Predicting the outcome of long-term care by clinical and functional indices: The role of nutritional status

Abstract

In elderly subjects, past researches have already underlined the role of nutritional status as a basic factor able to influence the prognosis either in acute wards or in rehabilitation and long-term care settings. Aim of the study is that of retrospectively verify, through a multivariate analysis, the factors able to condition mortality in long-term care, paying particular attention to the nutritional status.

Methods

The survey included 513 patients aged more than 65 years admitted to a long-term care unit during a three years period Exitus within the first three months of hospitalization was considered the outcome variable, while baseline functional, cognitive, clinical and nutritional status were considered the independent variables eventually related to mortality.

Results

The univariate analysis found that some variables were significantly correlated with the outcome: comorbidity, ADL, cognitive status, pressure sores, albumin, transferrin, CRP, mucoprotein, cholesterol, Cholinesterase, MAMC and MNA. The predictive value of the block model of the logistic regression analysis was 77.9% (specificity = 85.3%, sensitivity = 63.9%). With the forward stepwise analysis only MNA, Cholinesterase, CRP and mucoprotein were considered in the final model. In this case the predictive value of the model was 79.3% (specificity = 84.6%, sensitivity = 69.46%).

In elderly subjects, past researches have already underlined the role of nutritional status as a basic factor able to influence the prognosis either in acute wards or in rehabilitation and long-term care settings. The malnourished subject indeed has higher morbidity and mortality and higher incidence of adverse acute clinical events, harder recovery of his/her function deficits, easier cognitive impairment and, therefore, greater frailty 1, 2, 3, 4, 5.

On the other hand, the recovery of a well-nourished status is able to stop the subsequent spins of the “downward spiral” that triggers, when present malnutrition, and that leads to the exitus of the subject 6, 7, 8.

Aim of the study is that of retrospectively verify, through a multivariate analysis, the factors able to condition mortality in long-term care, paying particular attention to the nutritional status.

Methods

Sample

The survey included all patients admitted consecutively to the long-term care unit of the «Villa delle Querce» Clinical Rehabilitation Institute of Nemi (Rome - Italy) during a three years period: 513 patients aged more than 65 years: 274 women (54.4%) aged 82±8 years (65-100 years old) and 239 men (45.6%) aged 79±9 years (65-97 years old).

The local ethics committee approved the study.

At the admission the following were evaluated for each subject:

a. clinical status:

• •number of pathologies in place
• •
  pressure sores, as assessed by the classification of Shea (9):

  • -
    absence of lesions of degree ≤ 2
  • -
    presence of single or multiple lesions of degree 3 or 4);
• •number of drugs taken
• •
  comorbidity: by a modified version of the Geriatric Index of Comorbidity 10, 11 which is computed on the basis of severity of each 15 of most frequent chronic conditions. Five comorbidity classes were defined:

  • -
    Class I: patients with no symptomatic disease
  • -
    Class II: patients with symptomatic diseases under satisfactory control
  • -
    Class III: patients with only one uncontrolled disease
  • -
    Class IV: patients with two or more uncontrolled diseases
  • -
    Class V: patients with one or more diseases at their greater severity
    b. autonomy level and cognitive status
• •cognitive functions: by Short Portable Mental Status Questionnaire - SPMSQ (12)
• •functional status: by Activities of Daily Living - ADL (13)

  c. nutritional status:

• •
  Mini Nutritional Assessment (MNA) (14) which includes a comprehensive anthropometric assessment, data about general condition and dietary habits, and a self-evaluation of health and nutritional status. Patients were classified into three risk categories according to their MNA scores:

  • -
    well nourished: MNA ≥ 24
  • -
    at risk of malnutrition :17 ≤ MNA < 23.5
  • -
    malnourished: MNA < 17
• •biochemical parameters: total cholesterol, cholinesterase, CRP, α-1-glicoprotein acid, albumin, transferrin and lymphocytes count. The laboratory tests were performed at the Laboratory of “Villa delle Querce” Clinical Rehabilitation Institute. Peripheral venous blood was collected from antecubital vein after an overnight fast. Serum concentrations of the biological indices were determined by routine methods with conventional commercial kits obtained from ABX Italia (Rome, Italy). Laboratory tests were carried out using a COBAS-MIRA analyser and a Cell-Dyn 1700 Analyser (Abbott). The guidelines for the use of artificial nutrition in adult patients of the Italian Society for Artificial Nutrition (SINPE) provided the standard values of the biological indices (albumin = 3.5 g/l, transferrin = 2.0 g/l, lymphocytes count = 1500 #/mm³) (15).
• •anthropometric parameters: height, body weight, arm circumference (AC), calf circumference (CC) triceps skinfold thickness (TSF). Mid upper arm circumference (MAMC) was calculated from AC and TSF: MAC (cm) = AC (cm) - (π * TSF (cm)). Body Mass Index was calculated [BMI = weight (Kg) / (stature (m))²].

In bed-ridden subjects stature was estimated from knee-height using a validated equation for Italian elderly people (16).

The anthropometric measurements were performed by a single, trained operator according to the Standard Manual for Anthropometric Measures. The elderly subjects were measured barefoot with light-weight clothing. Weight was measured to the nearest 0.1 Kg on a SECA (Hamburg, Germany) weighing scale and stature was measured on a wall mounted stadiometer to the nearest 0.5 cm (SECA, Hamburg, Germany). To measure stature, the elderly subjects were requested to stand straight on an horizontal surface, heels together, the eyes straight forward. Circumferences were measured to the nearest 0.1 cm with a cloth tape, and TSF to the nearest 0.2 mm with a Harpenden skinfold calliper (British Indicators Ltd, St Albans, Herts, UK) on the dominant arm. The weighted mean of 10th percentile values for Italian samples enrolled in the SENECA study were used as lower limits of normality for anthropometric parameters. These were defined as: MAC = 22 cm for men and = 18.9 cm for women; TSF = 5.2 mm for men and = 9.7 mm for women 17, 18.

Data analysis

The exitus within the first three months of hospitalization was considered as the outcome variable: 0 when positive (patients still alive after three months), and 1 when negative (patients died within the first three months).

Baseline functional, cognitive, clinical and nutritional status were considered the independent variables eventually related to mortality.

Univariate analysis (t-test, χ2) was performed to describe the pattern of response of potential independent variables to the outcome variable (mortality). The odds ratio (OR) and the 95% confidence interval for the association of tested variables with the outcome variables were calculated.

Variables proven univariately correlated with the outcome were dichotomised and entered into a pool of potential contributors in the logistic regression analysis. The models were statistically evaluated using a block model where all variables were included, and a forward likelihood stepwise method (cut-off probability for entry: 0.05). With each added variable, the discriminant function was recalculated, and any variable that no longer met the significance level was removed from the equation (cut-off probability for removal: 0.1).

In order to assess how our model fit, the model predictions has been compared to the outcomes observed in the elderly subjects with a predicted probability greater or equal to 0.5 (overall percentage of correct classification, sensitivity, specificity, positive and negative predictive values).

Moreover, the area under the ROC (Receiver Operating Characteristics) curve was calculated. This is the graphic representation of sensitivity (probability to complete an error of first type - false negative) regarding the complement to 1 of the Specificity (esteem of the probability to complete an error of second type - false positive) of the logistic model to variation of the level of decisional threshold. The curve with ideal values of specificity and sensitivity more close has the course to the advanced side. The subtended area to such curve represents the probability that a subject correctly is classified by the logistic model and must have however a value higher than 0.5.

Statistical significance was set at p < 0.05 level. Data were analyzed using the SPSS for Windows 10.0 (SPSS Inc., 1989-1999) and the Win Episcope 2.0 [Facultad de Veterinaria di Saragozza (E), Wageningen University (N), University of Edinburgh (GB)] statistical software packages.

Results

Sample Description

The main characteristics of the examined sample are described in tab. 1. In particular, 57,3% of the sample, at the moment of admission in long-term care onset, was older than 80 years. More than 80% of the elderly subjects were dependent in more than two functions to the ADL and the cognitive status was seriously deteriorated in almost 40% of the cases.

Table 1.

Overall characteristics of the studied population at admission in long term care

		Male			Female
Subjects	number	239			274
Age	year		79±9			82±8
	Age 60-70	40 subj		16.7%	23 subj
8.4%
groups	70-80	84	35.1	72		26.3
(years)	80-90	93	38.9	143		52.2
	> 90	22	9.3	36		13.1
Pathology	number	4.1±1			4.0±1
(per patient)
per system	Cardiovascular	31.4%			33.3%
	Respiratory	9.2			7.3
	Central nervous system	16.4			19.6
	Osteoarticular	14.4			16.0
	Metabolic, endocrinologi	c 14.2			11.2
	Digestive	13.2			10.0
	Urinary	1.2			2.6
	Pressure Ulcers IVth	30 subj	12.6%	43 subj		15.7%
	degree
Drugs
(per patient)	number	4.2±2			4.4±2
Comorbidity	Class I	0%			0%
	II	11.2			13.3
	III	30.2			39.6
	IV	30.0			27.7
	V	28.8			19.4
Cognitive	Normal	35.2%			30.7%
functions	Mild impaired	28.6			28.0
(SPMSQ)	Severely impaired	36.5			41.3
Functional	Independent	10.9%			7.0%
status	Dependent in 1-2 ADLs	9.6			9.1
(ADL)	Dependent in > 2 ADLs	79.5			83.9
Outcome at	Exitus	91 sub	38.1%	79 sub		28.8%
3 months	Living	148	61.9	195		71.2

SPMSQ : Short Portable Mental Status Questionnaire; ADL : Activities of Daily Living.

As regards the clinical status, each patient was affected by 4 pathologies and for this reason he/she took more than 4 drugs/daily. At admission, pressure sores at IV degree were present in 12.6% of men and in 15.7% of women. Comorbidity level was at IV or V class (patients with two or more uncontrolled diseases or with one or more diseases at their greater severity, respectively) in 58.6% of men and in 47.1% of women.

Respect to mortality: the 38.1% of men and the 28.8% of women died within the first three months of hospitalization.

As regards the nutritional status, a high prevalence of malnutrition as either biochemical or anthropometric parameters detected was observed (tab 2).

Table 2.

Nutritional Status of the studied population at admission in long-term care

		Male	Female
BMI (kg/m2)	=> 18.5	15.9%	11.7%
	< 18.5	84.1	88.3
TSF (mm) Normal	85.8%	66.1%
Reduced: men:	< 5.2; women: < 9.7	14.1	33.9
AC (cm)	=> 22	69.5%	67.2%
	< 22	30.5	32.8
MAMC Normal	62.3%	61.7%
	(cm) Reduced: men: < 22; women: < 18.9	37.7	38.3
Albumin (g/l)	> 35	42.9%	42.7%
	31-35	32.8	37.2
	< 31	24.3	20.1
Transferrin (g/l)	=> 1.5	56.1%	71.2%
	< 1.5	43.9	28.8
Lymphocytes (#/mm3)	=> 1200	89.1%	89.4%
	< 1200	10.9	10.6
CRP (mg/l)	< 7	33.5%	46.4%
	7-20	25.5	25.1
	> 20	41.0	28.5
Mucoprotein (mg/dl)	< 160	74.9%	81.8%
	> 160	25.1	18.2
Cholesterol (mg/dl)	=> 150	59%	77.7%
	< 150	41	22.3
Cholinesterase (UI/l)	=> 3000	59%	66.4%
	< 3000	41	33.6
MNA Well nourished		4.6%	4.0%
Risk of Malnutrition		34.3	31.8
Malnutrition		61.1	64.2

CRP: C-reactive protein; BMI: body mass index; TSF : triceps skinfold thickness; AC: arm circumference; MAMC : mid upper arm muscle circumference; MNA: Mini Nutritional Assessment.

The data are confirmed by the MNA score that classified as malnourished the 61.1% of men and the 64.2% of women.

Univariate analysis (Table 3, Table 4)

Table 3.

Admission clinical parameters - correlation with 3-months survival

		A	D	p*	OR	CI 95%
Sex	Male	148	91	0.05	2	1.1-2.2
	Female	195	79
Age	years	80.3±9	81.9±8	0.08
	≤ 80 yrs	149	70	0.8	1.1	0.8-1.6
	> 80 yrs	194	100
Pressure	Absent	317	123	0.000	4.7	2.8-7.9
sores	Present	26	47
(IVth degree)
Comorbidity	Class II - III	193	51	0.000	3	2-4.4
	Class IV - V	150	119
Cognitive	Normal - impaired	132	37	0.02	2.2	1.5.3.4
functions	Mild - severe	211	133
(SPMSQ)	impairment (5-10 errors)
Functional	Dependent in 0-2	81	12	0.000	4.1	1.2-7.7
status (ADL)	functions
	Dependent in >2 functions	262	158

A: alive; D: dead; SPMSQ : Short Portable Mental Status Questionnaire; ADL : Activities of Daily Living. * Pearson's chi-square or Students's t test

Table 4.

Admission nutritional parameters - correlation with 3-months survival

		A	D	p*	OR	CI 95%
BMI (Kg/m2)	2	25.5±5	21.0±4	0.01
	> 18.5	62	8	0.06	4.47	2.1-9.6
	< 18.5	281	162
TSF (mm)	Normal	265	121	0.53	1.38	0.9-2-1
	Reduced*	78	49
AC (cm)		24.4±4	22.6±3	0.000
	≥ 22	258	92	0.001	2.57	1.7-3.8
	< 22	85	78
MAMC (cm)	Normal	251	67	0.000	4.19	2.8-6.2
	Reduced**	92	103
Albumin (g/l)		35.0±5	31.9±5	0.000
	> 31	289	111	0.000	2.85	1.8-4.4
	< 31	54	59
Transferrin (g/l)	1.79±0.5	1.52±0.5	0.000
	> 1.5	245	88	0.000	2.33	1.6-3.4
	< 1.5	98	82
Lymphocytes (#/mm3)		1916±617	1804±628	0.1
	>1200	313	145	0.29	1.8	1-3.2
	≤ 1200	30	25
CRP (mg/l)	< 7	177	30	0.000	4.98	3.2-7.8
	≥ 7	166	140
Mucoprotein (mg/dl)		103±36	151±47	0.000
	< 160	309	94	0.000	7.3	4.6-11.7
	≥ 160	34	76
Cholesterol (mg/dl)		175±45	155±44	0.000
	≥ 150	251	103	0.01	1.8	1.2-2.6
	< 150	92	67
Cholinesterase (UI/I)		3879±1315 2795±1272	0.000
	≥ 3000	261	62	0.000	5.5	3.7-8.3
	< 3000	82	108
MNA	score	16.2±5	9.9±5	0.000
Well nourished - Risk	174	17	0.000	9.3	5.4-16
of malnutrition
Malnutrition	169	153

^*TSF reduced : men: < 5.2 mm; women: < 9.7 mm; ** MAMC reduced: men: < 22 cm; women: < 18.9 cm; A: alive; D: dead; CRP: C-reactive protein; BMI: body mass index; TSF : triceps skinfold thickness; AC: arm circumference; MAMC : mid upper arm muscle circumference; MNA: Mini Nutritional Assessment; * Pearson's chi-square or Students's t test

Mortality was higher in males (38.1 vs 28.8%; χ2=3,7, p=0,05).

Comorbidity and pressure prejudiced mortality: 44.2% of elderly subjects with comorbidity level IV or V and 64.4% of patients with pressure sores level IV died within three months vs 33.1% of the entire sample (χ2=21.3 and χ2=29.2 respectively; p=0,000). OR for comorbidity level IV or V was 3 (CI: 2-4.4) while for the presence of pressure sores it was 4.7 (CI: 2.8-7.9).

Cognitive status (mild or severe impairment) and functional status (dependence on more than 2 ADLs) also influenced the outcome: respectively 38.7 and 37.6 mortality level vs 21.8 or 12.9% when cognitive and functional status were within the norm; (χ2=9.2 and 16; p=0,02 and 0.000). OR for impaired functional level was 4.1 (CI: 2.2-7.7) while for mild-severe cognitive impairment it was 2.2 (CI: 1.5-3.4).

A significant statistical correlation was evidenced for all nutritional parameters (both anthropometric and biochemical) except for TSF and lymphocytes count (tab. 4). In particular:

• -albumin: mortality was 27.8% and 52.2% in the sample respectively for normal/moderate reduction (> 31 g/l) and severe reduction (< 31 g/l) of albumin values (χ2=15, p=0,000). The mean value of albumin was different according to outcome: 31.9 ± 5 vs 35 ± 5 g/l respectively for negative or positive outcome (p = 0. 000). OR for reduced levels of albumin: 2.85 (CI: 1.8-4.4).
• -CRP: mortality was 14.5 and 45.8% in elderly subjects respectively with normal (< 7 mg/dl) or increased (> 7 mg/dl) levels (χ2=36,5 and p=0,000). OR for high levels of CRP: 4.98 (CI: 3.2-7.8).

Moreover cholesterol and cholinesterase levels were different in positive outcome compared to negative outcome: respectively 175+45 vs 155+44 mg/dl for cholesterol (p=0,000), 3879+1315 vs 2795+1272 U/l (p=0,000) for cholinesterase.

Anthropometric parameters showed similar results: BMI, AC and MAMC were significantly lower in elderly subjects died within the first three months in long-term care.

Finally MNA confirmed the trend: mortality was higher in elderly subjects classified as malnourished (47.5 vs 8.9%; χ²=58.3, p=0,000). The mean score of MNA was different according to the outcome: 9.9+5 vs 16.2+5 respectively for negative or positive outcome (p = 0.000). OR for malnutrition classification at MNA was: 9.3 (CI: 5.4-16).

Multivariate Analysis (table 5)

Table 5.

Mortality in long-term care - logistic regression analysis

	Variables in the model	OR	CI(95%)	Overall predictive value	Sensitivity	Predictive value Specificity	Positive predictive value	Negative predictive value	Area under the ROC curve
Block model	Comorbidity	1.56	0.7-3.4
	Functional status	2.36	0.58-9.7
	Pressure sores	1.27	04.8-3.37
	Cognitive status	0.95	0.39-2.35
	MNA	2.49	0.84-7.34
	MAMC	1.55	0.71-3.39	77.9	63.9	85.3	34.6	31.7	0.83
	Albumin	0.97	0.39-2.41						(SE. 0.065;
	Transferrin	0.71	0.31-1.64						CI95%:0.72-0.98)
	Cholesterol	0.7	0.3-1.62
	Cholinesterase	4.64	1.91-11.25
	CRP	2.29	0.95-5.55
	Mucoprotein	2.7	1.02-7.14
Forward stepwise	MNA	3.93	1.6-9.6	79.3	69.4	84.6	34.6	34.1
selection	Cholinesterase	3.83	1.78-8.24						0.83
	CRP	2.55	1.08-6.02						(SE: 0.057;
	Mucoprotein	2.53	1.03-6.2						CI 95%: 0.69-0.91)

SE: standard error; CI Confidence interval; ROC curve: receiver-operator-characteristic curve; MNA: Mini Nutritional Assessment, CRP: C-reactive protein, ADL: activities of daily living, MAMC: mid upper arm muscle circumference

The multivariate analysis was performed using only the independent variables significantly correlated with the outcome in the univariate analysis: comorbidity, ADL, cognitive status, pressure sores, albumin, transferrin, CRP, mucoprotein, cholesterol, cholinesterase, MAMC and MNA.

In the block model of the logistic regression analysis all the selected variables were included. The strength of association between mortality and independent variables was greater for cholinesterase (OR: 4.64; CI(95%): 1.91-11.25), mucoprotein (OR: 2.7; CI(95%): 1.02-7.14), MNA (OR: 2.49; CI(95%): 0.84-7.34), functional status (OR: 2.36; CI(95%): 0.58-9.67) and CRP (OR: 2.29; CI(95%): 0.95-5.55). The predictive value of the model was 77.9% (specificity = 85.3%, sensitivity = 63.9%, predictive positive value = 34.6% and predictive negative value = 31.7%). The area under the ROC curve (equal to 0,83) confirmed the validity of the model.

With the forward stepwise analysis only MNA, cholinesterase, CRP and mucoprotein were considered in the final model. The strength of association between mortality and independent variables was greater for MNA (OR: 3.93; CI (95%): 1.6-9.63) and cholinesterase (OR: 3.83; CI(95%): 1.78-8.24. In this case the predictive value of the model was 79.3% (specificity = 84.6%, sensitivity = 69.46%, positive predictive value = 34.6% and negative predictive value = 34,1%). In this case too the area under the ROC curve (equal to 0,83) confirmed the validity of the regression model.

Discussion

The main finding of the present study is the demonstration that patient's general status (i.e. comorbidity, pressure sores, cognitive and functional status) and in particular nutritional and inflammatory status (i.e. albumin, transferrin, CRP, mucoprotein cholesterol, cholinesterase, BMI, AC, MAMC and MNA score), reliably predicts the outcome of long-term care.

In institutionalized elderly subjects, few researches considered, with multivariate analyses, the predictive elements of mortality in nursing homes.

Levine SK et al. (19) developed a prognostic index for 1-year mortality of hospitalized older adults using standard administrative data in 6382 older adults discharged from general medicine services. Potential risk factors for 1-year mortality included: age, length of stay, discharge to nursing home, dementia, metastatic cancer and other comorbidities. Mortality at one year was 11% in the lowest-risk group and 48% in the highest-risk group. In this study, however, nutritional status was not considered.

The mortality at three months that has been detected in our sample stands at 33.1%. It represents a rate, while very high, consistent with reported data in other case studies.

Recently Mitchell SL et al (20) found a mortality of 54.8% over a period of 18 months in a sample of 323 nursing home residents with advanced dementia.

In another recent paper (21), the Authors found that the 1-year mortality rate was 29.3% in 82 disabled institutionalized elderly patients. In this case mortality was correlated with malnutrition (low levels of BMI and fat-free mass).

In a study aimed at the investigation of the effect of chronic diseases and disease combinations on 1-year mortality in nursing home residents (n=43.510), one year after baseline assessment through Minimum Data Set (MDS), 35% of the residents had died (22).

Despite the great diversity of the examined case studies, related to the organization of health systems as well as to cultural and social variables (easy access to nursing homes, social loneliness/isolation, expectations/wishes of the patient and the family), the mortality rate are sufficiently similar. In our, as well as in other studies, comorbidity, cognitive and functional status impairments, malnutrition were particularly severe at admission. These are all elements that in scientific literature justify, among others, an increased rate of mortality.

Functional decline and comorbidity are two of the main predictors of adverse outcomes in nursing home and of mortality in particular 23, 24, 25.

In the present study, functional impaired elderly subjects had an increased risk of mortality (OR: 4.1). Similarly, in a study aimed at the identification of intensive care unit (ICU), admission characteristics predictive of mortality among older nursing home residents, a higher APACHE III score (adj-RR 1.02; 95% CI: 1.01-1.04) and increasing functional dependency before ICU admission (adj-RR 1.6; 95% CI: 1.05-2.57, per ADL quartile) were independently associated with increased mortality rate within 90 days (26).

Several studies described the prognostic properties of comorbid illnesses in nursing home residents. In the study previously mentioned, (22) carried out on 43.510 nursing home residents, in addition to male sex and increasing age, the presence of dementia, stroke, cardiac diseases, malignancy, diabetes mellitus, and lower haemoglobin concentration were found to be negative prognostic indicators. The number of medical diagnoses and high scores on a comorbidity index also predicted mortality.

Older people are especially vulnerable to malnutrition as they frequently have multiple pathologies and impairments and poor nutritional intakes (6).

In our sample the prevalence of malnutrition resulted very high: 62.8% according to the MNA. Similarly, all parameters, either anthropometric or biochemical, were impaired in more than half of examined elderly subjects.

This data is neither more nor less in line with findings in other case studies. Two recently conducted nationwide prevalence surveys in the Netherlands and Germany reported prevalence rates among older hospitalised patients ranging from 32 to 56% 2, 27.

Previously published data reaffirm a prevalence of malnutrition of up to 85%, especially among older people resident in long-term care facilities, confirming that, unfortunately, in this field there is much to do yet 28, 29, 30, 31, 32, 33, 34, 35.

The importance of malnutrition in frail institutionalized elderly people has been emphasized in many papers since long time 36, 37, 38, 39, 40, 41.

In our study mortality was significantly related to nutritional status according either to univariate analyses or to multivariate model. In particular the MNA OR resulted to be the highest among the examined variables.

Kimyagarov S et al, in the paper cited before (21), found that 1-year mortality was associated in particular with low levels of BMI, fat-free body mass, skeletal muscle mass (OR respectively 1.73, 2.42 and 2,55 respectively).

A weight loss during the 6-months period was associated with a nearly two-fold increase in the likelihood of dying (adj-RR: 1.95, 95% CI 1.43 to 2.66) nursing home residents at high nutritional risk to determine (42).

Finally, in a recent systematic review of the literature (Medline search conducted from 1968 to 2007 under the search terms aging, nutrition, and nursing homes) it was seen that application of screening protocols in long-tem care institutions (oral cavity status examination, monitoring of weight and food intake, ..) and a precocious nutritional intervention (including enhancement of environment and increasing staff to assist with feeding) improved nutritional status and by this way reduced morbidity and mortality (43).

The link between malnutrition and mortality is justified by the fact that undernutrition is an important predictor of morbidity. It leads to a decline in functional status, exacerbates existing medical conditions and increases the risk of complications, (respiratory and cardiac problems, infections, deep venous thrombosis and pressure ulcers and multiorgan failure) compared to diagnostically comparable patients who are well nourished 4, 5, 33, 38, 44, 45, 46, 47, 48, 49.

In our sample comorbidity degree was particularly high as well as the levels of inflammatory parameters (CRP and mucoprotein). These two variables, as the cholinesterase, entered in the predictive final model with high OR.

Adverse effects linked to malnutrition are frequent in all age groups, but they are especially hazardous for older people, particularly those who reside in nursing care homes, who often exhibit high levels of frailty and disability. Furthermore, the effects of malnutrition may exacerbate each other, particularly in frail older people. For example, having an infection or a bone fracture may result in reduced appetite leading to a vicious cycle of infection/fracture and malnutrition. Therefore the treatment and prevention of malnutrition is an important challenge for the health care system but may significantly improve morbidity and mortality in elderly institutionalised people 32, 50, 51, 52, 53, 54, 55.

It is acknowledged that a main potential limitation should be considered in order to better assess our present data. The present study is based on a retrospective analysis of case report forms originally drawn up only for clinical purposes. For this reason some parameters, as quality of life, that were not included in the multidimensional geriatric evaluation which was performed routinely in our patients, have not been considered.

Finally, effects of specific disease combinations (malnutrition, comorbidity and functional decline) on mortality of the nursing home population have received little attention until now. Nutritional status remains, more than anything else, the element influencing mortality.