The present study examined the degree to which residential characteristics and social factors are associated with mortality, after controlling for personal characteristics, among adults with intellectual disabilities who have resided in nursing homes (facilities providing skilled care and related services) at baseline in the Chicago area. Initial assessments were conducted on 330 residents, and 2 follow ups were performed over 10 years. Seventy-one residents had died by the end of the study. The variables examined included personal characteristics of age, gender, intellectual level, health, and adaptive behavior; residential characteristics of type of residence, size of facility, attractiveness of physical environment (cleanliness, conditions, and aesthetic appeal), and diversity of physical environment (personalization of residents' rooms and distinctiveness of all living spaces); and social factors of family involvement and community integration. The results indicated, beyond age, adaptive behavior, and health, that size of residences, residential characteristics, and social factors are related to mortality.
Since the 1960s, many people with intellectual disabilities have moved from public residential institutions into community settings, as a result of institutional closures and phase downs. During this period, many of these residents have moved into inappropriate nursing home settings (facilities providing skilled care and related services; Mitchell & Braddock, 1990). Due to the passage of the Omnibus Budget Reconciliation Act (OBRA) in 1987, all nursing home (refers to nursing facilities only and not to intermediate care facilities for the mentally retarded [ICF/MR]) residents with developmental disabilities were required to be evaluated using a Pre-Admission Screening and Annual Resident Review (PASARR, which is part of OBRA) to determine if they needed nursing facility services. Residents who did not need nursing care were mandated to transfer out of the nursing home to community-based settings. Those residents who had resided in the nursing home for more than 30 months were given the option of staying or moving out of the nursing home. Between 1988 and 1998, the national number of residents with intellectual disabilities living in nursing homes declined from 53,230 to 35,887 (Braddock, Hemp, Parish, & Rizzolo, 2000). In 2000, there were 1,267 people with developmental disabilities residing in nursing homes in Illinois, a decline of 61% from 1986. By 2006, this number had risen to 1,535 individuals (Braddock, Hemp, & Rizzolo, 2008). This study examines the risk factors for mortality among adults with intellectual disabilities who either remained in nursing homes or who moved to community settings over a 10-year period.
Although transitions from institutions or nursing homes to community settings may result in short-term stress and risks, perhaps affecting mortality, the long-term survival rate may improve (Heller, 1985; Heller, Factor, Hsieh, & Hahn, 1998; Litzinger, Duvall, & Little, 1993). Long-term mortality has not been addressed in most studies of mortality (Eyman, Borthwick-Duffy, Call, & White, 1988; Eyman, Call, & White, 1989, 1991; Eyman, Chaney, Givens, Lopez, & Lee, 1986; Eyman, Grossman, Chaney, & Call, 1990, 1993; Eyman, Olmstead, Grossman, & Call, 1993; O'Brien & Zaharia, 1998; Strauss & Kastner, 1996; Strauss, Kastner, & Shavelle, 1998a; Strauss, Shavelle, Baumeister, & Anderson, 1998b). Furthermore, researchers have not examined long-term mortality rates among adults with intellectual disabilities who reside in nursing homes or who have transferred from nursing homes to other community settings. This study examines the personal, environmental, and social risk factors for mortality over 10 years among persons with intellectual disabilities who resided in nursing homes at baseline in the Chicago area.
Risk Factors of Mortality
Key Personal Characteristics
Key personal risk factors of mortality for people with intellectual disabilities have included age (Cooper, 1998), gender (e.g., Janicki et al., 1999; Mölsă, 1994), severe intellectual disability (Eyman et al., 1986, 1988, 1990; Strauss & Eyman, 1996), Down syndrome after Age 50 years (Eyman et al., 1989; Janicki, Dalton, Henderson, & Davidson, 1999), immobility (Eyman et al., 1990, Eyman, Grossman, et al., 1993; Lerman, Apgar, & Jordan, 2003; Miller & Eyman, 1978), low self-care, medical conditions (Lerman et al., 2003), epilepsy (Chaney & Eyman, 2000; Morgan, Baxter, & Kerr, 2003), and the presence of a feeding tube or poor eating skills (Eyman et al., 1988, 1989, 1991; Eyman, Olmstead, et al., 1993; Strauss & Kastner, 1996). Although most studies of both the general population and the population with intellectual disabilities have noted that women outlive men (e.g., Janicki et al., 1999; Mölsä, 1994), researchers studying a Finnish nationwide population found that gender had no influence on survival for adults with severe or profound intellectual disability (Patja, Iivanainen, Vesala, Oksanen, & Ruoppila, 2000). Age has been related to mortality (Eyman et al., 1989; Janicki, Dalton, Henderson, & Davidson, 1999) and to a number of medical disorders. Older individuals have more health problems than younger ones (Beange et al., 1995) and are more likely to develop hypertension and cardiovascular disease (Cooper, 1998).
Compared with the general population, some health problems are more prevalent in people with intellectual disabilities in conjunction with their pre-existing conditions. For example, people with Down syndrome present higher rates of hypothyroidism, Alzheimer's disease–dementia, osteoporosis, visual impairments, hearing impairments, and obesity (Bell & Bhate, 1992; Center, Beange, & McElduff, 1998; Dinani & Carpenter, 1990; Kapell et al., 1998; Schrojenstein Lantman-de Valk et al., 1997). People with cerebral palsy are more likely to have poor health and problems such as contractures of the lower and upper joints (Bax, Smyth, & Thomas, 1988), increased difficulty in swallowing (Ferrang, Johnson, & Ferrara, 1992), urinary tract infections, and incontinence (Mayo, 1992). Hence, age and health status need to be taken into account when assessing the association between risk factors and mortality.
Although some researchers have examined the effect of residential settings on the mortality rates of persons with intellectual disabilities, research results are inconsistent. Similar studies have indicated that institutional settings either had higher (Eyman et al., 1988; Miller & Eyman, 1978) or similar mortality rates (Eyman, Grossman, Tarjan, & Miller, 1987) compared with community settings, even after controlling for severity of disability. Other studies focused on comparable mortality rates between community settings and institutions (Eyman, Grossman, Tarjan, & Miller, 1987). Some studies in California have indicated that community settings have higher mortality rates than institutions (Strauss, Eyman, & Grossman, 1996; Strauss & Kastner, 1996; Strauss et al., 1998a, 1998b). On the other hand, Silverman, Zigman, and Silver (1992) found that movement of adults with profound intellectual disability and multiple disabilities from a service-intensive, skilled nursing facility (SNF) to small, community-based ICF/MRs did not increase mortality risk. In addition, Lerman et al. (2003), using a prospective matched control research design to examine mortality between institution residents (“stayers”) and those who moved out institution (“movers”), found that moving out of the institution to community settings was not associated with mortality. However, nursing home placement was a risk factor for mortality.
Research that used California's Department of Developmental Services system database comparing mortality in institutions with mortality in community settings raised a great deal of debate in this journal, Intellectual and Developmental Disabilities (the journal was named Mental Retardation when this topic was raised; 1998, Volume 36, Issue 5). Although the California studies have the advantage of a very large sample size, the data used in these studies (e.g., Eyman et al., 1988, 1990; Eyman, Grossman, et al., 1993; Eyman, Olmstead, et al., 1993; O'Brien & Zaharia, 1998; Strauss et al., 1996, 1998a, 1998b) were not sufficiently detailed to detect the impact that risk factors such as health status, quality of health care, and social factors have on the mortality of persons with intellectual disabilities. Several studies have examined the impact of residential settings or health status and health behaviors of residents with intellectual disabilities. A recent literature review (Hayden, Kim, & DePaepe, 2005) examining health status, health care use, and outcomes among individuals with intellectual disabilities identified that the overall health of residents was similar or improved (e.g., South Dakota Department of Human Services, 1996) after they moved from institutions into community settings (e.g., Conroy, Seiders, Jones, & Thomas, 1995). Other studies found that individuals living in community setting tended to have a higher risk of cardiovascular profiles, such as obesity, hypertension, and high cholesterol, and tended to lead unhealthy life styles, with less exercise, increased alcoholic intake, and more smoking than those in institutional residential settings (Rimmer, Braddock, & Fujiura, 1993, 1994; Rimmer, Braddock, & Marks, 1995).
In addition to type of setting, various features of the residence have been shown to affect the health and well-being of people with intellectual disabilities. Studies of community adjustment of people with intellectual disabilities who relocated to community settings showed that greater support services and smaller facility size related to more positive psychosocial outcomes (Haney, 1988; Rotegard, Hill, & Bruininks, 1983; Stancliffe, 1997; Tossebro, 1996). Heller et al.'s (1998) study of transitions out of nursing homes found that environmental diversity of residences was related to adaptive behavior and that type of residential setting was related to community integration at follow up. One study (Cagney, Egelston, & Rathouz, 2000) found that a higher staff:client ratio was related to lower mortality in nursing homes. Hence, beyond type of residence, we should take residential characteristics (i.e., staff:client ratio and environmental variables) and health care services into account in studying mortality of people with intellectual disabilities. Yet, to our knowledge, no studies to date have examined the relationship between residential characteristics (other than type) on mortality. In this study, we sought to understand the extent that residential characteristics impact the survival of individuals with intellectual disabilities living in various facilities. This information has implications for planning of residential facilities and supports.
Previous epidemiological studies of the general population suggest that having little participation in social activities and few close relationships are associated with increased mortality (Berkmen, 1986; Dalgard & Lund Haheim, 1998; Grand, Grosclaude, Bocquet, Pous, & Albarede, 1990; Mossey & Shapiro, 1982; Nakaniski et al., 1998; Warren & Knight, 1982). Epidemiological studies have shown evidence of causal links existing between risk factors and mortality in the population of elderly persons. In addition to the personal characteristics (i.e., age, male gender, self-rated health, degree of disability), social factors such as social support and social participation appear to be reliable predictive factors for overall mortality (Berkmen, 1986; Dalgard et al., 1998; Grand et al., 1990; Mossey & Shapiro, 1982; Nakaniski et al., 1998; Warren et al., 1982). The main strength of these studies is that most of these studies are large-scale, population-based studies. One, more recent study (Kiely & Flacker, 2003) noted that social engagement is a protective factor for mortality in a long-stay nursing home population. However, none of the mortality studies of persons with intellectual disabilities have examined the relationship between mortality and social factors. A study by Fujiura, Fitzsimons, Marks, and Chicoine (1997) found that social integration was the only factor associated with Body Mass Index (BMI) among adults with Down syndrome after controlling for diet, exercise, and disability status. Higher BMI levels have been associated with a number of chronic diseases, such as digestive, pulmonary, and gallbladder conditions and cardiac vascular disease (Bray, 1992). Hence, social integration of adults with intellectual disabilities might have an impact on mortality as well. In addition, Blacher (1998) suggested that family involvement should be taken into account while studying the mortality of people with intellectual disabilities.
The main purpose of this study was to assess the extent to which residential characteristics (mover status, facility size, environmental diversity, and physical attractiveness) and social factors (family involvement and community integration) are related to the mortality of adults with intellectual disabilities living in the Chicago area, after controlling for personal characteristics (i.e., age, gender, intellectual level, tube feeding, mobility, number of health conditions, and adaptive behavior).
A 10-year prospective cohort study design is used. The baseline assessments were conducted in 1989 to 1990 with a sample of 330. The Time 2 data were collected in 1991 and 1992. Two hundred and sixty-eight participants completed all protocols. The Time 3 follow-up data were gathered between 1997 and 1999.
The research participants resided in 18 nursing homes at baseline in the Chicago area. These 18 nursing homes represented all of the nursing homes serving this population in the Chicago area. These nursing homes were licensed either as skilled nursing facilities or intermediate care facilities prior to implementation of OBRA in 1987. After the passage of OBRA, under federal Medicaid guidelines, these facilities were called “nursing facilities” and are not to be confused with ICF/MRs that are licensed specifically for care of persons with intellectual disability. All participants had been screened during the OBRA Annual Resident Review process to determine their need for a nursing facility. At Time 2 and Time 3 follow ups, some participants moved to other community settings. Any residential facilities except nursing home facilities in the community were included as other community settings. These other community settings included foster homes, semi-independent units with supervisory staff in the building, group homes, one retirement home, and the ICF-MR/developmental disability. Those who moved to other community settings at Time 2 or Time 3 follow ups were referred to as “movers,” and those who stayed in nursing homes at Time 2 and Time 3 were referred to as “nonmovers.” Table 1 presents the dynamics of moving status among different residential settings. Sixty-three individuals had either died (n = 19) or were in facilities where the residential staff refused to participate (n = 44) at Time 2. Staff in participants' residences refused to participate (fill out the survey) due to their workload at their work siteTable 2.
Fifty-one of these 63 persons resided in nursing homes. Only 4 persons moved back from community settings to nursing homes at the second follow up. Nineteen percent of study participants stayed in nursing homes throughout the study period. By the end of the study, 40% of participants resided in settings of less than 15 residents.
The study participants in the current cohort study consisted of 330 adults with intellectual disabilities, Ages 30 years and older at baseline. The criteria for inclusion in this sample included the following: (a) participation in the Nursing Home Transition Study (Heller, Factor, & Hsieh, 1998; Heller, Miller, & Factor, 1998; Heller, Miller, & Hsieh, 2002) in which 352 residents who met the criteria were referred to the research staff and of which 330 consented (recruitment rate = 94%) to participate in the study at Time 1; (b) Age 30 years or older; (c) residence in a nursing home at the baseline data collection; and (d) living either in nursing homes or other facilities in community settings at follow ups. At the Time 2 follow-up assessment, data were available on 267 residents (201 nonmovers and 66 movers). Nineteen were deceased and the remaining 44 residents refused participation (either staff or resident refusals). At Time 3 follow up, 177 study participants completed the surveys, 52 residents had died, 23 refused to participate, and 14 were not located. Therefore, the total number of deaths was 71 and the total number of censors was 259 at the final stage of this study.
The participants included 171 men (52%) and 159 women (48%). The majority (79%) of participants were White, 21% were African American, and 1% was other. Their mean age was 46.95 years (SD = 13.98) and ranged from 30 to 89 years at baseline. Forty-five percent had a diagnosis of cerebral palsy, 44% had epilepsy (26% had both cerebral palsy and epilepsy), and 11% had Down syndrome. Only 4% had a feeding tube. Fifty-five percent were not mobile. Over half of the participants had speech impairments. Approximately two thirds of the participants had severe–profound intellectual disability.
Participants were identified and recruited by research staff, including nurses, social workers, psychology graduate students, and nursing students from 18 nursing homes in the Chicago metropolitan area in 1989. Prior to the baseline assessment, the participants with intellectual disabilities or their legal guardians signed a consent form.
Because most of the study participants did not verbally communicate (54%) and had severe to profound (65%) intellectual disability, an informant questionnaire (residential profile) was used to obtain information on participants' demographics, mental and physical health, and social integration. Research staff interviewed nursing home primary caregivers to obtain information about each study participant. In addition, medical records were reviewed by an advanced-practice nurse staff to examine the accuracy of the health data. In several cases, inconsistencies appeared between medical records and staff reports. In those cases, the reports were re-examined, and, if the inconsistencies could not be resolved, the data were scored as missing.
At each follow up, if a person in the cohort had died, we would determine where he or she died (nursing home or other community residence) and when the death occurred by searching the Social Security Death Index database through the Internet. We matched study participants by their names, states of residence, birth dates, and social security numbers. By the end of the study, a total of 71 deaths occurred. When three death certificates were not located from an Internet search, we estimated their death date by using a midpoint between two follow-up times. We requested 68 death certificates from the Illinois Department of Public Health, Division of Vital Records by providing the decedents' names, birth dates, social security numbers, residence cities/counties, and death dates.
Four measurement tools were used to collect data: (a) the Residential Profile (Heller et al., 1998), (b) the Inventory of Client and Agency Planning (ICAP; Bruininks, Hill, Weatherman, & Woodcock, 1986b), (c) the Nursing Home Residential Transition Study Site Survey (Heller, Miller, & Factor, 1998), and (d) death certificates (Vital Record, Illinois Department of Public Health). The first and second tools, the Residential Profile and the ICAP were collected at baseline and at two follow ups of the longitudinal Nursing Home Transition study. Death certificates were available only if residents died during the period of 1990–1998.
We asked primary caregivers, case managers, or residential staff who were most familiar with the resident to fill out the resident profile for each participant. This instrument contains a number of constructs and scales that describe the characteristics of participants in a number of areas. The tool also includes the disposition of the “Annual Resident Review” and whether care in a nursing facility was needed. The tool included residents' physical and mental health, medical conditions, medication, and social integration.
The ICAP (Bruininks et al., 1986b) provides measures that describe individual characteristics in a variety of areas. These areas include demographics, diagnosis, level of intellectual disability, functional status, broad independent behavior, maladaptive behavior, mental health, and therapeutic support services. It is mandated for staff to conduct an annual ICAP evaluation of each client with intellectual disabilities. Reliability and validity of the ICAP tool have been published (Bruininks, Hill, Weatherman, & Woodcock, 1986a).
The Nursing Home Residential Transition Site Survey (Heller et al., 1998) assessed residence environmental characteristics in a variety of areas. These areas included facility size, staff: client ratio, policymaking involvement for residents, community accessibility, environmental diversity, and physical attractiveness.
The state of Illinois Medical Certificate of Death was obtained from the Illinois Department of Public Health, Division of Vital Records. Death certificates provide information on the demographics of the deceased, parents–informant, disposition, death pronouncement, autopsy, and cause of death. Eight (12%) of 68 deaths had autopsies.
Personal characteristics—including age, gender, ethnicity, marital status, residential place, level of intellectual disability, primary and secondary diagnoses (presence of epilepsy, cerebral palsy, or Down syndrome), mobility, tube feeding, number of health conditions, adaptive behavior, and speech impairment—were obtained from the ICAP or Residential Profile. The ICAP program computed age based on the birth date and the date of the ICAP evaluation. Gender included male (1) and female (0). Ethnicity background included White or other ethnic group. Marital status was married (1) or not married (0).
Level of intellectual disability was based on the diagnosis in the ICAP (Bruininks et al., 1986b). It was coded from mild (1) to profound (4).
Mobility was based on the information in the ICAP, including (a) walks—with or without aids, such as crutches, walkers, and so forth; (b) usually in a wheelchair, or does not walk; (c) limited to bed most of the day; and (d) confined to bed for entire day. In the present study, mobility was defined as “walks with or without aids.” It was recoded from the ICAP Code (1) to 1 as mobile and ICAP Code (2–4) (i.e., “else”) to 0 as not mobile.
Presence of tube feeding was obtained from the Residential Profile. Tube feeding referred to the use of either a nasogastric or gastrostomy feeding tube. It was coded as yes (1) or no (0).
Number of health conditions was the total number of current chronic conditions based on the sum of any 31 illnesses and conditions. It was adapted from Older American Resources and Services (OARS) for physical health (Fillenbaum & Smyer, 1981). Twenty-two chronic conditions included cancer or leukemia, glaucoma, skin conditions, asthma, emphysema or chronic bronchitis, tuberculosis, high blood pressure, heart trouble, circulation trouble in arms or legs, anemia, diabetics, thyroid or other glandular disorders, ulcers (of the digestive system), other stomach or intestinal disorders or gall bladder problems, kidney disease, other urinary tract disorders, liver disease, arthritis or rheumatism, muscular dystrophy, and effects of polio, multiple sclerosis, and Parkinson's disease.
Adaptive behavior was defined by the ICAP Broad Independence score (Bruininks et al., 1986b), which measures ability to perform daily activities independently. It includes a 77-item scale of adaptive functioning that measures an individual's ability to perform daily activities in the four domains of motor skills, social and communication skills, personal living skills, and community living skills. On this scale, a higher score means a higher level of functioning (Cronbach's α = .98).
Presence of speech impairment was obtained from the Residential Profile. Speech impairment was defined as inability to speak. It was coded as yes (1) or no (0).
Mover status and residential characteristics
Mover status was coded as movers (1) for those who moved to other community settings at Time 2 or Time 3 follow ups and as nonmovers (0) for those who stayed in nursing homes or moved between nursing homes at Time 2 and Time 3. Residential characteristics included type and size of facility, environmental diversity, and physical attractiveness. The type of residential facility was divided into three categories: nursing homes, residences with less than 15 residents, and residences of more than 15 residents. The size of the residential facility was based on the number of residents for nursing homes and other facilities in community settings. Values of residential characteristics were defined based on the move status at all time points while the participants were in the study. Times of movers were calculated by subtraction from discharge date to admission date.
Environmental diversity assesses the physical environment. It was adapted from the Multiphasic Environmental Assessment Procedure (MEAP; Moos & Lemke, 1984) rating subscale called Environmental Diversity. The Environmental Diversity subscale includes nine items and assesses the variety and stimulation provided by the physical environment. The scale includes items such as “personalization of residents' rooms” and “distinctiveness of all living spaces,” which are rated on a scale of 0–3, with 3 indicating the greatest level of variety and stimulation in the physical environment. For example, a score of 3 on the “personalization of residents' rooms” is much personalization and a score of 0 is no personalization is evident. The possible score ranges from 0 to 27. The alpha reliability of this subscale for the present study was .80.
Physical attractiveness measures physical attractiveness of the residence. It was adapted from the 28-item Physical Attractiveness subscale of the MEAP Rating Scale (Moos & Lemke, 1984). This subscale assesses cleanliness, conditions, and aesthetic appeal of the facility. The possible score ranges from 0 to 84. The alpha reliability of this scale for the present study was .94.
Times of movers
Residential characteristics were defined based on the move status for all time points while participants were in the study. Time of movers was calculated by subtraction from discharge date to admission date. When a refusal (0.3%) or “unable to locate” occurred (4.2%) and admission or discharge date was not available, an estimate was made.
Community integration was assessed by how often the resident engaged in 12 community activities within the last month. The community activities included: (a) talked to family/friends on the phone, (b) visited relative outside of his/her residence, (c) visited with relative in his/her residence, (d) visited friends outside of his/her residence, (e) visited outside friends in his/her residence, (f) went to movies/concert, (g) went shopping, (h) went to restaurants/bars, (i) used recreation facilities, (j) did volunteer work, (k) went to church, and (l) did other activities outside the residence or day program. Community activities were ranked on a 4-point, Likert-type scale with a range from 1 (none) to 4 (two or more times per week). The alpha reliability for the present study was .86. Family involvement was assessed based on the frequency that the participant visited with a family member in the past year. It ranged from 1 (never) to 5 (monthly or more).
Time (days in this study) until a mortality event occurred was identified through staff report and death certificates. A dummy variable for occurrence of death was coded as 1 (yes) or 0 (no) for each participant.
Statistical software SPSS 14.0 (SPSS Inc., 2006) was used to conduct descriptive statistics. They included proportional frequency distributions, estimates of variance, skewness and kurtosis, and correlations among all the study variables. Tests of significance were based on an alpha level of .05. Baseline demographics were used to profile the survival and deceased groups and assess their comparability. In addition, comparisons on the baseline characteristics of persons who dropped out of the study with those who remained were made to assess any biases due to attrition. Chi-square tests were used to examine the distribution on demographic variables (i.e., gender, race, intellectual disability levels, and presence of cerebral palsy or Down syndrome) and functioning status (i.e., mobility, tube feeding). We used t tests to assess the differences between event (death) and censor (survivals, withdraws, and refusals) groups on measure variables (i.e., age, adaptive behavior). For the post hoc analyses, analyses of variance (ANOVAs) were conducted in group comparisons among three types of facilities (residences < 15 residents, residences > 15 residents, nursing homes) on environmental and social variables (facility size, environmental diversity, and family involvement). The nonparametric Kruskal-Wallis was used for community integration due to its skew distribution.
In the present study, to determine mortality risk factors, the extended Cox survival analyses were conducted. Cox models provide a parametric form for the effect of the covariate while leaving the underlying distribution unspecific or nonparametric. The Cox model is able to take into account the fact that some measures (e.g., moving status, health, adaptive behavior) are changeable in different time periods and to make appropriate use of right censored observations, including survivals at the end of study, persons lost to follow up, and withdrawals–refusals during the study. Therefore, we preferred the Cox model to the logistic regression model. SAS 9.1 (SAS, 2004) statistical software was used to conduct the Cox survival analyses.
The dependent measure was time until death (event). Survival time was constructed by subtraction from the end of the time point (dates of death, last date in the study, or end of study) to the beginning date in the study. The unit of time was calculated in days. Note that the midpoint of two time points was used only when informants refused to participate in the study when defining right-censored time. The effects on this outcome were measured in terms of hazard ratios (HRs) for the hazard of death. HR describes the relationship between survival time (time to mortality) and predictors after controlling for covariates. HR is a different measure from an odds ratio; however, it has similar interpretations of the strength of effects. A HR of 1, like an odds ratio of 1, means that there is no effect. A HR of 2, is interpreted like an odds ratio of 2; that is, the exposed group has two times the hazard (risk) of the unexposed group (Kleinbaum, 1996).
We conducted t tests or the Mann-Whitney test to determine whether the independent variables in the current study were significantly different for people who were deceased versus those who were right censored (survivals at the end of study, withdrawals or refusals, or unable to follow up). Seventy-one (22%) participants died (events) and 259 (78%) participants were right censored at the end of the investigation. At baseline, no differences were found between the two groups on the characteristics of gender; race; level of intellectual disability; diagnosis of cerebral palsy, Down syndrome, or epilepsy; mobility; speech impairment; adaptive behavior; and the number of selected health conditions. Differences were noted in age, t(97.99) = 3.78, p < .001, and tube feeding, χ(1, N = 330) = 10.99, p < .05. Those who died were 8 years younger and in poorer health at baseline than those who were right censored. Persons who died were more likely to have had tube feedings (11%) than those who were right censored (2%).
We also compared characteristics between those who refused or were unable to be located and those who completed the study. By the end of the study, 82 (25%) participants either refused to participate or research staff were unable to locate them at some point, and 177 (54%) completed the study. No significant differences between the two groups were identified on the following variables: age, race, gender, mobility, presence of Down syndrome, tube feeding, epilepsy, number of health conditions, and speech impairment. For the following variables, however, significant differences were observed: level of intellectual disability, adaptive behavior, and presence of cerebral palsy. Individuals who completed the study had a higher frequency of cerebral palsy (53% vs. 34%), more severe (31% vs. 21%) and profound (44% vs. 27%) levels of intellectual disability (there were fewer with a moderate level of intellectual disability; 9% vs. 17%), and a lower level of adaptive functioning, t(257) = −2.55, p < .05.
Cause of Death and Mortality Rate
There was a wide distribution in age at the time of death. The overall mean age of death was 57.6 years (SD = 15.6), with a range from 33 to 89 years. The mean age of death for women and men was 52.0 years (range = 34–89 years) and 61.7 years (range = 33–87 years), respectively, a difference of 9.7 years. The overall crude mortality rate (mortality rate from all causes of death for a population during a specific time period) over a 10-year period was 22%. The average yearly crude mortality rate was 2.41%, which is much lower than for the Illinois general population. The major cause of death was heart disease (41%). Acute myocardial infarction was the most common type of heart disease. Aspiration pneumonia (15%) was the second most common cause of death, followed by pneumonia (12%) and septicemia (10%).
The study variables were entered in the Cox proportional hazards model or Cox extended model, one at a time. The variables included the following: personal characteristics (age, gender, level of intellectual disability, tube feeding, number of health conditions, and adaptive behavior), residential characteristics (mover status, facility size, environmental diversity, and physical attractiveness), and social factors (family involvement and community integration). The findings of univariate analyses indicated that being a mover, having a higher level of environmental diversity, and having more involvement in community activities were protective factors against mortality. Older age, tube feeding, more health conditions, and larger residence size were risk factors for mortality (see Table 3).
Four hierarchical Cox extended models were used. Independent variables in Model 1 included age, intellectual disability level, mobility, tube feeding, number of health conditions, and adaptive behavior. In addition to the independent variables in Model 1, mover status was added into Model 2. Residential characteristics (facility size, environmental diversity) were added into Model 3. Social factors (community integration, family involvement) were added into Model 4.
Impact of Personal Characteristics
In examining the association between mortality and residential characteristics (age, gender, intellectual disability level, mobility, tube feeding, number of health conditions, and adaptive behavior), the first model (Model 1) was used. Wald chi-square tests demonstrated that higher mortality was significantly associated with lower adaptive behavior (HR = 0.990, 95% CI = 0.983–0.998) and tube feeding (HR = 3.241, 95% CI = 1.426–7.366, p < .01). Increased age (HR = 1.043, 95% CI = 1.023–1.063, p < .0001) was also a risk factor for mortality (Model 1, Table 4).
Impact of Mover Status and Residential Characteristics
A stepwise selection at the .05 level for removal from the model was used to select the other significant environmental variables. The physical attractiveness variable was removed from the model. Facility size and environmental diversity remained in the model. In examining the association between mortality and mover status, mover status was entered into the Cox extended model (Model 2, Table 4). After controlling for the personal characteristics, being a mover (HR = 0.273, 95% CI = 0.119–0.628, p < .01) to community settings was a protective factor against mortality. However, two environmental factors (facility size and environmental diversity) were added into the Cox extended model (Model 3, Table 4) after the mover-status variable, which made mover status no longer significant. The likelihood ratio for this model was 55.719, p < .0001; hence, Model 3 was significant. The Wald chi-square tests demonstrated that increased age (HR = 1.04, 95% CI = 1.015–1.056, p < .001) and having a feeding tube (HR = 3.686, 95% CI = 1.593–8.526, p < .01) were risk factors for mortality. Protective factors against mortality were higher adaptive behavior (HR = 0.99, 95% CI = 0.982–0.998, p < .05) and higher environmental diversity (HR = 0.94, 95% CI = 0.881–1.003, p < .07).
Impact of Social Factors
Social factors, including community integration and family involvement, were added into Model 4. Table 4 summarizes the results of the Cox extended models. After entering social factors, personal characteristics, mover status, and residential characteristics remained significant, as they did in Model 3. The significant variables in this full model were age, intellectual disability level, tube feeding, environmental diversity, and community integration. The findings indicated that age (HR = 1.027, 95% CI = 1.007–1.049, p < .01) and presence of a feeding tube (HR = 3.943, 95% CI = 1.703–9.130, p < .01) were risk factors for mortality, but lower level of intellectual disability (HR = 0.688, 95% CI = 0.520–0.935, p < .05), environmental diversity (HR = 0.929, 95% CI = 0.870–0.993, p < .01), and community integration (HR = 0.236, 95% CI = 0.085–0.653, p < .01) were protective factors against mortality.
Post Hoc Analyses
Post hoc analyses were conducted to examine differences in personal characteristics (age, gender, intellectual disability level, mobility, tube feeding, number of health conditions, and adaptive behavior) at baseline among movers and nonmovers. Three sets of comparisons of means between movers and nonmovers were made. The comparisons were as follows: nonmovers all the time versus movers at Time 2 or Time 3, nonmovers versus movers at Time 2, and nonmovers all the time versus movers at Time 3. Also, post hoc analyses were conducted to examine mean differences in residential characteristics (facility size and environmental diversity) and social factors (family involvement and community integration) among the types of residential groups.
Movers versus nonmovers
The mean scores of personal characteristics among nonmovers and movers are presented in Table 5. Overall, movers were younger, more mobile, and had a higher level of adaptive functioning at baseline than nonmovers. There were no differences in gender; presence of cerebral palsy, Down syndrome, or epilepsy; tube feeding; speech impairment; number of health conditions; and level of intellectual disability at baseline. Mean differences at baseline were found in age, t(327) = −3.20, p < .01, and adaptive behavior, t(328) = 2.65, p < .01, between nonmovers and movers. When comparing nonmovers with movers at Time 2, mean differences were noted in age, t(169) = −3.11, p < .01, and adaptive behavior, t(186) = −2.74, p < .01. In addition, mean differences were found with age, t(315) = −3.20, p < .01, and adaptive behavior, t(315) = 2.45, p < .05, in comparing nonmovers to movers at Time 3.
Comparisons of residential characteristics and social factors among residential types
One-way ANOVAs demonstrated significant group differences in facility size, F(2, 260) = 110.89, p < .001, at Time 2, and F(2,196) = 199.94, p < .001, at Time 3; environmental diversity, F(2, 265) = 18.06, p < .001, at Time 2, and F(2, 158) = 25.40, p < .001 at Time 3; family involvement, F(2, 161) = 4.90, p < .01; and community integration, F(2, 261) = 159.01, p < .001 at Time 2, and F(2, 167) = 20.49, p < .001, at Time 3. Scheffé tests were carried out to determine which group significantly differed from the other groups. At Time 3, the mean size of residences with more than 15 residents was close to that of nursing homes. The mean facility size at Time 1 was 222.50 residents (SD = 81.45), with a range of from 78 to 463. At Time 2, the mean facility sizes were, for nursing homes, 217.50 (SD = 107.90, range = 48–485); 5.20 for residences with less than 15 residents (SD = 2.71, range = 1–11); and 81.70 (SD = 55.70, range = 16–217) for residences with more than 15 residents. At Time 3, the mean facility sizes were 208.80 (SD = 85.80, range = 50–417), 4.69 (SD = 2.30, range = 1–15), and 189.10 (SD = 126.60, range = 18–382) for nursing homes, residences with less than 15 residents, and residences with more than 15 residents, respectively. The results of Scheffé tests also showed that homes with less than 15 residents had a higher mean of environmental diversity and community integration than homes with more than 15 residents and nursing homes at Time 2 and Time 3 (p < .05). Both nursing homes and residences of less than 15 residents had greater family involvement than residences with more than 15 residents, but no differences were found between nursing homes and residences with less than 15 residents at Time 2 or Time 3 (see Table 6).
The results of this study offer insight into the personal characteristics and residential and social factors associated with mortality over a 10-year period among adults with intellectual disabilities who had resided in nursing homes. After controlling for personal characteristics, findings indicated that higher environmental diversity and community integration were associated with lower mortality for these residents regardless of where they resided.
Several personal characteristics of the adults with intellectual disabilities, such as age, level of disability, and presence of a feeding tube, were significantly associated with mortality in the present study. A higher level of intellectual functioning was related to a low risk of mortality. Increased age and having a feeding tube, on the other hand, were a risk factor for mortality. These findings are consistent with previous studies (e.g., Cooper, 1998; Eyman et al., 1988, 1989, 1991, 1993b; Janicki, Dalton, Henderson, & Davidson, 1999; Strauss & Kastner, 1996).
Although moving from nursing homes to other community settings was associated with a lower risk of mortality compared with nonmovers, after adjusting for the personal characteristics, it was no longer significant after other environmental and social factors were taken into account. In the final model, the residential characteristics of environmental diversity and the social factor of community integration were protective factors against mortality among adults with intellectual disabilities after adjusting for personal characteristics. Therefore, whether residents lived in nursing homes or other community settings, those who had a greater variation in the physical environment and greater involvement in social activities had a lower risk of mortality. These findings are particularly notable given the controls that were used in the data analyses. These results differ from previous research findings indicating that individuals who moved from institutions to community based settings had a higher mortality rate than those who stayed in institutions (e.g., Strauss & Kastner, 1996; Strauss et al., 1996, 1998b). Most previous studies used either standard mortality rates or odds ratios to compare differences in mortality rates or risk ratios among different residential settings. Although those studies controlled for personal characteristics and medical conditions, other possible explanatory factors were not examined.
The findings of the current study highlight the importance of understanding how the programmatic structure and physical environment (other than just type and site of facility) affect the lives of residents, regardless of whether residents live in community settings or nursing homes. As Heller (1998) suggested, greater variety and stimulation in the physical and social environment might be related to psychological growth and/or cognitive and adaptive skills. On the other hand, individuals who are more isolated and less stimulated may experience a decline in cognitive and adaptive skills. Mitchell and Kemp (2000) have found that social participation and family involvement were related to life satisfaction among people who were elderly and lived in assisted living homes. The social factor of family involvement studied in the present study did not have a significant impact on mortality. Perhaps other domains of family involvement that were not measured in this study (e.g., distance from family member's residence) may have an impact on mortality. Another possible explanation is that most of the participants in the present investigation did not have families involved in their lives. Thus, there may not have been enough variance to detect the impact of family involvement on mortality as measured in this study.
Limitations in this study included sample size and attrition. Compared with the California studies, this sample was smaller, as it was limited to 18 nursing homes in the Chicago metropolitan area. In addition, this sample did not represent the general population of people with intellectual disabilities living in community based settings, because people referred from nursing homes are more likely to be in poorer health and have multiple disabilities and greater needs for medical supports. Thus, the results may not be generalizable to the population with intellectual disabilities in general and to other states. A replication of this study with a larger sample is needed. A larger sample with a variety of residential settings would provide sufficient statistical power to detect the association between mortality and residence size and other environmental factors. Another issue is the large amount of attrition that occurred in the 10-year period of this study. The problem with attrition (no data available for refusals or withdrawals) is that the total number of people at risk of death decreased when survival analyses were carried out. In addition, another limitation is that the data included in this study are 10 years old and changes may have occurred in this time period.
Although other residential characteristics beyond facility size and social factors were examined, some other important variables such as quality of life, residents' preferences or choices, and quality of services were not measured in the present study. Another potential difficulty of the current study is that information regarding the participant was provided by an informant. Consequently, differences in the information provided by staff from residential facilities may exist. Staff were from different residential facilities, and residents within different facilities may have been functioning at different levels. Thus, different staff may have a different frame of reference at different points in time, which could have resulted in a bias on measures of a participant at different time periods. Another limitation of the present investigation is that there were only two follow ups, with a long time interval between Time 2 and Time 3. When study participants had more than one move during that period, it was not possible to obtain the actual move dates for residents. Therefore, we need to interpret the results with caution.
One implication of these findings is that caregivers and health professionals can improve longevity by providing more personalized living environments and more community activities for this population. Many residential settings do not provide adequate diversification and personalization in living environments or a sufficient amount of social–community activities for residents. This study indicated that greater diversification–personalization of living environments and social–community activities may improve the quality of life for people with intellectual disabilities and increase their life expectancy. Despite facility size, staff or family members can be more creative in decorating living spaces and offering a more personalized living environment.
Another important implication of these findings is that staff and family members need to provide more opportunities for individuals with intellectual disabilities to participate in social activities. The results of post hoc analyses indicated that residents scored close to 1 (none) on all the community integration items at baseline. Yet, the mean score of community integration items only improved from 1.0 to 1.3 at the final follow up, which was less than once a month. Hence, when planning social activities, taking individuals' preferences into account is important. If financial resources are limited, staff–family members can use free community social activities, volunteer groups, low-cost park district programs, and senior citizen centers to provide additional social interaction for people with intellectual disabilities. For individuals living with family members, the advocacy and service network for family members and persons with disabilities may be one way to provide social activities for both family members and the person with a disability.
The results of this investigation suggest several directions for future research. The present study represents one of the first attempts to examine whether residential characteristics beyond facility size and social factors contributed to mortality using a robust statistical procedure to consider variables that differ over time. More in-depth investigation and replications are imperative. Future research could include measures of interactions between staff and clients, quality of care, access to specialized services, and residents' quality of life. Furthermore, developing valid and reliable assessments of health status for persons with intellectual disabilities may allow care providers to better identify residents at high risk.
Another direction for future research concerns the measurement of quality of life in different residential settings from the perspectives of persons with intellectual disabilities. Qualitative research may better assist researchers in understanding quality-of-life issues in the lives of persons with intellectual disabilities. An observational qualitative approach could help researchers understand the emotional reactions of persons with severe and profound intellectual disability to people in their environment.
In conclusion, the present study provides a deeper understanding of the relationship among residential characteristics, social factors, and mortality beyond personal characteristics for people with intellectual disabilities. In addition to comparing mortality rates in different settings, it is also important to identify and understand factors that contribute to the risk of mortality for people with intellectual disabilities. With promotion of community based alternatives for people with disabilities, initiatives to better support the health and safety of people with disabilities, wherever they choose to live, should be a national priority.
Preparation of this article was supported in part by the Rehabilitation Research and Training Center on Aging With Developmental Disabilities, University of Illinois at Chicago, through the U.S. Department of Education, National Institute on Disability and Rehabilitation Research (Grant H133B980046). The opinions contained in this article are those of the grantee and do not necessarily reflect those of the U.S. Department of Education.
Editor-in-Charge: Steven J. Taylor
Kelly Hsieh, PhD (firstname.lastname@example.org), Research Assistant Professor, University of Illinois at Chicago, Disability and Human Development, 1640 W. Roosevelt Rd., Room 416, Chicago, IL 60608. Tamar Heller, PhD, Head and Professor, University of Illinois at Chicago, Department of Disability and Human Development, Chicago, IL 60608-1336. Sally Freels, PhD, Associate Professor, University of Illinois at Chicago, School of Public Health, Chicago, IL 60612-4394.