The Comparative Value of 3 Electronic Sources of Medication Data

, , , , , , ,
The American Journal of Pharmacy Benefits, September/October 2014, Volume 6, Issue 5

Combining medication history information from multiple sources improves the completeness and accuracy of the medication information.

Transitions between care settings are a vulnerable time for patients. Patients are at high risk for adverse drug events (ADEs) during transitions of care, including many ADEs that are preventable.1,2 On average, patients admitted to a general hospital unit are at risk for 1.4 potential ADEs and the main cause is errors in the preadmission history.3 Another study found that 15 admitted patients out of every 100 experienced an ADE, 75% of which were preventable.4

One source of preventable ADEs are unintentional medication discrepancies,5 which are unexplained changes in a patient’s medication list. For example, discontinuation of medications for chronic disease during hospitalization is common and may increase the odds for patient harm.6 Unintentional discrepancies are often a result of incomplete or inaccurate medication histories.7 Omission of a medication has been found to be the most common error upon hospital admission,8 affecting 60% to 67% of patients.1 Generating an accurate medication list at the time of admission can help maintain medication accuracy throughout a patient’s hospital stay and decrease ADEs.9

Medication reconciliation is a process that can help eliminate unintentional medication discrepancies by creating an accurate list of all medications a patient is taking.10 The value of reconciliation is limited by the amount of information that is accurate, clinically relevant, and accessible at the time of medication reconciliation.9 Lack of integrated medication history information across settings and inadequate time to find the accurate information are important causes of medication discrepancies.11 Obtaining easily accessible, accurate medication history information is extremely desirable because it could reduce medication discrepancies and ultimately improve patient safety.12

Recently, both the Joint Commission and the Agency for Healthcare Research and Quality called for a way to reconcile medications more completely and accurately across the continuum of care.11 Today, healthcare is moving toward a greater level of sharing information across sites as a way to improve quality, safety, and efficiency.13 Indeed, the federal government has allocated over half a billion dollars through The State Health Information Exchange Cooperative Agreement Program to support state efforts to rapidly develop infrastructure to support health information exchange (HIE) within and between states.14 In addition to states and the federal government, communities and private funders are investing in technologies with the capability of collecting and sharing health information, including medication history information, across settings.14-16 Given the potential cost and resources associated with community HIE, understanding the value of combining medication history information sources from different sites is important to inform future investments and public health policies.17

OBJECTIVES

We undertook this case study to assess the accuracy and completeness of medication history information in 3 different electronic sources within a community. We also assessed the incremental value, defined as increased accuracy and completeness, of combining different sources of electronic medication history information.

METHODSStudy Design and Setting

We conducted this retrospective study from September 2010 through April 2011 at Albany Memorial Hospital, a 165-bed facility, and Samaritan Hospital, a 238-bed facility.

Both are located in upstate New York and are members of St. Peter’s Health Partners, a not-for-profit healthcare organization that also includes 7 ambulatory clinics. Notably, hospitalized patients are drawn primarily from this network of 7 clinics, which share a common electronic health record (EHR).

Medication History Information Sources

Medication history information was obtained from 3 sources: the healthcare organization’s EHR, a commercial medication database, and a community-wide HIE web portal.

The healthcare organization’s EHR is a commercially available system that contains electronic records for patients who visited the healthcare organization’s hospitals or 1 of its 7 ambulatory care clinics. The EHR was implemented in 1989 and has been customized to fit the hospital’s work flow. Iterative refinements have been made to improve the medication reconciliation process, including integrating electronic prescription or medication history information entered during visits to the local ambulatory care clinics or the hospital.

The commercial medication database contains data from a national e-prescribing network that provided prescription claims information and prescription fill information from dispensing pharmacy systems. As part of a pilot project to improve the medication reconciliation process, the commercial medication database was made available to pharmacists at Albany Memorial Hospital and Samaritan Hospital in 2010 and 2011. The commercial medication database was used to validate the existing medication history records and provide pharmacy fill information to help assess medication adherence among patients.

The community HIE portal was offered through a Regional Health Information Organization (RHIO). This HIE contained 1.6 million unique patient records at the time of the study. Patient records are constructed from information gathered from local hospitals, physician practices, national laboratories, imaging centers, nursing homes, home health agencies, and health plans participating in the RHIO. Some of the data in the hospital EHR are not transmitted as discreet data to the HIE but are instead stored in unstructured text fields. Information stored in unstructured fields (eg, a reconciled medication list) may not be shared with other providers via the HIE. Patient health information is only viewable to authorized users in organizations within the RHIO if patients consent to have their information accessed. In total, 2025 users from participating healthcare organizations had access to the portal during the study period, including more than 400 physicians, as well as nurses and pharmacists.

Study Population

We retrospectively reviewed records for patients who were aged more than 18 years, were admitted to 1 of the 2 hospitals, and had previously consented to have providers view their clinical information via the web portal. The institutional review boards at St. Peter’s Health Partners and Weill Cornell Medical College approved this study.

Data Collection

A trained research pharmacist from each hospital retrospectively reviewed the medication history information records in the 3 data sources for eligible patients. This included medication name, dose, frequency, route, and patient allergy information for both over-the-counter and prescription medications. The information in each source was compared with the gold standard—a validated list of the patient’s medications taken at home that had been generated as part of the routine intake medication reconciliation process. The medication reconciliation process is completed by pharmacists who interview patients to obtain a list of current medications. This list was then compared with all available information sources, including reviewing the healthcare organization’s EHR medical records and, if needed, calling the patient’s physician(s) and pharmacy(s) to confirm accuracy of the gold-standard medication list. Creation of a gold-standard medication list by a pharmacist has been used in a previous study.18

For the study, the research pharmacists evaluated the completeness and accuracy of each medication and allergy entry in the 3 sources of medication history information. Completeness was determined by assigning a score from 0 to 2 that represented the completeness of essential medication data for each patient on at least 1 essential medication within each data source. Essential medications were medications in 1 of 4 classes deemed fundamental for appropriate treatment of hospitalized patients: cardiovascular medications, anti-infectives, hypoglycemic agents, and anticoagulants. A score of 0 indicated that no essential medications were listed for that patient, a score of 1 indicated that some were listed, and a score of 2 indicated that all essential medications were listed.

Accuracy was assessed by determining the number of medications that each source correctly recorded compared with the gold standard as determined above. Accuracy was assessed both at the medication level for all medications and by class for the 11 most prevalent medication classes in the study. We assessed accuracy of allergy information by determining whether the medication source correctly recorded patients’ allergy information compared with the gold standard.

Data Analysis

We compared patient characteristics among the 2 hospital samples to confirm appropriateness of merging the data sources. We used a t test for age and x2 tests for sex and ethnicity.

For our analyses of medication information, we compared data from each of the 3 electronic sources with the gold standard. We evaluated 3 measures as percentages for each of the electronic sources: (1) the completeness of medication information for essential medications, (2) the accuracy of medication information, and (3) the accuracy of allergy information. Completeness of information for essential medications was collected at the patient level. For this measure, we reported the percentage of patients having no information on essential medications, partial information, or complete information in each source. Accuracy of medication information was collected at the medication level, whereby we determined whether each medication in the gold standard list was present in the electronic source and, if so, whether it was captured correctly. The accuracy of patient allergy information was assessed similarly, but at the patient level. Both accuracy measures were binary (accurate or inaccurate) at the patient or medication level, and are reported as percentages for each electronic medication source. We assessed medication accuracy overall and by class.

To assess the incremental value of combining electronic data sources, we pooled data from the 3 sources to obtain the most complete medication list. For each combination of sources, we considered the completeness of the essential medication information for each patient to be a maximum score (0, 1, or 2) for each individual source. Medications and allergies were considered accurate if at least 1 of the contributing data sources captured the information correctly. We examined the incremental value of adding (1) the commercial medication database to the community portal, (2) the community portal to the EHR, (3) the commercial medication database to the EHR, and (4) the community portal and the commercial medication database to the EHR. Because we reported for each source individually, results for the combined sources are expressed as percentages. All analysis was conducted using SAS version 9.2 (SAS Institute Inc, Cary, North Carolina).

Protection of Human and Animal Subjects

The study was performed in compliance with the World Medical Association Declaration of Helsinki on Ethical Principles for Medical Research Involving Human Subjects, and was reviewed by the institutional review boards at St. Peter’s Hospital and Weill Cornell Medical College.

RESULTS

A total of 858 patients were enrolled into this study (

Table 1

). Similar numbers of patients were seen at each hospital. The mean age of patients was 65 years (±18 years), 57% (n = 488) were female, and 87% (n = 744) were white.

Completeness of Essential Medication Information

The hospital EHR was most complete, capturing all essential medications for 71% (n = 611) of patients. The commercial medication database had all essential medications listed for 47% of patients (n = 400), and the community portal had all essential medication listed for 36% of patients (n = 312).

We assessed the incremental value of combining medication history information sources. By pooling information from all 3 electronic sources, we were able to increase the percentage of patients with all essential medications listed to 85% (n = 726). If only 2 sources of information were combined, the hospital EHR plus the commercial medication database contained all essential medications for 83% of patients (n = 715), the hospital EHR plus community portal contained all essential medications for 80% of patients (n = 682), and the community portal plus the commercial prescription source contained all essential medications for 52% of patients (n = 446) (

Figure 1

).

Accuracy of Medications

There were 7731 total medications among the 858 patients. The hospital EHR captured 80% (n = 6152) of medications accurately, the commercial medication database captured 45% (n = 3464), and the community portal captured 37% (n = 2838). Twenty-three percent of medications (n = 1781) were found accurately in all 3 sources, and 9% of medications (n = 734) were not found in any source.

When all 3 sources were pooled, accuracy was 91% (n = 6997). When pooling only 2 sources of data, the hospital EHR plus commercial medication database was the most accurate (89%, n = 6892) (

Figure 2

).

Accuracy of Medications by Class

We identified 11 classes of medications with more than 200 occurrences and quantified the percentage of medications captured accurately in each data source (

Table 2

). Overall, the medication accuracy in each of the top 11 most frequent drug classes was high, especially for the hospital EHR, which had more than 77% accuracy for every category except vitamins (69% accuracy). The level of accuracy by medication class was fewer for both the commercial prescription source and the community portal (range: 61%-13% and 50%-14%, respectively; Table 2), with both sources having the highest accuracy for diuretics and the lowest for vitamins.

The incremental value of adding sources of medication data by class was assessed. Adding medication history data from the commercial medication database and the portal to the hospital EHR did increase accuracy in all categories. This increase was at least 10% in the following classes: cardiovascular, central nervous system, autoimmune, diuretics, and narcotics. In all but 2 of the medication classes (electrolytes and vitamins medications), the hospital EHR plus the commercial medication database was equally or more accurate than the hospital EHR plus the community portal.

Accuracy of Allergy Information

The majority of patients’ allergies were accurately captured by the hospital’s EHR (94%, n = 806) and the community portal (90%, n = 769), but not the commercial medication database (5%, n = 46). The community portal captured 100% of the allergy information available in the commercial source. When we pooled allergy information, the hospital EHR plus the community portal was the most complete source of patient allergy information (97%, n = 829).

DISCUSSION

Our case study found that the most robust source of medication history information was the hospital EHR, which captured all essential medication information for 71% of patients and was accurate for 80% of medications. Patients’ allergy information was 97% complete when data from the hospital EHR and community portal were combined. Combing sources of medication history information improved the accuracy and completeness of essential medications and allergy information. To our knowledge, our study is the first to assess the incremental value of adding sources of medication history information to a hospital’s EHR system, and may offer insight to communities considering using HIE to increase medication information at the point of care.

Giving providers access to information that is accurate and complete during the prescribing process may improve patient safety by improving patients’ recall of their own medication history and allergy information, as well as by reducing medication discrepancies.19 Combining at least 2 sources of medication history information resulted in largely complete and accurate information for classes of medications identified by others as associated with high risk for preventable ADEs, including hypoglycemic and blood flow medications.3 Likewise, combining medication history information resulted in nearly complete allergy information, and allergic reactions are a common cause of ADEs.20 Reducing preventable ADEs by decreasing unintentional medication discrepancies and increasing knowledge of patients’ allergies may increase patient safety and decrease hospital costs.21

Of important incidental value was the pharmacy fill information contained in the commercial medication database. These data provided pharmacists with an objective marker of medication adherence. Indeed, evidence suggests that medication nonadherence is a common problem, is difficult to detect, and contributes to poorer healthcare outcomes and increased cost.22 In fact, one study found that 21% of preventable ADEs were due to nonadherance.23 The usefulness of identifying medications prescribed, but not filled, through different sources of medication information may have greater worth than what is suggested by the results of this analysis.

Poor care coordination across settings is a major concern,24 and inadequate clinician communication across settings is common.25 Further, preventable ADEs that occur at hospital discharge are more often the result of inaccurate medication reconciliation at hospital admission than inadequate reconciliation of hospital orders.3 Accurate medication reconciliation by a pharmacist using a medication list created from existing electronic sources is an important part of interventions to reduce medical errors.18 To create an accurate medication list upon admission, a “best possible medication history” should be obtained, which includes a patient interview and verification of the medication list from reliable sources of information.26 Through this study, we are able to provide information about the accuracy and completeness of various sources of information.

To increase patient safety, healthcare is moving toward sharing knowledge, including pharmacy fill information, across settings.13,27-29 Sharing clinical data has the potential to improve safety and lower costs.13,29-32 Other early adopters of HIT have also integrated medication history sources at the point of care. Partners Health Care in Boston created a medication reconciliation process that shares information across settings to facilitate reconciliation and reduce medication errors.32 For clinics affiliated with the University of Indiana, the Regenstrief Medication Hub was created to provide pharmacy dispensing information to providers who use their electronic prescribing system.33

Sharing information within communities has challenges, including identifying resources to support information sharing, uncertainly surrounding return on investment, and privacy and security concerns.34 Moreover, obtaining pharmacy fill information from commercial sources requires resources including payment of fees associated with purchasing access to a commercial database portal and integration of medication data into an existing HIE.

Despite these challenges, maximizing patient safety is essential both to protect the well-being of patients and to reduce healthcare costs.30 A study found that the increased adjusted cost of an ADE in a community hospital is $3420.35 Although acquiring the technology that facilitates HIE around medication safety may require a financial investment at the outset, reduction in ADEs may potentially reduce costs over the long term.

Limitations

Our case study has several limitations. First, the hospital’s EHR system is a long-standing commercial system with many iterative improvements made to increase the effectiveness of medication reconciliation, limiting generalizability. Second, this was a study of patients admitted to a single nonprofit healthcare system for which we have limited patient data. However, our overall sample of 858 patients is relatively large compared with previous work looking at medication discrepancies.2,3,18 Third, the majority of patients in the studied region stay within the same system for both their inpatient and outpatient care. Thus, our results may be less generalizable to regions with more fragmented healthcare utilization and without a unified EHR. Fourth, we used pharmacist medication review as the gold standard; however there may have been inaccuracies with the pharmacist reconciliation that we were unable to account for with our study design. Lastly, the value of data in any source is very dependent on the robustness of data available. Since the completion of our study, many more organizations have begun contributing data to the HIE, potentially closing the gap between the HIE and other health information technology as it relates to complete medication history information.

CONCLUSION

Given the significant financial investments being made in HIE to improve patient safety, our data support the value of combining medication history information from multiple sources to improve the completeness and accuracy of information. Communication of accurate medication history information across settings is essential to reduce medication discrepancies and prevent ADEs during transitions between care settings. Communities and policy makers can use information generated through this study in their decision-making processes about HIE investments.

Acknowledgments

We would like to acknowledge the contributions of Michael Fitzgerald, adoption specialist for the Healthcare Information Xchange of New York (Hixny) during the study period, who managed the tactical aspects of the project including project documentation, implementation, training, and end user support. We would also like to acknowledge Mike Dunay, Robert White, and Bob Duthe from the Management Information System at Northeast Health for their input on project design. Finally, thank you to the members of the Department of Pharmacy staff at Albany Memorial and Samaritan Hospitals, all of whom were involved in data collection.

The analysis of this study was conducted as part of the Health Information Technology Evaluation Collaborative (HITEC), an academic consortium designated by the state of New York as the evaluation entity for health information technology projects funded under the Health Care Efficiency and Affordability Law for New Yorkers (HEAL NY) capital grants program. HITEC is a formal academic collaborative of researchers at Weill Cornell Medical College, Columbia University, the University of Rochester, and the State University of New York at Albany and was supported by the New York State Department of Health (NYS contract number C023699).