Multimorbidity type, hospitalizations and emergency department visits among nursing home residents: A preliminary study

Abstract

Background

The burden of multimorbidity in institutionalized elderly is poorly investigated. We examined the associations of the type of multimorbidity (i.e., physical, mental or both) with the number of hospitalizations and emergency department (ED) visits in nursing home (NH) residents.

Methods

This is a cross-sectional study among NH residents. Information on residents' health, number of hospitalizations in the last 12 months and hospital department of admission (having been seen in ED vs. non) was recorded by NH staff of 175 French NHs (data was collected in 2011). Participants were screened for the presence of several mental (e.g., dementia) and physical conditions (e.g., diabetes).

Results

Data on hospitalization was available for 6076 NH residents. Compared to having no diseases, the concomitant presence of ≥ 2 physical conditions was the multimorbidity type more strongly associated with both the number of hospitalizations (incidence rate ratio (IRR) =1.93; 95% confidence interval (CI) =1.57 − 2.37) and ED visits (odds ratio (OR)= 1.79; 95% CI=1.24 − 2.58). The presence of a mental condition appeared to moderate the associations between physical conditions and hospitalizations, since the estimate effects were lower among people who had both physical and mental conditions, compared to those with only physical conditions. For example, compared to people with ≥ 2 physical conditions, those with multiple physical and mental conditions had lower IRR (IRR = 0.84; 95% CI=0.75 − 0.95) for the number of hospitalizations.

Conclusions

Mental diseases in very old and multimorbid NH residents probably moderate the associations between physical diseases and hospitalizations. To what extent this represents either a mirror of better clinical practice in NHs or the under-recognition from the NH staff of symptoms leading to justifiable hospitalizations remains unclear.

Introduction

Research on multimorbidity in older people is scarce (1). Recently, a panel of experts proposed a set of health outcomes to be addressed in multimorbid patients (2), and concluded that gaps remain regarding the evaluation of disease burden and cognitive function in such a complex population. Although some studies have investigated the burden of co-morbid diseases in older adults (1), most of them investigated the association between specific diseases, such as depression and diabetes or heart diseases 3, 4.

Despite its relevance for individuals’ health and the health system, multimorbidity in older adults living in NHs has been poorly investigated (5). Yet, NH residents are characterized by the presence of several concomitant diseases, multiple medications (6, 7, 8), and an increased use of health care services (9), which increase their risk of developing adverse health outcomes. To our knowledge, no study has quantified the risk of hospitalizations and emergency department (ED) visits according with the type of multimorbidity (i.e., mental, physical, or a mix between mental and physical diseases) in very old residents of NHs. We therefore investigated the burden of multimorbidity type by examining the associations of multimorbidity with the total number of hospitalizations and the probability of ED visits in a large sample of NH residents.

Methods

This work used the baseline data (cross-sectional design) from the IQUARE study; data was collected in 2011 and analyzed in 2012/2013. IQUARE's protocol was fully described elsewhere (7); it will be briefly reported herein. IQUARE is a multicentric individually-tailored controlled trial developed in NHs from Midi-Pyrénées, South-Western, France (trial registration number: NCT01703689). This is a 6-month intervention, with a 18-month follow-up, designed to improve NH quality indicators. NHs were allocated to one of the following two groups: 1) audit and feedback intervention on quality indicators associated to cooperative work meetings between hospital geriatricians and NH staff, or 2) audit and feedback only. IQUARE followed the principles of the Declaration of Helsinki and complied with ethical standards in France; study protocol was approved by the ethic committee of the Toulouse University Hospital and the Consultative Committee for the Treatment of Research Information on Health (CNIL: 07-438).

Since the functioning and integration of NHs in the whole health system vary among countries, we will briefly describe how it operates in France. In this country, each registered NH is required to have a coordinating physician among its staff members; this coordinating physician, generally a geriatrician, is responsible for comprehensive health evaluation of each NH resident and for health care coordination. Drug prescription remains under the responsibility of the resident's primary care physician.

Participants

A total of 6275 residents from 175 NHs participate in IQUARE. People were randomly selected (excepted for NHs that had ≤ 30 potential participants, in which all residents were included in the study) on the basis of alphabetical order within each NH: NHs having between 31 and 90 residents had to randomly select (at a pace of 1 / 2 persons) from 30 to 45 participants; NHs with > 90 residents had to randomly select from 45 to 60 participants (at a pace of 1 / 3 persons). Participants were 86 years-old (±8.2) and were mostly women (73.7%).

Outcomes

The main outcome measures were the number of hospitalizations (i.e., the number of times participants were seen in the hospital) in the last 12 months (discrete variable) and ED visits (binary variable: people who were not seen in ED in the last 12 months vs. people who were seen in ED).

Procedures

Participants were screened for the presence of: 1) physical conditions. cancer, diabetes, pain, stroke, myocardial infarction, congestive heart failure, peripheral vascular disease, and chronic pulmonary disease; 2) mental conditions. dementia, psychiatric diseases (other than depression), and depression. Information on residents’ health status, including diseases’ screening, was obtained through an on-line questionnaire completed by the NH medical staff, mainly the coordinating physician; this questionnaire was completed on the basis of information contained in the patients’ medical charts. These diseases were selected due to their high burden to the society according with information from a recent report on multimorbidity of the Institute of Medicine (1). The selection of diseases to be included was limited by data availability in the IQUARE study: therefore some diseases, such as osteoarthritis or atrial fibrillation, were not selected because there were no data on these conditions. Participants were then divided into one of the following “multimorbid groups”: none (none of the physical and mental diseases described above, n=495), only 1 physical condition (PC1, n=678), only 1 mental condition (MC1, n=960), ≥2 physical conditions (PC2, n=781), ≥2 mental conditions (MC2, n=497), 1 mental and 1 physical condition (PMC1, n= 971), or multiple physical and mental conditions (MPMC, n=1893. People in this group had both physical and mental conditions for a minimum of 3 diseases). “Multimorbid groups” constitutes our independent variable of interest.

Information on medical conditions, the number of hospitalizations and ED visits were reported by the NH coordinating physician. It is important to note that, for the purposes of public funding in France, residents’ diseases reported by the NH coordinating physician are regularly evaluated to confirm their exactness. This medical control, which is made by a physician from the Regional Agency of Health by consulting the medical charts of a fraction of the NH residents, supports the reliability of our data.

Confounders

Individual-related variables: age, sex, disability in activities of daily living (measured using a 6-item scale (e.g., bathing, toileting) – scores vary from 6 to 18, with higher scores indicating higher disability), use of antipsychotics, antidepressants, anxiolytics, hypnotics/sedatives, number of medications (summation of drugs other than those indicated before), history of fractures in vertebra, hip, femur, or wrist (single dichotomy: yes vs. no), falls in the last year (yes vs. no), living in a special care unit in the NH, weight change in the last two months (weight maintenance, loss, or gain), and the following diseases (each disease being a binary variable “yes” vs. “no”): hypertension, hemiplegia, peptic ulcer, liver disease, connective tissue disease, moderate/severe kidney failure, and epilepsy.

NH-related variables. ownership (public, private non-profit, or private for-profit), NH geographical localization (rural, low-urbanized, intermediate-urbanized, or highly urbanized), full-time equivalent per 100 NH beds for coordinating physician, for nurses, and for nurses’ aide.

Statistical Analysis

We reported incidence rate ratios (IRR) with 95% confidence intervals (CIs) after performing a mixed-effects Poisson regression on the number of hospitalizations; this procedure allowed us to adjust the model for the fact of living in a particular NH (variable entered as a random effect). For the binary variable “ED visits”, we reported odds ratios (OR) with 95% CI after performing a mixed-effects binary logistic regression (living in a particular NH entered into the model as a random effect). Potential multicollinearity was checked by using the variance inflation factor (VIF). Multivariate models contained the variable “multimorbid groups” and all the confounders. All analyses were performed using Stata version 11 (College Station, TX: StataCorp LP).

Results

Data on the number of hospitalizations were available for 6076 individuals (n=199 missing data). A total of 1944 subjects totalized 2921 hospitalizations in the last 12 months: 1119 persons had been admitted in the ED, 707 in other departments, and 4132 were not hospitalized (information about the department of admission was lacking for 118 hospitalized subjects). The distribution of hospitalizations according with multimorbid groups gave the following results (mean ± standard deviation): “no diseases” had an average of 0.2 (±0.6) hospitalizations, with 11.3% of participants (n=54) having been seen in ED; “PC1” 0.4 (±0.9) hospitalizations, with 16% (n=105) of ED visits; “MC1” 0.3 (±0.7) hospitalizations, with 13.1% (n=121) of ED visits; “PC2” 0.7 (±1.3) hospitalizations, with 22.7% (n=172) of ED visits; “MC2” 0.4 (±0.8), with 17.6% (n=85) of ED visits; “PMC1” 0.4 (±0.8), with 18.7% (n=176) of ED visits; “MPMC” 0.6 (±1.0), with 22.1% (n=406) of ED visits.

The prevalence of chronic conditions was: cancer – 12.7% (n=799), diabetes – 15.8% (n=992), pain – 23.4% (n=1467), stroke – 12.7% (n=795), myocardial infarction – 7.5% (n=472), congestive heart failure – 19.4% (n=1219), peripheral vascular disease – 17.9% (n=1122) and chronic pulmonary disease – 10.3% (n=645), dementia – 42.8% (n=2688), depression – 34.2% (n=2148) and psychiatric disease (other than depression) – 17.6% (n=1107). Participants were staying in the NH for longer than 41 months (median [25th – 75th percentiles]=1240 [527–2330] days).

With regards to the multivariate regression models, multicollinearity was probably not an issue since VIF values were < 2 for all independent variables in the models (mean VIF = 1.12). Table 1 shows the mixed-effects Poisson regression on the number of hospitalizations. As expected, adjusted analysis showed that the IRRs were higher for almost all multimorbid groups compared to “none” (except for the group MC1). Moreover, it is interesting to note that the IRRs for the “only physical conditions” groups (i.e., PC1 and PC2) were higher than the IRRs for the groups with a mix of physical and mental diseases (i.e., PMC1 and MPMC): people with one physical condition (PC1) had a IRR of 1.36 whereas those with both one physical and one mental condition (PMC1) had a IRR of 1.29; the same trend was found when comparing the group PC2 with the group MPMC.

Table 1.

Impact of multimorbidity on the number of hospitalizations among nursing home residents using mixed-effects Poisson regression

	Unadjusted model (n=6076)*		Adjusted model (n=6002)†
	IRR (95% CI)	P Value	IRR (95% CI)	P Value
Multimorbidity (ref. none)	-	-	-	-
PC1	1.58 (1.27 – 1.96)	< 0.001	1.36 (1.09 – 1.69)	0.006
MC1	1.19 (0.96 – 1.47)	0.117	1.01 (0.81 – 1.26)	0.91
PC2	2.57 (2.1 – 3.14)	< 0.001	1.93 (1.57 – 2.37)	< 0.001
MC2	1.68 (1.34 – 2.11)	< 0.001	1.41 (1.17 – 1.79)	0.004
PMC1	1.67 (1.35 – 2.05)	< 0.001	1.29 (1.05 – 1.60)	0.017
MPMC	2.35 (1.94 – 2.84)	< 0.001	1.63 (1.33 – 1.99)	< 0.001

^*The unadjusted model contained only the variable “multimorbid groups”, estimated as a fixed effect, and the fact of living in a particular nursing home, estimated as a random effect.

^†The adjusted model contained the variables of the unadjusted model and all the confounders. Note: CI, confidence interval; IRR, incidence rate ratio; PC1, only 1 physical conditions; MC1, only 1 mental condition; PC2, ≥ 2 physical conditions; MC2, ≥ 2 mental conditions; PMC, 1 physical + 1 mental conditions; MPMC, multiple physical and mental conditions.

Table 2 displays the results of the mixed-effects binary logistic regression on ED visits. Adjusted analysis showed that people in the multiple physical conditions group (PC2) and those in the multiple physical and mental conditions (MPMC) group had higher probabilities of having been seen in ED, compared to those in the “no diseases” group. As it was found for the outcome “number of hospitalizations”, PC2 group had a OR of having been seen in ED higher than the OR for the MPMC group.

Table 2.

Impact of multimorbidity on emergency department visits among nursing home residents using mixed-effects logistic regression

	Unadjusted model (n=6076)*		Adjusted model (n=6002)†
	OR (95% CI)	P Value	OR (95% CI)	P Value
Multimorbidity (ref. none)	-	-	-	-
PC1	1.55 (1.08 – 2.22)	0.018	1.33 (0.91 – 1.95)	0.15
MC1	1.22 (0.85 – 1.73)	0.28	0.98 (0.67 – 1.43)	0.92
PC2	2.38 (1.69 – 3.36)	< 0.001	1.79 (1.24 – 2.58)	0.002
MC2	1.78 (1.21 – 2.60)	0.003	1.50 (0.99 – 2.28)	0.054
PMC1	1.91 (1.36 – 2.68)	< 0.001	1.39 (0.96 – 2.00)	0.078
MPMC	2.33 (1.69 – 3.20)	< 0.001	1.48 (1.04 – 2.10)	0.029

^*The unadjusted model contained only the variable “multimorbid groups”, estimated as a fixed effect, and the fact of living in a particular nursing home, estimated as a random effect.

^†The adjusted model contained the variables of the unadjusted model and all the confounders. Note: CI, confidence interval; ED, emergency department; OR, odds ratio; PC1, only 1 physical conditions; MC1, only 1 mental condition; PC2, ≥ 2 physical conditions; MC2, ≥ 2 mental conditions; PMC, 1 physical + 1 mental conditions; MPMC, multiple physical and mental conditions.

On the basis of the unexpected results showing that people with only physical conditions had higher probabilities of hospitalizations and ED visits than people with both physical and mental conditions, we decided to further explore the differences in hospitalization-related variables according with multimorbidity type. For this, we ran exploratory analyses by performing the same adjusted regressions than before and changing the reference category to oppose the group PC1 against the group PMC1, and to oppose the group PC2 against the group MPMC. Mixed-effects Poisson regressions showed that, compared to PC1 (reference category), PMC1 group was not significantly associated with the number of hospitalizations (IRR = 0.95; 95% CI=0.81 – 1.12; p=0.56); however, the MPMC group had lower IRR (IRR = 0.84; 95% CI=0.75 – 0.95; p=0.005) for the number of hospitalizations, compared to the PC2 (reference category) group. For ED visits, mixed-effects logistic regressions showed the same pattern for the comparison between the groups PC1 (reference category) and PMC1 (OR = 1.04; 95% CI=0.77 – 1.41; p=0.77); the MPMC group (OR = 0.83; 95% CI=0.65 – 1.06; p=0.13) tended to have a lower probability of being seen in ED compared to the PC2 (reference category) group. We then examined if these results could be explained by a higher disease burden (as measured by the Charlson Comorbidity Index (10) in the groups PC1 vs. PMC and PC2 vs. MPMC: student t-test for independent samples showed that the groups PMC1 (1.9 ± 1.5) and MPMC (3.3 ± 1.9) had higher scores in the Charlson comorbidity index than the groups PC1 (1.6 ± 1.8) and PC2 (3.2 ± 2.3), respectively (statistically significant differences only for the PMC1 vs. PC1 comparison; p<0.001).

Discussion

This work showed, as expected, that multimorbidity in NH residents is associated with increased hospitalizations and higher probabilities of having been seen in ED, particularly when people had more than one physical condition; the new information added by this work is that these associations are dependent on the interaction of the type of the co-morbid conditions. Our results also suggest that the presence of mental diseases moderate the associations between physical conditions and hospitalizations since the estimate effects of the groups with both mental and physical conditions were lower than those of the groups with only physical conditions.

We quantified the burden of multimorbidity in NH residents: for example, having only one disease, regardless of its type (i.e., physical or mental condition), did not increase the probability of ED visits, whereas having more than one disease was associated with an up to 79% increased probability of ED visits. Although the associations between multimorbidity and hospitalization-related outcomes were expected, this study is the first to show the contributions of multimorbidity types. Interestingly, we expected that the higher the diversity of co-morbid conditions (i.e., concomitant presence of mental and physical conditions), the higher would be the rates of hospitalization, but it was not the case. Indeed, the groups “only one physical condition” (PC1) and “≥ 2 physical conditions” (PC2) had higher IRRs (number of hospitalizations – Table 1) and ORs (ED visit – Table 2) than the groups with concomitant physical and mental conditions (PMC1 and MPMC, respectively), with MPMC having a significantly lower IRR than PC2 for the number of hospitalizations. Exploratory analysis showed that these unexpected results were probably not dependent on disease burden since people with mixed conditions (PMC1 and MPMC) had higher scores in the Charlson co-morbidity index than people with only physical conditions (PC1 and PC2). Taken together, our findings suggest that mental conditions play a moderation role on the associations between physical conditions and hospitalizations. This moderation role is difficult to explain and needs further research to be confirmed. However, it is possible that reduced communication skills in people with mental disorders contribute to these results (11). Another potential explanation is that the NH staff avoids sending people with dementia (the most prevalent mental condition in our study, reaching 42.8% of our sample) to the ED, as found in a recent study among Swedish NH residents (11), for example, because a stay in the hospital can trigger or exacerbate individuals’ confusion (not rare in people with dementia) and accelerate functional declines (12). Indeed, there is only minimal guidance for dealing with dementia in the ED (13). To what extent the reduced number of hospitalizations and decreased probability of ED visits in people with mental diseases represent either a good recognition of the disease-related symptoms (e.g., agitation, which is a common symptom in people with dementia, is an important reason for hospital admission) 14, 15, leading to a better care of the patient, or the under-recognition from the NH staff of a condition deserving justifiable hospitalizations (e.g., due to the patient limitation in communicating his/her problems) is unclear.

This study has important strengths. First of all, our main finding, i.e., the fact that mental diseases probably play a moderation role in the associations between physical diseases and hospitalizations, are novel and add important information to the fields of geriatrics and public health. Moreover, we investigated several of the most prevalent and burdensome diseases in a large sample of very old and vulnerable NH residents. The main limitations of the present work are its cross-sectional design, which did not allow us to investigate incident hospitalizations, and the lack of assessment with regards to diseases’ severity, which may have accounted for some residual confounding. Furthermore, our retrospective approach on hospitalization-related outcomes only included current survivors, which means that NH residents with severe conditions (and maybe with several hospitalizations) could have died and did not have the chance to be selected. Finally, physical and mental diseases were screened from patients’ medical charts, which could lead to the underestimation of some diseases.

This work stimulates the debate in the neglected field of multimorbidity in the institutionalized elderly. Future studies should investigate to what extent hospitalizations of patients with single-type (i.e., only physical or only mental) or mixed (i.e., both physical and mental conditions) multimorbidity are inappropriate or avoidable. Research should also focus on the development of more complex and better adapted models of care that meet the real needs of multimorbid older adults.