Methods

2.5.1 Study setting

The study was conducted on a 28-bed adult general medical ward of a 600-bed NHS teaching hospital. Medications were generally administered at four scheduled drug round times each day: 08:00, 12:00, 18:00, and 22:00 hours. Nurses administered medications for the patients they were looking after against medication orders prescribed on handwritten paper drug charts. Non-IV patient-specific medications were stored in individual patient bedside medication lockers, and ward stock was stored in separate stock cupboards located in a treatment room at one end of the ward. Local trust policy allowed some patient-specific medications to be kept at the bedside rather than in the bedside medication locker; these

were insulin, creams, and inhalers. Controlled drugs were stored separately in an automated drug storage unit (Pyxis MedStation™) that was accessed either via fingerprint recognition or individual user log in by two trained and registered nurses; the system was implemented in 2006 (Franklin et al., 2010).

The ward operated a patients’ own drugs (PODs) and one-stop dispensing (OSD) scheme, as endorsed nationally (Audit Commission, 2001; Department of Health, 2000b). In the PODs scheme, patients were encouraged to bring their medicines into hospital to facilitate accurate medicines reconciliation, minimise the risk of missed doses due to medication not being available, and thus reduce waste, in addition to increasing safety. The OSD scheme involved pharmacy staff dispensing 28-day inpatient-specific supplies labelled with instructions on how to take the medicine; these were intended for both inpatient administration and given to patients at discharge thus minimising repeat dispensing, and reducing waiting time for medication supplies. Medicines that were unlikely to be continued post discharge were not dispensed as OSD; instead ward stock or non-OSD inpatient supplies (medications labelled without directions) were used. During pharmacy opening hours, nurses ordered medications via the ward pharmacist and/or by going to the pharmacy dispensary. Outside pharmacy opening hours, an emergency on-call resident pharmacist was available who could be contacted for obtaining medicines if necessary.

2.5.2 Data collection

Non-IV drug rounds were observed by MM and data recorded on pre-piloted data collection forms (appendix 1). Nurses were informed of the study objectives during staff meetings and given opportunities to ask questions about the study. MM arrived approximately 1-2 hours prior to the drug round time as pilot work revealed nurses sometimes started the drug round up to two hours early. Verbal consent from each nurse was obtained prior to the start of any

observations. In general, the ward was divided into four sections during the 08:00, 12:00 and 18:00 hour drug rounds and two sections for the 22:00 drug round. As the nurse looking after each section often started their respective drug rounds at similar times, only one section of the ward was observed at any one scheduled drug round time. Observations were organised to ensure all four scheduled drug round times were observed on all seven days of the week over 20 days; these were approximately equally distributed across all sections of the ward.

All MAEs observed, defined as “any dose of medication administered (or omitted) that deviates from the patient’s medication order”(Allan & Barker, 1990), were documented; each were categorised according to table 2.1. Consistent with the approach used in previous observational MAE studies, MM intervened whenever there was a risk that the patient would be harmed as a result of the MAE. Errors prevented by MM or the patient were included as MAEs, those prevented by other health care professionals were not. The clinical appropriateness of the prescription was not assessed. All regular, “when required” and “once only” non-IV medication orders that were due on the scheduled drug round observed were included. Doses given in between the four daily scheduled drug rounds were excluded;

controlled drugs were therefore also excluded as these were generally prepared between drug rounds when two nurses were available. In addition, any dose that could not be observed, for example, rectal administrations were also excluded. Medical gases, dietary supplements, and thromboembolic deterrent stockings were excluded from the study.

Confirmation of a patient’s identity was recorded if the nurse visibly checked the patient’s identity band against the details on the drug chart and/or asked the patient to confirm their name and date or birth prior to drug administration. Availability of the medication at the patient’s bedside was taken to include successful dose retrieval from the patient’s bedside medication locker, and/or any area around the bedside. A dose taken by a patient was considered observed by the nurse if the nurse remained at the patient’s bedside while the

patient took the dose and/or if the dose was administered directly to the patient by the nurse.

Five types of administration documentation were recorded: (1) dose was administered and signed for, (2) dose was administered but not signed for, (3) dose was not administered and a reason documented, (4) dose was not administered but signed to suggest it was administered, and (5) dose not administered and not signed.

Timing of each drug round observed started when the nurse picked up the first drug chart for medication administration and stopped after the last dose was administered or drug administration was documented (whichever was the last task). The prescribed time for each dose observed was also recorded.

The number of interruptions during each drug round observed was recorded. An interruption was defined as any action(s) from a person (other than the patient to whom medication is being administered) that prevents the nurse from continuing with the drug round. For example, an interruption included the nurse being called away to answer a telephone call, being asked by a different patient to do something which takes the nurse’s attention away from the task of medication administration. An interruption did not include any interaction between the nurse and the patient, to whom medication was being administered, for example, talking to the patient, answering questions from the patient.

2.5.3 Sample size

A sample size calculation was made on the assumption that an intervention would be made.

Based on a normal approximation to the binomial distribution, a sample of 634 observed dose administrations before and after an intervention was required to provide a power of 80% to detect a reduction in MAEs from 7% (Franklin et al., 2007) to 3.5% based on a two-sided test with an α of 0.05. The reported baseline MAE rate from Franklin et al (2007) was used for the

following reasons: (1) many of the methods used in the current study were adapted from Franklin et al (2007), (2) the study was conducted at the same hospital and thus the MAE rate identified might be expected to be comparable, and (3) this was the most recent UK study of MAEs identified at the time. Data were collected until this sample was achieved.

2.5.4 Data analysis

All data were analysed using descriptive statistics. In addition, a cumulative quality filter of six medication administration related measures was constructed which comprised: (1) timeliness, calculated as the percentage of doses administered on a drug round that started and finished within each of 1, 1.5, and 2 hours of scheduled round time, (2) percentage of doses where the patient’s identity was checked prior to administration, (3) percentage of doses that were given and/or omitted for a therapeutic reason, (4) percentage of doses given that were administered correctly, (5) percentage of doses where the nurse observed patient taking the medication, and (6) percentage of doses that were correctly documented. The cumulative quality filter was produced by adapting the approach used by (Garfield et al., 2009). This involved identifying the key medication process steps, superimposing the compliance rate at each process step and then calculating the cumulative compliance rate by aggregating the compliance rate at each step with the preceding compliance rates; all compliance rates for the quality filter were calculated using the same denominator. In the present study, the reliability of each of the six tasks was determined and then combined in a stepwise manner to produce an overall percentage of ‘compliance with standard good practice’. Additionally, an overall MAE rate was calculated using the total number of MAEs identified divided by the total number of opportunities for error, multiplied by 100. The timing of actual drug round start times was also compared with prescribed times of observed doses.

2.5.5 Ethical considerations

Research ethics approval was not required as the local ethics committee considered this study to be service evaluation.