This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Medical Informatics, is properly cited. The complete bibliographic information, a link to the original publication on http://medinform.jmir.org/, as well as this copyright and license information must be included.
Patients with multiple conditions have complex needs and are increasing in number as populations age. This multimorbidity is one of the greatest challenges facing health care. Having more than 1 condition generates (1) interactions between pathologies, (2) duplication of tests, (3) difficulties in adhering to often conflicting clinical practice guidelines, (4) obstacles in the continuity of care, (5) confusing self-management information, and (6) medication errors. In this context, clinical decision support (CDS) systems need to be able to handle realistic complexity and minimize iatrogenic risks.
The aim of this review was to identify to what extent CDS is adopted in multimorbidity.
This review followed PRISMA guidance and adopted a multidisciplinary approach. Scopus and PubMed searches were performed by combining terms from 3 different thesauri containing synonyms for (1) multimorbidity and comorbidity, (2) polypharmacy, and (3) CDS. The relevant articles were identified by examining the titles and abstracts. The full text of selected/relevant articles was analyzed in-depth. For articles appropriate for this review, data were collected on clinical tasks, diseases, decision maker, methods, data input context, user interface considerations, and evaluation of effectiveness.
A total of 50 articles were selected for the full in-depth analysis and 20 studies were included in the final review. Medication (n=10) and clinical guidance (n=8) were the predominant clinical tasks. Four studies focused on merging concurrent clinical practice guidelines. A total of 17 articles reported their CDS systems were knowledge-based. Most articles reviewed considered patients’ clinical records (n=19), clinical practice guidelines (n=12), and clinicians’ knowledge (n=10) as contextual input data. The most frequent diseases mentioned were cardiovascular (n=9) and diabetes mellitus (n=5). In all, 12 articles mentioned generalist doctor(s) as the decision maker(s). For articles reviewed, there were no studies referring to the active involvement of the patient in the decision-making process or to patient self-management. None of the articles reviewed adopted mobile technologies. There were no rigorous evaluations of usability or effectiveness of the CDS systems reported.
This review shows that multimorbidity is underinvestigated in the informatics of supporting clinical decisions. CDS interventions that systematize clinical practice guidelines without considering the interactions of different conditions and care processes may lead to unhelpful or harmful clinical actions. To improve patient safety in multimorbidity, there is a need for more evidence about how both conditions and care processes interact. The data needed to build this evidence base exist in many electronic health record systems and are underused.
Patients affected by multiple diseases are acknowledged to be one of the greatest challenges for modern health care, especially as populations age [
Estimates of the prevalence of multimorbidity emanate from countries with detailed primary care records. A national population study carried out in the Netherlands estimated an overall prevalence of 29.7%, ranging from 10% in those younger than 20 years to 78% in those older than 80 years [
The model of care in multimorbidity is changing, from a disease- and organization-centered approach [
The presence of simultaneous care plans for multiple conditions leads to confusion and, in turn, generates safety hazards. Clear care plans, blending clinical care with self-management, are essential in multimorbidity [
Clinical research processes tend to focus narrowly on a single disease, mechanism, or treatment. This parsimony is reflected in the production of clinical practice guidelines; therefore, interactions between diseases are barely touched upon in care pathways (even if they are referred to as “integrated”) [
Even the most primary care-focused of health care systems, such as the NHS [
Multimorbid health care requires complex communication, analysis, summarization, and presentation of heterogeneous clinical information from multiple sources. It is acknowledged that electronic health records (EHRs), especially in primary care, require enhanced functionality to support decisions in these complex care processes [
Previous reviews have investigated specific aspects of CDS in multimorbidity; for example, prescribing in the elderly [
The aim was to review the current state of the art of CDS in multimorbidity. The objectives were to review the aspects of decision support target, contextual information about patients/practitioners/services, decision support technology, user interface considerations, decision maker(s), diseases, and evaluation. These aspects were analyzed to identify what works and what does not in CDS for multimorbidity, why systems failed to produce the expected outcomes, and what solutions might be adopted to address the problems.
This review follows the guidelines from Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) framework [
Studies that linked the concepts of multimorbidity, comorbidity, or polypharmacy to the concept of CDS, referring to the definitions provided previously, were selected from the literature.
The studies included in this literature review are articles about CDS systems that (1) address general issues about the multimorbid population, (2) support care for a particular subpopulation of multimorbid patients, (3) manage comorbidities related to a main disease, (4) deal with multiple concurrent medications in multimorbid population, and (5) describe statistical or machine-learning methods for clinical prediction in which the multimorbid patients’ data feed the modeling/learning and an holistic approach is adopted.
Studies excluded from this literature review were about (1) CDS characteristics in general, without describing a CDS system in detail; (2) economic evaluations of CDS; (3) CDS systems in which multimorbidity was not a key feature; (4) social and operational research into CDS with no reference to clinical outcomes; (5) statistical or machine-learning approaches in which comorbidities were part of the model, but the patient-centered approach was not considered; and (6) systems that checked drug-drug interactions by means of simple rules, without taking into account multimorbidity or comorbidities.
MEDLINE and Scopus [
Some target studies could only be found in the grey literature, such as theses and conference proceedings. Scopus allows search restrictions to some categories of grey literature, such as conference proceedings. This wider searching aimed to reduce publication bias.
The searches were performed in December 2013 and January 2014 without any restriction in the publication date.
For the search, we followed 3 key points from the Cochrane Handbook [
Searches should seek high sensitivity—this may result in poor precision.
Too many different search concepts should be avoided, but a wide variety of search terms should be combined with “or” within each concept.
Both free-text and subject headings should be used (eg, MeSH) [
The focused clinical question that drove this systematic review was: What is the current level of adoption of CDS in multimorbidity? To answer this question, 3 different search concepts were selected:
Decision support: it has many related MeSH descriptors, such as “decision support systems, management” or “decision support techniques.” Examples of individual hyponyms manually selected are “clinical decision support system,” “decision support software,” and “decision support tool.”
Multimorbidity: it has zero related MeSH descriptors. Semantically, the closed concept comorbidity has 1 MeSH descriptor. Examples of synonyms manually selected are “concurrent conditions,” “multiple chronic diseases,” and “multiple pathologies.”
Polypharmacy: it has just 1 MeSH descriptor and it should not be confused with the concept polypharmacology. Examples of synonyms manually selected are “several prescriptions,” “poly-prescriptions,” and “multiple medications.”
In essence, the search created for the focused clinical question that drove this systematic review was based on 3 different search concepts and the hyponyms and synonym terms combined with “or.” Conceptually, our clinical query was the following:
In Scopus, the query created imposed that the relevant terms selected appear in the title, abstract, or keywords. The search yielded 954 articles (see
Because multimorbidity is underrepresented in MeSH (ie, no MeSH descriptor), we created a PubMed query that looked for the relevant terms selected in the title/abstract. The search created yielded 10,223 articles (ie, 10 times more document results than in Scopus). We investigated the origin of this high number by looking at the query as it appeared under search details when using the PubMed search engine. Some of the synonyms manually selected for multimorbidity were not recognized; thus, they were split up automatically by PubMed [
Knowing other researchers who were also conducting systematic reviews in the area of CDS, we thought of a search intended for a global evidence map [
Review flow diagram.
For the PubMed article excerpts retrieved out of the broad query, we modified the manual approach to screening citations for systematic reviews and adopted some automation. In the area of automated document classification, there is an emerging body of research that uses machine-learning methods to help with the process of citation screening (eg, [
The annotation was performed using a controlled vocabulary (ie, the list of the hyponyms and synonym terms manually created for our clinical query). This annotation can coexist with native annotations from PubMed article excerpts based on MeSH and/or authors’ keywords. The concrete details of the annotation process are out of the scope of this paper. Once the annotation was performed, a selection of articles were selected based on our clinical query “<decision support> AND (<multimorbidity> OR <polypharmacy>).” Thus, only article excerpts with at least 1 term in title/abstract related to decision support and at least 1 term in title/abstract related to multimorbidity or polypharmacy were identified as related to our clinical question.
Articles obtained by the preceding procedure were combined with the ones from the Scopus search and, after removing duplicates, screened based on title and abstract. Relevant articles were assessed through full-text analysis to select the articles to be included in the systematic review.
A careful selection of relevant features was agreed on by the authors (PF, JA, and IB) and data on the following aspects were collected. A summary was generated for each data item and study.
This included clinical tasks supported by the CDS system: prevention, diagnosis, care pathway guidance (ie, management of patients according to clinical practice guidelines), medication (eg, prescription, medication review), patient education, patient self-management, and care continuity (supporting communication between health care professionals involved in multimorbid patients).
Information was collected regarding the context processed or taken into account by the system to provide support: patient clinical notes (ie, demographics or family history), laboratory results, comorbidities, medications, clinical practice guidelines, and clinicians’ knowledge.
These data included:
Mode of delivery: type of technical solution used to deliver the system: desktop application, Web application, and mobile application.
Methodology: methods used to perform the CDS intervention: data visualization techniques [
User interface considerations: reported considerations about techniques to enhance and make easier user utilization of the system: interactivity, user-centered design, summarization, and workflow graphs.
Decision maker(s): user(s) of the CDS system: nurse, specialist doctor, pharmacist, generalist doctor (ie, general practitioner or family doctor), and patient.
Diseases/conditions: CDS target conditions: obesity, diabetes mellitus, cardiovascular diseases, chronic respiratory diseases, chronic kidney disease, neurological conditions, mental health disorders, chronic musculoskeletal diseases, etc.
Evaluation: type of evaluation of system’s effectiveness: uncontrolled impact studies (eg, surveys or health services measurements before/after CDS), controlled comparisons (eg, comparing new vs old/no CDS), and no evaluation.
The results of the review were summarized in a table. The table was organized by the aspects of CDS defined previously and provides a qualitative summary for each included study. An additional quantitative summary to highlight general trends over time and patterns of evidence is also provided.
The search via Scopus retrieved 954 articles. We retrieved 17,145 articles via PubMed by using the broad search and annotation introduced previously; 79 results were recalled. After removing duplicates and screening the title and abstract, 50 articles were selected for in-depth analysis of the full text. A total of 20 studies were included in the review. The PRISMA process was followed and is reported in
Summary of collected items for included studies.
Authors | Decision support target | Contextual information | Decision support methods/delivery | User interface considerations | Decision maker/diseases | Evaluation |
Abidi [ |
Pathways (merging clinical practice guidelines for different diseases into 1 personalized guideline) | Patient clinical notes & clinical practice guidelines | Knowledge-based system (ontology) & international standards/— | — | —/— | — |
Abidi et al [ |
Diagnosis & pathways (alignment of care pathways in a patient-specific comorbid combination) & patient education | Patient clinical notes & clinical practice guidelines | Knowledge-based system (ontology based)/desktop application | Interactivity & summarization | Generalist doctor/chronic cardiovascular diseases | Controlled comparison-expert panel (revision by 2 generalist doctors and 1 specialist doctor) |
Bindoff et al [ |
Medication (review) | Patient clinical notes & medications & laboratory results | Knowledge-based system (rule based)/ — | — | Pharmacists/— | Controlled comparison-human vs system comparison (system identified more problems) |
Dassen et al [ |
Medication (prescription) | Patient clinical notes & medications & clinical practice guidelines & comorbidities & laboratory results | Knowledge-based system (ontology based) & international standards/desktop application | Interactivity & workflow graphs | Specialist doctor/cardiovascular diseases | — |
de Wit et al [ |
Medication (review) | Patient clinical notes & clinical practice guidelines & clinician knowledge & laboratory results | Knowledge-based system (rule based)/— | — | Nurses/other (home care for the elderly) | No evaluation |
Duke et al [ |
Medication (review) | Medications & clinician knowledge | Knowledge-based system & data visualization techniques & natural language processing/Web platform | Interactivity & summarization | Specialist doctor & generalist doctor/— | Controlled comparison-new vs old system (same accuracy but decreasing in time of 60%) |
Farkas et al [ |
Diagnosis (comorbidities) | Patient clinical notes | Natural language processing/— | — | —/obesity | Controlled comparison-simulations (Fβ=1 score of 97% for classification based on textual evidence and 96% for intuitive judgments; Fβ=1 score of 76% for classification based on textual evidence and 67% for intuitive judgments) |
Georg et al [ |
Medication (prescription) | Patient clinical notes & clinical practice guidelines | Knowledge-based system (rule based)/— | — | Generalist doctor/cardiovascular diseases | — |
Grando et al [ |
Medication (prescription) | Patient clinical notes & clinical practice guidelines | Knowledge-based system (ontology based)/— | — | Generalist doctor/chronic respiratory diseases & diabetes & cardiovascular diseases & chronic musculoskeletal diseases & others | — |
Jafarpour et al [ |
Pathways (merging clinical practice guidelines for different diseases into 1 personalized guideline) | Patient clinical notes & clinical practice guidelines & clinician knowledge | Knowledge-based system (ontology based)/— | — | Generalist doctor/cardiovascular diseases | No evaluation |
Martínez-García et al [ |
Care continuity & pathways | Patient clinical notes & clinical practice guidelines & clinician knowledge | International standards & social network techniques/Web application (linked to electronic health record) | — | Nurse, generalist doctor, specialist doctor/— | Controlled comparison-survey (positively judged) |
Michel et al [ |
Medication (prescription) | Patient clinical notes & clinical practice guidelines & clinician knowledge & medications & laboratory results & comorbidities | Knowledge-based system & data visualization techniques & international standards/desktop application (linked to electronic health record) | Summarization | Generalist doctor/chronic pain (opioid treated) | — |
Naureckas et al [ |
Diagnosis & pathways | Patient clinical notes & clinical practice guidelines | Knowledge-based system & data visualization techniques/desktop application (linked to electronic health record) | User-centered design | Generalist doctor/child obesity and related diseases (eg, diabetes, cardiovascular diseases, chronic kidney disease) | Impact evaluation-service performance metrics & survey |
Riaño et al [ |
Diagnosis & medication (prescription) & pathways (developing a personalized treatment) & prevention | Patient clinical notes & clinical practice guidelines & clinician knowledge | Knowledge-based system (ontology based) & international standards/desktop application (linked to electronic health record) | — | Generalist doctor/home care in long-term conditions (eg, obesity, diabetes, cardiovascular diseases, chronic respiratory diseases, chronic kidney disease, neurological conditions, mental health disorders, chronic musculoskeletal diseases) | Controlled comparison-survey (positively judged) |
Riaño et al [ |
Medication (prescription) | Patient clinical notes & clinician knowledge | Knowledge-based system (rule based)/— | — | Generalist doctor/cardiovascular diseases & diabetes | Controlled comparison-expert panel (results validated by a generalist doctor) |
Suojanen et al [ |
Diagnosis | Patient clinical notes & clinician knowledge | Machine learning/— | — | Specialist doctor/chronic neurological diseases | Controlled comparison-simulation (out of 30 cases: false positive rate=19%; false negative rate=23%) |
Vallverdú et al [ |
Medication (prescription) | Patient clinical notes & clinician knowledge | Knowledge-based system (rule based)/desktop application | — | Generalist doctor/cardiovascular diseases & diabetes | Controlled comparison-expert panel (agreement with output from the system 100%-20/20) |
Wicht et al [ |
Diagnosis (comorbidities) | Patient clinical notes & clinician knowledge | Knowledge-based system + data visualization techniques/Web platform | Interactivity | Specialist doctor/other (cancer) | Controlled comparison-expert panel (agreement with output from the system 84%-26/31) |
Wilk et al [ |
Pathways (merging clinical practice guidelines for different diseases into 1 personalized guideline) | Patient clinical notes & clinical practice guidelines & clinician knowledge | Knowledge-based system (rule-based constraint logic programming)/— | Workflow graphs | Generalist doctor/other (duodenal ulcer, transient ischemic attack) | — |
Wilk et al [ |
Pathways (alerting physicians about possible adverse interactions between 2 concurrent clinical practice guidelines) | Patient clinical notes & clinical practice guidelines | Knowledge-based system (rule-based constraint logic programming [ |
— | Specialist doctor & generalist doctor/chronic neurological & gastrointestinal diseases | — |
Most articles reviewed focused on 1 of 3 clinical tasks: medication (n=10), clinical guidance (n=8), and diagnosis (n=6). From a methodological point of view, knowledge-based systems were the most frequently used (n=17). To further illustrate this, Riaño et al [
Abidi et al [
In the articles reviewed, medication was the main theme by far. This clinical task had the most contextualized input data and appeared as prescription (n=7) and medication review (n=3). Michel et al [
Another prevalent theme was the possible interaction between concurrent clinical practice guidelines for multimorbid patients. For example, Abidi et al [
Some studies addressed the diagnosis of comorbidities for patients affected by an index condition/disease. For example, Farkas et al [
Synthesis of occurrences’ numbers and references for collected data items.
Theme and category | Frequency | References | |
|
|
|
|
|
Prevention | 1 | [ |
|
Diagnosis | 6 | [ |
|
Pathway | 8 | [ |
|
Medication | 10 | [ |
|
Patient education | 1 | [ |
|
Continuity of care | 1 | [ |
|
Self-management | 0 | — |
|
|
|
|
|
Data visualization techniques | 4 | [ |
|
Social network techniques | 1 | [ |
|
International standards | 5 | [ |
|
Machine learning | 1 | [ |
|
Natural language processing | 2 | [ |
|
Knowledge-based system | 17 | [ |
|
Mobile technologies | 0 | — |
|
|
|
|
|
Patient clinical notes | 19 | [ |
|
Laboratory results | 4 | [ |
|
Comorbidities | 2 | [ |
|
Medications | 4 | [ |
|
Clinician knowledge | 11 | [ |
|
Clinical practice guidelines | 13 | [ |
|
|
|
|
|
Nurse | 2 | [ |
|
Specialist doctor | 6 | [ |
|
Generalist doctor | 13 | [ |
|
Pharmacist | 1 | [ |
|
Patient | 0 | — |
|
Not specified | 2 | [ |
|
|
|
|
|
Obesity | 3 | [ |
|
Diabetes | 5 | [ |
|
Cardiovascular diseases | 9 | [ |
|
Chronic respiratory diseases | 2 | [ |
|
Chronic kidney diseases | 2 | [ |
|
Chronic neurological conditions | 3 | [ |
|
Mental health disorders | 1 | [ |
|
Chronic musculoskeletal diseases | 2 | [ |
|
Other | 8 | [ |
|
Not specified | 4 | [ |
|
|||
|
|
|
|
|
Interactivity | 4 | [ |
|
User-centered design | 1 | [ |
|
Summarization | 3 | [ |
|
Workflow graphs | 2 | [ |
|
Not specified | 13 | [ |
|
|
|
|
|
Impact evaluation (service performance metrics) | 1 | [ |
|
Impact evaluation (survey) | 1 | [ |
|
Controlled comparison (expert panel) | 4 | [ |
|
Controlled comparison (survey) | 2 | [ |
|
Controlled comparison (simulation) | 2 | [ |
|
Controlled comparison (human vs system) | 1 | [ |
|
Controlled comparison (new vs old system) | 1 | [ |
|
No evaluation | 2 | [ |
|
Not specified | 7 | [ |
For the decision makers, generalist doctors were the most cited users of the CDS systems (n=13) followed by specialist doctors (n=6). No articles reported the patient as the decision maker. The system that appeared to involve the largest number of decision makers was described by Martinez-Garcia et al [
For disease, many articles considered multiple diseases (eg, [
For user interface considerations, most articles (n=13) did not provide details about the user interface. Where this information was provided, interactivity (n=4) [
Regarding type of evaluation, some articles reported effectiveness objectively, including controlled comparisons (n=9) or impact evaluations (n=1). The articles that conducted surveys about their systems achieved positive judgments about the outcome provided [
This literature review found a modest number of articles addressing CDS and multimorbidity—an evidence base disproportionately small in comparison to the need for decision support.
Most articles dealt with CDS targets that (1) were narrowly defined in terms of comorbidities around an index condition or (2) considered patient comorbidities only during prescription for a specific condition. Thereby, only a few of the studies reviewed referred to multimorbidity using a patient-centered approach, which is the ideal [
An important challenge of multimorbidity in CDS is the combination of clinical practice guidelines in a nonharmful way [
Discontinuity of care between different health professionals is an important source of safety problems, which is highly relevant to multimorbidity considering the large numbers of professionals involved. Yet only 1 article [
Self-management is key in multimorbidity [
From a methodological point of view, knowledge-based systems were most commonly reported. Data-driven methods, such as machine-learning techniques, were barely used in the reviewed studies, with just 1 study adopting them [
Multimorbidity is composed of interacting variables; therefore, systems need to be aware of as many contextual factors as possible to deliver relevant support and information [
Evaluations of usability and effectiveness of systems are key to avoiding patient harm and waste in health care systems [
This review has several limitations. First, only Scopus and PubMed sources were searched—other relevant material may exist in the grey literature. Second, the titles and abstracts of the articles selected are anchored to the terms included in the 3 thesauri—some articles may have been missed if other synonyms were used. Third, it was not possible to find studies covering all aspects of CDS we considered—some aspects, such as the evaluation of the effectiveness and usability, were quite sparsely covered, but this is a general weakness of the CDS literature [
This review shows how multimorbidity is understudied in CDS, yet this is an area of public health and clinical importance that should be a prime target for CDS research.
There are already many technologies in health care and industry relevant to dealing with the complexity of multimorbid decision support. Kawamoto et al [
Multimorbidity is a relatively new field of clinical research and more evidence is needed to support CDS in this area. This underpinning knowledge is, however, challenging. For example, patients with multiple conditions or on multiple medications are often excluded from clinical trials [
Patients with multiple conditions are one of the most important groups for health care systems to understand and evolve to serve [
electronic health record
clinical decision support
National Institute for Health and Care Excellence
Funded by the national Institute for Health Research Greater Manchester Primary Care Patient Safety Translational Research Centre (NIHR GM PSTRC) and the MRC Health eResearch Centre, Farr Institute, UK (MR/K006665/1). The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health.
None declared.