Design, setting and participants
A before-after study (sequential design) was conducted on a geriatric ward of a university hospital in Belgium. The e-learning intervention was implemented over 3 months between 2 periods of data collection i.e. the non-intervention patient cohort (before group, consisting of usual care; enrolled during 4 months) and the intervention patient cohort (after group; enrolled during 4 months). Both cohorts had a follow-up of 12 months from time of admission to the geriatric ward. Dutch speaking patients who were 70 years or older and consecutively admitted to the geriatric ward, were eligible for inclusion. Patients with severe hearing or visual problems, very poor health condition (e.g. palliative patients, patients with unstable cardiac or respiratory problems), isolation because of infectious disease, or those unable to hold a conversation were excluded. Patients who were readmitted during the study period, or had an expected discharge within 24 h after admission were also excluded. Furthermore, all nurses of the geriatric ward were eligible for inclusion. The study was approved by the Medical Ethics Committee of the University Hospitals Leuven, and written informed/proxy consent was obtained in each patient before inclusion.
Intervention
An on-line self-directed nursing staff educational program on delirium was developed by the research team (ED, FD, EJ, KM). This e-learning tool consists of 11 modules including information about delirium specifics, prevention and treatment strategies for delirium (e.g. including a checklist of 12 risk factors), and information about screening tools for the detection of delirium (with possibility to download the instruments). To help translate new knowledge into practice, the tool incorporates textual information in combination with audio-visual materials, case studies and tests for self-assessment with feedback. The e-learning tool is freely accessible at www.deliriummodule.be. Details about the content, development and feasibility testing of the tool have been described elsewhere [25, 29].
The intervention included (1) a live information session (one hour at the geriatric ward) to offer nurses oral and written information about navigation through the e-learning program, and (2) the completion of six compulsory modules (e.g. ‘occurrence and consequences’, ‘clinical presentation’, ‘exercises in delirium recognition’, ‘predisposing and precipitating risk factors’, ‘screening for delirium, and ‘prevention of delirium’) during a 3-month learning period. The five other modules could be completed on a voluntary basis. The e-learning tool remained available until the end of the study. Participants could access the modules at any time using their personal log-in code. It takes between 5 and 15 min to complete one module. Nurses who did not complete the six compulsory modules within two months were encouraged by the head nurse to complete the course. Additionally, a poster was displayed at the geriatric ward to act as a prompt and further enable knowledge translation.
Variables and measurements
Baseline data
Patient baseline data collected included age, gender, social living circumstances, education level, main diagnosis, number of medications prescribed, number of comorbidities, premorbid functional status, cognitive functioning, confirmed diagnosis of dementia and history of delirium. The number of comorbidities was retained from the modified Charlson Comorbidity Index, and varies between 0 and 13 [30]. The premorbid functional status was evaluated using the Katz Index of activities of daily living (ADL) [31], indicating the level of independence in performing the following six activities scored on a 3-point scale (0 = independent; 1 = partly dependent; 2 = dependent): bathing, dressing, feeding, continence, transfer and toileting. Total score ranges between 0 and 12, with higher scores indicating more dependency. Cognitive functioning was evaluated with the 12-item Mini-Mental State Examination (MMSE) [32]. Total scores vary between 0 and 12, with higher scores indicating better cognitive functioning. Patient baseline data were collected through patient interview, requested from a family member, or based on the medical or nursing records.
Nurse characteristics were collected at the start of the intervention implementation period and included age, gender, work experience as a nurse, percentage employment, day- or night work, highest level of education and delirium education attended in the 5 years prior to the start of the study.
Primary outcomes
In-hospital prevalence of delirium was measured with the Confusion Assessment Method (CAM) [33, 34], which was scored after a structured interview including the 12-item Mini-Mental State Examination (MMSE) [32]. Accordingly, delirium was diagnosed when the criteria “(acute onset OR fluctuation), inattention, AND (disorganized thinking OR altered level of consciousness)” were rated as positive on at least one of the measurement points (see procedure).
Duration of delirium was defined as the number of days on which a positive CAM score was obtained.
Severity of delirium was assessed with the 7-item Delirium Index (DI) [35], including inattention, disorganized thinking, altered level of consciousness, disorientation, memory impairment, perceptual disturbance, and disorder of psychomotor activity. Each item was scored on a scale from 0 (absent) to 3 (present and severe) resulting in a total score varying between 0 and 21, with higher scores indicating greater severity.
Secondary outcomes
Patients’ in-hospital mortality is defined as the number of deaths occurring while being hospitalized at the geriatric unit. Twelve-month mortality includes all patients that died within 12 months after admission, including cases of in-hospital mortality.
Delirium recognition in nurses was assessed with standardized ‘cases vignettes’ [36], including validated cases about hospitalized patients with dementia, hypoactive delirium, hyperactive delirium, hypoactive delirium superimposed on dementia (DSD) or hyperactive DSD. Before as well as after the e-learning intervention, four slightly different case vignettes were used to avoid recall bias (i.e. dementia, hypoactive delirium, hyperactive delirium and, hyperactive DSD or hypoactive DSD). The behavioral symptoms described in each case had to be scored as dementia, delirium, DSD, normal ageing, depression or none of the options, with each case having only one correct answer. Total delirium recognition (DR) was defined as the number of case vignettes answered correctly (range 0 to 4).
Delirium knowledge in nurses was assessed with the 35-item true-false Delirium Knowledge Questionnaire (DKQ) [25, 37]. Ten items are related to the presentation, symptoms and consequences of delirium, 11 items to the causes and risk factors of delirium, and 14 items to the prevention and management strategies of delirium. The total DKQ score was defined as the number of questions answered correctly and ranged from 0 to 35.
Completion of the e-learning tool in nurses
The number of e-learning modules finalized by each nurse was recorded and ranged from 0 to 11.
Procedure
Patient baseline data, premorbid functional status, number of comorbidities, cognitive functioning, delirium and delirium severity were assessed on the first day after admission to the geriatric ward. In addition, delirium and delirium severity were evaluated on the third, fifth and seventh day after admission to the geriatric ward, and on the day before discharge. From the seventh day after admission, delirium and delirium severity were assessed weekly (e.g. 14th, 21th, day) until hospital discharge. If the patient had delirium on one of the measurement points, the patient was followed up daily until a negative CAM score was obtained. Mortality was recorded during hospitalisation and twelve-month mortality was checked by telephone contact with the patient or his proxy. Procedures for patient assessments in the non-intervention and intervention cohorts were identical. There were no service changes or changes to protocol during the entire study period.
Six study nurses with a master degree performed all assessments. They were trained (i.e. theoretical training of 4 h) by two experts in delirium (ED and KM) according to criteria set in the manuals of MMSE and CAM [33, 34], including evaluation of four clinical cases at the bedside and follow-up discussions. Inter-rater reliability for CAM was κ = 1.00, indicating perfect agreement (inter-rater reliability refers to the agreement of CAM scoring for each study nurse compared with CAM scoring of one of the investigators (ED), and calculated two by two in a random sample of 18 paired observations of enrolled patients).
At the beginning of the one-hour live information session before implementation of the intervention and at the end of the study, nurses received the three questionnaires to assess their baseline data, their knowledge about delirium (DKQ) and their ability to recognize delirium (case vignettes), as described above. Returning a completed questionnaire was considered as informed consent.
Sample size
According to a power analysis for two cohorts using a two-tailed test of significance with an alpha of 0.10, a beta of 0.30 and an estimated proportion of delirium of 30% for the control cohort [38,39,40], a sample size of 71 participants was required in each cohort to detect a difference of 50% in prevalence of delirium.
Blinding
Although patients were blinded to the intervention, nurses and research nurses (data collectors) could not be blinded because of the nature of this study.
Analysis
Descriptive analysis (i.e. means/median, standard deviations/interquartile ranges, or absolute numbers and percentages) for patients in the control and intervention cohorts, as well as for all included nurses were calculated as appropriate.
A chi square test was used to compare in-hospital prevalence of delirium in the control and intervention cohort. This difference was further explored using a logistic regression model in which a random effect for patient was modelled to account for clustering. Duration of delirium (in days) was compared with the Mann-Whitney U-test. Severity of delirium in the two cohorts was compared using a linear mixed model with a random effect accounting for clustering. The mortality risk was explored with a logistic regression model in all patients and in the subgroup of delirious patients. To correct for baseline differences between both cohorts, baseline functional status score and gender were included in all logistic regression and linear mixed models.
Both in the logistic regression and linear mixed models, a time by group interaction was tested first, and a main effect is estimated in case of a non-significant interaction effect. Non-linear trends of time are considered using quadratic and cubic splines-based trends. The models are likelihood-based and therefore provide valid results in case of a random drop-out pattern, this is when the drop-out chance may be associated with previous observations or covariates in the model [41]. Linear mixed models were performed by using the measurement data on the first, third, fifth, seventh, fourteenth, twenty-first days after admission and those of the day before discharge.
In nurses, delirium recognition scores and delirium knowledge scores before and after introduction of the e-learning intervention were compared using paired t-tests for normally distributed data and the Wilcoxon Signed Rank test for non-normally distributed data. McNemar’s tests were used to test differences in proportions of correct answers on the four ‘case vignettes’.
All tests were two-sided, with p-values < 0.05 considered as significant. All analysis were performed on intention-to-treat principle using SPSS, version 21 (SPSS Inc., Chicago, IL) and SAS System for Windows version 9.2 (SAS Institute Inc., Cary, NC, USA).