Article Text

Medication safety program reduces adverse drug events in a community hospital
Free
1. M M Cohen,
2. N L Kimmel,
3. M K Benage,
4. M J Cox,
5. N Sanders,
6. D Spence,
7. J Chen
1. Missouri Baptist Medical Center, BJC HealthCare, St Louis, Missouri, USA
1. Correspondence to:  Dr M M Cohen  Chief Medical Officer, NYU Hospitals Center, 550 First Avenue, HCC 1514, New York, NY 10016, USA; max.cohenmed.nyu.edu

Abstract

Background: There is widespread interest in improving medication safety, particularly in the hospital setting. Numerous suggestions have been made as to how this should be done, but there is a paucity of data demonstrating the effectiveness of any of the interventions that have been proposed.

Objectives: To assess the impact of a wide ranging, community hospital based patient safety program on patient harm as measured by the rate of adverse drug events.

Design: An audit of discharged hospital patients was conducted from January 2001 to December 2003. Baseline data were collected for the first 6 months and multiple drug protocols and other interventions were instituted on the nursing units and in the pharmacy department over the subsequent 9 months (transition period). These interventions were largely based on information about medication risks acquired from internal medication event reporting. Each month of the study adverse drug events (ADE) were sought from a random sample of inpatient charts. A trigger tool was used to detect clues to ADEs, the presence of which was confirmed or excluded by detailed manual chart review. The severity of these events was categorized using the classification system of the National Coordinating Council for Medication Error and Reporting and Prevention.

Main outcome measures and results: Median ADEs per 1000 doses of medication dispensed declined significantly from 2.04 to 0.65 (p<0.001). Median ADEs per 100 patient days declined significantly from 5.07 to 1.30 (p<0.001). The proportion of inpatients with one or more ADE in the baseline period was 31% and declined threefold (p<0.001). The severity of reported medication events also declined. The number of ADEs associated conclusively with patient harm was 1.67 per total doses delivered in the baseline period and declined eightfold (p<0.001).

Conclusion: The implementation of a carefully planned series of low cost interventions focused on high risk medications, driven by information largely from internal event reporting, and designed to improve a hospital’s medication safety leads to a significant decrease in patient harm.

• medication safety program

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Awareness and scrutiny of medical errors, particularly in the hospital setting, has increased substantially in recent years. Attention has focused on medication errors because these represent the largest single component of all medical errors and many are life threatening.1 A wide range of relatively simple steps has been proposed that hospitals can take to reduce the likelihood of medication errors,2–4 but these interventions—while common sense and intrinsically sound—have not been subjected to formal study and there are few data to support their effectiveness. Seventy nine suggested safety practices were recently reviewed5 and 11 of these were highly rated for the weight of evidence in their support. Of these 11, only one was specifically designed to reduce the likelihood of a medication error.

More sophisticated and substantially more expensive solutions to the problem of medication errors—such as computer assisted management,6 computer based alerts,7 computerized physician order entry,8 advanced monitoring,9 bar coding10 and robotics11—have been proposed, but their impact on the incidence of adverse drug events has not been consistently and reliably documented.12,13

At Missouri Baptist Medical Center we launched our patient safety program by first addressing the culture of our organization and changing it from a punitive one to a just and fair one. We hired a full time patient safety specialist and, over a period of about a year among a broad menu of interventions, established executive safety rounds, an educational program for the entire staff, anonymous and simplified reporting of errors and events, feedback concerning interventions, and a rewards program for safety ideas. The success of our patient safety program, as measured primarily by the rate of error reporting, has been previously reported.14 Armed with detailed knowledge of the medication errors that were being made, we were able to prioritize the most appropriate (based on frequency and risk) and potentially most cost effective corrective interventions. We then instituted a broad program of practical, low cost measures, including evidence based protocols, to improve medication safety at our hospital. These were largely directed at high risk medications. Concurrently, a failure modes and effects analysis (FMEA)15 was conducted of the pharmacy dispensing system to prevent dispensing errors before they occurred.

The primary purpose of this paper is to describe the impact of the medication safety component of our overall patient safety program on the harm caused to patients by medication errors.

METHODS

Setting

An audit of adverse drug events (ADE) was performed at Missouri Baptist Medical Center (MBMC) in St Louis, Missouri. MBMC is a not-for-profit 489 bed non-teaching suburban community hospital which is part of BJC HealthCare, a 13 hospital integrated healthcare delivery system.

Design of study

The institutional review board provided a waiver of this study. This audit was performed from January 2001 until December 2003. The study was divided into three time periods: baseline period from January 2001 until June 2001, transition period from July 2001 until March 2002, and post-intervention period from April 2002 until December 2003. The transition period was defined as starting with the appointment of the hospital’s full time patient safety specialist and lasting 9 months during which most of the key interventions were put in place. The dependent variables were ADEs per 1000 doses dispensed and ADEs per 100 patient days.

Intervention

Structural changes were made: a patient safety council was formed in May 2001, a full time patient safety specialist was hired in July 2001, and new event reporting systems were put in place.16 Most interventions were based on the information generated by these event reporting systems. Medication error reports originated from a variety of sources. These have been previously described14 and are listed in box 1.

Box 1 Sources of error reports

• Medication event check box reporting form

• Pharmacy staff written reports

• Hotline calls

• Executive rounds

• Safety briefings on nursing units

• Direct calls to patient safety specialist

As a result of intensive work on culture change, medical error reporting increased significantly (p<0.001) from 35 per 1000 patient days in 2001 to 132 per 1000 patient days in 2003.14 All errors were entered into a single database. Reports run from the database identified high risk medications (such as insulin, narcotics, anticoagulants and antibiotics). Reports also identified which process step most frequently failed (prescribing, transcription, dispensing, administering, or monitoring). We were thus able to identify those drugs and/or processes that were most frequently associated with reported events. While event reporting encompassed all medical errors, this paper reports exclusively on medication errors and includes errors occurring at any step in the medication delivery system.

This resulted in the introduction of a number of drug protocols as shown in box 2. These protocols are available from the authors.

Box 2 Protocols to improve medication safety

• Weight based heparin

• Warfarin

• Lepirudin (Refludin) dosing

• Sedation

• Intravenous potassium

• Intravenous phosphate

• Sliding scale insulin

• Hypoglycemia

• Venous thromboembolism prophylaxis screening

• Clinical pharmacokinetics

• Enteral nutrition

• Total parenteral nutrition

A variety of additional interventions were instituted based on the recommendations of the Institute for Safe Medication Practices (ISMP), the American Society of Health System Pharmacists (ASHP), the Institute for Healthcare Improvement (IHI), the Joint Commission for the Accreditation of Healthcare Organizations (JCAHO), and the Agency for Healthcare Research and Quality (AHRQ). These are shown in box 3.

Box 3 Other interventions to improve medication safety

• Drotrecogin Alfa (Xigris) prescribing and dosing protocol

• Standardized PCA orders

• Standardized postoperative nausea and vomiting orders

• Antibiotic conversion from IV to PO protocol

• Sub-acute rehabilitation weekly medication profile audits

Ten to twenty randomly (using a table of random numbers) selected charts of discharged inpatients were reviewed monthly from January 2001 until December 2003 to assess patient harm caused by medication errors. No patient was sampled more than once. Each chart was audited by two reviewers (a clinical pharmacist and a nurse manager) using an ADE trigger tool designed and tested by the IHI.17 An ADE was defined using the WHO definition: “a response to a drug which is noxious and unintended and which occurs at doses normally used in man for prophylaxis, diagnosis or therapy of disease, or the modification of physiological function”. Medication errors include more events than ADEs as they fail to account for unintended effects of drugs given appropriately. ADEs include any and all results that place patients at risk or expose them to harm. This instrument employs 24 triggers or clues suggestive of patient harm. If, on initial review, a trigger was identified by the reviewers, a more detailed review of the chart was performed to determine if there was an ADE that could reasonably be attributed to a medication error. If the reviewers reached different conclusions, this was resolved by obtaining the opinion of a critical care physician. Rozich et al17 used a sample size of 10 charts. This same sample size was used no matter the size of the hospital being studied. They found that increasing the sample size higher than 20 did not improve the reproducibility or reduce the variability of the data (personal communication). The majority of our monthly samples comprised 20 charts but on several occasions during the latter part of 2001 only 10 charts were reviewed because of staffing problems. The sample size was not based on a fixed percentage of total patients. The hospital has approximately 1700 discharges per month. Harm was defined as temporary or permanent impairment of physical or psychological body function or structure and includes transfers to a higher level of care or admission to a hospital as a result of the harm. The severity of harm of every ADE was scored using categories E to I of the National Coordinating Council for Medication Error Reporting and Prevention (NCCMERP) severity scoring scale18 (table 1).

Table 1

NCCMERP index for categorizing medication errors

These chart reviews for ADE, while designed as the main outcome measure of the impact of the medication safety program, also identified additional opportunities for improvement in the use of insulin (trigger: serum glucose less than 50 mg/dl) and in narcotic management (trigger: oversedation, lethargy or fall). These opportunities were consistent with the process deficiencies that were identified by the evaluation of reported errors. Adverse drug reactions were excluded from the study.

FMEA15 conducted on the pharmacy dispensing system revealed a substantial number of opportunities for improving the safety of medication dispensing and all of these were instituted (box 4). Details of these interventions are available from the authors.

Box 4 Interventions implemented as result of FMEA of the medication dispensing system

• Decrease in not using the patient’s profile to obtain medication from automated dispensing cabinet.

• Nursing units provided with larger refrigerators equipped with separate sections for each patient.

• Pharmacy staff picked up discontinued IVs and drugs three times daily.

• Safety checklist created to verify correct storage of IV fluids.

• Intravenous medication times printed on medication administration record (MAR).

• TALL-man multi colored lettering used when appropriate on medication packaging and on shelves in pharmacy.

• Installation of bar coded dispensing process for the automated dispensing cabinets (Pyxis ParX).

• Separation of “sound-a-likes” in Pyxis drawers.

• All subcutaneous doses greater than 1 ml drawn up by IV room in a single syringe.

• Use of color coded CADD pumps for patient controlled analgesia or epidural administration of narcotics.

• Dispensing of all first dose antibiotics in green bag for easy identification.

• Standardization of PCA concentrations.

• Epinephrine 1 mg/ml ampoules in a ziplock bag with label stating “not for IV use, subcutaneous use only”.

Data analysis

The rate of ADEs per 1000 doses dispensed by the hospital pharmacy and the rate of ADEs per 100 hospital days during the three time periods (baseline, transition and post-intervention) were compared using the non-parametric Kruskal-Wallis test. The numbers of doses dispensed were the total number of medication doses dispensed to the patients randomly selected each month. These data were obtained from the pharmacy management system. The numbers of hospital days were the total number of patient days for the randomly selected patients each month. These data were obtained from the hospital’s health information management department.

The change in the proportion of ADEs per 1000 doses dispensed by the pharmacy or 100 hospital days, and the change in the proportion of patients with an ADE during hospitalization in the three time periods were compared using the χ2 test for trend. Relative risks and 95% confidence intervals were calculated using EpiInfo Version 6. All other statistical tests were performed using SPSS 12.0 (SPSS Inc, Chicago, IL).

All data were tracked using statistical process control charts using three standard deviations to set the upper and lower control limits. The first center line value was selected in December 2001 using the first 11 points available. This gave a mean value of 1.79 ADE per 1000 doses and an upper control limit of 4.82. By July 2002 there were eight points on one side of the center line (Rule #2 in Nelson’s test) and at that point a new center line was calculated as 0.69 ADE per 1000 doses with an upper control limit of 2.36. These was no special cause variation from that time through December 2003.

RESULTS

Examples of triggers identified and the associated ADE found on chart review include the following:

• A patient on coumadin with INR greater than 5 (trigger) subsequently developed a gastrointestinal bleed (ADE).

• A patient receiving two oral hypoglycemic medications and with severe hypoglycemia (trigger) subsequently required transfer to ICU (ADE).

• A patient in whom visual disturbance was noted (trigger), found to have a digoxin level twice the upper end of therapeutic range and an active order for digoxin (ADE).

The median ADE rates per 1000 doses delivered (interquartile range) were 2.04 (1.79–2.70) in the baseline period, 1.26 (0.21–1.53) in the transition period, and 0.65 (0.41–0.87) in the post-intervention period (p = 0.001). Comparison of the proportion of ADEs per total number of doses delivered in the three time periods showed a 3.6-fold lower risk of ADEs during the post-intervention period (p<0.001, χ2  =  30.253, table 2). The statistical process control chart (fig 1) for these data illustrates the time sequence of the reduction in ADE.

Table 2

ADE rates for baseline, transition, and post-intervention period

Figure 1

Statistical process control chart for adverse drug events (ADEs) per 1000 doses of medication dispensed during the entire period of audit from January 2001 until December 2003. The solid line represents the mean ADE rate and the dotted line represents the upper control limit, defined as three standard deviations above the mean. Each point represents the result of a singe month’s audit.

The median (interquartile range) ADE rates per 100 patient days were 5.07 (3.79–6.02) in the baseline period, 3.19 (0.58–5.03) in the transition period, and 1.30 (0.87–1.71) in the post-intervention period (p = 0.001). Comparison of the proportions of ADE in the three time periods per 100 patient days also showed a 3.7-fold reduction in risk of ADE during the post-intervention period (p<0.001, χ2  =  34.115, table 2).

The proportion of patients with ADEs in the baseline period (31%) showed a 3.0-fold reduction in risk of an ADE in the post-intervention period (p<0.001, χ2  =  25.000, table 2).

Severity of events

The number of ADEs associated conclusively with patient harm (rated F–I) was 1.67 per total doses delivered in the baseline period and declined eightfold in the post-intervention period (p<0.001, χ2  =  17.734, table 2). No patient deaths attributable to medication error were detected by the review of patient charts during this study. There were two life threatening events detected during the baseline period but none during the transition or post-intervention periods.

Acknowledgments

The authors are grateful to the staff of the Institute for Healthcare Improvement and the hospital participants in the patient safety collaborative for freely sharing their ideas on patient safety with us, and to Margie Olsen PhD, MPH for statistical consultation.