Alessandra Marengoni MD, Division of Internal Medicine I, Department of Medical and Surgery Sciences, University of Brescia, Spedali Civili, Piazzale Spedali Civili 1, 25123 Brescia, Italy. (fax: +39 030 396011; e-mail: email@example.comfirstname.lastname@example.org).
Objective. We aimed to disentangle the effect of chronic multimorbidity and disability on 3-year functional decline and survival in the elderly.
Design. Prospective cohort study with a mean of follow-up of 2.8 years.
Setting. Swedish elderly persons from the Kungsholmen Project (1987–2000).
Subjects. A total of 1099 subjects, 77–100 years old, living in the community and institutions.
Main outcome measurements. Medical diagnoses (based on clinical examination, drug use, medical records and blood tests), and functional assessment (according to Katz Index) at baseline were investigated in relation to functional decline and death occurring during follow-up.
Results. At baseline, 12.1% of participants had disability, and 52.3% were affected by multimorbidity. During follow-up, 363 persons died and 85 worsened in functioning. The number of chronic conditions incrementally increased the risk of functional decline [hazard ratio (HR) increased from 1.5 in subjects with one disease to 6.2 in persons with 4+ diseases]. However, this was not the case for mortality, as the HR of death was the same for people with one disease as well as 4+ diseases (HR = 2.3). Baseline disability had the highest impact on survival, independently of number of diseases [HR = 8.1; 95% confidence interval (CI) = 4.8–13.7 in subjects with one disease and HR = 7.7; 95% CI = 4.7–12.6 in those with 2+ diseases].
Conclusions. In the elderly subjects, chronic disability rather than multimorbidity emerged as the strongest negative prognostic factor for functionality and survival.
Chronic morbidity and disability represent major challenges both for clinicians and health systems in a world where thousands of persons turn 65 years of age every day [1, 2]. The ageing of the populations primarily draws the attention of many researchers to chronic conditions and their coexistence – referred to as multimorbidity [3, 4]. Cross-sectional studies have shown a close association between the number of chronic diseases and disability [5–8], confirmed by longitudinal observations that reported an increasing risk of mobility loss and incident functional dependency by number of chronic diseases [9–12]. In turn, disability has been consistently found as a good prognostic indicator for mortality [13–15], whereas the association between multimorbidity and survival in old age is still controversial [15–20].
Very few population-based studies have based the ascertainment of health status on clinical examination, or covered most of the chronic disorders affecting old persons . Most studies were based on interview or self-administered questionnaires, which could be a less reliable method of detecting morbidity in the ageing population. Indeed, older age is associated with less accurate self-reports of diseases, independently of cognitive status , and having multiple diseases decreases the agreement between self-reports and medical records . Using a complete ascertainment of health status through different sources of medical information, we recorded current chronic diseases affecting elderly subjects living both in the community and institutions to define multimorbidity . In addition, to further explore chronicity, functional level was assessed directly by trained nurses.
This study aims at disentangling the effect of chronic multimorbidity and disability on 3-year functional decline and survival in a very old cohort.
Data were gathered from the longitudinal data set of the Kungsholmen Project (1987–2000). The Kungsholmen Project is a community-based, prospective study on ageing and dementia. Briefly, all registered inhabitants who were living in the Kungsholmen district of central Stockholm, Sweden, and were 75 years and older in October 1987 (n = 2368) were initially invited to participate in the project . Informed consent was requested from all subjects. The research follows the guidelines of the Swedish Council for Research in Humanities and Social Sciences , and has been approved by the Ethics Committee of the Karolinska Institutet. Of the 1700 elderly who agreed to participate in the first examination (1987–1989), 429 died and 172 moved or refused the first follow-up (1991–1993). Thus, 1099 subjects participated at first follow-up and constituted the baseline population of the present study. At this phase a clinical examination performed by doctors was available for all participants. During the 3-year follow-up, 363 persons died, eight moved and 48 refused to participate. Thus, 680 persons were reassessed in 1994–1996.
At baseline all participants were examined following a standardized protocol, including a social interview, a neuropsychological battery and a medical examination lasting about 3 h. The elderly and their next-of-kin were interviewed by trained nurses using a structured questionnaire on educational background, living conditions and social factors.
Chronic disease assessment
Different sources of medical diagnoses were employed to reduce potential ascertainment biases. The clinical examination included geriatric, neurological and psychiatric assessment, as well as medication use. Doctors asked the participants to show prescription forms and/or drug containers of the used drugs. Drugs were registered according to the Anatomical Therapeutic Chemical (ATC) classification system . The International Classification of Diseases – Ninth Revision (ICD-9)  was used for all diagnoses, except for the following disorders: (i) diagnosis of major depression was made according to DSM-IV criteria  using data from the Comprehensive Psychopathological Rating Scale ; (ii) DSM-III-R diagnostic criteria for the clinical diagnosis of dementia and different dementia types ; (iii) diagnosis of anaemia following the WHO criteria ; and (iv) obesity defined as a body mass index (BMI) ≥30 kg m−2. The information on BMI was available for 768 persons. Chronic disorders were diagnosed by the examining doctor (based on clinical examination, medical history, laboratory data and current drug use), or detected in the computerized Stockholm Inpatient Register during the baseline ascertainment period. This register encompasses all hospitals in the Stockholm area since 1969, and records up to six diagnoses at discharge. A disease was classified as chronic if one or more of the following characteristics were present: (i) being permanent, (ii) caused by nonreversible pathological alteration or (iii) requiring a long period of care or rehabilitation . Amongst all the chronic diseases detected in the population , 22 conditions with a proven impact on disability and mortality [5, 7, 9, 10, 17–19] were included in the calculation of multimorbidity in the present study (Table 1). Multimorbidity was defined as any co-occurrence of two or more chronic conditions in the same individual.
Table 1. Chronic diseases included in the calculation of multimorbidity grouped according to prevalence range (per 100). Single disorders are listed in ascending order of prevalence
Prevalence per 100
Reprinted with kind permission from the American Journal of Public Health.
Functional status was measured using the Katz index of ADL , which is a scale formed by dependency in the following six activities: bathing, dressing, going to the toilet, transferring, continence and feeding. Functional independence was defined as no need for assistance or need for assistance in one activity, partial disability was defined as need of assistance in two to four activities, and total disability was defined as need of assistance in five or six activities. Data on functional status were collected by a nurse by questioning and observing the subjects both at baseline and at follow-up. Seventeen subjects had missing data on functional status at baseline and nine at follow-up.
Functional decline was defined as a change in functional status – from being independent or partially dependent to becoming partially or totally dependent during the follow-up period. Only three subjects improved and were included in the group of persons with no change in functioning. Due to this finding, we are confident in considering disability in this population as chronic. In order to be more certain in using this categorization, we cross-checked the entire analysis defining no disability as having no basic function lost and decline in functioning as any basic function lost during the follow-up, obtaining similar results (data not shown).
Mortality data for all participants were derived monthly from the Swedish National Death Register.
Cross-sectional associations between chronic multimorbidity and disability were tested using prevalence figures and logistic regression models. Cox regression analysis was employed to analyse the association between increasing number of diseases (no disease was the reference group) and functional decline and mortality during the 3-year follow-up. The models included the following covariates: age (85+ vs. 77–84 years), gender (female versus male), education (≤7 vs. ≥8 years) and disability at baseline (defined as either partial or total disability). To further explore the correlation between multimorbidity, disability and mortality, we examined the combined effect of disability and number of chronic diseases on functional decline and survival, using no disease/no disability as the reference group. Having chronic morbidity (1 disease) with or without disability and chronic multimorbidity (2+ diseases) with or without disability were the subgroups included in the analysis. In the models for testing functional decline, persons with total disability were excluded. Finally, to confirm the reliability of previous analyses, a multinomial logistic regression model with either functional decline or death as dependent variables was created, obtaining similar results (data not shown). Subjects with missing data or loss at follow-up were excluded from all the analysis.
Of the 1099 subjects who were examined at baseline, the majority were women (77%), who were older [mean 84.8 (4.6) vs. 83.7 (4.2) years, P = 0.002) and less educated (55.0% vs. 39.2% with 2–7 years of education, P < 0.001) than men. Fifty-two per cent of participants had multimorbidity (53.6% of women and 48.0% of men). At baseline, most of the subjects were functionally independent: older persons (85+ years old) and women were more likely to be disabled (Table 2). Persons with no disease were completely independent, whereas the prevalence of disability increased with increasing number of chronic conditions, from 14.5% in subjects with one chronic disease to 17.0% in persons with 4+ conditions (Table 2). After adjusting for socio-demographic characteristics, the logistic regression analysis showed a 30% increased odds for disability for each additional chronic condition (OR = 1.2; 95% CI = 1.1–1.5 in men and OR=1.7; 95% CI = 1.2–2.5 in women, P for gender difference = 0.145).
Table 2. Socio-demographic and clinical characteristics of the study population by functional status at baseline, functional decline and death during follow-up
Missing data on education for seven women and on functional status at baseline for four men and for 13 women. aPersons with total disability at baseline were excluded. **P < 0.01, ***P < 0.001.
8 + years
Number of diseases
During the 3-year follow-up (average 2.8 ± 1.1 years), 363 persons died (26.4% of those with no disability at baseline in comparison with 76.3% of persons with disability, P < 0.001) and 85 persons worsened in functional status (7.8% of persons without baseline disability versus 12.9% of those with partial disability, P = 0.007). The proportion of subjects with functional decline at follow-up was higher in older subjects, and increased with increasing number of chronic conditions (Table 2). Amongst older persons with no disease, 13.5% died during the follow-up: 23.3% amongst men and 8.8% amongst women (P = 0.007). About one-third of diseased participants died during the follow-up independently of the number of conditions affecting the subjects at baseline (Table 2).
After adjusting for socio-demographic characteristics and baseline disability, the Cox regression model showed that the number of diseases incrementally increased the risk of functional decline, but not mortality. When the number of diseases was included as a linear variable, each additional condition increased the risk of functional decline by 50% [hazard ratio (HR) = 1.5; 95% CI = 1.3–1.8], with no differences between men and women. Figure 1 contrasts the HR of functional decline and death due to each additional chronic conditions, showing a clear divergence in the association pattern of multimorbidity with functional decline (chi-square for trend = 18.1, P < 0.001) and mortality (chi-square for trend = 0.14, P = NS) respectively. The HR for functional decline increased from 1.5 (95% CI = 0.6–3.7) in persons affected by one disease to 6.2 (95% CI = 2.5–15.2) in those with 4+ conditions. In contrast, having one disease showed the same risk of dying of having 4+ conditions (HR = 2.3; 95% CI = 1.5–3.7 and HR = 2.3; 95% CI = 1.5–3.9 respectively).
When examining the combined effect of chronic diseases and disability, the subgroup of persons with disability had a much higher risk of dying independently of the number of diseases (Table 3, Fig. 2). Older age was associated with both functional decline and death: men and lower educated persons had a higher risk of dying (Table 3).
Table 3. Hazard ratios (HR) and 95% confidence intervals (95% CI) for functional decline and mortality during follow-up due to morbidity (one disease) or multimorbidity (2+ diseases) and presence of disability at baseline. HR adjusted for age, gender and education
HR for functional decline
HR for death
85+ vs. 77–84 yrs
Female versus male
2–7 vs. 8+ years
No chronic morbidity and no disability
Chronic morbidity and no disability
Chronic morbidity with disability
Chronic multimorbidity and no disability
Chronic multimorbidity with disability
In this old population, chronic multimorbidity and disability emerged as negative prognostic factors for functioning and survival. Over a period of 3 years, old persons with chronic multimorbidity experienced functional decline with a 50% increased risk for each additional disease. However, multimorbidity did not affect survival, whereas baseline disability highly increased the risk of dying, independently of number of diseases. In spite of the differential gender distributions of baseline multimorbidity and mortality, the effect of multimorbidity and disability on functional decline and survival was similar in men and women.
In agreement with previous research [5, 6], the results of this study underline the close association between number of diseases and disability in old age. In fact, persons without chronic conditions at baseline did not have disability, whereas the prevalence of disability increased up to 17% in those with 4+ disorders. On the contrary, the majority of persons with disability were affected by multimorbidity (63%). Moreover, although only 8.4% of the baseline population worsened in functioning during follow-up, the number of diseases was a good indicator of functional decline. Persons with multimorbidity not only developed new dependency, but those with partial disability further declined (26.5% of persons with 2+ diseases versus 6.9% of those with 0–1 disease). This is in contrast to previous reports that chronic disorders are relevant for determining new disability, but not for worsening functional status .
Previous research into multimorbidity and mortality in the elderly has reported inconsistent results [15–17, 20]. In this population, having an increasing number of chronic diseases did not affect survival. We can reasonably hypothesize that disease severity and interactions between acute and chronic conditions are probably much more important than the simple count of chronic morbidities in increasing mortality risk. On the contrary, baseline disability emerged as a strong risk factor for mortality independently of the number of diseases. In agreement with previous findings , the large majority of old persons with partial or total dependence did not survive, supporting the hypothesis that disability in basic activities of daily living is a good indicator of disease severity.
Some limitations need to be discussed. First, a potential limit of using the sum of diseases as ordinal scale without taking into account severity is that moving from zero to one disease could realize the majority of the multimorbidity effect, with additional disease having a minor impact . Although the information on disease severity was not available in the Kungsholmen Project, we limited the diagnoses included in the definition of multimorbidity to those conditions with potential of severity that could affect either disability or mortality. Moreover, we used disability in activities of daily living as a quantified sum score of the impairment associated with chronicity. This choice was supported by the finding that people with no diseases had no disability. In contrast, using the count variable approach has the advantages of expressing multimorbidity in an additive form, and differentiating people at each level of morbidity. Moreover, the contribution of any disease, expressed in the multimorbidity term, can be estimated, avoiding problems of insufficient statistical power . A second limitation of the study is the possible underestimation of some disorders such as musculoskeletal diseases , probably decreasing the strength of association between multimorbidity and disability. However, different sources of medical diagnoses were employed, including direct clinical examination, which reduced potential ascertainment biases that often affect the accuracy of the assessment of multimorbidity in the elderly . This allowed the detection of disorders, which are usually under-reported (e.g. mental diseases). Finally, the present study has the advantage of using a community-based population, including subjects living both at home and in institutions, and of having a limited number of dropouts both at baseline and first follow-up.
The Kungsholmen Project cohort consisted of old individuals, living in a geographically defined central area of Stockholm. One special feature of this population is that it included only two persons who had been farmers in their main occupation. Thus, some caution is needed when generalizing the findings to younger persons or those living in rural areas. However, findings from studies carried out in this population may be generalized to old urban populations in Western countries.
In conclusion, this study represents one of the first attempts to disentangle the effect of multimorbidity and disability on the health of elderly persons living in community and institutions using direct assessment of health status and physical functions. Our findings have relevant implications both for clinicians and public health organizers , as they showed that a count of medical conditions is not a good indicator of survival, but allows the identification of subjects particularly at risk of functional disability. Elderly persons affected by disability and multimorbidity should represent a specific target for secondary prevention to stop or delay the progression to severe and chronic disability.
Conflict of interest
The authors do not have any conflict of interest.
This study was funded by research grants from the Swedish Council for Working Life and Social Research. The authors’ work was independent of the funders.