certain other high-risk medication classes possibly is associated with reduced health care utilization; this should be studied further.

Little is known about factors that influence medication reconciliation. The current study shows that a quality improvement program that provides medication reconciliation performance feedback to individual clinicians and compares them with their peers (initiated in 2012 at both study institutions [separately]) is associated with a significant increase in reconciliation. This finding demonstrates a systemwide effect on medication reconciliation rates and provides a blueprint for successful implementation. A higher frequency of primary care visits also is linked to higher rates of medication reconciliation likely because every visit provides an additional opportunity for reconciliation (which in this analysis was evaluated over a 6-month period). On the other hand, several demographic characteristics of patients, specifically white race and English as the primary language, were associated with lower medication reconciliation rates. The reasons for this finding are uncertain and deserve additional investigation.

Risk of ED visits and hospitalizations among patients with diabetes may be affected by multiple factors in addition to medication reconciliation. One such factor is the risk of hypoglycemia associated with medications used to treat diabetes. In this study, the number of diabetes medications the patient was taking was inversely related to frequency of both ED visits and hospitalizations. Because most diabetes medications, apart from sulfonylureas and insulin, have a lower risk of hypoglycemia, this finding may be an indirect indication of the additional risk of adverse events associated with hypoglycemia-prone medications. On the other hand, the analysis did not have sufficient power to establish whether the observed decrease in health care utilization with higher diabetes medication reconciliation rates was specifically due to diabetes-related events (i.e., whether it was due to a change in the incidence of hypoglycemia). As a result, the nature of the links between evidence of reconciliation and the medical outcomes the analysis uncovered is not fully understood.

The current study had a number of strengths. It was a large analysis of >30,000 demographically diverse patients. Furthermore, the analysis included periods when medication reconciliation was not explicitly encouraged (before its introduction as a National Patient Safety Goal by The Joint Commission in 2005), was encouraged at the institutional level (2005–2011), and was encouraged at the individual provider level (after 2012), allowing for greater generalizability across various health care settings.

The study's findings, however, must be interpreted in light of its limitations. As a retrospective analysis, the study cannot provide definitive evidence of a causative relationship but only an association. Its conclusions are necessarily tentative and need confirmation. Data available for analysis only included hospitalizations and ED visits at institutions within a single health care system. Health care utilization outside this system was not included, and as a result, the reported frequencies of ED visits and hospitalizations are likely underestimates. We also did not have information on whether medication reconciliation was accurate, other patient safety metrics, and other processes of care relevant to patients with diabetes, such as foot or eye examinations. The study did not include an econometric analysis and, therefore, could not determine cost-effectiveness of medication reconciliation. Finally, because the study was limited to two academic medical centers in eastern Massachusetts, the findings may not be generalizable outside these settings. Future directions of research could include investigations of whether other approaches to medication reconciliation (e.g., attestation to reconciliation of an entire list of medications by means of a single checkbox) have a similar relationship with patient outcomes. An important area of study would be whether reconciliation of other classes of medications (e.g., anticoagulants, opioids) has a similarly pronounced relationship with ED visits or hospitalizations.

In conclusion, this analysis shows that higher rates of reconciliation of diabetes medications are associated with lower health care utilization in the acute hospital setting, the most expensive kind. At the same time, the finding of a significant increase in medication reconciliation rates after institution of individual performance feedback demonstrates how these higher rates can be achieved. These results, therefore, provide actionable information to both individual providers and health care system administrators about reconciliation of diabetes medications.

Funding. Funding was provided by the Patient-Centered Outcomes Research Institute (CE1304-6756).

The funding source had no direct effect on the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.T. supervised the study and contributed to the acquisition of data. A.T., O.S., H.Z., and M.S. contributed to the study concept and design and analysis and interpretation of the data. A.T., O.S., H.Z., M.S., S.P.D., D.C.S., and M.A.T. contributed to the critical revision of the manuscript for important intellectual content. A.T. and H.Z. drafted the manuscript. O.S., H.Z., and M.S. contributed to the statistical analysis. D.C.S. and M.A.T. obtained funding. A.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 76th Scientific Sessions of the American Diabetes Association, New Orleans, LA, 10–14 June 2016.

1.
Cornish
PL
,
Knowles
SR
,
Marchesano
R
, et al
.
Unintended medication discrepancies at the time of hospital admission
.
Arch Intern Med
2005
;
165
:
424
429
[PubMed]
2.
Gleason
KM
,
Groszek
JM
,
Sullivan
C
,
Rooney
D
,
Barnard
C
,
Noskin
GA
.
Reconciliation of discrepancies in medication histories and admission orders of newly hospitalized patients
.
Am J Health Syst Pharm
2004
;
61
:
1689
1695
[PubMed]
3.
Pippins
JR
,
Gandhi
TK
,
Hamann
C
, et al
.
Classifying and predicting errors of inpatient medication reconciliation
.
J Gen Intern Med
2008
;
23
:
1414
1422
[PubMed]
4.
Tam
VC
,
Knowles
SR
,
Cornish
PL
,
Fine
N
,
Marchesano
R
,
Etchells
EE
.
Frequency, type and clinical importance of medication history errors at admission to hospital: a systematic review
.
CMAJ
2005
;
173
:
510
515
[PubMed]
5.
Wong
JD
,
Bajcar
JM
,
Wong
GG
, et al
.
Medication reconciliation at hospital discharge: evaluating discrepancies
.
Ann Pharmacother
2008
;
42
:
1373
1379
[PubMed]
6.
Azzi
M
,
Constantino
M
,
Pont
L
,
Mcgill
M
,
Twigg
S
,
Krass
I
.
Medication safety: an audit of medication discrepancies in transferring type 2 diabetes mellitus (T2DM) patients from Australian primary care to tertiary ambulatory care
.
Int J Qual Health Care
2014
;
26
:
397
403
[PubMed]
7.
Rubin
DJ
,
Donnell-Jackson
K
,
Jhingan
R
,
Golden
SH
,
Paranjape
A
.
Early readmission among patients with diabetes: a qualitative assessment of contributing factors
.
J Diabetes Complications
2014
;
28
:
869
873
[PubMed]
8.
Persell
SD
,
Bailey
SC
,
Tang
J
,
Davis
TC
,
Wolf
MS
.
Medication reconciliation and hypertension control
.
Am J Med
2010
;
123
:
182.e9
182.e15
9.
Mueller
SK
,
Sponsler
KC
,
Kripalani
S
,
Schnipper
JL
.
Hospital-based medication reconciliation practices: a systematic review
.
Arch Intern Med
2012
;
172
:
1057
1069
[PubMed]
10.
Kwan
JL
,
Lo
L
,
Sampson
M
,
Shojania
KG
.
Medication reconciliation during transitions of care as a patient safety strategy: a systematic review
.
Ann Intern Med
2013
;
158
:
397
403
[PubMed]
11.
Mekonnen
AB
,
Abebe
TB
,
McLachlan
AJ
,
Brien
JA
.
Impact of electronic medication reconciliation interventions on medication discrepancies at hospital transitions: a systematic review and meta-analysis
.
BMC Med Inform Decis Mak
2016
;
16
:
112
[PubMed]
12.
Leotsakos
A
,
Zheng
H
,
Croteau
R
, et al
.
Standardization in patient safety: the WHO High 5s project
.
Int J Qual Health Care
2014
;
26
:
109
116
[PubMed]
13.
The Joint Commission. National Patient Safety Goals [Internet], 2018. Available from https://www.jointcommission.org/standards_information/npsgs.aspx. Accessed 6 June 2018
14.
Bedell
SE
,
Jabbour
S
,
Goldberg
R
, et al
.
Discrepancies in the use of medications: their extent and predictors in an outpatient practice
.
Arch Intern Med
2000
;
160
:
2129
2134
[PubMed]
15.
Kaboli
PJ
,
McClimon
BJ
,
Hoth
AB
,
Barnett
MJ
.
Assessing the accuracy of computerized medication histories
.
Am J Manag Care
2004
;
10
:
872
877
[PubMed]
16.
Frank
C
,
Godwin
M
,
Verma
S
, et al
.
What drugs are our frail elderly patients taking? Do drugs they take or fail to take put them at increased risk of interactions and inappropriate medication use
?
Can Fam Physician
2001
;
47
:
1198
1204
[PubMed]
17.
Nassaralla
CL
,
Naessens
JM
,
Chaudhry
R
,
Hansen
MA
,
Scheitel
SM
.
Implementation of a medication reconciliation process in an ambulatory internal medicine clinic
.
Qual Saf Health Care
2007
;
16
:
90
94
[PubMed]
18.
Weingart
SN
,
Cleary
A
,
Seger
A
, et al
.
Medication reconciliation in ambulatory oncology
.
Jt Comm J Qual Patient Saf
2007
;
33
:
750
757
[PubMed]
19.
Varkey
P
,
Cunningham
J
,
Bisping
DS
.
Improving medication reconciliation in the outpatient setting
.
Jt Comm J Qual Patient Saf
2007
;
33
:
286
292
[PubMed]
20.
Bayoumi
I
,
Howard
M
,
Holbrook
AM
,
Schabort
I
.
Interventions to improve medication reconciliation in primary care
.
Ann Pharmacother
2009
;
43
:
1667
1675
[PubMed]
21.
Cox
AR
,
Ferner
RE
.
Prescribing errors in diabetes
.
Br J Diabetes Vasc Dis
2009
;
9
:
84
88
22.
Gurwitz
JH
,
Field
TS
,
Harrold
LR
, et al
.
Incidence and preventability of adverse drug events among older persons in the ambulatory setting
.
JAMA
2003
;
289
:
1107
1116
[PubMed]
23.
van der Hooft
CS
,
Sturkenboom
MC
,
van Grootheest
K
,
Kingma
HJ
,
Stricker
BH
.
Adverse drug reaction-related hospitalisations: a nationwide study in The Netherlands
.
Drug Saf
2006
;
29
:
161
168
[PubMed]
24.
Budnitz
DS
,
Shehab
N
,
Kegler
SR
,
Richards
CL
.
Medication use leading to emergency department visits for adverse drug events in older adults
.
Ann Intern Med
2007
;
147
:
755
765
[PubMed]
25.
Budnitz
DS
,
Lovegrove
MC
,
Shehab
N
,
Richards
CL
.
Emergency hospitalizations for adverse drug events in older Americans
.
N Engl J Med
2011
;
365
:
2002
2012
[PubMed]
26.
Keogh
C
,
Kachalia
A
,
Fiumara
K
,
Goulart
D
,
Coblyn
J
,
Desai
SP
.
Ambulatory medication reconciliation: using a collaborative approach to process improvement at an academic medical center
.
Jt Comm J Qual Patient Saf
2016
;
42
:
186
194
[PubMed]
27.
Simes
RJ
.
An improved Bonferroni procedure for multiple tests of significance
.
Biometrika
1986
;
73
:
751
754
28.
Hochberg
Y
.
A sharper Bonferroni procedure for multiple tests of significance
.
Biometrika
1988
;
75
:
800
802
29.
Barnsteiner
JH
.
Medication reconciliation: transfer of medication information across settings-keeping it free from error
.
Am J Nurs
2005
;
105
(
Suppl.
):
31
36
; quiz 48–51
[PubMed]
30.
Pronovost
P
,
Weast
B
,
Schwarz
M
, et al
.
Medication reconciliation: a practical tool to reduce the risk of medication errors
.
J Crit Care
2003
;
18
:
201
205
[PubMed]
31.
Vira
T
,
Colquhoun
M
,
Etchells
E
.
Reconcilable differences: correcting medication errors at hospital admission and discharge
.
Qual Saf Health Care
2006
;
15
:
122
126
[PubMed]
32.
Murphy
EM
,
Oxencis
CJ
,
Klauck
JA
,
Meyer
DA
,
Zimmerman
JM
.
Medication reconciliation at an academic medical center: implementation of a comprehensive program from admission to discharge
.
Am J Health Syst Pharm
2009
;
66
:
2126
2131
[PubMed]
33.
Schenkel
S
.
The unexpected challenges of accurate medication reconciliation
.
Ann Emerg Med
2008
;
52
:
493
495
[PubMed]
34.
Miller
SL
,
Miller
S
,
Balon
J
,
Helling
TS
.
Medication reconciliation in a rural trauma population
.
Ann Emerg Med
2008
;
52
:
483
491
[PubMed]
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.