Setting and Study Population
All research was undertaken in high-income–country settings, including the United States (5/17, 29%), the United Kingdom (3/17, 18%), Canada (3/17, 18%), Sweden (2/17, 12%), Belgium (1/17, 6%), France (1/17, 6%), Norway (1/17, 6%), and the Netherlands (1/17, 6%).
The study population varied greatly from study to study. In studies involving care or nursing homes residents, the largest study population was 6161 residents [21], and the smallest study population was 52 residents in the study by Walker et al [24]. In studies involving health care professionals, the largest number of participants was 27 staff members in the study by Coulongeat et al [10], and the smallest number of participants was 14 registered nurses in the study by Johansson-Pajala et al [16].
RQ1: What Has Been the Purpose of Clinical Decision Support Tools? Which Professionals Are the Intended Users?
There were 8 different conditions or purposes supported through the use of clinical decision support tools. They included medication management (5/17, 29%) [12,13,15-17], pressure ulcer prevention (4/17, 24%) [9,11,14,21], dementia management (3/17, 18%) [18-20], falls prevention (3/17, 18%) [21,23,24], hospitalization (2/17, 12%) [21,22], malnutrition prevention (1/17, 6%) [14], urinary tract infection (1/17, 6%) [25], and COVID-19 infection (1/17, 6%) [10].
In total, 65% (11/17) of the studies defined the professional user group using the decision support tool, whereas 35% (6/17) did not indicate who the intended users of the tools were [11,13,17,21,24,25]. Where stated (5/11, 45%), the most commonly specified professionals were either care home staff or nursing home staff [10,14,18,22,23]. Of these 5 studies, 4 (80%) specified that the clinical decision support tool that was evaluated was used by nurses; 50% (2/4) of these studies provided further detail on the types of nursing staff, including registered nurses, special needs educator, and nurse aides [14] or directors, physicians, and nurses [10]. Of the 11 studies that defined the professional user group using the decision support tool, 1 (9%) specified that the users of the decision support tool were pharmacists [12], and 1 (9%) stated that the tool was used by health professionals [16].
RQ2: What Evidence Base Was Used to Develop the Clinical Decision Support Tools?
Overall, 88% (15/17) of the studies provided further information on how the support tools were developed. Among these 15 studies, 9 (60%) stated that the tools were developed using clinical guidelines [13,15,17,20,25-29], 11 (73%) stated that the decision support tools were developed through users’ opinions [13,17,20,27-33] (eg, through the Delphi method [9,28] or other ways of stakeholder involvement [17,20,27,29,33]), 2 (13%) stated that the support tools were developed using systematic reviews of scientific evidence [34,35], and 2 (13%) used data analysis to understand the factors associated with the investigated outcome [30,32].
Of the 17 included studies, 11 (65%) evaluated the effectiveness of the decision support interventions, but no clear consensus could be arrived at on the association between the effectiveness and involvement of stakeholders or use of clinical guidelines in their development and effectiveness. Of these 11 studies, 2 (18%) evaluating tools that involved stakeholders in their development achieved significantly positive results [9,21], whereas 4 (36%) achieved mixed results [16,17,23,24], and 3 (27%) obtained nonsignificant results if the tools involved stakeholders in the development [11,20,22]. When it came to using clinical guidelines for developing decision support tools, none of the 5 studies concerned reported significant results: 4 (80%) demonstrated mixed results [16,17,23,25], and 1 (20%) obtained nonsignificant results [20].
RQ3: How Are Clinical Decision Support Tools Used in Adult Long-term Care Facilities? What Is the Effectiveness of These Tools?
Medication Management
Of the 17 included studies, 5 (29%) focused on medication management (12,13,15-17).
De Wit et al [12] evaluated a clinical decision support system that was designed for medication management. The system operates by extracting medication data of residents and 2 weeks’ worth of historical laboratory data from electronic health records for all residents. Since 2008, a total of 39 clinical rules have been created in the system. If a laboratory value is deemed to be abnormal, in combination with the appropriate drug, the system generates an alert. The system then helps with dosage adjustments in accordance with various conditions, such as decreased renal function or electrolyte dysfunction [12].
The study showed that only 3 clinical rules had an efficiency of >10% (phenytoin with hypoalbuminemia, bisphosphonates dosage regime, and ceftazidime with decreased renal function). Most of the clinical rules demonstrated efficiencies of <10%, and the efficiency of 2 rules was 0% (oral oncolytics and stop dates and methotrexate dosage regime). The efficiency was calculated by dividing the number of actions for both new alerts and repeat alerts by the total number of new and repeat alerts [12]. As this was a retrospective analysis of a database, there was no control group. This means that it is difficult to evaluate whether the system had a significant effect on improved medication management.
Dorfman et al [13] investigated the potential benefits of a clinical decision support system identifying drug-gene interactions in nursing home residents who were being treated with multiple medications. They tested the system on 987 residents at 4 nursing homes. The pharmacogenetic (PGx) system uses residents’ medication data and electronic health record information, together with genetic information stored in the electronic health records, to produce information regarding drug-drug interactions and other potentially dangerous drug therapy problems. On the basis of the information in the health record systems and algorithms built into the PGx system, the PGx system offers guidance to nurses and pharmacists. The study concluded that the intervention has the potential to be useful for nurses when obtaining a profile of patients’ medication regarding drug-drug interactions, therapeutic duplications, and warnings for unsuitable drugs [13]. However, this is a qualitative study; therefore, no statistical significance was explored.
Johansson et al [15] evaluated the LIFe-reader, which is a PDA with a mobile medical decision support system, that was developed for safer medication management in nursing homes. The tool is used to scan the European Article Number codes on drug packages, through which the LIFe-reader generates alerts for inappropriate drugs and drug combinations: drug-drug interactions, therapeutic duplications, and warnings for drugs unsuitable for older adults. In addition to the aforementioned features, the tool includes Microsoft Word, Microsoft Excel, email, calendar, calculator, and phone. Through interviewing 22 registered nurses at various care homes, the evaluation study found that the scanning function was easy and time saving, and the LIFe-reader was useful and user friendly. However, the users requested more content and functions on the device [15].
Johansson-Pajala et al [16] studied the use of a web-based computerized decision support system that was designed for drug prescribing and medication reviews. The system is linked to electronic medical records and evaluates the quality of drug treatments based on national indicators and potential adverse drug reactions based on the residents’ symptoms. The system produces warnings and explanations about inappropriate drugs, drug-drug interactions, drug use in decreased renal function, and possible adverse drug reactions. The evaluated system includes 2 widely used criteria: screening tool of older people’s prescriptions, screening tool to alert to right treatment, and Beers criteria.
Kane-Gill et al [17] evaluated a clinical decision support system called TheraDoc, a clinical surveillance system containing predeveloped alerts and customizable alerts to detect potentially inappropriate prescribing, which is integrated into electronic health records. More information can be found on the TheraDoc website [36]. Alerts were created for high-risk medications, laboratory monitoring alerts, and antibiotic-stewardship–related alerts, all developed with the purpose of preventing adverse drug events. The alerts are delivered in real time [17].
The tool is reported to have been developed using medical guidelines and users’ opinions; however, the study does not state exactly who the intended users of the system are. The evaluation was undertaken at 4 nursing homes, with 2127 nursing home residents as participants [17].
Of the aforementioned 5 studies focusing on medication management, only 2 (40%) were comparative studies. The study by Johansson-Pajala et al [16] showed the intervention to report significantly more adverse drug reactions and more drug-drug interactions than registered nurses. There was no significant difference between reports of inappropriate drugs and drug duplications when comparing the intervention with the actions of registered nurses. The study also investigated the nurses’ views on drug management; however, the results were nonconclusive. More specifically, the registered nurses did not find that the decision support system significantly affected their drug management methods; however, many saw potential benefits of using the system [16].
The study by Kane-Gill et al [17] showed that the intervention group had significantly lower incidence of alert-specific adverse drug events than usual care. There was no statistically significant difference between the groups for all-cause hospitalizations and 30-day readmissions [17].
Pressure Ulcer Management and Nutrition
Overall, 24% (4/17) of the papers focused on pressure ulcer management and nutrition [9,11,14,21].
Beeckman at al [9] evaluated an electronic clinical decision support system called PrevPlan that generates a resident-tailored protocol for pressure ulcer prevention. After data entry to the system regarding the availability of preventive materials and residents’ characteristics (manual entry), the protocol included recommendations regarding skin observation, the use of support surfaces, repositioning, and heel elevation. The evaluation, which involved 464 nursing home residents and 118 health care professionals, reported that the participants had more positive than negative attitudes regarding the decision support tool, with the difference being statistically significant. It was also found that nurses with specific training regarding pressure ulcer management and higher education levels had more positive attitudes than nurses who were not experts in pressure ulcer management or were in the early years of studying to become nurses. The study found that the experimental group had significantly lower pressure ulcer incidence than the control group [9]. Further information regarding the support system can be found on the PrevPlan website [37]; however, the website was last updated in 2011 and is not available in English.
Fossum et al [14] investigated the Risk Assessment Pressure Sore (RAPS) scale for pressure ulcer risk screening and the Mini Nutritional Assessment (MNA) tool for screening nutritional status.
Both Olsho et al [21] and Davidson et al [11] evaluated On-Time Pressure Ulcer Prevention, which uses risk reports embedded in electronic health records to identify recent changes in risks for developing pressure ulcers. Although both studies evaluated the same decision support system, it can be assumed that these studies are not related. The system gathers information on residents’ nutritional status, incontinence issues, and recent pressure ulcer history. The documentation is then used to produce 4 weekly core reports, identifying residents at high risk for pressure ulcer formation, enabling monitoring of weekly changes in risk. On-Time relies on staff communication across disciplines and documentation by certified nursing assistants. A certified nursing assistant is a person who helps patients with activities of daily living and other health care needs under the direct supervision of a registered nurse. A change team incorporates the reports from On-Time into clinical workflow and identifies which changes in care are required to manage the risk of developing pressure ulcers [11,21].
Olsho et al [21] carried out the evaluation, which used interrupted time-series design, at 25 nursing homes with 6161 nursing home residents as participants. The study found that the intervention components reduced pressure ulcer incidence individually and in combination [21].
Davidson et al [11] evaluated the system at 47 nursing homes; however, the study did not specify the number of participants. For their evaluation method, Davidson et al [11] used difference-in-differences design and investigated the scalability of the intervention. In spite of the large number of nursing homes involved in the study, the authors did not find whether the tool played a significant role in improving pressure ulcer prevention [11].
Of the 4 studies that investigated clinical decision support systems that aimed to help prevent pressure ulcer formation, 2 (50%) found that the incidence of pressure ulcers significantly decreased in the intervention group [9,21]; however, the remaining 2 (50%) studies found no statistically significant differences in terms of pressure ulcer incidence [11,14]. Only the study by Fossum et al [14] used a validated tool (RAPS) for pressure ulcer management. In their study, Fossum et al [14] did not find significant differences in nutritional status of residents between the intervention and control groups when using the MNA.
Dementia Management
In total, 18% (3/17) of the studies focused on evaluating decision support tools developed for dementia management. Keenan et al [18] and Moniz-Cook et al [20] evaluated 2 different decision support systems that aimed to help care home staff support residents with commonly occurring challenging behaviors. The system included assessment tools that collected relevant information regarding the residents and then applied logic-based algorithms that generated biopsychosocial action plans that the staff could implement [18,20].
The study by Keenan et al [18] was a qualitative study looking at the contextual and organizational mechanisms of, as well as barriers and facilitators for, the intervention. Four mechanisms of implementation of the intervention were identified: (1) access to, and use of, care homes; (2) resources in terms of IT for e-learning activity; (3) demonstrating capacity to apply action care planning in care practice; and (4) receptivity of care home staff to e-learning and the individually tailored action care planning that followed [18].
The study by Moniz-Cook et al [20] was a cluster randomized trial undertaken at 63 care homes. In total, 658 nursing home residents and 436 care home staff members took part in the study. It was found that there was no statistically significant difference in the number of incidents of challenging behavior between the intervention and control groups. The intervention did not significantly affect the experience of staff members with regard to the prescription of psychotropic medication. The quality of life of residents was not measured because of the large amount of missing data [20].
Kovach et al [19] evaluated a decision support tool called the Serial Trial Intervention, which aimed to help with assessment and treatment of pain and other physical problems of residents with advanced dementia who are unable to report symptoms clearly or consistently. The tool is a 9-step assessment and treatment process, previously evaluated as a 5-step tool. If an assessment is negative, or if interventions fail to decrease symptoms, the nurse moves to the next step. The study, undertaken at 12 nursing homes with 125 nursing home residents, compared the effectiveness of the 2 versions of the protocol. It found that the residents being treated using the 9-step intervention received more assessment-driven treatment and evaluation-driven follow-up. It was also found that these residents had less static and dismissive care than those treated using the 5-step intervention [19].
Falls Prevention
Of the 17 included studies, 2 (12%) focused on evaluations of interventions that aimed to support the management or prevention of falls. Tzeng et al [23] carried out a quality improvement project to evaluate the impact of the Fall Tailoring Interventions for Patient Safety program on preventing falls and fall-related injuries among residents. This program was developed to help staff modify falls prevention interventions based on daily assessments, and it can be used as a personalized falls prevention plan displayed on screens placed at residents’ bedsides. The program was developed to help nursing staff identify evidence-based interventions for each area of risk. In the paper, Tzeng et al [23] report that after implementing the intervention the reduction in the average monthly fall rate was clinically significant: the average monthly fall rate reduced from 10.07 falls to 7.95 falls. However, no statistical significance was reported [23].
Walker et al [24] evaluated the Guide to Action Care Home falls prevention program, which consists of posters and paper-based decision support tools in the form of a checklist that helps to identify risk factors associated with falls and suggests actions to reverse or modify these falls risk factors. Walker et al [24] found that the fall rates were lower, and there were nearly twice as many general practitioner visits at control homes than at intervention homes over 6 months of follow-up. That being said, no statistical significance was reported because of the small number of falls [24].
Other Purposes: Hospitalization, Urinary Tract Infection, and COVID-19 Infection
Pasay et al [25] evaluated a decision support tool based on the principles of building a culture of safe, effective, and sustainable antimicrobial use for urinary tract infection. The intervention consisted of 4 parts: education of physicians, nursing staff, families, and caregivers; posters with myths and facts regarding the diagnosis and treatment of urinary tract infection; a pamphlet for family and caregivers; and a clinical tool to help with behavioral changes in residents (drugs, eyes and ears, low oxygen states, infection, retention of urine or stool, ictal, underhydration or undernutrition, metabolic, and subdural [DELIRIUMS] tool). The evaluated decision support tool is a checklist that guides staff to identify urinary tract infections based on clinical symptoms, to collect a urine culture only when indicated, and to review antimicrobial therapy if prescribed. The checklist also acted as an interprofessional communication tool [25].
Pasay et al [25] found that there was a statistically significant reduction in urine testing in the intervention group compared with the control group. There was also a statistically significant reduction in the rates of antimicrobial prescribing in the intervention group compared with the control group. There were no differences in admissions to acute care or the emergency department between the 2 groups [25].
Tena-Nelson et al [22] evaluated a program called Interventions to Reduce Acute Care Transfers New York, which consisted of six parts: (1) the Situation, Background, Assessment, Recommendation (SBAR) tool, which was designed to help with communication among medical professionals; (2) the Early Warning Tool (Stop and Watch), which was designed to help in recognizing significant change in residents early; (3) a hospital transfer review tool to guide retrospective review of hospitalizations; (4) a standardized patient transfer form and a checklist; (5) care paths to guide treatment options for common conditions; and (6) advance care planning tools to guide decision-making and communication about end-of-life care.
SBAR is a structured communication format that enables health information to be transferred between individuals and institutions. It aims to convey critical information understandably, clearly, and succinctly [38].
Stop and Watch is a tool to help spot the signs warning that a person’s condition is deteriorating. The poster helps staff to recognize signs and take steps to reduce a person’s risk of morbidity, further disability, organ failure, and mortality [39].
According to the study by Tena-Nelson et al [22], there were no statistically significant changes in hospitalization rates between before and after the intervention. No statistically significant factors were found to be associated with the changes in hospitalization rate. The authors stipulated that the program’s effectiveness could be improved by including participant recommendations on planning, staff and stakeholder engagement, implementation, training, and sustainability [22].
Coulongeat et al [10] investigated a local support platform that aimed to help nursing homes manage their cases of COVID-19 infection. Although the other decision support tools included in the review consisted mostly of 2 actors (a human and a computer [or paper in some cases]), the COVID-19–infection management tool used multiple human actors as the decision support. The reason for this might be that because COVID-19, at the time of tool development, was a very new disease, there was very little evidence available to support the development of a computer program that could help manage the condition. The decision support aspect consisted of a multidisciplinary team, a specialist phone hotline, and mobile geriatric medicine teams, all reachable through information and communication technologies. The intervention helped to satisfactorily address some issues that were revealed by the COVID-19 pandemic. These issues were as follows: limiting the feeling of isolation, getting the health professionals’ questions answered, providing solutions to individual problems, and reassurance of the nursing home staff regarding the optimal treatments for residents. The intervention was less effective in improving the quality of life for residents or staff at nursing homes with a COVID-19 cluster [10]; however, why this was the case is not explained in the paper.