Janique Jessurun

At least 1 in 20 patients in healthcare settings is affected by preventable harm, most frequently caused by medication-related incidents. In hospitals, many interventions to improve medication safety have been explored, but errors (with and without harm) still occur throughout the medication process. The associated patient and economic burden emphasises the necessity to find ways to prevent errors related to medication practices.

The studies presented in this thesis focus on estimating the prevalence of intravenous admixture preparation errors as well as medication administration errors in a hospital setting. An additional focus was identifying possible determinants of these errors, which may help to design effective preventive strategies. We also studied the effect of two promising strategies to prevent medication errors in a hospital setting. First, a pharmacy-based intravenous admixture service, and second, centralised automated unit dose dispensing combined with barcode-assisted medication administration.

Part I. Intravenous admixture preparation and safety

The first part of this thesis focuses on the prevalence and determinants of intravenous admixture preparation errors in a hospital setting as well as the effect of a pharmacy-based intravenous admixture service on the prevalence of these errors.

In Chapter 2, we describe an observational study performed in five clinical wards in a Dutch university hospital aimed to assess the prevalence and determinants of intravenous admixture preparation errors. Data on medication preparation were collected by observation of nursing staff in daily clinical practice. Two pharmacists independently assessed deviations from the medication order or local protocols. Hygiene and documentation errors fell outside the study scope. We identified one or more errors in 367 of 614 admixtures (59.8%) prepared; in total 533 errors were observed. One out of nine errors had the potential to lead to patient harm. The most prevalent error type was inadequate mixing after adding the medication solution to the infusion fluid (n=240). For determinant analysis, multivariable mixed-effects logistic regression analysis was used. The following variables were associated with the occurrence of errors: multistep versus single-step preparations (odds ratio [OR] 4.08, 95% confidence interval [95% CI] 2.27–7.35); interruption versus no interruption (OR 2.32, 95% CI 1.13–4.74); weekend versus weekdays (OR 2.12, 95% CI 1.14–3.95); time window 2 p.m.-6 p.m. versus 7 a.m.-10 a.m. (OR 3.38, 95% CI 1.60–7.15); and paediatric versus adult wards (OR 0.14, 95% CI 0.06–0.37). We concluded that the high rate of intravenous admixture preparation errors warrants implementation of effective systemic defences to prevent these errors. Preventive strategies in clinical wards should focus on working conditions and complex preparations in particular.

In Chapter 3, we describe an uncontrolled before-and-after study aimed to examine the effect of a pharmacy-based centralised intravenous admixture service for multistep preparations on the prevalence of intravenous admixture preparation errors in a university hospital. Data were collected in clinical wards and in the hospital pharmacy. Pharmacy staff used preparation software fully integrated with the computerised physician order entry system. Pharmacy staff prepared medication in cleanrooms. Nursing staff had access to local electronically available protocols and prepared medication on workbenches in a medication room. Data collection methods and analysis methods were similar to those in Chapter 2. In this study, the vast majority of included admixtures were antibiotics (82%). We identified one or more errors in 14 of 543 admixtures (2.6%) in the pharmacy and in 148 of 282 admixtures (52.5%) in the clinical wards (OR 0.02, 95% CI 0.004-0.05). No potentially harmful errors occurred in the pharmacy opposed to 22 (7.8%) in the clinical wards. Several disinfection procedures were better adhered to in the pharmacy compared to clinical wards. Nursing staff was quite satisfied with the centralised service with a score of 77 on a 100-point scale. All findings taken together, a pharmacy-based centralised intravenous admixture service is an important strategy to prevent intravenous admixture preparation errors, including clinically relevant errors.

The study in Chapter 4 examined the clinical relevance of inadequate mixing during preparation, a prevalent error type. In this experimental study, concentrations of four unmixed antibiotic infusions (cefuroxime, meropenem, flucloxacillin, and vancomycin) were compared at three intervals during infusion (beginning, middle, and end) with the declared concentration and between preparation sites (hospital pharmacy versus clinical ward). Real-life administration of infusion bags was mimicked using infusion pumps at the patient bedside. The median concentrations of the four antibiotics were comparable to the declared concentration at the three time points. All median concentrations deviated less than 20% from the declared concentration, except for pharmacy-prepared cefuroxime at the end of infusion, which was 30.8% lower than the declared concentration. No significant differences were found between preparation sites. This study adds insights into this topic by showing that mixing errors are not clinically relevant for the studied antibiotics.

Part II. In-hospital medication administration and safety

The second part of the thesis focuses on the prevalence and determinants of medication administration errors in a hospital setting with different electronic support systems already in place and on the effect of centralised automated unit dose dispensing combined with barcode-assisted medication administration on the prevalence of these errors.

The study in Chapter 5 examined the prevalence and determinants of medication administration errors in two hospitals in the Netherlands (six wards in a university hospital and five wards in a teaching hospital). Both hospitals had the following systems in place: electronic medical record, computerised physician order entry, and electronic medication administration registration. The electronic medication management systems and medication cart filling procedures (pharmacy staff versus nursing staff) as well as the presence of barcode technology for parenteral medication (only present in the teaching hospital) differed between the hospitals. We prospectively collected data on medication administration errors by observation of nursing staff in daily practice. Pharmacists assessed whether a medication administration error had occurred, defined as a deviation from the medication order or local protocol. Multivariable mixed-effects logistic regression analysis was used for determinant analysis. Medication administration errors occurred in 352 of 2576 medication administrations (13.7%). Of all errors (n=380), the most prevalent types were omission (n=87) and wrong medication handling (n=75). Forty-five errors (11.8%) were potentially harmful. The pharmaceutical forms oral liquid (OR 3.22, 95% CI 1.43-7.25), infusion (OR 1.73, 95% CI 1.02-2.94), injection (OR 3.52, 95% CI 2.00-6.21), ointment (OR 10.78, 95% CI 2.10-55.26), suppository/enema (OR 6.39, 95% CI 1.13-36.03), and miscellaneous (OR 6.17, 95% CI 1.90-20.04) were more prone to errors compared to oral solid forms. Errors were more likely to occur when administered between 10 a.m.-2 p.m. (OR 1.91, 95% CI 1.06-3.46) and 6 p.m.-7 a.m. (OR 1.88, 95% CI 1.00-3.52) compared to 7 a.m.-10 a.m. Errors were less likely to occur in the teaching hospital compared to the university hospital (OR 0.17, 95% CI 0.08-0.33). Day of the week, patient-to-nurse ratio, interruptions, and nurse characteristics other than educational level were not associated with the occurrence of errors. In conclusion, this study showed that medication administration errors are prevalent in modern clinical practice with supportive electronic medication systems in place. The identified determinants suggest that interventions focusing on complex pharmaceutical forms, error-prone administration times, and complex patient populations may help prevent these errors.

In Chapter 6, we describe a prospective uncontrolled before-and-after study performed in six clinical wards in a university hospital aimed to investigate the effectiveness of centralised automated unit dose dispensing combined with barcode-assisted medication administration. Before intervention, electronic medical record, computerised physician order entry, and electronic medication administration registration were already in place. Data collection and analysis methods were similar to those in Chapter 5. In this study, one or more medication administration errors occurred in 291 of 1490 administrations (19.5%) pre-intervention and in 258 of 1630 administrations (15.8%) post-intervention (OR 0.70, 95% CI 0.51-0.96). The rate of omission (i.e. not administering medication) fell from 4.6% to 2.0% and the rate of wrong dose from 3.8% to 2.1%, while rates of other error types were unchanged. The rate of potentially harmful errors fell from 3.0% (n=44) to 0.3% (n=5). Adherence to scanning of patient and medication barcode post-intervention needed improvement, as the scanning rates were, respectively, 13.6% and 55.9%. In the post-implementation period, error rates were lower for scanned medication compared to non-scanned medication (13.0% versus 19.5%). As expected, scanned medication had lower rates of different error types (i.e. omission, unordered drug, wrong dosage form, and wrong dose) compared to non-scanned medication, except for wrong administration technique and wrong medication handling. The median overall satisfaction score of the nurses with the medication administration system (on a 100-point scale) was 70 (interquartile range 63-75, n=193) pre-intervention and 70 (interquartile range 60-78, n=145) post-intervention (P=0.626). In conclusion, centralised automated unit dose dispensing combined with barcode-assisted medication administration is a valuable strategy to reduce the number of medication administration errors and therefore to improve patient safety.

Chapter 7 reports on the cost-effectiveness of this combined intervention from the hospital perspective. An economic evaluation was conducted alongside the effectiveness study described in Chapter 6. The total costs were the sum of costs for pharmaceutical service, nurse medication handling, wastage, and materials related to the dispensing device. Based on 2,260,870 administered medications in the entire hospital annually, a total of 102,210 medication administration errors and 59,830 potentially harmful medication administration errors were estimated to be avoided. The intervention was associated with a cost-effectiveness ratio of €17.69 per avoided error and €30.23 per avoided potentially harmful error.

General discussion and conclusion

Finally, in Chapter 8, we discuss the main findings in this thesis in a wider perspective and address recommendations for clinical practice and future perspectives. The studies of this thesis highlighted that preparation and administration errors often occur in hospitals and that there are several ways to reduce the frequency of these errors; i.e. with the examined interventions and interventions arising from identified determinants. These studies focused primarily on the frequency of medication errors and secondarily on potentially harmful errors. If feasible, future studies should predominantly focus on clinically relevant endpoints, as improving patient-related outcomes is the main goal of patient safety interventions. Because of practical and feasibility considerations, we used the prospective uncontrolled before-and-after study design for the intervention studies, a relatively weak method to assess a cause-effect relation, instead of a more robust design with control groups. Nonetheless, the studies had large representative cohorts and bias was minimised as much as possible, which rendered useful findings. Studies in this thesis also showed that the implementation of the interventions could be improved. Implementation studies are recommended to explore factors influencing the successes, failures, and the efficiency of these interventions. The centralised pharmacy systems may affect different aspects of nursing practice, including safety, efficiency, and experience with several tasks. Several findings in this thesis (e.g. low compliance with scanning) highlighted the importance of adequate technology, involving nurses and end-users when designing and implementing interventions, and tailoring interventions to the workflow of nurses. Another point to take into account is that centralised pharmacy systems may hamper patient engagement and efforts should be made to facilitate patient engagement (also with barcode scanning) because of its assumed long-term safety benefits. Policy makers have to make a thorough selection of the interventions to be implemented, especially considering the scarce resources in healthcare. With regard to identified determinants of medication errors, the effectiveness of interventions arising from these should be explored, because robust evidence is scarce and effects are difficult to predict. Further research should also focus on more direct determinants of medication errors (e.g. workload) instead of relatively indirect determinants (e.g. time-related determinants) in order to increase its applicability.

This thesis highlighted that despite the many medication safety initiatives in hospitals, medication preparation and administration errors still frequently occur. This emphasises the necessity to identify and implement additional interventions to prevent these errors. This thesis showed that the two studied interventions were effective in reducing the occurrence of medication errors, including those that can lead to patient harm. In addition, insights were provided into the types of errors and determinants of errors, which may elicit new preventive strategies.