Implementing the best available evidence in early delirium identification in elderly hip surgery patients


Correspondence: Ms Kathleen Ann Russell-Babin, Meridian Health, Institute for Evidence-Based Care, Neptune, NJ 07753, USA. Email:



Delirium is a frequent complication in the surgical experience of elderly hip surgery patients. Its impact can be severe and may even include death. Implementation of a delirium predictor tool might focus attention on early recognition of delirium, thereby potentially decreasing its impact. A related aim is to evaluate best practices in implementation strategies in this project.


After an exhaustive search of the literature, no consensus was found regarding delirium predictors for the elderly hip surgery patient. A local research study was implemented to determine factors that may predict delirium in this population. With evidence secured, a multidisciplinary implementation project augmented by ongoing audit was instituted. A variety of social diffusion and education tools were used. Implementation was guided by the use of the Promoting Action on Research Implementation in Health Services framework assessment tool and the Alberta Context Tool, as well as traditional performance improvement tools, such as fishbone charting. Audit identified the rate of use of the predictor tool and pre- and post-rates of delirium. This project was part of the Joanna Briggs Institute Signature Project, an implementation project consisting of six teams, each representing a different organisation. This overall project was supported by experts in the field of translation and implementation science internationally.


Initial compliance to the use of the predictor tool was assessed at 54% within 3 months of implementation and increased to 56% in the ensuing months. Before the study use of the predictor tool, the delirium rate was 10.4% (12 of 115 patients). An interim analysis 4 months after implementation identified a 20% delirium rate (18 of 70 patients) and an updated analysis 8 months into the project showed a 16.3% delirium rate. Delirium predictor tool use was associated with a lower delirium rate (9/76, 11.84%) than no delirium predictor tool (13/60, 21.67%), but the difference was not statistically significant with a sample size of 133 (P = 0.122).


The delirium predictor tool shows promise as a prompt for best practices in prevention of delirium. This study showed a change in delirium rates as a result of its use. Although the results were not statistically significant, they may be clinically meaningful. Comprehensive assessment and implementation planning by a multidisciplinary team contributed to only 56% compliance in use. Despite this low rate, delirium identification rates were higher.


Elderly hip surgery patients often experience increased delirium rates. Such patients may have delirium rates as high as 61%.[1] Delirium is associated with higher rates of complications,[2, 3] longer length of stay,[3, 4] discharge to institutional care[2-4] and even death.[4] Delirium assessment, prediction and care are the subject of much attention in the literature over several decades. It remains an area ripe for improvement. This article will trace the journey of evidence assessment and implementation on delirium prediction in hip surgery patients.

The hospital involved in this project is a Nursing Improving Care for Healthsystem Elders facility and its nurses have been identified for their excellence by the American Nurses Credentialing Center Magnet Recognition program (Magnet). Magnet designation recognises organisations throughout the world for demonstrating superior patient care and practice environments that support nurse development. The hospital is a member of a prominent health system that maintains high emphasis on research, education, shared governance and clinical excellence. It maintains a staff of five nurse scientists and a separate interdisciplinary institute for evidence-based care.

This project was initiated as part of knowledge translation workgroup sponsored through the Joanna Briggs Institute. International leaders in evidence translation and implementation guided six healthcare organisations through a knowledge translation project entitled the Signature Project. Members of the team from this single hospital in central New Jersey focused on the subject of delirium in elderly hip surgery patients.

This article will

  • Outline the research secured to initiate an evidence-based delirium predictor project
  • Describe the translation team's assessment of readiness for evidence-based practice change
  • Delineate the implementation team activities
  • Share the data collection process before and after predictor tool implementation
  • Discuss the results, conclusions and implications

In the literature, the experience of delirium is seen since the days of Hippocrates,[5] yet it remains elusive even today. The roots of delirium have been traced to the Latin word meaning ‘off the track’.[6] Delirium is an abrupt, transient and fluctuating disturbance in consciousness, cognitive function or perception. The symptoms can vary from lethargy (silent or hypoactive) to hyperagitation.[7] The surgical geriatric patient is especially prone to the development of delirium, with the highest reports in orthopaedic surgery and vascular surgery.[8] It is noted early in the postoperative period, typically on the first or second day.

A review of the literature was performed by the authors to determine the state of the science in delirium prediction, with a special focus on elderly hip surgery (both elective and emergency patients). Over 40 predictors of postoperative delirium (POD) were identified (see Table 1 for list and sources). Little consensus was found, although several studies posed emergency surgery as a predictor variable. The rigor of the studies available varied tremendously, with most categorised as either prospective cohort or observational studies. Studies often used convenience samples. The patients that were included and excluded had no consistency. At times, studies excluded variables that were predictive in other investigations. The available evidence was considered insufficient to direct predictors in this population.

Table 1. Sources of evidence for predictors of delirium
  1. BMI, body mass index; CV, cardiovascular; WBC, white blood cell.
AgeMarcantonio et al.,[3] Galanakis et al.,[9] Freter et al.,[10] Kalisvaart et al.,[11] Vaurio et al.,[12] Priner et al.,[13] Ansaloni et al.[14]
Emergent statusBowman,[15] Duppils and Wiklblad,[16] Kalisvaart et al.,[11] Dasgupta and Dumbrell,[17] Galanakis et al.[9]
CV diseaseDuppils and Wiklblad[16]
Delay in surgeryDuppils and Wiklblad[16]
Hearing impairmentDuppils and Wiklblad[16]
Vision impairmentFreter et al.[10]
HypotensionEdlund et al.[2]
Dependent livingDuppils and Wiklblad[16]
Male genderFisher and Flowerdew,[18] Edlund et al.[2]
Reduced clock drawingFisher and Flowerdew[18]
Substance issues, alcohol or benzodiazepinesFreter et al.[10]
Cognitive impairmentMarcantonio et al.,[3] Freter et al.,[10] Duppils and Wiklblad,[16] Kalisvaart et al.[11]
Reduce activities of daily livingMarcantonio et al.,[3] Freter et al.[10]
DepressionGalanakis et al.,[9] Juliebø et al.[19] Leung et al.[20]
Lower educational levelGalanakis et al.[9]
Lower sodiumMarcantonio et al.,[3] Galanakis et al.,[9] Zakriya et al.[21]
APACHE II levelKalisvaart et al.[11]
Attention deficitLowery et al.[22]
Lower potassiumMarcantonio et al.[3]
Lower glucose levelMarcantonio et al.[3]
Faecal incontinenceShuurmans et al.[23]
Inability to bathe selfShuurmans et al.[23]
Comorbid psychiatricShuurmans et al.[23]
Multiple comorbilitiesShuurmans et al.[23]
Preoperative painVaurio et al.[12]
Increased painVaurio et al.[12]
Parenteral pain medicationVaurio et al.[12]
Normal WBCZakriya et al.[21]
ASA classZakriya et al.[21]
Fracture indoorsJuliebø et al.[19]
BMI < 20Juliebø et al.[19]
Cumulative illness ratingAnsaloni et al.[14]
High glucoseAnsaloni et al.[14]
HADS (depression)Ansaloni et al.[14]
Mental statusAnsaloni et al.[14]
More than three medicationsInouye[24]
Foley catheterInouye[24]
Iatrogenic eventInouye[24]
IQCODEPriner et al.[13]
Blood lossPriner et al.[13]
Psychotropic drugsPriner et al.[13]
Mini-Mental Status EvaluationPriner et al.[13]

A local retrospective chart review of randomly selected elderly patients from 2010 who underwent hip surgery was conducted. This research identified three factors that correlated with POD: benzodiazepine use on day 1 postop, low haematocrit on day 1 postop and history of depression. Surprisingly, emergency surgery did not appear. Because of its prominence as a predictor in five studies, the research team retained it as a predictor for continued testing. A final predictor tool of these four variables was adopted for use. This article will detail the course of this evidence production.

Translating evidence into practice is social, political and educational experience. Accordingly, a variety of social diffusion and education tools were used to support the implementation of this tool into practice. The organisation's experience in evidence implementation will be shared.

Aims/objectives of the project

The overriding goal of this knowledge translation project was to reduce the incidence and impact of delirium on elders experiencing hip surgery. Improving identification of delirium was considered paramount to ensuing patient safety. The premise of the project is that implementation of an evidence-based delirium predictor tool (DPT) might focus attention on early recognition of delirium, thereby potentially decreasing its impact. When the DPT was assessed as positive, it was expected that the nurse would adopt early intervention strategies to prevent or lessen delirium. Among these strategies were diligent use of the confusion assessment method (CAM),[25] collaboration with the attending physician and interdisciplinary team, and implementation of a delirium prevention plan of care.


Setting and population

Ocean County is second only to most counties in Florida for its dominant elderly population. Ocean Medical Centre is a key location for care of this population and a local leader in orthopaedic care. Ocean Medical Centre is a 241-bed acute care facility containing specialised units for acute elderly care and orthopaedics.

Overview of key stages in the project

Stage 1: evidence review

As was previously noted, a thorough review of the literature was conducted in search of a composite of predictors of POD. The initial focus of the query was based upon elderly hip surgery patients. This produced a limited amount of articles, so the broader categories of orthopaedic surgery and then general surgery with at least 40% of the patients in orthopaedic were reviewed. The goal of the literature review was to locate a predictor tool already in use and to replicate its use. What resulted was a larger, diverse set of potential predictors studied across various institutions globally.

Multiple literature sources (as previously noted in Table 1) documented a variety of potential predictors of delirium. Works of Inouye and Charpentier[25] and Marcantonio et al.[3] in this area are prominent, although work of Inouye and Charpentier was with a medical population and work of Marcantonio et al. was with a surgical population. No consensus on predictors was found. The opportunity to perform local research became evident.

Stage 2: local evidence production

This section further summarises the production of local evidence for a predictor tool in elderly hip surgery patients. All data collection was approved through the local institutional review board. Data collection was accomplished by the authors. A standard data collection was created and approved for use by institutional review process. Prior to the full data collection, the nurse researchers reviewed a sample of charts to attain consistency in chart review. There were no discrepancies, so that the data collection method was consistent and accurate through the process. Every other chart was accessed pre- and post-implementation of the predictor tool. In the pre-implementation period, a list of patients who had undergone hip surgery (identified by corresponding ICD9 codes for the hip procedure and delirium) in 2010 at Ocean Medical Center was obtained from the data warehouse. If a patient had multiple hip surgery dates, only the first date was used. A chart review was done (every other chart on the list due to available resource reasons) for evidence of any predictors that were noted in the literature. The predictors were divided into three categories: preoperative, intraoperative and postoperative variables (examined on postop day 1, 2 and 3). Based on the 40 predictors found in the literature, a total of 59 data points were collected during the patient's hospital stay.

The inclusion criteria were adults over the age of 65 who had hip surgery and completed their postoperative course in the general ward. Since admission to a critical care area carries a higher risk of delirium related to additional potential causative factors, those patients were excluded. Patients with dementia were also excluded as clinicians in this system often continue to confuse delirium and dementia. The focus was to limit as many confounding variables as possible. A total of 178 charts were examined for manual data abstraction. A total of 63 charts were not analysed due to the exclusion criteria above. A final total of 115 charts were used for analysis. Delirium was considered to be present when two separate entries of confusion were evident in the record. At the time of the study, use of CAM was in its infancy so the gold standard of identification could not be relied upon. Furthermore, the desire was to ensure that any incidence was found, ICD-9 coded or not. The same nurses reviewed the charts preoperatively and postoperatively.

Of the preoperative factors identified (such as significant comorbidities, place of residence, vital signs and activities of daily living), only a history of depression yielded a significant association with POD. A total of 12 patients developed confusion indicating delirium. Twenty-four patients had a history of depression. Of these, six developed delirium indicating a 25% incidence. This compared with six patients with delirium out of 91 who had no history of depression or a 6.6% incidence. This difference was significant with the Fisher's exact test (P-value = 0.018) and logistic regression (P-value = 0.014), showing a 4.72 increase in odds of delirium if the patients had a history of depression. No intraoperative factors yielded any significant relationship to POD.

Postoperative variables proving useful were benzodiazepine use on day 1 postoperatively and low haematocrit on day 1 postoperatively. A multivariate analysis using haematocrit as a dichotomous variable was performed. It was found that with each unit decrease in haematocrit, the odds of developing delirium increased 23%. This result was statistically significant. Day 1 and day 2 use of benzodiazepines were highly correlated at the univariate level (P = 0.009), but only day 1 was significant on multivariate analysis. A total of 15 patients received routine oral benzodiazepines the first day, reflecting a 33.3% incidence of delirium compared with only 7.1% incidence in patients who did not receive a benzodiazepine (multivariate 6.40 (1.32, 30.93), P = 0.0210).

The final predictor tool included the following variables: benzodiazepine use on day 1 postop, low haematocrit on day 1 postop and history of depression. Furthermore, the variable of emergency surgery was added for further testing.

During the course of this analysis, a pre-evaluation of delirium rate was assessed. Delirium rate was found to be 10.4%. This low rate seemed logical given the trend of lower emergency cases and strong practice patterns surrounding the care of these patients. These practices included such items as prompt removal of urinary catheters, early ambulation and adequate pain management.

Stage 3: translation team assessment

The authors enlisted the assistance of a staff nurse in active practice to assess the potential for a successful change in practice to prevent delirium. This small core team assessed the barriers to implementation of the newly created predictor tool using a cause and effect diagram. Three factors were determined and included expected resistance to change, potential lack of interest and lack of education. Lack of education was deemed the most significant of these. Strengths assessed included supportive leadership, staff who are highly motivated to improve care for the elderly and available resources to support the project. An action plan was derived to provide small group education utilising case studies to personalise the message of care of the patient threatened by delirium. This group also used the Promoting Action on Research Implementation in Health Services (PARiHS) self-assessment tool to further determine readiness for evidence implementation.[26] Evidence strength was moderately rated, while organisational strength was highly rated.

Stage 4: use of a project team

Prior to this initiative, few efforts to address delirium existed and no organised group was responsible for improvement in care related to delirium. A project team was set up to function as a steering committee. Various methods to gain participation in evidence-based change have been studied and include education, opinion leader work, computer hard-wiring, quality improvement data feedback, sophisticated marketing and incentives. An underlying theory for much of this work is Rogers' Diffusion of Innovation.[27] Rogers emphasises the importance of social interactions within change. Key stakeholders and opinion leaders must be identified and involved in the process. Opinion leaders may be effective in supporting evidence implementation.[28] The steering committee was set up as an interdisciplinary team with high expertise and interest in the topic and key placement in various points in the care of the hip surgery patient. The team involved a geriatrician, a geriatric clinical nurse specialist, a geriatric nurse practitioner, a nurse coordinator for orthopaedics, a charge nurse for orthopaedics, a staff nurse from post-anaesthesia care, a nurse leader for perioperative care, a physical therapist serving the orthopaedic population, an evidence review expert and a statistician. This team assumed responsibility for instituting the Alberta Context Tool.[29] The Alberta Context Tool provides insight into the organisational environment for evidence-based change. Forty-five team members completed the tool. Results confirmed earlier assessments by the core team. Leadership was highly rated. The organisation seeks best practice, sponsors staff development, is highly patient centred and has strong teamwork. Social capital was among the highest areas rated. Team feedback was moderately high in rating. Furthermore, the team assumed responsibility for supporting all aspects of their implementation plan.

Stage 5: description of best practice

In the absence of any literature to guide the process of predictor tool use, the local steering committee identified the steps in the process of introducing and using the tool. Refinements to the appearance of the hard copy predictor tool were made based upon team feedback (see Fig. 1). The team supported a prompt for further action in delirium prevention as part of the tool. This included the reminder to perform a CAM, a prompt to collaborate with the physician in charge of the case and encouragement to enact the standard care plan for delirium prevention available within the clinical information system. The team created a small group education package using a standard PowerPoint presentation by the nurse coordinator for orthopaedic services. A project initiation date was chosen that was assessed to be free of major concurrent pressures. Audit and feedback were planned to further support the implementation of evidence.[30]

Figure 1.

Delirium predictor tool.

Description of the implementation methods

The steering committee approved a process of predictor tool use involving both the post-anaesthesia unit and the orthopaedic unit. The process involved the following:

  • Peri-anaesthesia (post-anaesthesia care unit (PACU)) nurse case identification
  • Volunteer posting of manual form to charts
  • Communication handoff of case identification between PACU and orthopaedic unit
  • Orthopaedic unit 7 am dual shift rounding to complete the predictor tool
  • Communication of results to physician
  • Prompting for documentation of delirium status in the existing CAM chapter of the clinical information system on the DPT documentation form

Post-data collection and analysis

Post-implementation, the data elements collected were pared down to 13. These included the following:

  • Age; sex
  • Diagnosis
  • Emergency surgery status
  • Haematocrit value
  • History of depression
  • Use of benzodiazepines during stay
  • CAM completed
  • CAM documented delirium presence
  • Narrative chart documentation of delirium
  • Plan of care instituted for delirium in high-risk patients
  • Length of stay
  • Discharge disposition

Data were collected after the June 2011 implementation phase-in through January 2012, in a similar fashion as the pre-implementation data collection. The analysis primarily focused on what the incidence of delirium was and if the predictor variables remained significant.


Initial compliance to use of the predictor tool was assessed at 54% within 3 months of implementation and increased to 56% in the ensuing 3 months with revisions to weekend procedures for tool placement in the chart. Before the study use of the predictor tool, the delirium rate was 10.4% (12 of 115 patients).

An interim analysis 4 months after implementation identified a 20% delirium rate (18 of 70 patients) and an updated analysis 9 months farther into the project in February 2012 showed a 15.8% (25 of 158 patients). Those who used DPT had 0.475 times odds to develop delirium compared with those who did not use DPT; in other words, those who used DPT have half the odds to develop delirium than those who did not use it.

Ongoing assessment of the fit of the local delirium predictor variables was performed. In the post-implementation analysis, the only significant variables for delirium prediction in the local population were age and emergency surgery. The original variables did not hold true in this second cohort. The post-implementation group was older than the pre-implementation group. Those experiencing delirium in the pre-analysis group were 77.3 years of age, while those experiencing delirium in the post-analysis group were an average of 83.22 years of age. Emergency status was 40% in the pre-implementation group and 37.3% in the post-implementation group.

Despite these findings, the DPT did demonstrate some overall predictive power. An analysis of 80 patients who had a DPT used showed a sensitivity of 88.8% (proportion of delirious patients correctly identified) and a specificity of 49.3% (proportion of non-delirious patients correctly identified). The positive value predictive value (or the percentage of patients at high risk of delirium as predicted by the DPT who really developed delirium) result was 18.2%. The negative value predictive value (or the percentage of patients not at risk for delirium as predicted by the DPT who did not develop delirium) was 97.2%. The Fisher's exact test for the difference in delirium rates between those identified by DPT and those not was significant at a P-value of 0.037. In those at high risk, as determined by the DPT, 8/44 of 18.2% developed delirium, where those who were not at high risk as depicted by DPT experienced a rate of only 2.78%.

Documentation of CAM was confounded throughout the study as the nurses were allowed to document condition unchanged and not systematically evaluate each component of the CAM assessment. While CAM was ‘documented’ 95% of the time, if one accepts the status unchanged indicator, chart narratives more accurately depicted the presence of delirium both before and after the project.

Delirium predictor tool use was designed to prompt a delirium care plan. This was part of the educational component to the nurses and the tool clearly re-emphasised this. There were 110 instances of where either the DPT was not used or the DPT results were negative. This left 47 cases where DPT data were available (one missing case existed). Of these 47 cases where DPT was available and positive on more or more variables, delirium was evident in eight cases (10.6%). Care plans were initiated in 17 cases or only 36% of the time. The statistical analysis showed no significant relationship between the rate of delirium and the plan of care initiated.


The use of a predictor tool for delirium has two reasonable outcomes: (i) delirium is prevented and the rate is lower or (ii) delirium is increased as the identification is greater. The latter conclusion is seen in this performance improvement project. A change in the results was evidenced despite only 56% compliance in the use of the tool. Results were clinically meaningful although not statistically significant.

The PARiHS model calls for both evidence and context to be highly effective in order to best move evidence into practice. A major challenge in this project was the evidence quality. Initial expectations were that with the large amount of research performed on this subject, consensus on predictors or risk factors would be easily seen. This was not the case. The team then generated local research to identify their own predictors for their population. Three predictors were found. Sustained success was not within their grasp, however, as the original predictors did not remain stable in subsequent follow up. Modest predictive ability exists in the original tool. All in all, evidence strength followed a jagged course. This factor was the major contributor to the lack of success in the use of the predictor tool.

Confusion assessment method documentation and care plan use remain confounded by a myriad of practical issues. The clinical value of the current CAM documentation has led the steering committee to adopt a pilot policy on this unit that CAM elements must be documented, in full, each shift. The element ‘no change from prior shift’ will no longer be possible. Care plans are currently being updated on the topic to ensure maximum relevance.

Delirium remains a less than optimally managed condition. Delirium experts see delirium as an indicator of the overall quality of care rendered in hospitals today.[31] The literature indicates that clinicians do not recognise delirium.[32] Healthcare professionals are challenged to make incremental improvements in care of patients at risk for delirium given its profound impacts.

A primary limitation of this work was that both the evidence and the implementation were carried out on one unit in one hospital, in central New Jersey. Efforts are underway to assess predictors in other facilities of this healthcare system.

The role of predictor tools may go beyond their face value as they may serve as a reminder tool for best practice. That may be the clearest lesson learned through this work to date. Care improvement often needs to begin at the front end of the process, in the awareness and assessment stages. This study endeavoured to impact this fundamental phase of the care process. A modest result was seen despite the use of best practices in evidence introduction.


The authors acknowledge the contributions of Ms Amy Wozniak, MS, biostatistician, Clare Tang, MS, biostatistician and Ms Sharon Lubeck, RN at the local level and Susan Salmond, Cheryl Holly, Alison Kitson, Rick Weichula, Tiff Conroy and Tim Schultz at the Signature Project level.