bs_bs_banner
Original article
Acute hospital admission for nursing home residents without cognitive impairment with a diagnosis of cancer J. DRAGESET, PHD, RN, POSTDOCTOR, Faculty for Health and Social Science, Bergen University College, Bergen, Norway, G.E. EIDE, PHD, BIOSTATISTICIAN, PROFESSOR, Centre for Clinical Research, Western Norway Health Region Authority, Bergen, and Research Group for Lifestyle Epidemiology, Department of Public Health and Primary Health Care, University of Bergen, Bergen, Norway, C. HARRINGTON, PHD, RN, PROFESSOR, Department of Social & Behavioral Sciences, School of Nursing, University of California, San Francisco, CA, USA, & A.H. RANHOFF, PHD, MD, PROFESSOR, Kavli Research Centre for Ageing and Dementia, Haraldsplass Hospital, Bergen, and Institute for Internal Medicine, University of Bergen, Bergen, Norway DRAGESET J., EIDE G.E., HARRINGTON C. & RANHOFF A.H. (2015) European Journal of Cancer Care 24, 147–154 Acute hospital admission for nursing home residents without cognitive impairment with a diagnosis of cancer Studies of hospitalisation of cognitively intact nursing home (NH) residents with cancer are scarce. Knowledge about associations between socio-demographic, medical and social support variables and hospital admissions aids in preventing unnecessary admissions. This is part of a prospective study from 2004 to 2005 with follow-up to 2010 for admission rates. We studied whether residents with cancer have more admissions and whether socio-demographic and medical variables and social support subdimensions are associated with admission among cognitively intact NH residents with (n = 60) and without (n = 167) cancer aged ≥65 years scoring ≤0.5 on the Clinical Dementia Rating Scale and residing ≥6 months. We measured social support by face-to-face interview. We identified all respondents through NH medical records for hospital admission, linking their identification numbers to the hospital record system to register all admissions. We examined whether socio-demographic and medical variables (medical records) and social support subscales were associated with the time between inclusion and first admission. Residents with cancer had more admissions (25/60) than those without (53/167) (odds ratio 1.7). Social integration was correlated with admission (P = 0.04) regardless of cancer diagnosis. Residents with cancer had more hospital admissions than those without. Higher social integration gave more admissions independent of cancer diagnosis.
Keywords: nursing home, cancer, social support, hospital admission.
INTRODUCTION Older people who need residential care have a high risk of hospitalisation (Konetzka et al. 2008; Phelan et al. 2012). Among nursing home (NH) residents, acute illness is common (Alessi & Harker 1998; Murray & Laditka 2010) Correspondence address: Jorunn Drageset, Faculty of Health and Social Sciences, Bergen University College, Møllendalsv 6-8, N-5005 Bergen, Norway (e-mail: [email protected]).
Accepted 3 April 2014 DOI: 10.1111/ecc.12205 European Journal of Cancer Care, 2015, 24, 147–154
© 2014 John Wiley & Sons Ltd
and hospitalisation is frequent (Grabowski et al. 2008). The benefits of the hospitalisation of NH residents are often questioned because of the high risk of complications (Sager et al. 1996; Inouye et al. 2006) and lack of evidence indicating that hospitalisation leads to better outcomes (Konetzka et al. 2008). Several studies from various countries and healthcare systems have investigated admission from NH to hospital (Grabowski et al. 2008). One systematic review showed that men who are older and have a shorter length of stay in the NH are more likely to be hospitalised. Physical and cognitive disability and specific health conditions such as
DRAGESET ET AL.
congestive heart failure and infections increased the risk for hospitalisation (Grabowski et al. 2008). The most frequent admission diagnoses are found to be infections (Alessi & Harker 1998; Kruger et al. 2011), fractures, cardiovascular and gastrointestinally related diagnoses (Kruger et al. 2011), respiratory diseases, fall-related diagnoses and circulatory diseases (Graverholt et al. 2011). Like NH residents without cancer, NH residents with cancer have several chronic conditions and comorbid conditions (Jordhoy et al. 2003; Rodin 2008), but they tend to have symptoms that differ from the symptoms of the others (Brandt et al. 2005). Symptoms such as pain, shortness of breath, vomiting, weight loss and diarrhoea have been shown to be more prevalent among residents with cancer than among those without cancer (Duncan et al. 2009). Compared with people with dementia, those with cancer report more pain, dyspnoea, constipation and weight loss (Rodin 2008). In studies from Norway and the Netherlands, people with cancer frequently reported fatigue, anorexia, nausea and vomiting (Jordhoy et al. 2003; Brandt et al. 2005). Based on the knowledge of their symptoms it is reasonable to believe that NH residents with cancer are more likely to be hospitalised. Only a few studies of hospital admission of NH residents with cancer have been published. A study of NH residents who were admitted to departments of internal medicine in Spain between 2005 and 2008 showed that cancer was a significant factor together with higher age, female sex, a previous feeding tube, acute infectious disease, acute respiratory failure and nosocomial pneumonia in predicting in-hospital mortality (Barba et al. 2012). Another study reported that people with a diagnosis of cancer in NH often presented drug-related problems and, avoidable complications that resulted in rehospitalisation or extended post-acute care stays. These events were related to transferring people between settings, particularly the NH and the hospital (Rodin 2008). A follow-up study after surgery among people with colorectal cancer who had metastases (Woo et al. 1991) showed frequent intermediate admissions to an acute hospital and that most died in an acute hospital. A retrospective study of potential risk factors for hospitalisation of 50 NH residents who were hospitalised during days 4 to 30 of rehabilitation were compared with those of a matched group of 50 people in rehabilitation after hospitalised but without subsequent hospitalisation. The strongest predictors of hospitalisation were a history of a malignant solid tumour, recent hospitalisation for gastrointestinal conditions and low serum albumin (Grabowski et al. 2008). Most studies of the hospital admission of NH residents with a diagnosis of cancer have focused on demographic variables and diseases (Woo et al. 1991; Barba et al. 2012; 148
Dombrowski et al. 2012), whereas little attention has been paid to psychosocial factors as social support in relation to hospital admission. People with a cancer diagnosis are known to have specific disabilities and symptoms caused by the disease and problems in engaging in social activities and social relationships (Jordhoy et al. 2003; Duncan et al. 2008; Foster et al. 2009). Social support has shown to be important for health and well-being among NH residents (Drageset et al. 2009) and to contribute positively for people with cancer (Yildirim & Kocabiyik 2010). Based on the literature review and our previous findings, we wanted to investigate whether residents with a cancer diagnosis have more hospital admissions than other NH residents and whether demographic (sex, age group, marital status, educational level), medical diagnosis and social support variables (attachment, nurturance, reassurance of worth and social integration) are associated with hospital admission among cognitively intact NH residents with and without cancer. We used a model of social support developed by Weiss. Weiss’s (1973, 1974) model of social provisions includes all the primary components of the most current conceptualisations of social support (Cutrona & Russell 1987) and describes six categories of relational provisions: attachment (emotional closeness from which people derive a sense of security); social integration (relationships in which people share concerns and common interests); opportunities for nurturance (being responsible for providing care to others); reassurance of worth (having a sense of competence and esteem); reliable alliances (people can count on assistance in times of need) and guidance (having relationships with people who can provide knowledge and advice). Each provision is associated with a specific type of relationship. People’s needs for specific relational provisions may differ according to age, life situation and environmental situation (Weiss 1974), as studies among older people have shown (Tiikkainen & Heikkinen 2005; Lyyra & Heikkinen 2006). The subscales ‘guidance’ and ‘reliable alliance’ are highly correlated with the other subscales (Cutrona & Russell 1987), and we therefore omitted them (Mancini & Blieszner 1992; Andersson & Stevens 1993). The three specific aims of the study were: • to investigate whether NH residents with a diagnosis of cancer have a higher risk of hospital admission than those without a diagnosis of cancer; • for residents with at least two hospital admissions, to investigate whether the same diagnosis occurred at both the first and second hospital admissions more often for those with a cancer diagnosis than for those without; and © 2014 John Wiley & Sons Ltd
Acute hospital admission for nursing home residents
• to investigate whether sex, age, marital status, educational level, medical diagnoses and social support (attachment, nurturance, reassurance of worth and social integration) were associated with transfer to hospital among NH residents with a cancer diagnosis. METHODS Design and population A cross-sectional design was used at baseline, with 5-year follow-up for hospital admission. All long-term care residents from the 30 NHs in Bergen, Norway (n = 2042) were potential participants and included between 15 January 2004 and 31 May 2005, with follow-up in 2010 (Drageset et al. 2009). The inclusion criteria were that the residents were aged 65 years or older, had resided for at least 6 months and were cognitively intact. A resident was defined as cognitively intact when she or he scored 0.5 or less on the Clinical Dementia Rating Scale (CDR) (Hughes et al. 1982) and was capable of carrying out a conversation. CDR was developed as an instrument for determining the staging of dementia and is scored as no (0), questionable (0.5), mild (1), moderate (2) and severe (3) dementia. The overall level of dementia is derived by using a standard algorithm (Morris 1993). A study in Norway (Nygaard & Ruths 2003) showed that CDR staging is a valid substitute for a dementia work-up among NH residents. Trained nurses who had observed the NH residents for at least 4 weeks assessed CDR before inclusion. The nurses were instructed to base their CDR scoring upon mental functioning and not to include physical frailty. The CDR has high interrater reliability for physicians and clinical nursing specialists (McCulla et al. 1989). The main purpose was to study health-related quality of life (HRQOL) and survival. This article focuses on comparing the risk of hospital admission between residents with and without cancer at the time of inclusion. The same groups have previously been compared with respect to different aspects related to HRQOL and survival (mortality) after 5 years (Drageset et al. 2012, 2013a,b,c). Of 2042 NH residents, 252 fulfilled the inclusion criteria and were invited by a primary care nurse to participate. We obtained the data on Social Provision Scale (SPS) through face-to-face interviews. We conducted the interviews in the respondent’s room or at another appropriate place in the NH. The principal investigator recorded the demographic information and conducted the interviews by reading the questions to the participants and circling the indicated answer. This was required, since many NH residents have problems holding a pen and have impaired © 2014 John Wiley & Sons Ltd
vision. We gave each participant a large-type version of the questionnaire so that they could follow the questions. The principal investigator ensured that the residents understood the questions. Of the 252, 25 (10%) declined to participate. The sample comprised 227 residents for data collection and analysis, of which 60 had a cancer diagnosis and 167 were without cancer diagnosis at inclusion. The 227 participants provided informed consent. The Western Norway Regional Committee for Medical and Health Research Ethics and the Norwegian Social Science Data Services approved the study (REK.Vest no. 162.03/ 2009/1550). We identified all 227 respondents through NH medical records for hospital admission. Their personal identification numbers were linked to the hospital records to assure that no admissions were missing. We recorded the information about diagnoses at discharge from hospital, frequencies and length of stay in hospital and mortality.
Outcome variables The primary outcome was time from inclusion in the sample in 2004–2005 until the first hospital admission. For residents with no hospital admission after inclusion, we recorded the time until the end of follow-up (in 2010). We defined such times as censored admission times. Further, for people dying in the NH before the first hospital admission, we recorded the time until death and treated it as a censored admission time.
Explanatory variables Demographic variables: sex and having a cancer diagnosis at inclusion (no information about demographic variables and cancer diagnoses was obtained from the medical records and included active cancer as well as previously treated cancer of all types. Medical diagnoses: we used Groll’s index (Functional Comorbidity Index, FCI) to classify medical diagnoses. The FCI is a clinically based measure that includes 18 categories of diagnoses scored present and not present and has a maximum score of 18 (Groll et al. 2005). The diagnoses listed in the index match the general International Classification of Diseases, 10th revision (ICD-10) diagnosis codes except for obesity. The FCI predicts physical functioning as an outcome more accurately than the Charlson Comorbidity Index (Groll et al. 2005). The FCI includes mental and physical diagnoses such as depression and osteoporosis but not social health and has been used previously in studies of NH residents (Drageset et al. 2013a,b,c). 149
DRAGESET ET AL.
Social support: we used the revised SPS (Cutrona et al. 1986; Cutrona & Russell 1987) to assess social support. The form was originally designed in 1984 (Russell et al. 1984) and later revised (Cutrona & Russell 1987). It has six subscales: attachment, social integration, nurturance, reassurance of worth, reliable alliance and guidance. We selected the first four subscales for this study of older NH residents because they are highly correlated and the most important to older people (Cutrona & Russell 1987; Bondevik & Skogstad 1998). Four statements assessed each subscale. We used the total score for each subscale in the SPS (four items) to assess the level of attachment, social integration, nurturance and reassurance of worth (ranges 0–4). The revised SPS has shown high reliability when used among older people living in the community (Andersson & Stevens 1993; Bondevik & Skogstad 1996, 1998; Saevareid et al. 2012) and in NHs (Bondevik & Skogstad 1998; Drageset 2002; Drageset et al. 2009) and good validity (Bondevik & Skogstad 1996, 1998; Drageset 2002). We used the ICD-10 for the primary diagnosis at discharge from hospital. We retrieved this through hospitalbased patient records and presented in aggregated levels of the main chapters of the ICD-10. We calculated the length of stay using admission and discharge dates. We confirmed the date of death by link between the patient’s journal system and the National Population Registry. The incidence of hospital admission, diagnosis at discharge, length of stay and in-hospital mortality were based on this information.
Statistical analysis Residents with and without cancer were compared with respect to categorical variables (sex, age group, marital status, education and length of stay categorised in 3-year groups) with the chi-square test (Lydersen et al. 2009) and continuous variables (FCI) with Gosset’s t-test (Student 1908). Descriptive statistics (means, medians and standard deviations, interquartile range and proportions) were used for describing the distribution of the recorded variables, and Pearson’s chi-square test was used to test whether residents with a cancer diagnosis had more hospital admissions than residents without a cancer diagnosis. We analysed the time from inclusion (2004–2005) to the first hospital admission with survival analysis (time-toevent analysis) defining time to death and end of follow-up (2 February 2010) as censored observation times. Unadjusted survival analysis was performed using the Kaplan– Meier procedure (Kaplan & Meier 1958; Mantel 1966) comparing residents with and without cancer with the log-rank test (Mantel 1966) and a Cox simple regression 150
model analysis (Cox 1972). Finally, we performed backward stepwise Cox regression analysis to adjust for potential confounding variables. The fully adjusted model (first step) and final model (last step) are reported and the interaction between cancer and social integration was tested. The criterion for statistical significance was P ≤ 0.05. We performed the statistical analysis using SPSS (version 18).
RESULTS The baseline characteristics of the study sample have been reported previously (Drageset et al. 2012). The most frequent types of cancer were prostate cancer (12%) among men and breast cancer (20%) among women. Common chronic diseases were equally prevalent between residents with and without cancer except for osteoporosis (P < 0.001), which was more common among residents with cancer. In total, 81 residents of 227 (36%) were hospitalised. Of the 81, 26 (43%) had a cancer diagnosis at the time of the interview. Residents with a cancer diagnosis had more hospital admissions (26 of 60) than residents without (55 of 167) (log-rank test, P = 0.04) (Fig. 1; Table 1). Among the 54 residents that had at least two hospital admissions, 17 of 34 residents without cancer (50%) were in the same diagnosis group at both admissions. Among the 20 residents with cancer with at least two hospital admissions, five had the same diagnosis group (25%) [odds ratio (OR) = 0.33; 95% confidence interval (CI): (0.10, 1.12); chi-square test, P = 0.09]. Table 2 shows the fully adjusted analysis of hospital admissions. Residents with cancer had more admissions than residents without cancer (P = 0.04). Hospital admission occurred more frequently for residents with the highest education (P = 0.03) and higher social integration (P = 0.02). In the final model, a cancer diagnosis and social integration were statistically significant (P = 0.03 and P = 0.008 respectively). Residents with a cancer diagnosis at inclusion had 1.7 times higher risk for hospital admission than residents without, but cancer did not interact with the social integration. Education was no longer statistically significant (P = 0.07), but modelling it as a dichotomy, those with ≥3 years of education had a significantly higher risk of hospitalisation than those with less education. The estimates for cancer and social integration did not change substantially with adjustment for education.
DISCUSSION This study showed that NH residents with cancer diagnoses had a 1.7 times higher risk for hospital admission than © 2014 John Wiley & Sons Ltd
Acute hospital admission for nursing home residents
Figure 1. Kaplan–Meier curve for time from inclusion to first hospital admission for 227 cognitively intact nursing home residents in Bergen, Norway included in 2004–2005.
Table 1. Primary diagnosis at discharge of 81 hospitals admissions among cognitively intact nursing home residents in Bergen, Norway included in 2004–2005 Primary diagnosis at discharge of 81 admission ICD-10 codes Main chapter Infectious diseases (A00-B99) Cancer (C00-D48) Haematological diseases (D 50-89) Mental disorders (F00-99) Diseases of circulatory system (I00-99) Diseases of respiratory system (J00-99) Diseases of digestive system (K00-99) Skin diseases (L00-99) Musculo-skeletal disorders (M00-99) Injuries and intoxications (S00-99) Diseases of genitourinary system (N00-N99) Other and not specified diagnosis (R00-99, Z00-99) Total number of admissions
Residents with cancer
Residents without cancer
(%)
n
(%)
1 8 1 1 5 3 1 2 0
(3.8) (30.8) (3.8) (3.8) (19.2) (11.5) (3.8) (7.7) (0.0)
4 0 3 0 13 13 4 3 4
(7.3) (0.0) (5.5) (0.0) (23.6) (23.6) (7.3) (5.5) (6.2)
4 0
(15.4) (0.0)
5 6
(9.1) (10.9)
n
26
55
ICD, International Classification of Diseases, 10th revision.
residents without cancer. More frequent hospitalisation among NH residents with a cancer diagnosis has been shown previously (Woo et al. 1991) in a more limited cancer population (colorectal cancer with metastasis), whereas this study included all types of cancer. In this study, we have no information about the symptoms leading to admission in hospital, but NH residents with cancer diagnoses often have generalised weakness, loss of © 2014 John Wiley & Sons Ltd
appetite, vomiting and nausea (Brandt et al. 2005; Duncan et al. 2008) and cancer-related pain (Bourbonniere & Van Cleave 2006; Rodin 2008) and may therefore need more advanced palliative care than NHs usually can offer (Ranhoff & Linnsund 2005). Other explanations could be insufficient healthcare in NHs (Konetzka et al. 2008), lack of end-of-life decisions (Murray & Laditka 2010) and unawareness of residents’ preferences (Grabowski et al. 2008). 151
DRAGESET ET AL.
Table 2. Proportional hazards regression analysis of time to first hospital admission for 227 cognitively intact nursing home residents in Bergen, Norway in 2004–2005 with 5 years of follow-up
Sex Women Men Age group (years) 65–74 75–84 85–94 ≥95 Educational level Primary school
Original article
Acute hospital admission for nursing home residents without cognitive impairment with a diagnosis of cancer J. DRAGESET, PHD, RN, POSTDOCTOR, Faculty for Health and Social Science, Bergen University College, Bergen, Norway, G.E. EIDE, PHD, BIOSTATISTICIAN, PROFESSOR, Centre for Clinical Research, Western Norway Health Region Authority, Bergen, and Research Group for Lifestyle Epidemiology, Department of Public Health and Primary Health Care, University of Bergen, Bergen, Norway, C. HARRINGTON, PHD, RN, PROFESSOR, Department of Social & Behavioral Sciences, School of Nursing, University of California, San Francisco, CA, USA, & A.H. RANHOFF, PHD, MD, PROFESSOR, Kavli Research Centre for Ageing and Dementia, Haraldsplass Hospital, Bergen, and Institute for Internal Medicine, University of Bergen, Bergen, Norway DRAGESET J., EIDE G.E., HARRINGTON C. & RANHOFF A.H. (2015) European Journal of Cancer Care 24, 147–154 Acute hospital admission for nursing home residents without cognitive impairment with a diagnosis of cancer Studies of hospitalisation of cognitively intact nursing home (NH) residents with cancer are scarce. Knowledge about associations between socio-demographic, medical and social support variables and hospital admissions aids in preventing unnecessary admissions. This is part of a prospective study from 2004 to 2005 with follow-up to 2010 for admission rates. We studied whether residents with cancer have more admissions and whether socio-demographic and medical variables and social support subdimensions are associated with admission among cognitively intact NH residents with (n = 60) and without (n = 167) cancer aged ≥65 years scoring ≤0.5 on the Clinical Dementia Rating Scale and residing ≥6 months. We measured social support by face-to-face interview. We identified all respondents through NH medical records for hospital admission, linking their identification numbers to the hospital record system to register all admissions. We examined whether socio-demographic and medical variables (medical records) and social support subscales were associated with the time between inclusion and first admission. Residents with cancer had more admissions (25/60) than those without (53/167) (odds ratio 1.7). Social integration was correlated with admission (P = 0.04) regardless of cancer diagnosis. Residents with cancer had more hospital admissions than those without. Higher social integration gave more admissions independent of cancer diagnosis.
Keywords: nursing home, cancer, social support, hospital admission.
INTRODUCTION Older people who need residential care have a high risk of hospitalisation (Konetzka et al. 2008; Phelan et al. 2012). Among nursing home (NH) residents, acute illness is common (Alessi & Harker 1998; Murray & Laditka 2010) Correspondence address: Jorunn Drageset, Faculty of Health and Social Sciences, Bergen University College, Møllendalsv 6-8, N-5005 Bergen, Norway (e-mail: [email protected]).
Accepted 3 April 2014 DOI: 10.1111/ecc.12205 European Journal of Cancer Care, 2015, 24, 147–154
© 2014 John Wiley & Sons Ltd
and hospitalisation is frequent (Grabowski et al. 2008). The benefits of the hospitalisation of NH residents are often questioned because of the high risk of complications (Sager et al. 1996; Inouye et al. 2006) and lack of evidence indicating that hospitalisation leads to better outcomes (Konetzka et al. 2008). Several studies from various countries and healthcare systems have investigated admission from NH to hospital (Grabowski et al. 2008). One systematic review showed that men who are older and have a shorter length of stay in the NH are more likely to be hospitalised. Physical and cognitive disability and specific health conditions such as
DRAGESET ET AL.
congestive heart failure and infections increased the risk for hospitalisation (Grabowski et al. 2008). The most frequent admission diagnoses are found to be infections (Alessi & Harker 1998; Kruger et al. 2011), fractures, cardiovascular and gastrointestinally related diagnoses (Kruger et al. 2011), respiratory diseases, fall-related diagnoses and circulatory diseases (Graverholt et al. 2011). Like NH residents without cancer, NH residents with cancer have several chronic conditions and comorbid conditions (Jordhoy et al. 2003; Rodin 2008), but they tend to have symptoms that differ from the symptoms of the others (Brandt et al. 2005). Symptoms such as pain, shortness of breath, vomiting, weight loss and diarrhoea have been shown to be more prevalent among residents with cancer than among those without cancer (Duncan et al. 2009). Compared with people with dementia, those with cancer report more pain, dyspnoea, constipation and weight loss (Rodin 2008). In studies from Norway and the Netherlands, people with cancer frequently reported fatigue, anorexia, nausea and vomiting (Jordhoy et al. 2003; Brandt et al. 2005). Based on the knowledge of their symptoms it is reasonable to believe that NH residents with cancer are more likely to be hospitalised. Only a few studies of hospital admission of NH residents with cancer have been published. A study of NH residents who were admitted to departments of internal medicine in Spain between 2005 and 2008 showed that cancer was a significant factor together with higher age, female sex, a previous feeding tube, acute infectious disease, acute respiratory failure and nosocomial pneumonia in predicting in-hospital mortality (Barba et al. 2012). Another study reported that people with a diagnosis of cancer in NH often presented drug-related problems and, avoidable complications that resulted in rehospitalisation or extended post-acute care stays. These events were related to transferring people between settings, particularly the NH and the hospital (Rodin 2008). A follow-up study after surgery among people with colorectal cancer who had metastases (Woo et al. 1991) showed frequent intermediate admissions to an acute hospital and that most died in an acute hospital. A retrospective study of potential risk factors for hospitalisation of 50 NH residents who were hospitalised during days 4 to 30 of rehabilitation were compared with those of a matched group of 50 people in rehabilitation after hospitalised but without subsequent hospitalisation. The strongest predictors of hospitalisation were a history of a malignant solid tumour, recent hospitalisation for gastrointestinal conditions and low serum albumin (Grabowski et al. 2008). Most studies of the hospital admission of NH residents with a diagnosis of cancer have focused on demographic variables and diseases (Woo et al. 1991; Barba et al. 2012; 148
Dombrowski et al. 2012), whereas little attention has been paid to psychosocial factors as social support in relation to hospital admission. People with a cancer diagnosis are known to have specific disabilities and symptoms caused by the disease and problems in engaging in social activities and social relationships (Jordhoy et al. 2003; Duncan et al. 2008; Foster et al. 2009). Social support has shown to be important for health and well-being among NH residents (Drageset et al. 2009) and to contribute positively for people with cancer (Yildirim & Kocabiyik 2010). Based on the literature review and our previous findings, we wanted to investigate whether residents with a cancer diagnosis have more hospital admissions than other NH residents and whether demographic (sex, age group, marital status, educational level), medical diagnosis and social support variables (attachment, nurturance, reassurance of worth and social integration) are associated with hospital admission among cognitively intact NH residents with and without cancer. We used a model of social support developed by Weiss. Weiss’s (1973, 1974) model of social provisions includes all the primary components of the most current conceptualisations of social support (Cutrona & Russell 1987) and describes six categories of relational provisions: attachment (emotional closeness from which people derive a sense of security); social integration (relationships in which people share concerns and common interests); opportunities for nurturance (being responsible for providing care to others); reassurance of worth (having a sense of competence and esteem); reliable alliances (people can count on assistance in times of need) and guidance (having relationships with people who can provide knowledge and advice). Each provision is associated with a specific type of relationship. People’s needs for specific relational provisions may differ according to age, life situation and environmental situation (Weiss 1974), as studies among older people have shown (Tiikkainen & Heikkinen 2005; Lyyra & Heikkinen 2006). The subscales ‘guidance’ and ‘reliable alliance’ are highly correlated with the other subscales (Cutrona & Russell 1987), and we therefore omitted them (Mancini & Blieszner 1992; Andersson & Stevens 1993). The three specific aims of the study were: • to investigate whether NH residents with a diagnosis of cancer have a higher risk of hospital admission than those without a diagnosis of cancer; • for residents with at least two hospital admissions, to investigate whether the same diagnosis occurred at both the first and second hospital admissions more often for those with a cancer diagnosis than for those without; and © 2014 John Wiley & Sons Ltd
Acute hospital admission for nursing home residents
• to investigate whether sex, age, marital status, educational level, medical diagnoses and social support (attachment, nurturance, reassurance of worth and social integration) were associated with transfer to hospital among NH residents with a cancer diagnosis. METHODS Design and population A cross-sectional design was used at baseline, with 5-year follow-up for hospital admission. All long-term care residents from the 30 NHs in Bergen, Norway (n = 2042) were potential participants and included between 15 January 2004 and 31 May 2005, with follow-up in 2010 (Drageset et al. 2009). The inclusion criteria were that the residents were aged 65 years or older, had resided for at least 6 months and were cognitively intact. A resident was defined as cognitively intact when she or he scored 0.5 or less on the Clinical Dementia Rating Scale (CDR) (Hughes et al. 1982) and was capable of carrying out a conversation. CDR was developed as an instrument for determining the staging of dementia and is scored as no (0), questionable (0.5), mild (1), moderate (2) and severe (3) dementia. The overall level of dementia is derived by using a standard algorithm (Morris 1993). A study in Norway (Nygaard & Ruths 2003) showed that CDR staging is a valid substitute for a dementia work-up among NH residents. Trained nurses who had observed the NH residents for at least 4 weeks assessed CDR before inclusion. The nurses were instructed to base their CDR scoring upon mental functioning and not to include physical frailty. The CDR has high interrater reliability for physicians and clinical nursing specialists (McCulla et al. 1989). The main purpose was to study health-related quality of life (HRQOL) and survival. This article focuses on comparing the risk of hospital admission between residents with and without cancer at the time of inclusion. The same groups have previously been compared with respect to different aspects related to HRQOL and survival (mortality) after 5 years (Drageset et al. 2012, 2013a,b,c). Of 2042 NH residents, 252 fulfilled the inclusion criteria and were invited by a primary care nurse to participate. We obtained the data on Social Provision Scale (SPS) through face-to-face interviews. We conducted the interviews in the respondent’s room or at another appropriate place in the NH. The principal investigator recorded the demographic information and conducted the interviews by reading the questions to the participants and circling the indicated answer. This was required, since many NH residents have problems holding a pen and have impaired © 2014 John Wiley & Sons Ltd
vision. We gave each participant a large-type version of the questionnaire so that they could follow the questions. The principal investigator ensured that the residents understood the questions. Of the 252, 25 (10%) declined to participate. The sample comprised 227 residents for data collection and analysis, of which 60 had a cancer diagnosis and 167 were without cancer diagnosis at inclusion. The 227 participants provided informed consent. The Western Norway Regional Committee for Medical and Health Research Ethics and the Norwegian Social Science Data Services approved the study (REK.Vest no. 162.03/ 2009/1550). We identified all 227 respondents through NH medical records for hospital admission. Their personal identification numbers were linked to the hospital records to assure that no admissions were missing. We recorded the information about diagnoses at discharge from hospital, frequencies and length of stay in hospital and mortality.
Outcome variables The primary outcome was time from inclusion in the sample in 2004–2005 until the first hospital admission. For residents with no hospital admission after inclusion, we recorded the time until the end of follow-up (in 2010). We defined such times as censored admission times. Further, for people dying in the NH before the first hospital admission, we recorded the time until death and treated it as a censored admission time.
Explanatory variables Demographic variables: sex and having a cancer diagnosis at inclusion (no information about demographic variables and cancer diagnoses was obtained from the medical records and included active cancer as well as previously treated cancer of all types. Medical diagnoses: we used Groll’s index (Functional Comorbidity Index, FCI) to classify medical diagnoses. The FCI is a clinically based measure that includes 18 categories of diagnoses scored present and not present and has a maximum score of 18 (Groll et al. 2005). The diagnoses listed in the index match the general International Classification of Diseases, 10th revision (ICD-10) diagnosis codes except for obesity. The FCI predicts physical functioning as an outcome more accurately than the Charlson Comorbidity Index (Groll et al. 2005). The FCI includes mental and physical diagnoses such as depression and osteoporosis but not social health and has been used previously in studies of NH residents (Drageset et al. 2013a,b,c). 149
DRAGESET ET AL.
Social support: we used the revised SPS (Cutrona et al. 1986; Cutrona & Russell 1987) to assess social support. The form was originally designed in 1984 (Russell et al. 1984) and later revised (Cutrona & Russell 1987). It has six subscales: attachment, social integration, nurturance, reassurance of worth, reliable alliance and guidance. We selected the first four subscales for this study of older NH residents because they are highly correlated and the most important to older people (Cutrona & Russell 1987; Bondevik & Skogstad 1998). Four statements assessed each subscale. We used the total score for each subscale in the SPS (four items) to assess the level of attachment, social integration, nurturance and reassurance of worth (ranges 0–4). The revised SPS has shown high reliability when used among older people living in the community (Andersson & Stevens 1993; Bondevik & Skogstad 1996, 1998; Saevareid et al. 2012) and in NHs (Bondevik & Skogstad 1998; Drageset 2002; Drageset et al. 2009) and good validity (Bondevik & Skogstad 1996, 1998; Drageset 2002). We used the ICD-10 for the primary diagnosis at discharge from hospital. We retrieved this through hospitalbased patient records and presented in aggregated levels of the main chapters of the ICD-10. We calculated the length of stay using admission and discharge dates. We confirmed the date of death by link between the patient’s journal system and the National Population Registry. The incidence of hospital admission, diagnosis at discharge, length of stay and in-hospital mortality were based on this information.
Statistical analysis Residents with and without cancer were compared with respect to categorical variables (sex, age group, marital status, education and length of stay categorised in 3-year groups) with the chi-square test (Lydersen et al. 2009) and continuous variables (FCI) with Gosset’s t-test (Student 1908). Descriptive statistics (means, medians and standard deviations, interquartile range and proportions) were used for describing the distribution of the recorded variables, and Pearson’s chi-square test was used to test whether residents with a cancer diagnosis had more hospital admissions than residents without a cancer diagnosis. We analysed the time from inclusion (2004–2005) to the first hospital admission with survival analysis (time-toevent analysis) defining time to death and end of follow-up (2 February 2010) as censored observation times. Unadjusted survival analysis was performed using the Kaplan– Meier procedure (Kaplan & Meier 1958; Mantel 1966) comparing residents with and without cancer with the log-rank test (Mantel 1966) and a Cox simple regression 150
model analysis (Cox 1972). Finally, we performed backward stepwise Cox regression analysis to adjust for potential confounding variables. The fully adjusted model (first step) and final model (last step) are reported and the interaction between cancer and social integration was tested. The criterion for statistical significance was P ≤ 0.05. We performed the statistical analysis using SPSS (version 18).
RESULTS The baseline characteristics of the study sample have been reported previously (Drageset et al. 2012). The most frequent types of cancer were prostate cancer (12%) among men and breast cancer (20%) among women. Common chronic diseases were equally prevalent between residents with and without cancer except for osteoporosis (P < 0.001), which was more common among residents with cancer. In total, 81 residents of 227 (36%) were hospitalised. Of the 81, 26 (43%) had a cancer diagnosis at the time of the interview. Residents with a cancer diagnosis had more hospital admissions (26 of 60) than residents without (55 of 167) (log-rank test, P = 0.04) (Fig. 1; Table 1). Among the 54 residents that had at least two hospital admissions, 17 of 34 residents without cancer (50%) were in the same diagnosis group at both admissions. Among the 20 residents with cancer with at least two hospital admissions, five had the same diagnosis group (25%) [odds ratio (OR) = 0.33; 95% confidence interval (CI): (0.10, 1.12); chi-square test, P = 0.09]. Table 2 shows the fully adjusted analysis of hospital admissions. Residents with cancer had more admissions than residents without cancer (P = 0.04). Hospital admission occurred more frequently for residents with the highest education (P = 0.03) and higher social integration (P = 0.02). In the final model, a cancer diagnosis and social integration were statistically significant (P = 0.03 and P = 0.008 respectively). Residents with a cancer diagnosis at inclusion had 1.7 times higher risk for hospital admission than residents without, but cancer did not interact with the social integration. Education was no longer statistically significant (P = 0.07), but modelling it as a dichotomy, those with ≥3 years of education had a significantly higher risk of hospitalisation than those with less education. The estimates for cancer and social integration did not change substantially with adjustment for education.
DISCUSSION This study showed that NH residents with cancer diagnoses had a 1.7 times higher risk for hospital admission than © 2014 John Wiley & Sons Ltd
Acute hospital admission for nursing home residents
Figure 1. Kaplan–Meier curve for time from inclusion to first hospital admission for 227 cognitively intact nursing home residents in Bergen, Norway included in 2004–2005.
Table 1. Primary diagnosis at discharge of 81 hospitals admissions among cognitively intact nursing home residents in Bergen, Norway included in 2004–2005 Primary diagnosis at discharge of 81 admission ICD-10 codes Main chapter Infectious diseases (A00-B99) Cancer (C00-D48) Haematological diseases (D 50-89) Mental disorders (F00-99) Diseases of circulatory system (I00-99) Diseases of respiratory system (J00-99) Diseases of digestive system (K00-99) Skin diseases (L00-99) Musculo-skeletal disorders (M00-99) Injuries and intoxications (S00-99) Diseases of genitourinary system (N00-N99) Other and not specified diagnosis (R00-99, Z00-99) Total number of admissions
Residents with cancer
Residents without cancer
(%)
n
(%)
1 8 1 1 5 3 1 2 0
(3.8) (30.8) (3.8) (3.8) (19.2) (11.5) (3.8) (7.7) (0.0)
4 0 3 0 13 13 4 3 4
(7.3) (0.0) (5.5) (0.0) (23.6) (23.6) (7.3) (5.5) (6.2)
4 0
(15.4) (0.0)
5 6
(9.1) (10.9)
n
26
55
ICD, International Classification of Diseases, 10th revision.
residents without cancer. More frequent hospitalisation among NH residents with a cancer diagnosis has been shown previously (Woo et al. 1991) in a more limited cancer population (colorectal cancer with metastasis), whereas this study included all types of cancer. In this study, we have no information about the symptoms leading to admission in hospital, but NH residents with cancer diagnoses often have generalised weakness, loss of © 2014 John Wiley & Sons Ltd
appetite, vomiting and nausea (Brandt et al. 2005; Duncan et al. 2008) and cancer-related pain (Bourbonniere & Van Cleave 2006; Rodin 2008) and may therefore need more advanced palliative care than NHs usually can offer (Ranhoff & Linnsund 2005). Other explanations could be insufficient healthcare in NHs (Konetzka et al. 2008), lack of end-of-life decisions (Murray & Laditka 2010) and unawareness of residents’ preferences (Grabowski et al. 2008). 151
DRAGESET ET AL.
Table 2. Proportional hazards regression analysis of time to first hospital admission for 227 cognitively intact nursing home residents in Bergen, Norway in 2004–2005 with 5 years of follow-up
Sex Women Men Age group (years) 65–74 75–84 85–94 ≥95 Educational level Primary school
