Abstract
INTRODUCTION
Dementia is a major public health concern with well-known detrimental consequences such as high risk for institutionalization and poor survival [1, 2]. However, the effects of dementia on health care consumption have been investigated less thoroughly, and most such studies have used data from convenience samples of individuals tracked in claims files. These studies found that elderly people with dementia have 2 to 4 times higher risk for all-cause hospitalization [3, 4] or unplanned hospitalization [5] than elderly people without dementia. It is important to reduce hospitalizations to lower the formal costs of dementia care and to prevent the further functional decline that is associated with prolonged hospital stays. Some hospitalizations are expected because they are part of diseases treatment, but others, such as hospitalizations for primary care-sensitive conditions (PCSCs), may be avoidable. The concept of PCSCs is widely used to assess health care performance or the quality of diagnosis and treatment in primary care [6]. The concept usually covers chronic and acute health conditions that should be managed in primary care, such as diabetes, heart failure, and pyelonephritis. Hospitalizations for these conditions can be avoided by preventing the onset of these diseases, controlling the acute episodic illness, or managing the chronic condition more effectively [7]. Even the early stages of cognitive impairment alter people’s ability to make health care decisions [8]. Thus, cognitive impairment can affect both timely access to primary care and the process of care, increasing avoidable hospitalizations [9]. Dementia has been associated with increased rates of hospitalizations for PCSC and with all-cause hospitalization in two population-based studies. One study investigated hospitalizations from dementia incidence to death, potentiating the hazard of assessing events in the late stages [10]. The second study focused only on prevalent dementia cases [11]. To our knowledge, no studies have addressed the influence of incipient dementia on hospitalization for PCSCs. Given the high number of undiagnosed dementia cases in the general population of elderly people [12], especially at early stages [13], a better understanding of the association between incipient dementia and hospitalization outcomes especially hospitalizations for PCSCs would be helpful in promoting dementia screening. In the present study, we aimed to use data from a population-based cohort study of Swedish older adults to test the hypothesis that incipient dementia is associated with an increased likelihood of hospitalization for PCSCs.
MATERIALS AND METHODS
Study population
Data were derived from the population-based Swedish National study on Aging and Care in Kungsholmen (SNAC-K) [14]. Study participants were randomly selected adults aged ≥60 years living at home or in institutions in Kungsholmen, a central district of Stockholm, Sweden. They were divided into 11 age cohorts. Follow-up examinations were conducted at six-year intervals for the younger cohorts (60 through 72 years) and at three-year intervals for the older cohorts (≥78 years). The baseline examination was carried out between March 2001 and June 2004. A total of 3,363 (73.3%) participated in the baseline examination. Written informed consent was obtained from all participants at baseline or from their next of kin if the participant had cognitive impairment. SNAC-K was approved by the Ethics Committee at Karolinska Institutet and by the Regional Ethical Review Board in Stockholm. A total of 3,000 of the 3,363 baseline participants lived in the community and did not have dementia. At the time of the follow-up, 257 of the 3,000 declined to be assessed, 377 had died, 98 had moved or were not reachable, and 2,268 were re-examined (Fig. 1). Of those who were re-examined, 175 developed dementia and 2,093 did not.
Data collection
The baseline survey collected information about present status and past events through interviews, clinical examinations, and psychological testing by trained staff, including nurses, psychologists, and physicians, all of whom followed structured protocols (available at http://www.snac.org). Assessments comprised a social interview to collect socio-demographic data, evaluation of physical function performed by a nurse, a clinical examination performed by a physician, and a psychological test battery administered by a psychologist. Sociodemographic characteristics, including age, sex, education, and living arrangements were recorded during the nurse interview. Education was measured as the highest level of formal schooling and was categorized as elementary school level (grades 1 through 9), high school level (grades 10 through 12), or university level or above. Participants were divided by living arrangements into those who lived alone and those who did not. Medical conditions (diabetes, congestive heart failure, coronary heart disease, cancer, chronic obstructive pulmonary disease (COPD), hypertension, and depression, and stroke) were diagnosed by the examining physicians on the basis of the clinical examination, self-reported medical history, and laboratory data (see Supplementary file for more information). Information on medical history was also obtained from the computerized Swedish National Patient Register. This register, which has covered all hospitals in Stockholm since 1967, records diagnoses of diseases and other health problems in accordance with the International Classification of Diseases, ninth and tenth revisions (ICD-9 and ICD-10). Mini-Mental State Examination (MMSE) scores were used to control for global cognitive abilities [15]. Basic activities of daily living (ADL) and instrumental activities of daily living (IADL) were assessed. ADL disability was defined as any restriction in bathing, dressing, toileting, mobility, continence, or feeding [16]. IADL disability was defined as the inability to carry out at least one of the following four activities: Using the telephone, using public transportation, handling finances, and shopping [17]. Inappropriate drug use (IDU) was defined as the use of drugs with anticholinergic properties (urinary and gastrointestinal antispasmodics, anticholinergic antiemetics, class Ia antiarrhythmics, anticholinergic antiparkinsonian drugs, low-potency antipsychotics, tricyclic antidepressants, and first-generation antihistamines), use of long-acting benzodiazepines (diazepam, nitrazepam, or flunitrazepam), propiomazine, or of tramadol [18]. Participants were also asked for their use of current health care assistance (e.g., help with injections, taking medicines, or putting on compression stockings).
Diagnosis of dementia
At baseline and at the follow-up, the diagnosis of dementia followed a three-step procedure and was conducted in accordance with the revised edition of the Diagnostic and Statistical Manual of Mental Disorders-IV criteria. Two preliminary diagnoses were made by physicians working independently; a third opinion was requested in case of disagreement. Because we were interested in the incipient stage of dementia (i.e., the period preceding dementia occurrence), we excluded all prevalent cases of dementia. Thus, the incipient stage of dementia was defined as the period between baseline and follow-up.
Outcomes
The primary outcome, hospitalization for PCSCs, was studied as 1) the presence of at least one hospitalization and 2) the number of hospitalizations between baseline and follow-up. Data on hospital admissions were derived from the National Patient Register. The results of an external review from 2011 suggest that the overall accuracy of diagnoses in this register is about 85% to 95% [19]. Hospitalization for PCSCs was defined as any admission for any condition categorized as an ambulatory care sensitive condition by the National Board of Health and Welfare and Swedish Association of Local Authorities and Regions [20]. This definition has been used in Sweden for several years. The following acute and chronic conditions are included in the Swedish definition: Anemia, asthma, diabetes, congestive heart failure, hypertension, COPD, angina, bleeding stomach ulcer, diarrhea, pelvic inflammatory disease, pyelonephritis, and ear, nose, and throat infections (see Appendix 1). Secondary outcomes included 1) at least one all-cause hospitalization (hospital admission for any cause requiring an overnight stay) and 2) the number of all-cause hospitalizations.
Statistical analyses
Baseline characteristics of participants with incipient dementia (those who developed dementia) and participants without incipient dementia (those who did not develop dementia) were compared using χ2 tests and student’s t-tests separately for the unmatched sample and the matched sample. To investigate the association between dementia status and number of hospital admissions, we performed zero-inflated Poisson regressions and reported incident rate ratios (IRR) with 95% confidence intervals (CIs). This approach deals with overdispersion and zero-inflation, allowing for excess zeros in count models under the assumption that the population is characterized by two groups: One in which members always have zero counts, and one in which members have zero or positive counts. The likelihood of having at least one hospitalization for a PCSC or at least one all-cause hospitalization was analyzed using logistic regressions. In all analyses, the first model was controlled for age, sex, and educational level. We further controlled for diabetes, congestive heart failure, coronary heart disease, cancer, COPD, hypertension, depression, stroke, MMSE score, ADL and IADL disability, inappropriate drug use, and use of health care assistance. Rates of hospitalization were computed as the total number of admissions in each group divided by the delay between baseline and follow-up. To further account for differences between participants who developed dementia and those who did not, we conducted a case-control comparison study, matching each participant with incipient dementia to two control participants for age, gender, number of medical conditions, and living arrangements. Because the delay between baseline and follow-up varied by age group (around 6 years for those ≤72 years and 3 years for those ≥78 years at baseline), we added an offset term that took the interval delay in all analyses into account. We performed several sensitivity analyses. First, we excluded participants who died during the year after follow-up to ensure that the findings were independent of the proximity of death. Second, all analyses were rerun: 1) on the basis of a fixed period of 2 years preceding follow-up and 2) separately for those ≤72 years and those ≥78 years at baseline. Finally, to quantify the impact of incipient dementia on the risk of hospitalization for PCSCs, we calculated the population attributable fraction (PAF). The PAF for dementia and its 95% CI were estimated with the following formula: P*(OR-1)/(1+ P*[OR-1]). In this formula, P represents the proportion of the source population exposed to incipient dementia and OR is the adjusted odds ratio for the association between dementia and hospitalization for PCSCs [21]. All analyses were carried out using SAS software, version 9.3 (SAS Institute Inc.).
RESULTS
Characteristics of participants
Baseline sociodemographic, clinical, and care use characteristics differed significantly between participants who developed dementia and those who did not (Table 1). Participants who developed dementia were older than those who did not (mean±SD: 82.0±8.3 versus 71.4±9.6) and more likely to be women, to have a lower level of education, and to live alone. Participants who developed dementia were also more likely to have congestive heart failure, coronary heart disease, or stroke than those who did not. The two groups did not differ in terms of diabetes, cancer, COPD, hypertension, or depression.
Hospitalization outcomes
The mean interval between baseline and follow-up was 4.7 years (range 2 to 7.5 years). Hospitalizations for PCSCs were rare events, and very few participants experienced one or more (n = 164; 7.2%) or two or more (n = 57; 2.5%) such events. The rate of hospitalization for PCSCs was 25.6 per 1000 person-years in participants who did not develop dementia and 88.2 in participants who did (Table 2). The latter were more prone to have at least one event (19.9% versus 6.3% , p < 0.0001). After adjustment for age, sex, and education, those who developed dementia still had a higher risk of hospitalization for PCSCs (OR = 2.37, 95% CI 1.51–3.73) than those who did not. Further adjustment for health conditions and care use did not affect the OR. There was no association between dementia status and the rate of hospitalizations for PCSCs. Matched analyses slightly attenuated these findings (Table 2).
When we calculated the fraction of PCSC hospitalizations that was potentially attributable to dementia status, we found that 9.6% (95% CI 9.1–10.1) may have been attributable to incipient dementia.
Few participants were hospitalized for acute PCSCs: 1.2% of participants who did not develop dementia and 7.4% of those who did. After adjustment for age, sex, and education, incipient dementia was associated with an increased risk of at least one hospitalization (OR = 3.43, 95% CI 1.61–7.27; Table 3). Further adjustment for health conditions and care use reduced the association (OR = 2.99, 95% CI 1.29–6.92). A total of 5.7% of the participants were hospitalized for chronic PCSCs: 5.2% of those who did not develop dementia and 12% of those who did. In the fully adjusted model, incipient dementia was not associated with hospitalization for chronic PCSCs (OR = 1.63, 95% CI 0.86–3.11). Matched analyses confirmed these findings. Hospitalizations for pyelonephritis, diabetes, and congestive heart failure accounted for 41% of all PCSC hospitalizations. The ORs (95% CI) for hospitalizations for incipient dementia were 6.08 (95% CI 2.30–16.04) for pyelonephritis, 7.25 (95% CI 1.46–36.13) for diabetes, and 2.38 (95% CI 1.08–5.02) for congestive heart failure.
All-cause hospitalizations were frequent; approximately half of all participants (n = 1126, 49.6%) experienced at least one event. However, there was no difference in IRR (0.98, 95% CI 0.83–1.17) or in risk (OR = 1.07, 95% CI 0.73–1.55) between participants who developed dementia and those who did not (Table 4). The results of the matched analyses confirmed these findings (Table 4).
Sensitivity analyses
Over the year after follow-up, 78 participants died. After excluding these participants, we found a stronger association between incipient dementia and risk of hospitalization for PCSCs (OR = 2.66, 95% CI 1.52–4.64). The association between incipient dementia and increased risk for hospitalization for PCSCs remained consistent when we performed analyses for the fixed 2-year period preceding follow-up (OR = 2.82, 95% CI 1.46–5.42), when we included only those 72 years or less (OR = 4.47, 95% CI 1.58–12.62), or only those ≥78 years (OR = 1.89, 95% CI 1.02–3.52). Finally, when we repeated the analyses using the number of hospitalizations for PCSCs and all-cause hospitalization as the outcomes, we obtained similar results to those of the initial analyses.
DISCUSSION
In this large population-based cohort of commu-nity-dwelling elderly people, we found that incipient dementia was independently associated with an approximately 2.3 times higher odds of hospitalization for PCSCs. Importantly, this association remained robust even after taking into account indicators of greater patient complexity that are known to influence hospitalization, such as age, functional status, medical conditions, or inappropriate drug use [22]. However, neither the number of hospitalizations for PCSCs nor all-cause hospitalization differed by dementia status. The literature on the relationship between hospitalization and dementia is limited, especially on the early stage of dementia. Our findings, which suggesting that incipient dementia may influence the care process, represent a first step forward. Most previous studies have used medical records registries to track dementia, which result in the identification of non-representative cases of predominantly moderate to severe dementia. Moreover, classification of cases and controls was highly questionable since about half of dementia cases are not diagnosed which contributes to miss dementia cases who could be considered as non-demented in control groups [12]. One study followed incident dementia cases screened in a population based cohort until the end of follow-up and found a 78% higher rate for PCSC and 41% for all-cause hospitalizations [10]. Nevertheless, in that study, higher mortality rates in those with dementia explained half the difference in all-cause admissions; after excluding follow-up periods in which people died, the rate fell to 21% . In the present study, we did not find any association between all-cause hospitalization and incipient dementia. Our finding is in line with the findings of two previous studies that investigated health care consumption in the year preceding dementia diagnosis, even though in those studies, dementia cases were identified in medical records [23] or, in another population-based cohort, 2 years before dementia diagnosis [24]. One possible explanation is that at such an early stage, dementia status may not be associated with an overall difference in health status that could lead to differences in all-cause admissions.
Although hospitalization for PCSCs is a validated indicator of impeded access to primary and preventive care services, there is currently no international standard that dictates which diagnoses and acute events should be included in the definition of PCSCs, and this limits comparison between countries [25]. Rates of hospitalizations for PCSCs were lower in our cohort (88 admissions per 1000 person-years in those who developed dementia and 26 in those who did not) than in Phelan et al.’s study (116 admissions per 1000 person-years in those who developed dementia and 37 in those who did not) [10]. Phelan et al. used the PCSC definition of the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Projects, which encompass more conditions than the Swedish definition. Thus, the small difference in PCSC admissions rates may be explained by baseline characteristics of participants (younger in Phelan et al.’s study) and the fact in the previous study, the participants were members of an integrated health care plan with a proactive approach that aimed to anticipate diseases complications.
In about one out of ten subjects experiencing a PCSC hospitalization it may be attributable to incipient dementia in our cohort of non-demented participants. Because hospitalization for PCSCs may imply a breakdown in the quality of and/or access to primary care [26, 27] or care continuity [28], our findings suggest that more systematic cognitive screening (notably screening for mild cognitive impairment) and close medical monitoring may have the potential to prevent a non-negligible proportion of hospitalizations for PCSCs. Even though the conditions included in the Swedish definition of PCSCs are principally manageable in primary care, a variety of complex causes contribute to the hospitalization process. Some causes originate at the organizational level; that is, at the level of primary care delivery (suboptimal clinical monitoring, accessibility of general practitioners). Others may be more attributable to the patient’s centered level (peoples’ self-care behavior, judgment of their health symptoms, and propensity to seek medical attention). Both levels are affected by the dementing process. Although there has been a decades-long debate over whether to screen older people for dementia, cognitive impairment is far from systematically screened for in primary care, and dementia still frequently goes undiagnosed [29]. The onset of dementia is insidious, and the prodromal phase can last for more than a decade [30]. The progressive loss of cognitive and intellectual functioning alters medical decision-making in an early manner [8, 31]. Consequently, people with dementia are more prone to be diagnosed at advanced stages when they seek medical attention for functional decline or behavioral troubles; thus the formal recognition of symptoms by health professionals predominantly occurs at the time of complications. Moreover, the diagnosis and management of comorbid conditions can be influenced by dementia, as dementia dominates clinical encounters and shifts attention away from other health problems [32, 33]. Thus, delayed diagnosis could postpone implementation of timely person centered-care management. We found that incipient dementia was associated with an increased risk for hospitalization for only three PCSCs. Two of these, diabetes and congestive heart failure, are major predictors of hospitalization in elderly people. Among the 28 hospitalizations for diabetes, and 69 ones for congestive heart failure, 21 (75%) and 27 (39%) respectively, concerned participants with incipient dementia. Nevertheless, we acknowledge that the probability of loss of statistical power caused by the low prevalence of events cannot be ruled out.
Strengths of this study include the large population-based sample of elderly people initially free of dementia, the availability of the Swedish register-based data on hospitalization, and the thorough ascertainment of medical conditions from several sources of medical diagnoses, including clinical examination. Dementia was diagnosed through a standardized procedure, which allowed us to investigate the incipient stage of dementia using real incident dementia cases rather than prevalent claims-based cases. Moreover, numerous sensitivity analyses and adjustment for cognitive and functional performance in fully-adjusted models increased our confidence that our findings are related to dementia status rather than to baseline differences between groups. However, some limitations need to be pointed out. Some participants dropped out of the study (mainly because of death); thus incipient dementia could have been developed in some individuals and missed in our analyses. Second, some participants who did not have dementia at follow-up may have developed it later. Thus, including these individuals in our analyses might have resulted in an underestimation of the influence of incipient dementia on hospital outcomes. Third, the severity of some medical conditions could also confound the relationship between dementia status and hospital outcomes. Moreover, cognitive decline associated with dementia is surely part of global aging process that cannot be properly taken into account by adjustment for individual diseases or disabilities. Fourth, although hospitalization for PCSCs is an important quality indicator, the extent to which enhanced quality of primary care could prevent such hospitalizations remains unclear. Caution is thus needed when considering the implications of our findings. Finally, our study included participants from the center of an urban area with a high density of primary care services and easily accessible hospitals, so the generalizability of our findings is questionable. Nevertheless, it is important to note that deprivation and care accessibility are strong determinants of PCSC admissions; the magnitude of the association between hospitalization for PCSCs and incipient dementia could be even greater in more deprived settings [9, 35]. Finally, as in any observational study, there is a possibility of residual confounding.
In conclusion, we found that incipient dementia was associated with an increased risk of hospitalization for PCSCs but not with all-cause hospitalization. Future research should seek to determine the specific mechanisms through which cognitive impairment leads to hospitalization for PCSCs and to assess the degree to which cognitive screening has the potential to reduce these mechanisms and their associated outcomes. Furthermore, it would be relevant to characterize which subgroups of elderly people would require cognitive testing to prevent deleterious events.
Footnotes
ACKNOWLEDGMENTS
Research grants were received from the Swedish Council for Working Life and Social Research and the Swedish Research Council in Medicine. This study was also supported in part by funds from the Loo and Hans Ostermans Foundation and the Foundation for Geriatric Diseases at Karolinska Institutet, the Gamla Tjänarinnor Foundation, the Gun och Bertil Stohnes Foundation, Demensfonden, Tornspiran Foundation, and Lindhes Advokatbyrå AB (Sweden).
The sponsors had no role in study design, data collection, data analysis, data interpretation, the writing of the report, or the decision to submit the paper for publication.
