Introduction
Advances in medical care and public health policies lead to an ageing population resulting in an increasing rate and burden of chronic diseases (Mathers and Loncar, Reference Mathers and Loncar2006) along with a rise in healthcare costs (Soual, Reference Soual2017).
The concept of multimorbidity was published for the first time in 1976 (Brandlmeier, Reference Brandlmeier1976). According to the “World Health Organization (WHO), multimorbidity is defined as the coexistence of two or more chronic health conditions in the same individual” (WHO, 2008).
The term ‘condition’ led to criticisms and different interpretations, yet many attempts to refine this definition have been unsuccessful (van den Akker et al., Reference van den Akker, Buntinx and Knottnerus1996; Almirall and Fortin, Reference Almirall and Fortin2013). In particular, this definition failed to assess which items were suitable for the prevention of severe issues or were useful for handling multimorbidity in primary care consultations (Muth et al., Reference Muth, Beyer, Fortin, Rochon, Oswald, Valderas, Harder, Glynn, Perera, Freitag and Kaspar2014a; Muth et al., Reference Muth, van den Akker, Blom, Mallen, Rochon, Schellevis, Becker, Beyer, Gensichen, Kirchner and Perera2014b).
When MM is defined by two or more conditions, almost half of the patients in primary care turn to be multimorbid (Harrison et al., Reference Harrison, Britt, Miller and Henderson2014). Family physicians (FPs) are in the first-line for care of multimorbid patients (Huber et al., Reference Huber, Knottnerus, Green, Van Der Horst, Jadad, Kromhout, Leonard, Lorig, Loureiro, Van Der Meer and Schnabel2011; Willadsen et al., Reference Willadsen, Bebe, Koster-Rasmussen, Jarbøl, Guassora, Waldorff, Reventlow and de Fine Olivarius2016). They are thus concerned by the challenging concept of multimorbidity’s definition that relates to a global view of the patient in the scope of family medicine (FM) according to the World Organization of National Colleges, Academies and Academic Associations of General Practitioners/FPs (WONCA) (Allen et al., Reference Allen, Gay, Crebolder, Heyrman, Svab and Ram2002). Investigating the complexity of multimorbidity may improve FPs’ ability to detect a patient’s frailty as well as to prevent complications, that is hospitalization or death.
The concept of multimorbidity aroused considerable interest in the research agenda from the European General Practitioners Research Network (EGPRN) (Hummers-Pradier et al., Reference Hummers-Pradier, Beyer, Chevallier, Eilat-Tsanani, Lionis, Peremans, Petek, Rurik, Soler, Stoffers and Topsever2010). A research team including 9 European countries involved in the EGPRN aimed to clarify the concept of multimorbidity (Le Reste et al., Reference Le Reste, Nabbe, Lygidakis, Doerr, Lingner, Czachowski, Munoz, Argyriadou, Claveria, Calvez and Barais2012). In 2012, a comprehensive definition of multimorbidity in FM and long-term care was presented by the EGPRN (Le Reste et al., Reference Le Reste, Nabbe, Manceau, Lygidakis, Doerr, Lingner, Czachowski, Munoz, Argyriadou, Claveria and Le Floch2013). It was the first attempt to define multimorbidity from the FP’s pragmatic point of view (Le Reste et al., Reference Le Reste, Nabbe, Lazic, Assenova, Lingner, Czachowski, Argyriadou, Sowinska, Lygidakis, Doerr and Claveria2016). According to the EGPRN, multimorbidity is defined as any combination of chronic disease with at least one other disease (acute or chronic) or psychosocial factor (associated or not) or somatic risk factor. Any biopsychosocial factor, any somatic risk factor, the social network, the burden of diseases, the health care consumption and the patient’s coping strategies may function as modifiers (of the effects of multimorbidity). Multimorbidity modifies health outcomes and leads to increased disability or a decreased quality of life or frailty. Thirteen themes were outlined and translated into ten European languages in order to enable and standardize collaborative studies (Le Reste et al., Reference Le Reste, Nabbe, Rivet, Lygidakis, Doerr, Czachowski, Lingner, Argyriadou, Lazic, Assenova and Hasaganic2015a).
Nevertheless, the latter exhaustive definition is too broad and thus limits the identification of patients at risk of severe outcomes. The prevention of acute hospitalization or death being a major concern of FM (Galvin et al., Reference Galvin, Gilleit, Wallace, Cousins, Bolmer, Rainer, Smith and Fahey2017), the EGPRN considered that the highest priority should be given to identify which variables could prevent these outcomes (Le Reste et al., Reference Le Reste, Nabbe, Rivet, Lygidakis, Doerr, Czachowski, Lingner, Argyriadou, Lazic, Assenova and Hasaganic2015b). Such variables could be integrated within the FPs’ medical software and become a useful tool to follow-up multimorbid patients (Lussier et al., Reference Lussier, Richard, Glaser and Roberge2016; Lee et al., Reference Lee, Heckman, McKelvie, Jong, D’Elia and Hillier2015).
We previously conducted a feasibility cohort study in 2014 including patients from primary care in Western Brittany, France, meeting the EGPRN definition of multimorbidity and followed up for 6 months (Le Reste et al., Reference Le Reste, Nabbe, Billot Grasset, Le Floch, Grall, Derriennic, Odorico, Lalande, Le Goff, Barais and Chiron2017). The present study aimed to continue this study and to follow-up patients for 24 months in order to assess which criteria from the EGPRN concept of multimorbidity could predict outpatients at risk of decompensation (death or acute hospitalization) at 24 months of follow-up.
Materials and methods
Ethics statement
The authors confirm that all methods were carried out in accordance with relevant national guidelines and regulations. The study was approved by the ethics committee of the “Université de Bretagne Occidentale” Faculty of Medicine, Brest, France. All participants signed a written informed consent.
The detailed method was published previously (Le Reste et al., Reference Le Reste, Nabbe, Billot Grasset, Le Floch, Grall, Derriennic, Odorico, Lalande, Le Goff, Barais and Chiron2017).
Participant selection
Patients were recruited by 31 FPs in Western Brittany, France, selected from the Clinical Teacher list of Brest University.
Included patients met the EGPRN definition of multimorbidity, that is, having a chronic disease, with at least one other disease (acute or chronic) or a biopsychosocial factor (associated or not) or a somatic risk factor.
Exclusion criteria were: criteria of the multimorbidity’s definition unmet, nursing home residents, inability to be followed for the duration of the study, being under legal protection, and having a life-threatening condition or a life expectancy estimated to be less than three months.
Data collection
Participants were given full information on the study by the FP and signed informed consent at the first visit. A questionnaire was filled out by the FPs to assess the risk factors of decompensation among themes and subthemes of multimorbidity (Le Reste et al., Reference Le Reste, Nabbe, Billot Grasset, Le Floch, Grall, Derriennic, Odorico, Lalande, Le Goff, Barais and Chiron2017). The questionnaire was designed by the research group with reference to the definition of multimorbidity validated by the scientific committee of the research team and tested by FPs and medical students (Le Reste et al., Reference Le Reste, Nabbe, Billot Grasset, Le Floch, Grall, Derriennic, Odorico, Lalande, Le Goff, Barais and Chiron2017). It includes 52 questions covering each item of the definition of multimorbidity, according to a clinical and anamnestic approach of the FP.
According to the results from the feasibility study, a revised questionnaire was used for patients included in the study from September to November 2015. The rank of questions was modified to improve its administration. No question could be asked before the previous one had been answered in order to avoid missing data. The FP was only asked question number 40 if he/she had answered “yes” to question 39 related to organized or individual screening, in to avoid errors in completion.
Biopsychosocial risk (defined as psychological and psychosocial risk factors), lifestyle, socio-demographic characteristics (age, gender), psychological distress, ageing, beliefs and expectations of patients as well as data regarding physiology and pathophysiology were collected. The somatic risk was assessed according to the cardiovascular risk factor, risk factors for falls (calculated with the CETAF score) (Bongue et al., Reference Bongue, Hugues, Achour, Colvez and Sass2016) and assessment of hygiene, nutrition and physical activity assessed by the FP. Although the CETAF score was not validated for patients under 65-year-olds, we assumed that it would not be high for people under 65 years old and would not have impacted our result. Therefore, the CETAF score was calculated for every patient (HAS, 2013).
According to the previous feasibility study, the following irrelevant variables were not considered in the present study: chronic condition redundant with chronic disease or psychological risk factor, cost of care (impossible to estimate given the time and resources dedicated to the study), disability (disability/impairment, quality of life and health outcomes concerning the consequences rather than the characteristics of the multimorbidity), frailty (absence of consensual definition, criterion assessed by study and methodologically impossible to assess at the beginning of the study), physiology (broad notion, impossible to evaluate), disease and assessment (present in the theoretical definition but missing from the coding book and not found in the transcripts), demography and aging (duplicates of sociodemographic characteristics)
Follow-up data
Information on the status of patients was collected by FPs (contacted by email or phone) 24 months after inclusion. According to a consensus from a peer group of physicians, residential students and researchers in family practice, status was defined as: “decompensation” (D) or “nothing to report” (NTR). Decompensation was defined as any hospitalization with a duration of at least 7 days (the mean duration for hospitalization in the European Union is 6.7 days) or death, within 24 months of follow-up (DRESS, 2017).
Data analysis
Data from questionnaires were computed by the online survey software EVA-LANDGO®. About 102 chronic diseases were reported. Chronic diseases were referenced using the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10).
Missing data were replaced by the median value to be considered in the statistical analysis.
The status of patients (“decompensation” and “nothing to report”) at 24 months were compared using a bi-dimensional analysis for each variable.
Quantitative variables were compared using Fischer’s exact or Chi-2 test. Qualitative variables were compared using Shapiro–Wilk’s test (comparisons of medians) with an alpha-risk at 5 percent.
A logistic regression was used to predict the patient’s status at the sixth and ninth month of follow-up with the status as the dependent variable (D versus NTR). A Cox model was performed for each patient. The aim was to find the best subgroup of variables for predicting and explaining the patient’s status at 24 months.
The overall survival was compared for each variable between the two groups D and NRT, using a non-parametric estimation of Kaplan–Meier and a Logrank test (Figure 1). Then, the survival function using semi-parametric models was estimated. Univariate and multivariate analyses using Cox’s regression model were presented as hazard ratios (HRs), estimating the association between a variable and the risk of decompensation. Significant variables in univariate analysis were selected by an expert consensus to be integrated into the multivariate model.
Figure 1. Overall survival curve of the cohort.
For all analyses, a P-value ≤ 0.05 was considered statistically significant.
Results
Participants
About 137 patients were included by 31 FPs, among whom 6 were excluded (incomplete questionnaires or duplicates). The status at 24 months was collected for all 131 patients. 11 of them were lost to follow-up for change of FP or the FP ceased his activity (Figure 2).
Figure 2. Flow chart.
Comparison of patients characteristics at 24 months
At 24 months, 44 patients (36.6%) were in the group D and 76 (63.4%) in the NTR group. Comparisons of v-were presented in Table 1. Compared to patients from the group NRT, patients from group D were significantly more likely to suffer from postural instability (73% vs. 49%, P = 0.018) and to be single or widowed (52% vs. 30%, P = 0.028), to benefit from readily available equipment and human resources at their home (39% vs. 13%, P = 0.003% and 52% vs. 26%, P = 0.008, respectively); and had a more detailed and complex medical history (93% vs. 68%, P = 0.004). Among the quantitative variables, six were significantly different between the two groups. Compared to the group NRT, patients in the group D were older (80 vs. 69 years, P < 0.001), had more chronic and acute conditions (7 vs. 6, P = 0.016), had more frequently visited their FPs (12 vs. 4, P = 0.010) and were exposed to more medications per day (8 vs. 7, P = 0.003). The mean number of FPs’ visits per year was higher for patients from group D compared to the group NRT (12.52 vs. 9.08, P < 0.001).
Table 1. Themes and subthemes of the multimorbidity according to EGPRN’s definition of multimorbidity
Survival analysis
Overall survival at 24 months was compared for each variable between the D and NTR groups (Figure 1).
The probability of decompensation at 24 months was 36.7% (95% CI 27.4%–44.7%). Two variables had a significant protective effect from decompensation: excess weight (P = 0.038) and not being single or widowed (P = 0.015). Four variables were significantly associated with decompensation: a detailed and complex medical history (P = 0.003), human support/caregivers at home (P = 0.002), equipment to help at home (P-value < 0.001) and patients with multiple complaints (P = 0.03).
Table 2 presents the variables associated with the risk of decompensation in univariate analysis (Cox regression). Twenty-five variables were associated with the risk of decompensation, yet only 16 were statistically significant. Four (excess weight, being in a relationship, global vision of patient’s diseases, good quality of communication) had a protective effect and twelve were predictive of a decompensation. For determining which combination of variable could be clinically relevant for the multivariate analysis, the research team used a formalized consensus method (using a Delphi method) with the international group of expert, we issued from the EGPRN from the beginning of the survey. The following question was used: as a clinician which factors would you consider as interesting to predict decompensation (ie death or acute hospitalization) in the following factors? Consequently, the protective factors did not meet any interest for the experts who reached an agreement with: the number of FPs’ visits per year, total number of diseases, CETAF score and multiple complaints. According to the multivariate analysis, the number of FPs’ visits per year and the total number of diseases were significantly associated with the risk of decompensation (HR = 1.06, 95% CI 1.03–1.10, P < 0.001 and HR = 1.12, 95% CI 1.01–1.25, P = 0.039, respectively).
Table 2. Characteristics of D group and NTR group for each variable as described by FPs
NI: not interpretable. Bold: significantly results.
Discussion
Statement of principal findings
Among criteria from the EGPRN definition of multimorbidity, the number of visits to FPs and the total number of diseases predicted patients at risk of decompensation at 24 months of follow-up. The cut-offs found were 12 FP’s visits per year and 7 diseases. This significant difference is not due to an intervention. The whole survey is observatory, and FPs did not know in advance who will decompensate and who will not. They were not aware of which factors we were collecting and furthermore, they were in usual care with no intervention at all from the research team. Consequently, we are extremely confident that our results were the consequences of a decompensation and not from any intervention. Both variables are subthemes of “health care consumption” in the EGPRN definition of multimorbidity. Previous studies failed to assess the meaning or the intensity of the relationship between multimorbidity and health care consumption (Speechley and Tinetti, Reference Speechley and Tinetti1991; Winograd, Reference Winograd1991). Both variables could simplify the prediction for decompensation in practice.
Our results are concordant with previous findings of an association between multimorbidity and the number of FPs’ visits (Le Reste et al., Reference Le Reste, Nabbe, Rivet, Lygidakis, Doerr, Czachowski, Lingner, Argyriadou, Lazic, Assenova and Hasaganic2015; Palladino et al., Reference Palladino, Tayu Lee, Ashworth, Triassi and Millett2016; Salisbury et al., Reference Salisbury, Johnson, Purdy, Valderas and Montgomery2011). The number of FP’s visits per year had also been found to be a risk factor for decompensation at 6, 9, 12, 15 and 18 months (Le Reste et al., Reference Le Reste, Nabbe, Billot Grasset, Le Floch, Grall, Derriennic, Odorico, Lalande, Le Goff, Barais and Chiron2017). In our previous study at 6 months of follow-up, we found that “age”, “number of visits to FPs” and “family problems” were associated with the risk of decompensation (Le Reste et al., Reference Le Reste, Nabbe, Billot Grasset, Le Floch, Grall, Derriennic, Odorico, Lalande, Le Goff, Barais and Chiron2017). Age being a non-modifiable factor, identification of “family problems” and “number of FPs visits” could represent preventive measures of decompensation. In the present study, age was significantly associated with decompensation in the univariate analysis but was not selected for the final model in multivariate analysis by the expert group as it was a well-known factor of decompensation. In contrast with our previous findings, “family problems”, which is part of the psychosocial risk factor theme, were not significantly associated with the risk of decompensation in the univariate analysis at 24 months. Such difference may be explained by the timeline which tends to soften family problems.
Strengths and weaknesses of this study
To ensure accurate reporting, this qualitative study was developed and reported according to the STROBE guidelines.
To the best of our knowledge, this is the first study to have assessed which items from the EGPRN’s definition of multimorbidity predicted the risk of decompensation among outpatients at 24 months.
The present study has several limitations. First, multimorbid patients were selected by FPs aware of the study’s aim. Therefore, patients with a high risk of decompensation may have been rather selected although this bias was minimized by the exclusion of patients with estimated survival of less than 3 months. Second, recruiting FPs were mainly clinical teachers, and their patients may differ from that general population (Letrilliart et al., Reference Letrilliart, Rigault-Fossier, Fossier, Kellou, Paumier, Bois, Polazzi, Schott and Zerbib2016; Peto et al., Reference Peto, Coulter and Bond1993), which limits the generalization of our results. Third, missing data were limited by the impossibility to move on without having answered to one question, yet, missing or inconsistent data remained but were replaced by medians in the analysis to reduce information bias. Fourth, chronic diseases were reported by FPs but the clustering of chronic diseases was assessed by the scientific committee using the ICD-10 to limit information bias. At last, given the high number of variables compared to the small number of patients included, the redundant and insignificant variables were removed from the analysis, according to the peer group.
Implications for practice, teaching and future research
In everyday practice, FPs and trainees in FM should consider the risk of decompensation of multimorbid patients who frequently visit them. This represents an efficient measure to monitor the risk of decompensation in primary care.
In teaching, it could be of use in both initial and continuous medical education to alert students and FPs about these two risk factors and their cut-offs.
In research, this study was a part of an EGPRN project the results need to be confirmed in a large-scale European study.
Conclusion
The number of FPs’ visits and the total number of diseases were the two independent risk factors from the EGPRN’s definition of multimorbidity for decompensation among multimorbid patients at 24 months.
A more efficient understanding of the concept of multimorbidity and the identification of risk factors of decompensation should enable optimized management of multimorbid patients as well as a reduction of healthcare costs. The two risk factors found in this survey are pragmatics, easy to find and easy to follow by FPs. This should enable a more efficient follow-up of these patients, as FPs are always short of time and need clear concepts for clinical work.
Nevertheless, the limitations of the survey are numerous and the results should be confirmed. On one hand, a large-scale causality study with a simplified questionnaire could confirm these results. On the other hand, an alternative way could use deep learning to assess a large regression model with less patients and all the variables to confirm the result. The research group has now to choose between these two options.
Data availability statement
Data supporting the findings of this study are available from the corresponding author [PA].
Acknowledgments
The authors are grateful to the FPs of the RICPRPG (réseau d’investigation Clinique en prévention des risques en population générale) for their contribution to the present survey. We also thank Ms A Gillman for her translation in English and Ms N. Marpillat for her proofreading. This article is supported by the French network of University Hospitals HUGO (‘Hôpitaux Universitaires du Grand Ouest’).
Author contributions
Conceptualization: AP, NP, LS, LGD, VJ, FJ, TAL, GF, LRJY. Data curation: LRJY. Formal analysis: AP, NP, LS, LGD, VJ, FJ, TAL, GF, LRJY. Investigation: AP, NP, LS, LGD, VJ, FJ, GF, LRJY. Methodology: AP, NP, LS, LGD, VJ, FJ, TAL, GF, LRJY. Project administration: LRJY. Resources: LRJY. Supervision: LRJY. Validation: LRJY. Writing, original Draft: AP. Writing, review & editing: AP, NP, LS, LGD, VJ, FJ, TAL, GF, LRJY. All Authors read and approved the final version of the manuscript.
Funding statement
None.
Competing interests
All authors declare no conflict of interest related to the content of this article.
Ethical standards
The study was approved by the Ethical Committee of the ‘Université de Bretagne Occidentale’. All participants have given their written informed consent.
Open access
