- Email: [email protected]
Patients' diagnosis decisions in Alzheimer's disease: The influence of family factors
Social Science & Medicine 118 (2014) 9e16
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Patients' diagnosis decisions in Alzheimer's disease: The influence of family factors Thomas Rapp, PhD ^le de Toulouse, France Universit e Paris Descartes (LIRAES) & G erontopo
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 December 2013 Received in revised form 10 July 2014 Accepted 23 July 2014 Available online 24 July 2014
It is surprising to observe that the number of patients receiving a late diagnosis for Alzheimer's disease (AD) remains high even in countries promoting early diagnosis campaigns. We explore the impact of family history and family support on the risks of receiving a delayed diagnosis. We use French data of 1131 patients diagnosed between 1991 and 2005. We find that the presence of AD history in the family increased the risks of receiving a delayed diagnosis. This was true especially when AD history involved brothers, sisters and other relatives (uncles or cousins). The presence of an informal caregiver at the time of the first warning signs reduced the risks of receiving a late diagnosis, regardless of the informal caregiver concerned (spouse, son, daughter etc.). We identify several opportunities for early detection campaigns. Families with history of disease should be targeted. Campaigns should also target isolated patients, who do not benefit from informal care. Our results underline the importance of improving the diagnosis access for old patients and for men. © 2014 Elsevier Ltd. All rights reserved.
Keywords: France Diagnosis Information Informal care
1. Introduction The implementation of early detection campaigns in Alzheimer's disease (AD), which aim at minimizing the negative effects of the disease and preventing disease-related complications, is a global challenge. Indeed, the number of patients receiving a late diagnosis for AD remains high even in countries that promote early diagnosis campaigns. In France, it has been estimated that only 50% of AD patients are diagnosed, and that the average time from first warning signs to AD diagnosis is 2 years (Dartigues, 2011). Similar results have been observed even in countries that have developed integrated care systems, such as Canada (Carpentier et al., 2010). Surprisingly, several surveys underline that people generally would prefer to learn their risks of having AD in the future. In the United-States, two surveys provided evidence that more than 70% of people would be willing to get a diagnosis test for AD if it was available, regardless of its accuracy (Neumann et al., 2012, 2001). Similarly, a French survey underlined that 90% of people declare willing to get a diagnosis test before the first warning signs for AD if such a test was available, 78% of them because the benefits
E-mail address: [email protected]. http://dx.doi.org/10.1016/j.socscimed.2014.07.052 0277-9536/© 2014 Elsevier Ltd. All rights reserved.
associated with the news are expected to be greater than the costs (TNS-SOFRES, 2013). In other words, there is a disconnection between the willingness to get diagnosed for AD (as expressed in surveys), and the actual diagnosis rates observed when people actually face the first warning signs of the disease. This difference underlines the complexity of the AD diagnosis decision. Indeed, several economic, social, psychological and disease-related factors can influence help-seeking behaviors, leading previous research to find a lot of heterogeneity in decisions (Neumann et al., 2012). Gender, education, income, family history, age and healthy behaviors were associated with differences in diagnosis decisions, and were found to influence the willingness to pay for a diagnosis test (Lin et al., 2013; Neumann et al., 2012, 2001). For many people, the associated social stigma and psychosocial consequences of the AD diagnosis make it not worth pursuing, and contribute to make the inclination to search for a diagnosis lower for AD than for other conditions such as arthritis or prostate cancer (Neumann et al., 2012). In consequence, diagnosis decisions must be further explored. In the economic theory, the patient's attitude towards the risks and the uncertainty associated with her future health status, e.g. when she does not know with certainty the outcomes about her health state, can explain the diagnosis decisions. There is evidence that people would prefer risks over uncertainty (Ellsberg, 1961) and would present an aversion to ambiguous situation, e.g. when the
10
T. Rapp / Social Science & Medicine 118 (2014) 9e16
probabilities of different outcomes are unknown or uncertain (Viscusi et al., 1991). Following these evidences, early AD diagnosis choices could be interpreted as choices made by people disliking the ambiguity coming with the first warning signs, and preferring negative news to an ambiguous situation (Neumann et al., 2012). Previous research showed that patients' disease risks awareness is a crucial variable to model the diagnosis-seeking process (Clare, 2003; Clare et al., 2008; Hutchinson et al., 1997). In addition, the demand for tertiary prevention strongly depends on the sick individuals' perception of its degree of efficiency: individuals will not make the diagnosis test if they are too pessimistic (Etner and Jeleva, 2013). Factors such as age, sex, and health status contribute to explain why people' perceptions of their risks of death or disease may differ from their objective risks (Andersson and Lundborg, 2007; Slovic, 2000). Beliefs about future health states may also contribute to reduce the willingness to get diagnosed. Specifically, anxiety (or aversion to information) can be an inhibiting factor that can lead patients avoiding a diagnosis in the presence of suspicious symptoms, when they anticipate that their health state undiagnosed and treated without a medical treatment is greater than their health state with the diagnosis and the associated treatment (Koszegi, 2003). Finally, previous research has explored the economic value of a diagnosis test, and provided evidence of the importance of the benefits and costs of potential treatments (Eeckhoudt et al., 1984). In AD, potential medical treatments have a low efficacy, which may reduce the incentive to get a diagnosis. Patients' attitude towards risks and uncertainty is likely to be influenced by two family factors: disease history in the family and the presence of informal caregiver when the patients experience the first warning signs. First, there is evidence that individuals' attitudes towards a disease change when a close a relative was previously affected by this disease. For instance, previous research underlined that smoking intensity was complementary to newly diagnosed non-smoking-related family cancers (Ganz, 2001). In New-Zealand, the presence of family disease history was associated with delayed physicians visits among a breast cancer population (Meechan et al., 2002). Moreover, there is evidence that when past experience is composed by the decision relevant events, it directly influences insurance decisions (Cohen et al., 2008). Following these evidences, it can be assumed that AD history in patients' close family is likely to influence the behavior of the person experiencing the first warning signs. For instance, it is likely that individuals with a family history of AD do not feel the need for diagnosis since they have observed that AD treatments remain minimally efficacious. Moreover, it can be assumed that patients with AD history in the family have worked out familial strategies from prior dealings with the condition, which could increase the odds of a delayed diagnosis. Second, diagnosis choices can be explained by interactions between individuals. In particular, previous research underlined the central role of informal caregivers (in general, spouses or partners) in the management of AD (Wimo et al., 2002). It is likely that the siblings of someone experiencing AD signs play a role in the diagnosis seeking decision. In France, 38% of people would ask advice to a close family member before deciding to get a diagnosis (TNS-SOFRES, 2013). Focusing on the initial phases of AD patients' care trajectory, previous research also provided strong evidence that family members, friends and neighbors play a central role in the recognition of the disease, which is crucial to implement interventions for early detection (Carpentier et al., 2010). Finally, there has been evidence that family members often were the first to express concerns about the patient's health, and played a key role in the initiation of the diagnosis (Hansen et al., 2008). Following these evidences, it is expected that the presence of informal caregivers would impact the diagnosis decision.
In this paper, we explore AD diagnosis decisions from an empirical perspective. Using data from a population of communitydwelling French patients diagnosed between 1991 and 2005, we explore to what extend family factors influence the AD diagnosis delays. Specifically, we have two objectives. First, we explore whether the presence of AD history in the family is associated with risks of a late diagnosis. Second, we explore to what extend the presence of informal caregivers at the date of first signs is associated with risks of a late diagnosis. Providing empirical evidence that family factors are influencing the diagnoses would be very important from a health policy perspective, as AD detection campaigns usually involve informal caregivers.
2. Research design 2.1. The PLASA study Our sample was drawn from the PLASA study, which design, inclusion/exclusion criteria, demographics, and methodology are detailed in previous publications (Nourhashemi et al., 2010, 2008). It is a French sample of 1131 community-dwelling patients recruited between 2003 and 2005 nationwide. The study was funded by a public grant from the French government. In the overall study, patients were randomized in two arms, one receiving the intervention and the other usual care. Ethical procedures were followed in the study through an internal review board agreement. Consent processes following the French Law were used. The Institutional Review Boards of the University of Toulouse approved methods in May 2003. The aim of the PLASA study was to explore the impact of a multicomponent intervention designed to decrease the rate of functional decline in patients with mild to moderate AD compared with usual care in memory clinics. Note that the intervention design is not relevant in our article, since we used data collected prior to the intervention. We used historical data collected from the PLASA participants, and our analyses focus on events that occurred prior to inclusion or at inclusion when the diagnosis was provided when the patient entered the study. Patients were recruited if they had a diagnosis for mild to moderate AD. The diagnosis was given by a doctor using the criteria of the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer, Disease and Related Disorders Association or probable or possible Alzheimer's disease (McKhann et al., 1984), with a Mini Mental Status Examination (MMSE) score between 12 and 26. Patients were recruited with age 60þ, but there were no age limits. Patients were recruited if they had a primary informal caregiver: patients were asked if a relative, co-resident or not, was providing assistance for performing activities of daily living. The primary informal caregiver was asked to participate in the study. Informal caregivers were family members or close relatives in charge of providing the general support to the patient. Specifically, the primary informal caregiver was defined as the person mainly in charge of helping the patient with her activities of daily living, instrumental activities of daily living, and involved in her supervision. Patients and caregivers were recruited directly from memory clinics. Note that in France, access to healthcare is subject to a minimal out-of-pocket contribution. The memory clinics were sampled regardless of convenience sample considerations, but were chosen to have sufficient expertise in both diagnosis and management of AD. Written informed consent from both the patients and their caregivers were obtained at inclusion, and the study was reported “according to the consolidated standards of reporting trials statement and its extensions to cluster randomized trials and to nondrug interventions” (Nourhashemi et al., 2010).
T. Rapp / Social Science & Medicine 118 (2014) 9e16
2.2. Analysis sample Because of missing observations for the variable recording the date of the first warning signs, our final sample included 995 observations. That the removed observations did not differ from the remaining observations was checked using t-tests. We found that there were statistical differences for the following variables: the removed observations had more patients in the job category “employee” (p ¼ 0.0134), more patients with an informal caregiver (p ¼ 0.0260), and a lower rate of late diagnoses (p ¼ 0.0386). For all other variables, we did not find statistical differences. 3. The econometric model 3.1. Dependent variable We created a categorical measure of diagnosis delay, using the time span observed between the year of first warning signs and the diagnosis year. To define relevant categories, we explored when AD history and informal care presence were more likely to influence delayed diagnoses using a non-parametric survival model. The temporal variable measured the difference between the date of first signs (origin) and the diagnosis date (failure). This time span variable was measured in years because the day and month of these events were not recorded in the study design. We used the (Kaplan and Meier, 1958) estimator as a non-parametric estimate of the survivor function S(t) that measured the probability of being diagnosed after year t. In our sample given by t1.….tk diagnosis dates (k ranging from 0 to 18 years), this estimate was given by:
b SðtÞ ¼
Y jjtj t
nj dj nj
!
where nj was the number of patients who experienced the first signs at time tj, and where dj was the number of diagnoses for AD at time tj. We first compared KaplaneMeier curves using a dichotomous variable representing the presence of AD history in the patients' family. Second, we compared our survival estimations using two groups that had or not the presence of informal caregivers between the date of first signs and the diagnosis date. We tested the presence of statistical differences by groups for all nonparametric estimations using the Wilcoxon's test provided by (Breslow, 1970; Gehan, 1965). The KaplaneMeier functions (curves not reported, but available upon request) showed that 4 years after the first signs the proportion of patients with AD diagnosis was below 15%, and that the differences between groups were larger within the first 4 years. In other words, the presence of AD history in the patients' family and the presence of informal caregivers seemed to explain delays in diagnosis within the first 4 years after the first signs occurrence, and the largest differences occurred before the fifth year after the date of first signs. Therefore, we created a categorical measure of the time span between the date of the first warning signs and the diagnosis date (0, 1, 2, 3, and 4þ years). 3.2. The models We ran ordered logistic models where the dependent variable was the categorical measure for diagnosis delay. Our models can be presented as latent-variable models for which the structural model was defined as follows:
11
y*i ¼ a þ bXi þ εi where y* was the latent variable ranging from ∞ to þ∞, i was the observation, a was a constant, Xi a vector of patients i's characteristics, and ε an error term. The latent variable was divided into 5 ordinal categories defined as follows:
8 0 > > > > > <1 yi ¼ 2 > > > 3 > > : 4
if t0 if t1 if t2 if t3 if t4
¼ ∞ y*i < t1 y*i < t2 y*i < t2 y*i < t4 y*i < t5 ¼ þ∞
where the cut points t1et4 were estimated. The category 0 represented the patients diagnosed the same year as the year of first signs, the category 1 represented the patients diagnosed one year after the year of first signs, the category 2 represented the patients diagnosed two years after the year of first signs, the category 3 represent the patients diagnosed three years after the year of first signs, and the category 4 represented the patients diagnosed four (or more) years after the year of first signs.
3.3. Specifications We specified two models. In Model 1, our two control variables of interest were: a dichotomous variable showing whether AD history was found in the patients' family, and a dichotomous variable showing whether informal caregivers were involved in patients' care provision between the date of first troubles and the diagnosis date. In Model 2, we replaced the variable measuring the presence of AD history in the patients' family by three variables measuring whether the patients' brother/sisters had AD, whether the patients' parents had AD, and whether the patients' other family relatives (cousins or uncles) had AD. We also replaced the variable showing the presence of an informal caregiver by a variable describing the nature of the relationship between this caregiver and the patient: spouse, son, daughter, and other. Models 1 and 2 also controlled for several economic, demographic and clinical variables that were likely to explain late diagnoses. We controlled for patients' socioeconomic and demographic characteristics: patients' age at the diagnosis date, gender, monthly income categories (less than V760, V760eV1525, V1525eV2300, V2300eV3000, and greater than V3000), education level (greater than baccalaureate or not), and job category (trader, worker, employee, farmer, and executive). We controlled for two additional dichotomous variables measuring the occurrence of clinical events that could have impacted the diagnosis date: the presence of a surgery between the trouble date and the diagnosis date, and the presence of a diagnosis for depression between the trouble date and the diagnosis date. We controlled for two informal caregivers' demographic characteristics: age at the diagnosis year and gender. Finally, we introduced a variable measuring the difference between the patients' subjective quality of life and her quality of life estimated by her primary informal caregiver. We used the Quality of Life in Older Adults with Cognitive Impairment (QoL-AD) questionnaire. The questionnaire has 13-items covering physical health, energy, mood, living situations, memory, family, marriage, friends, chores, fun, money, self and life as a whole. Using coefficient alpha, previous research underlined the high internal consistency reliability of patient and caregiver reports on the QoL-AD: alpha values ranged from 0.83 to 0.90 (Logsdon et al., 2002). Moreover, Pearson
12
T. Rapp / Social Science & Medicine 118 (2014) 9e16
correlation coefficients provided evidence of a significant association between associations between patient and caregiver-reported QoL-AD scores and several measures representing hypothesized QoL domains (Logsdon et al., 2002). Because of additional missing variables dealing with quality of life assessments, these analyses were run using a sample of 883 observations. We ran three additional sets of ordered logistic regressions. Model 3 controlled for the QoL difference, using the sample of 883 patients who had or not an informal caregiver between the first signs date and the diagnosis date. Model 4 controlled for the QoL difference among a subgroup of 372 patients who had an informal caregiver between the first signs date and the diagnosis date. Model 5 controlled for the QoL difference among a subgroup of 511 patients who did not have an informal caregiver between the first signs date and the diagnosis date. These Models did not control for education, job category and income, which were found to be non-significant in Models 1 and 2. 3.4. Post-estimation tests and sensitivity analyses
Table 1 Descriptive statistics (n ¼ 995). Variable
Mean
Standard deviation
Time to diagnosis y* Presence of AD history Involving brother/sister Involving parent Involving another relative Presence of informal caregiver Spouse Son Daughter Other Income < 760 euros/month 760 < Income < 1525 euros/month 1525 < Income <2300 euros/month 2300 < Income < 3000 euros/month 3000 < Income Farmer Traders Other Non executive employee Worker Executive Baccalaureate Patient is male Patient's age at diagnosis Caregiver's age at diagnosis Surgery after warning signs Depression diagnosis after warning signs
2.294 1.983 0.311 0.075 0.144 0.091 0.431 0.180 0.057 0.149 0.045 0.101 0.429 0.224 0.116 0.098 0.073 0.142 0.080 0.362 0.144 0.071 0.151 0.308 78.689 61.150 0.251 0.152
2.117 1.293 0.463 0.264 0.351 0.288 0.495 0.384 0.233 0.356 0.208 0.301 0.495 0.417 0.320 0.298 0.261 0.349 0.272 0.481 0.351 0.258 0.358 0.462 6.027 13.542 0.434 0.359
Min
Max
0 0
18 4
51 23
96 90
Ordered logistic models rely on the strong assumption that the odds between diagnosis dates categories are parallels. We tested the relevance of this assumption using the Wald test defined by (Brant, 1990). We found that the parallel regression assumption was not violated. The p-values of the Brant tests in Models 1, 2, 3, 4, and 5 were respectively: p ¼ 0.254, p ¼ 0.137, p ¼ 0.307, p ¼ 0.413, and p ¼ 0.907. We further explored the robustness of our results using different specifications of our dependent variable. The time span between the date of the first warning signs and the diagnosis date being measured as follows: 0, 1, 2, 3þ years, and 0, 1, 2, 3, 4, and 5þ years. We also controlled for regional dummies to capture potential geographical differences, but none of these variables were significant. These analyses, not reported in the manuscript but available upon request, provided similar results. For all Models, we computed odds ratios (OR) and robust (White corrected) standard errors.
between patients with AD history in the family and patients with no AD history in the family (p ¼ 0.0108). Table 2 also show that the presence of an informal caregiver improved the chances of an early diagnosis. The mean time to diagnosis was 1.55 year when the patient had an informal caregiver, and 2.85 years when the patient did not have an informal caregiver (p < 0.01).
4. Results
4.3. Correlation between diagnosis delay and AD history
4.1. Sample description
Table 3 provides the results of the ordered logistic models. In Model 1 the presence of AD history in the family was associated with a higher risk of receiving a delayed diagnosis (OR ¼ 1.50; p < 0.01). In Model 2, we found that patients whose brother/sister had AD faced a higher risk of getting a late diagnosis, compared with patients with no family history of disease (OR ¼ 1.63; p ¼ 0.016). We found a similar relationship when comparing patients with AD family history involving other relatives with patients with no AD history in the family (OR ¼ 1.67; p ¼ 0.013). However, we found that having a parent with AD was not associated with differences in the odds of receiving a late diagnosis (p ¼ 0.119).
Table 1 provides descriptive statistics of our sample. On average, patients experiencing first warning signs spent 2.2 years (SD: 2.1) before the AD diagnosis. In our sample, 31.1% of the patients had AD history running in their family (SD: 46.3), involving brothers/sisters (7.5%), parents (14.4%) or other relatives (9.1%). Less than half of the patients had an informal caregiver involved between the date of first signs and the diagnosis date (43.1%): 18% were spouses, 5.7% were sons, 14.9% were daughters and 4.5% were other relatives (friends, cousins or uncles). The mean informal caregivers' age at the patients' diagnosis date was 61. The mean patients' age at diagnosis was 78, and 30.8% of the patients were males. Only 15.1% of patients had a diploma higher than the baccalaureate. When they were active, most of patients were non-executive employees (36.2%), workers (14.4%) or traders (14.2%). Executive employees represented 7.1% of our sample. A quarter of our sample received surgery between the dates of first signs and diagnosis, and 15.2% received a diagnosis for depression within that time span.
4.4. Correlation between diagnosis delay and informal care Table 3 also provides the results of the association between diagnosis delay and the presence of an informal caregiver at the
Table 2 Wilcoxon tests (n ¼ 995).
4.2. Wilcoxon tests Table 2 provides results of the Wilcoxon test for equality of means. The mean diagnosis time for the patients with no AD history in the family (2.19 years) was lower than the mean diagnosis time of the group defined by patients with AD history in the family (2.51 years). Results confirmed the presence of statistical differences
Mean Presence of AD history No Yes Presence of informal caregiver No Yes
2.19 2.51 2.85 1.55
p-value 0.0108
<0.01
T. Rapp / Social Science & Medicine 118 (2014) 9e16
13
Table 3 Determinants of a delayed diagnosis for Alzheimer's disease (AD).a Dependent variable: 0,1,2,3,4þ years of delay
Model 1
Model 2
Odds ratio
P-value
% Change
1.502
<0.01
50.2
Odds ratio
P-value
% Change
1.638 1.327 1.670
0.016 0.119 0.013
63.8 32.7 67.0
0.249 0.211 0.211 0.202 Reference 0.985 1.104 1.170 1.002 Reference 1.286 1.204 0.978 1.380 1.437 1.408 1.290 1.044 0.982 1.763 1.761
<0.01 <0.01 <0.01 <0.01
75.1 78.9 78.9 79.8
0.932 0.640 0.535 0.993
1.5 4.2 5.1 0.2
0.338 0.372 0.933 0.060 0.120 0.102 0.078 <0.01 <0.01 <0.01 <0.01
28.6 20.4 2.2 38.0 43.7 40.8 29.0 4.4 1.8 76.3 76.1
Presence of AD history Involving brother/sister Involving parent Involving another relative Presence of informal caregiver Spouse Son Daughter Other Income <760 euros/month 760 < Income < 1525 euros/month 1525 < Income < 2300 euros/month 2300 < Income < 3000 euros/month 3000 < Income Executive Farmer Traders Other Non executive employee Worker Baccalaureate Patient is male Patient's age at diagnosis Caregiver's age at diagnosis Surgery after warning signs Depression diagnosis after warning signs
Reference 0.978 1.110 1.176 0.997 Reference 1.292 1.212 0.980 1.376 1.448 1.386 1.316 1.043 0.983 1.776 1.761
t1 t2 t3 t4
0.166 1.977 3.225 4.083
0.107 1.918 3.168 4.028
Pseudo R2 Brant test Number of observations
0.0684 0.254 995
0.0689 0.137 995
0.226
<0.01
77.4
0.902 0.623 0.521 0.990
2.2 11.0 17.6 0.3
0.322 0.353 0.937 0.061 0.109 0.113 0.055 <0.01 <0.01 <0.01 <0.01
29.2 21.2 2.0 37.6 44.8 38.6 31.6 4.3 1.7 77.6 76.1
a This table presents the results of the ordered logit models where the dependent variable measured 5 time span categories: 0,1,2,3, 4þ years of delay between the date of first warning signs and the diagnosis date.
time of the first warning signs. In Model 1, the presence of primary informal caregivers reduced the risk of a late diagnosis (OR ¼ 0.22; p < 0.01). In Model 2, we see that this effect was found when the caregiver was either the patient's spouse, son or daughter, or any other family relative (p < 0.01 for all variables). 4.5. Additional determinants of AD diagnosis delays Table 3 provides results of additional determinants of diagnosis delays. We found that income categories were not associated with differences in the odds of receiving a late diagnosis. Compared with executives, the non-executive job category was associated with higher odds of receiving a delayed diagnosis (OR ¼ 1.37; p ¼ 0.061). Education was not associated with diagnosis delay (p ¼ 0.113). An increase in patients' age (atdiagnosis) was associated with an increase in the odds of receiving a late diagnosis (OR ¼ 1.04; p < 0.01). Compared to females, male patients had higher odds of being diagnosed with delay (Model 1: OR ¼ 1.31; p ¼ 0.055; Model 2: OR ¼ 0.29; p ¼ 0.078). The presence of a depression diagnosis after the date of first warning signs was a strong predictor for late diagnosis receipt (OR ¼ 1.76; p < 0.01). Patients receiving surgery after the date of warning signs had a higher risk of having their diagnosis date postponed, controlling or not for the presence of confusion events (OR ¼ 1.76; p < 0.01). Finally, an increase in informal caregivers' age was associated with a decrease in the risks of a late diagnosis (OR ¼ 0.98; p < 0.01 in Models 1 and 2).
4.6. Association with QoL differences In Table 4, we found that a greater QoL difference was associated with higher late diagnosis risks (Model 3: OR ¼ 1.02; p ¼ 0.036). In these analyses, the positive role of informal caregivers was also confirmed (OR ¼ 0.23; p < 0.01). Models 3, 4, and 5 also confirmed that the presence of AD history in the family increased the risks of a late diagnosis. When running these analyses among patients with informal care, we found that the effect of the QoL difference was no longer significant (p ¼ 0.372). When running these analyses among patients with no informal care, we found that QoL difference increased the odds of a late diagnosis (OR ¼ 1.02; p ¼ 0.028). All other variables had similar effects and significance than in Models 1 and 2. 5. Discussion 5.1. Key findings Our principle findings were that family history slows diagnosis, but informal care speeds it up. These results must be carefully discussed, as they were not intuitive. We first found that the presence of AD history in the family increased the risks of receiving a delayed diagnosis. This result can be explained by the fact that patients with AD history in the family already have some knowledge about the disease and their future states of health, and that the diagnosis news are not going to move
14
T. Rapp / Social Science & Medicine 118 (2014) 9e16
Table 4 Differences in quality of life estimations and odds of a delayed diagnosis.a Dependent variable: 0,1,2,3,4þ years of delay
Model 3 Odds ratio
Presence of AD history Presence of informal caregiver Patient is male Patient's age at diagnosis Caregiver's age at diagnosis Surgery after warning signs Depression diagnosis after warning signs Quality of life difference
t1 t2 t3 t4 Pseudo R2 Brant test Number of observations
1.555 0.233 1.362 1.039 0.983 1.744 1.788 1.022
Model 4
Model 5
P-value
% Change
Odds ratio
P-value
% Change
Odds ratio
<0.01 <0.01 0.042 <0.01 <0.01 <0.01 <0.01 0.036
55.8 76.2 44.5 3.8 1.6 72.5 81.6 2.2
1.713
0.015
65.3
1.421
0.026
48.3
1.167 1.038 0.986 1.482 1.949 1.015
0.534 0.018 0.099 0.146 <0.01 0.372
25.1 3.8 1.5 55.2 92.9 1.1
1.465 1.039 0.980 1.949 1.629 1.029
0.013 0.018 0.013 <0.01 0.032 0.028
60.1 3.8 1.6 80.5 67.4 2.9
0.078 1.787 3.038 3.881
1.141 2.785 4.076 4.881
0.433 2.019 3.277 4.145
0.0695 0.307 883
0.0398 0.413 372
0.0344 0.585 511
P-value
% Change
a This table presents the results of the ordered logit models where the dependent variable measured 5 time span categories: 0,1,2,3, 4þ years of delay between the date of first warning signs and the diagnosis date. Model 4 was run on a subsample of 372 patients who had an informal caregiver between the date of first warning signs and the diagnosis date. Model 5 was run on a subsample of 511 patients who did not have informal caregivers.
their beliefs by much. Indeed, previous evidence suggests that disease knowledge is more likely to be related to the family experiences than “outside” information: while there is evidence that dementia risk factors provided by scientific evidence has little impact the general public knowledge (Low and Anstey, 2007), there has been empirical evidence that families' past experience and knowledge of the disease must be considered when exploring diagnosis-seeking in AD (Carpentier et al., 2010). Intuitively, patients with family history faced the dramatic consequences of the disease, and are aware of the large uncertainty surrounding the diagnosis, because of the absence of an effective medical treatment. Therefore, they prefer to remain undiagnosed rather than diagnosed with a medical treatment that has a reduced and uncertain impact on their future well-being. This intuition was strengthened by the fact that the effect was higher when the AD history involved a brother or sister, and not significant when AD history involved a parent. In the case of brothers or sisters, it can be assumed that the patient has observed the low efficacy of current AD treatments. We then found that the presence of an informal caregiver reduced the risks of receiving a late diagnosis, regardless of the informal caregiver concerned (spouse, son, daughter, cousin or friend). Previous research underlined the impact of health risk perception on diagnosis decisions (Cohen et al., 2008; Etner and Jeleva, 2013), and provided evidence of the central role of informal caregivers on the perception of the disease risks in dementia (Hansen et al., 2008). Following these theoretical and empirical evidences, our interpretation is that siblings involved in informal caregiving increase the chances of an early diagnosis by influencing patients' risk perception. This result confirms previous empirical evidence showing the importance of caregivers networks in diagnosis decisions (Carpentier et al., 2010). Indeed, a recent French survey showing that more than 85% of people declare that they would encourage their relative (spouse or parent) getting diagnosed (TNS-SOFRES, 2013). This result underlines the importance of informal care networks at the early phases of AD management. 5.2. Additional factors influencing diagnosis seeking We found four additional results that must be interpreted. First, we provided evidence that differences in subjective QoL estimations were associated with greater risks of a delayed diagnosis,
underlining the need to inform patients and informal caregivers about AD symptoms. In France, a priority was given to inform and educate the public to AD warning signs to improve early diagnosis. Specifically, the 9th objective of the 2008e2012 Alzheimer Plan (Plan-Alzheimer, 2012) was focused on that matter trough the implementation of a dedicated hotline, a dedicated website, the implementation of regional conferences, and research program focusing on the public's perception of the disease. The AD diagnosis announcement is strictly regulated and monitored (HAS, 2011). It has to be provided by a specialist in a dedicated place allowing intimacy and confidentiality. Moreover, the diagnosis can be progressively provided to the patient, following a set of several meetings with the specialist in charge of the diagnosis. Our findings might be used to enhance diagnostic practice. Following previous research (Hansen et al., 2008), we recommend the development of education programs for general practitioners dealing with the benefits of a formal diagnosis of dementia and the need to improve patients' awareness. Second, demographic variables were strongly associated with diagnosis delays, confirming previous evidence (Andersson and Lundborg, 2007; Slovic, 2000). The difference between men and women confirmed a result that has been observed in other diseases prevention campaigns. For instance, previous research has documented that gender was a determinant of late diagnosis in HIV (Kivela et al., 2010). This could be explained by overconfidence rates within that population: previous results obtained in behavioral finance showed that for some decisions men are more subject to overconfidence than women (Barber and Odean, 2001; Bernasek and Shwiff, 2001). According to these articles, gender would explain specific attitudes towards risks and specifically that men are more likely to adopt risky financial investment behaviors. Our interpretation is that similar effects may also be relevant for diagnosis decisions. The effect of patients' age confirmed the presence of a potential bias against the elderly people's value of life, which was described in previous AD research (Fillit et al., 2010). It can also be explained by the fact that in France, after 70 years old, only 9% of people are willing to get an AD diagnosis (TNS-SOFRES, 2013). Third, the economic variables had a reduced impact, while previous evidence showed that diagnosis seeking decisions are likely to be associated with economic variables such as coverage (David et al., 2012). The absence of effect related to income and education and the small statistical effect of job categories can be
T. Rapp / Social Science & Medicine 118 (2014) 9e16
explained by the French health expenditures coverage system. In France, the decision to postpone AD diagnosis should not be explained by budgetary constraint reasons, as AD medical expenditures are fully reimbursed by the National Health Insurance System. Moreover, public financial support is provided to patients with functional and cognitive declines to finance professional and informal care use. Fourth, our results underlined that the presence of surgery and depression diagnosis were associated with diagnosis delays. The effect of these two variables shows that care producers were unable to determine the patients' propensity of having AD, raising the key role of the physician in the diagnosis decision. Further research will have to explore the behavior of care producers facing patients at risks of AD, and specifically their propensity to provide a diagnosis and to treat patients. From a theoretical perspective, this case study is interesting because in AD the medical decision-makers simultaneously face both diagnostic ambiguity and therapeutic ambiguity. The overall effect of diagnostic ambiguity and therapeutic ambiguity on treatment rates is difficult to predict. In the presence of ambiguity aversion, ambiguity regarding the diagnosis of a patient will increase the physician's propensity to provide a treatment for to the patient, while treatment ambiguity will reduce the probability of treatment (Berger et al., 2013). In our sample, 80% of the patients started to use anti-Alzheimer therapies the same year as the diagnosis date, which was high given the small clinical effect of these drugs. These large treatment rates at the time of diagnosis could be explained by the presence of diagnosis ambiguity and ambiguity aversion among physicians in our sample, even if using the year of treatment and the year of diagnosis did not allow for accurate explorations. 5.3. Limitations Our study was faced with some limitations. Cancer data analyses provide evidence that the presence of private insurance has a significant role to explain differences in information seeking behaviors (Walsh et al., 2012). We were not able to control for that parameter in our analyses. However, the French national insurance system reduces the likelihood of such a mechanism in our sample. The generalizability of our results also is of concern. The PLASA study was designed to test the impact of an innovative intervention in AD. It is well known that clinical trials participants usually are more likely to be care seekers. In consequence, the patients that participated in the PLASA study could have some specific (unobserved) characteristics that could impact their history of disease and the outcome that was explored in our paper. The absence of precise records (day/month/year) for the date of first signs and diagnosis date did not allow us running parametric survival analyses, which could have increased the accuracy of our results. However, the use of ordered logistic regression was relevant given the distribution of our outcome, and it was confirmed by the non-significant Brant tests. Finally, our sample excluded 136 patients with missing information for one or more variables. We found that these excluded patients had lower late diagnosis rates and had higher rates of informal caregiving use. The exclusion of this subsample may have leaded us to underestimate the effect of the informal care variable. However, our sensitivity analyses (analyses not reported in the paper but available upon request) confirmed the robustness of our results. First, we used a sample selection model to explore whether there was a correlation between the unobserved factors associated with the probability of having non-missing values and the unobserved factors associated with the probability of receiving an early diagnosis. We controlled in the first stage for a dichotomous variable indicating the presence of missing values for the job variable.
15
We did not find the presence of a significant correlation (Chi2 ¼ 0.32, p ¼ 0.482) and therefore concluded that the potential selection bias was reduced. Second, we found similar results when running a more restricted Model that did not control for income but used more observations of 1005 patients. 5.4. Health policy implications Our results have central health policy implications. Being able to improve early detection in AD is expected to have important economic benefits. There is evidence that the cost of AD dramatically increases with the severity of cognitive, behavioral and functional symptoms, and that being able to manage the disease when the patient has moderate symptoms can lead to large savings (Rapp et al., 2012). Moreover, there is evidence that ambiguity about health status has an impact on quality of life, which can be stronger than physical problems (Graham et al., 2011). Knowing the most important determinants of late diagnoses in AD is therefore crucial for early detection campaigns. We identify several opportunities for early detection campaigns. Families with history of disease should be targeted, as we confirmed the presence of avoidance behaviors among patients with disease awareness. Campaigns should also target isolated patients, who do not benefit from informal care. We recommend that public policies encourage the organization of informal care networks when the patients experience the first warning signs. Our results underlined the importance of improving diagnosis for old patients and for men. We found that early detection can be blurred by clinical events such as surgery or depression diagnosis, revealing that care suppliers' behaviors can also explain delayed diagnoses. Specific campaigns should also be designed to help care producers identifying AD symptoms. Acknowledgments s d’e te This paper was presented at Sciences Po Lille (Universite Alzheimer), and at the Ethics, Economics and Health Seminar of APHP. I wish to thank Bruno Vellas, Nicolas Sirven and Lise Rochaix, and the three anonymous reviewers for providing helpful comments on the manuscript. All errors remain mine. References Andersson, H., Lundborg, P., 2007. Perception of own death risk. An analysis of roadtraffic and overall mortality risks. J. Risk Uncertain. 34, 67e84. Barber, B., Odean, T., 2001. Boys will be boys: gender, overconfidence, and common stock investment. Q. J. Econ. 116, 261e292. Berger, L., Bleichrodt, H., Eeckhoudt, L., 2013. Treatment decisions under ambiguity. J. Health Econ. 32, 559e569. Bernasek, A., Shwiff, S., 2001. Gender, risk, and retirement. J. Econ. Issues 35, 345e356. Brant, R., 1990. Assessing proportionality in the proportional odds model for ordered logistic regression. Biometrics 46, 1171e1178. Breslow, N.E., 1970. A generalized Kruskal-Wallis test for comparing K samples subject to unequal patterns of censorship. Biometrika 57, 579e594. Carpentier, N., Bernard, P., Grenier, A., Guberman, N., 2010. Using the life course perspective to study the entry into the illness trajectory: the perspective of caregivers of people with Alzheimer's disease. Soc. Sci. Med. 70, 1501e1508. Clare, L., 2003. Managing threats to self: awareness in early stage Alzheimer's disease. Soc. Sci. Med. 57, 1017e1029. Clare, L., Rowlands, J., Bruce, E., Surr, C., Downs, M., 2008. ‘I don't do like I used to do’: a grounded theory approach to conceptualising awareness in people with moderate to severe dementia living in long-term care. Soc. Sci. Med. 66, 2366e2377. Cohen, M., Etner, J., Jeleva, M., 2008. Dynamic decision making when risk perception depends on past experience. Theory Decis. 64, 173e192. Dartigues, J.F., 2011. Alzheimer's disease: early diagnosis. Rev. Prat. 61, 926e930. David, G., Saynisch, P., Acevedo-Perez, V., Neuman, M.D., 2012. Affording to wait: medicare initiation and the use of health care. Health Econ. 21, 1030e1036. Eeckhoudt, L.R., Lebrun, T.C., Sailly, J.C., 1984. The informative content of diagnostic tests: an economic analysis. Soc. Sci. Med. 18, 873e880. Ellsberg, D., 1961. Risk, ambiguity, and the Savage axioms. Q. J. Econ. 75, 643e669.
16
T. Rapp / Social Science & Medicine 118 (2014) 9e16
Etner, J., Jeleva, M., 2013. Risk perception, prevention and diagnostic tests. Health Econ. 22, 144e156. Fillit, H., Cummings, J., Neumann, P., McLaughlin, T., Salavtore, P., Leibman, C., 2010. Novel approaches to incorporating pharmacoeconomic studies into phase III clinical trials for Alzheimer's disease. J. Nutr. Health Aging, 640e647. Ganz, M.L., 2001. Family health effects: complements or substitutes. Health Econ. 10, 699e714. Gehan, E.A., 1965. A generalized Wilcoxon test for comparing arbitrarily singly censored samples. Biometrika 52, 203e223. Graham, C., Higuera, L., Lora, E., 2011. Which health conditions cause the most unhappiness? Health Econ. 20, 1431e1447. Hansen, E.C., Hughes, C., Routley, G., Robinson, A.L., 2008. General practitioners' experiences and understandings of diagnosing dementia: factors impacting on early diagnosis. Soc. Sci. Med. 67, 1776e1783. HAS, 2011. Diagnostic et prise en charge de la maladie d'Alzheimer et des maladies es: recommandation 71. Haute Autorite de Sante , Saint-Denis. apparente Hutchinson, S.A., Leger-Krall, S., Wilson, H.S., 1997. Early probable Alzheimer's disease and awareness context theory. Soc. Sci. Med. 45, 1399e1409. Kaplan, E.L., Meier, P., 1958. Nonparametric estimation for incomplete observations. J. Am. Stat. Assoc. 53, 457e481. Kivela, P.S., Krol, A., Salminen, M.O., Ristola, M.A., 2010. Determinants of late HIV diagnosis among different transmission groups in Finland from 1985 to 2005. HIV Med. 11, 360e367. Koszegi, B., 2003. Health anxiety and patient behavior. J. Health Econ. 22, 1073e1084. Lin, P.J., Cangelosi, M.J., Lee, D.W., Neumann, P.J., 2013. Willingness to pay for diagnostic technologies: a review of the contingent valuation literature. Value Health 16, 797e805. Logsdon, R.G., Gibbons, L.E., McCurry, S.M., Teri, L., 2002. Assessing quality of life in older adults with cognitive impairment. Psychosom. Med. 64, 510e519. Low, L.F., Anstey, K.J., 2007. The public's perception of the plausibility of dementia risk factors is not influenced by scientific evidence. Dement. Geriatr. Cogn. Disord. 23, 202e206. McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., Stadlan, E.M., 1984. Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work
Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology 34, 939e944. Meechan, G., Collins, J., Petrie, K., 2002. Delay in seeking medical care for selfdetected breast symptoms in New Zealand women. N. Z. Med. J. 115, U257. Neumann, P.J., Cohen, J.T., Hammitt, J.K., Concannon, T.W., Auerbach, H.R., Fang, C., et al., 2012. Willingness-to-pay for predictive tests with no immediate treatment implications: a survey of US residents. Health Econ. 21, 238e251. Neumann, P.J., Hammitt, J.K., Mueller, C., Fillit, H.M., Hill, J., Tetteh, N.A., et al., 2001. Public attitudes about genetic testing for Alzheimer's disease. Health Aff. (Millwood) 20, 252e264. Nourhashemi, F., Andrieu, S., Gillette-Guyonnet, S., Giraudeau, B., Cantet, C., Coley, N., et al., 2010. Effectiveness of a specific care plan in patients with Alzheimer's disease: cluster randomised trial (PLASA study). BMJ 340, c2466. Nourhashemi, F., Gillette-Guyonnet, S., Andrieu, S., Rolland, Y., Ousset, P.J., Vellas, B., 2008. A randomized trial of the impact of a specific care plan in 1120 Alzheimer's patients (PLASA Study) over a two-year period: design and baseline data. J. Nutr. Health Aging 12, 263e271. Plan-Alzheimer, 2012. Plan Alzheimer 2008e2012. Rapp, T., Andrieu, S., Molinier, L., Grand, A., Cantet, C., Mullins, C.D., et al., 2012. Exploring the relationship between Alzheimer's disease severity and longitudinal costs. Value Health 15, 412e419. Slovic, P., 2000. The Perception of Risk. Earthscan Publications Ltd, London and Sterling. l'anticipation de la maladie d'Alzheimer. In: T. TNS-SOFRES, 2013. Les français face a SOFRES (Ed.). TNS SOFRES, Paris. Viscusi, W., Magat, W., Huber, J., 1991. Communication of ambiguous risk information. Theory Decis. 31, 159e173. Walsh, B., Silles, M., O'Neill, C., 2012. The role of private medical insurance in socioeconomic inequalities in cancer screening uptake in Ireland. Health Econ. 21, 1250e1256. Wimo, A., von Strauss, E., Nordberg, G., Sassi, F., Johansson, L., 2002. Time spent on informal and formal care giving for persons with dementia in Sweden. Health Policy 61, 255e268.
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Patients' diagnosis decisions in Alzheimer's disease: The influence of family factors Thomas Rapp, PhD ^le de Toulouse, France Universit e Paris Descartes (LIRAES) & G erontopo
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 December 2013 Received in revised form 10 July 2014 Accepted 23 July 2014 Available online 24 July 2014
It is surprising to observe that the number of patients receiving a late diagnosis for Alzheimer's disease (AD) remains high even in countries promoting early diagnosis campaigns. We explore the impact of family history and family support on the risks of receiving a delayed diagnosis. We use French data of 1131 patients diagnosed between 1991 and 2005. We find that the presence of AD history in the family increased the risks of receiving a delayed diagnosis. This was true especially when AD history involved brothers, sisters and other relatives (uncles or cousins). The presence of an informal caregiver at the time of the first warning signs reduced the risks of receiving a late diagnosis, regardless of the informal caregiver concerned (spouse, son, daughter etc.). We identify several opportunities for early detection campaigns. Families with history of disease should be targeted. Campaigns should also target isolated patients, who do not benefit from informal care. Our results underline the importance of improving the diagnosis access for old patients and for men. © 2014 Elsevier Ltd. All rights reserved.
Keywords: France Diagnosis Information Informal care
1. Introduction The implementation of early detection campaigns in Alzheimer's disease (AD), which aim at minimizing the negative effects of the disease and preventing disease-related complications, is a global challenge. Indeed, the number of patients receiving a late diagnosis for AD remains high even in countries that promote early diagnosis campaigns. In France, it has been estimated that only 50% of AD patients are diagnosed, and that the average time from first warning signs to AD diagnosis is 2 years (Dartigues, 2011). Similar results have been observed even in countries that have developed integrated care systems, such as Canada (Carpentier et al., 2010). Surprisingly, several surveys underline that people generally would prefer to learn their risks of having AD in the future. In the United-States, two surveys provided evidence that more than 70% of people would be willing to get a diagnosis test for AD if it was available, regardless of its accuracy (Neumann et al., 2012, 2001). Similarly, a French survey underlined that 90% of people declare willing to get a diagnosis test before the first warning signs for AD if such a test was available, 78% of them because the benefits
E-mail address: [email protected]. http://dx.doi.org/10.1016/j.socscimed.2014.07.052 0277-9536/© 2014 Elsevier Ltd. All rights reserved.
associated with the news are expected to be greater than the costs (TNS-SOFRES, 2013). In other words, there is a disconnection between the willingness to get diagnosed for AD (as expressed in surveys), and the actual diagnosis rates observed when people actually face the first warning signs of the disease. This difference underlines the complexity of the AD diagnosis decision. Indeed, several economic, social, psychological and disease-related factors can influence help-seeking behaviors, leading previous research to find a lot of heterogeneity in decisions (Neumann et al., 2012). Gender, education, income, family history, age and healthy behaviors were associated with differences in diagnosis decisions, and were found to influence the willingness to pay for a diagnosis test (Lin et al., 2013; Neumann et al., 2012, 2001). For many people, the associated social stigma and psychosocial consequences of the AD diagnosis make it not worth pursuing, and contribute to make the inclination to search for a diagnosis lower for AD than for other conditions such as arthritis or prostate cancer (Neumann et al., 2012). In consequence, diagnosis decisions must be further explored. In the economic theory, the patient's attitude towards the risks and the uncertainty associated with her future health status, e.g. when she does not know with certainty the outcomes about her health state, can explain the diagnosis decisions. There is evidence that people would prefer risks over uncertainty (Ellsberg, 1961) and would present an aversion to ambiguous situation, e.g. when the
10
T. Rapp / Social Science & Medicine 118 (2014) 9e16
probabilities of different outcomes are unknown or uncertain (Viscusi et al., 1991). Following these evidences, early AD diagnosis choices could be interpreted as choices made by people disliking the ambiguity coming with the first warning signs, and preferring negative news to an ambiguous situation (Neumann et al., 2012). Previous research showed that patients' disease risks awareness is a crucial variable to model the diagnosis-seeking process (Clare, 2003; Clare et al., 2008; Hutchinson et al., 1997). In addition, the demand for tertiary prevention strongly depends on the sick individuals' perception of its degree of efficiency: individuals will not make the diagnosis test if they are too pessimistic (Etner and Jeleva, 2013). Factors such as age, sex, and health status contribute to explain why people' perceptions of their risks of death or disease may differ from their objective risks (Andersson and Lundborg, 2007; Slovic, 2000). Beliefs about future health states may also contribute to reduce the willingness to get diagnosed. Specifically, anxiety (or aversion to information) can be an inhibiting factor that can lead patients avoiding a diagnosis in the presence of suspicious symptoms, when they anticipate that their health state undiagnosed and treated without a medical treatment is greater than their health state with the diagnosis and the associated treatment (Koszegi, 2003). Finally, previous research has explored the economic value of a diagnosis test, and provided evidence of the importance of the benefits and costs of potential treatments (Eeckhoudt et al., 1984). In AD, potential medical treatments have a low efficacy, which may reduce the incentive to get a diagnosis. Patients' attitude towards risks and uncertainty is likely to be influenced by two family factors: disease history in the family and the presence of informal caregiver when the patients experience the first warning signs. First, there is evidence that individuals' attitudes towards a disease change when a close a relative was previously affected by this disease. For instance, previous research underlined that smoking intensity was complementary to newly diagnosed non-smoking-related family cancers (Ganz, 2001). In New-Zealand, the presence of family disease history was associated with delayed physicians visits among a breast cancer population (Meechan et al., 2002). Moreover, there is evidence that when past experience is composed by the decision relevant events, it directly influences insurance decisions (Cohen et al., 2008). Following these evidences, it can be assumed that AD history in patients' close family is likely to influence the behavior of the person experiencing the first warning signs. For instance, it is likely that individuals with a family history of AD do not feel the need for diagnosis since they have observed that AD treatments remain minimally efficacious. Moreover, it can be assumed that patients with AD history in the family have worked out familial strategies from prior dealings with the condition, which could increase the odds of a delayed diagnosis. Second, diagnosis choices can be explained by interactions between individuals. In particular, previous research underlined the central role of informal caregivers (in general, spouses or partners) in the management of AD (Wimo et al., 2002). It is likely that the siblings of someone experiencing AD signs play a role in the diagnosis seeking decision. In France, 38% of people would ask advice to a close family member before deciding to get a diagnosis (TNS-SOFRES, 2013). Focusing on the initial phases of AD patients' care trajectory, previous research also provided strong evidence that family members, friends and neighbors play a central role in the recognition of the disease, which is crucial to implement interventions for early detection (Carpentier et al., 2010). Finally, there has been evidence that family members often were the first to express concerns about the patient's health, and played a key role in the initiation of the diagnosis (Hansen et al., 2008). Following these evidences, it is expected that the presence of informal caregivers would impact the diagnosis decision.
In this paper, we explore AD diagnosis decisions from an empirical perspective. Using data from a population of communitydwelling French patients diagnosed between 1991 and 2005, we explore to what extend family factors influence the AD diagnosis delays. Specifically, we have two objectives. First, we explore whether the presence of AD history in the family is associated with risks of a late diagnosis. Second, we explore to what extend the presence of informal caregivers at the date of first signs is associated with risks of a late diagnosis. Providing empirical evidence that family factors are influencing the diagnoses would be very important from a health policy perspective, as AD detection campaigns usually involve informal caregivers.
2. Research design 2.1. The PLASA study Our sample was drawn from the PLASA study, which design, inclusion/exclusion criteria, demographics, and methodology are detailed in previous publications (Nourhashemi et al., 2010, 2008). It is a French sample of 1131 community-dwelling patients recruited between 2003 and 2005 nationwide. The study was funded by a public grant from the French government. In the overall study, patients were randomized in two arms, one receiving the intervention and the other usual care. Ethical procedures were followed in the study through an internal review board agreement. Consent processes following the French Law were used. The Institutional Review Boards of the University of Toulouse approved methods in May 2003. The aim of the PLASA study was to explore the impact of a multicomponent intervention designed to decrease the rate of functional decline in patients with mild to moderate AD compared with usual care in memory clinics. Note that the intervention design is not relevant in our article, since we used data collected prior to the intervention. We used historical data collected from the PLASA participants, and our analyses focus on events that occurred prior to inclusion or at inclusion when the diagnosis was provided when the patient entered the study. Patients were recruited if they had a diagnosis for mild to moderate AD. The diagnosis was given by a doctor using the criteria of the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer, Disease and Related Disorders Association or probable or possible Alzheimer's disease (McKhann et al., 1984), with a Mini Mental Status Examination (MMSE) score between 12 and 26. Patients were recruited with age 60þ, but there were no age limits. Patients were recruited if they had a primary informal caregiver: patients were asked if a relative, co-resident or not, was providing assistance for performing activities of daily living. The primary informal caregiver was asked to participate in the study. Informal caregivers were family members or close relatives in charge of providing the general support to the patient. Specifically, the primary informal caregiver was defined as the person mainly in charge of helping the patient with her activities of daily living, instrumental activities of daily living, and involved in her supervision. Patients and caregivers were recruited directly from memory clinics. Note that in France, access to healthcare is subject to a minimal out-of-pocket contribution. The memory clinics were sampled regardless of convenience sample considerations, but were chosen to have sufficient expertise in both diagnosis and management of AD. Written informed consent from both the patients and their caregivers were obtained at inclusion, and the study was reported “according to the consolidated standards of reporting trials statement and its extensions to cluster randomized trials and to nondrug interventions” (Nourhashemi et al., 2010).
T. Rapp / Social Science & Medicine 118 (2014) 9e16
2.2. Analysis sample Because of missing observations for the variable recording the date of the first warning signs, our final sample included 995 observations. That the removed observations did not differ from the remaining observations was checked using t-tests. We found that there were statistical differences for the following variables: the removed observations had more patients in the job category “employee” (p ¼ 0.0134), more patients with an informal caregiver (p ¼ 0.0260), and a lower rate of late diagnoses (p ¼ 0.0386). For all other variables, we did not find statistical differences. 3. The econometric model 3.1. Dependent variable We created a categorical measure of diagnosis delay, using the time span observed between the year of first warning signs and the diagnosis year. To define relevant categories, we explored when AD history and informal care presence were more likely to influence delayed diagnoses using a non-parametric survival model. The temporal variable measured the difference between the date of first signs (origin) and the diagnosis date (failure). This time span variable was measured in years because the day and month of these events were not recorded in the study design. We used the (Kaplan and Meier, 1958) estimator as a non-parametric estimate of the survivor function S(t) that measured the probability of being diagnosed after year t. In our sample given by t1.….tk diagnosis dates (k ranging from 0 to 18 years), this estimate was given by:
b SðtÞ ¼
Y jjtj t
nj dj nj
!
where nj was the number of patients who experienced the first signs at time tj, and where dj was the number of diagnoses for AD at time tj. We first compared KaplaneMeier curves using a dichotomous variable representing the presence of AD history in the patients' family. Second, we compared our survival estimations using two groups that had or not the presence of informal caregivers between the date of first signs and the diagnosis date. We tested the presence of statistical differences by groups for all nonparametric estimations using the Wilcoxon's test provided by (Breslow, 1970; Gehan, 1965). The KaplaneMeier functions (curves not reported, but available upon request) showed that 4 years after the first signs the proportion of patients with AD diagnosis was below 15%, and that the differences between groups were larger within the first 4 years. In other words, the presence of AD history in the patients' family and the presence of informal caregivers seemed to explain delays in diagnosis within the first 4 years after the first signs occurrence, and the largest differences occurred before the fifth year after the date of first signs. Therefore, we created a categorical measure of the time span between the date of the first warning signs and the diagnosis date (0, 1, 2, 3, and 4þ years). 3.2. The models We ran ordered logistic models where the dependent variable was the categorical measure for diagnosis delay. Our models can be presented as latent-variable models for which the structural model was defined as follows:
11
y*i ¼ a þ bXi þ εi where y* was the latent variable ranging from ∞ to þ∞, i was the observation, a was a constant, Xi a vector of patients i's characteristics, and ε an error term. The latent variable was divided into 5 ordinal categories defined as follows:
8 0 > > > > > <1 yi ¼ 2 > > > 3 > > : 4
if t0 if t1 if t2 if t3 if t4
¼ ∞ y*i < t1 y*i < t2 y*i < t2 y*i < t4 y*i < t5 ¼ þ∞
where the cut points t1et4 were estimated. The category 0 represented the patients diagnosed the same year as the year of first signs, the category 1 represented the patients diagnosed one year after the year of first signs, the category 2 represented the patients diagnosed two years after the year of first signs, the category 3 represent the patients diagnosed three years after the year of first signs, and the category 4 represented the patients diagnosed four (or more) years after the year of first signs.
3.3. Specifications We specified two models. In Model 1, our two control variables of interest were: a dichotomous variable showing whether AD history was found in the patients' family, and a dichotomous variable showing whether informal caregivers were involved in patients' care provision between the date of first troubles and the diagnosis date. In Model 2, we replaced the variable measuring the presence of AD history in the patients' family by three variables measuring whether the patients' brother/sisters had AD, whether the patients' parents had AD, and whether the patients' other family relatives (cousins or uncles) had AD. We also replaced the variable showing the presence of an informal caregiver by a variable describing the nature of the relationship between this caregiver and the patient: spouse, son, daughter, and other. Models 1 and 2 also controlled for several economic, demographic and clinical variables that were likely to explain late diagnoses. We controlled for patients' socioeconomic and demographic characteristics: patients' age at the diagnosis date, gender, monthly income categories (less than V760, V760eV1525, V1525eV2300, V2300eV3000, and greater than V3000), education level (greater than baccalaureate or not), and job category (trader, worker, employee, farmer, and executive). We controlled for two additional dichotomous variables measuring the occurrence of clinical events that could have impacted the diagnosis date: the presence of a surgery between the trouble date and the diagnosis date, and the presence of a diagnosis for depression between the trouble date and the diagnosis date. We controlled for two informal caregivers' demographic characteristics: age at the diagnosis year and gender. Finally, we introduced a variable measuring the difference between the patients' subjective quality of life and her quality of life estimated by her primary informal caregiver. We used the Quality of Life in Older Adults with Cognitive Impairment (QoL-AD) questionnaire. The questionnaire has 13-items covering physical health, energy, mood, living situations, memory, family, marriage, friends, chores, fun, money, self and life as a whole. Using coefficient alpha, previous research underlined the high internal consistency reliability of patient and caregiver reports on the QoL-AD: alpha values ranged from 0.83 to 0.90 (Logsdon et al., 2002). Moreover, Pearson
12
T. Rapp / Social Science & Medicine 118 (2014) 9e16
correlation coefficients provided evidence of a significant association between associations between patient and caregiver-reported QoL-AD scores and several measures representing hypothesized QoL domains (Logsdon et al., 2002). Because of additional missing variables dealing with quality of life assessments, these analyses were run using a sample of 883 observations. We ran three additional sets of ordered logistic regressions. Model 3 controlled for the QoL difference, using the sample of 883 patients who had or not an informal caregiver between the first signs date and the diagnosis date. Model 4 controlled for the QoL difference among a subgroup of 372 patients who had an informal caregiver between the first signs date and the diagnosis date. Model 5 controlled for the QoL difference among a subgroup of 511 patients who did not have an informal caregiver between the first signs date and the diagnosis date. These Models did not control for education, job category and income, which were found to be non-significant in Models 1 and 2. 3.4. Post-estimation tests and sensitivity analyses
Table 1 Descriptive statistics (n ¼ 995). Variable
Mean
Standard deviation
Time to diagnosis y* Presence of AD history Involving brother/sister Involving parent Involving another relative Presence of informal caregiver Spouse Son Daughter Other Income < 760 euros/month 760 < Income < 1525 euros/month 1525 < Income <2300 euros/month 2300 < Income < 3000 euros/month 3000 < Income Farmer Traders Other Non executive employee Worker Executive Baccalaureate Patient is male Patient's age at diagnosis Caregiver's age at diagnosis Surgery after warning signs Depression diagnosis after warning signs
2.294 1.983 0.311 0.075 0.144 0.091 0.431 0.180 0.057 0.149 0.045 0.101 0.429 0.224 0.116 0.098 0.073 0.142 0.080 0.362 0.144 0.071 0.151 0.308 78.689 61.150 0.251 0.152
2.117 1.293 0.463 0.264 0.351 0.288 0.495 0.384 0.233 0.356 0.208 0.301 0.495 0.417 0.320 0.298 0.261 0.349 0.272 0.481 0.351 0.258 0.358 0.462 6.027 13.542 0.434 0.359
Min
Max
0 0
18 4
51 23
96 90
Ordered logistic models rely on the strong assumption that the odds between diagnosis dates categories are parallels. We tested the relevance of this assumption using the Wald test defined by (Brant, 1990). We found that the parallel regression assumption was not violated. The p-values of the Brant tests in Models 1, 2, 3, 4, and 5 were respectively: p ¼ 0.254, p ¼ 0.137, p ¼ 0.307, p ¼ 0.413, and p ¼ 0.907. We further explored the robustness of our results using different specifications of our dependent variable. The time span between the date of the first warning signs and the diagnosis date being measured as follows: 0, 1, 2, 3þ years, and 0, 1, 2, 3, 4, and 5þ years. We also controlled for regional dummies to capture potential geographical differences, but none of these variables were significant. These analyses, not reported in the manuscript but available upon request, provided similar results. For all Models, we computed odds ratios (OR) and robust (White corrected) standard errors.
between patients with AD history in the family and patients with no AD history in the family (p ¼ 0.0108). Table 2 also show that the presence of an informal caregiver improved the chances of an early diagnosis. The mean time to diagnosis was 1.55 year when the patient had an informal caregiver, and 2.85 years when the patient did not have an informal caregiver (p < 0.01).
4. Results
4.3. Correlation between diagnosis delay and AD history
4.1. Sample description
Table 3 provides the results of the ordered logistic models. In Model 1 the presence of AD history in the family was associated with a higher risk of receiving a delayed diagnosis (OR ¼ 1.50; p < 0.01). In Model 2, we found that patients whose brother/sister had AD faced a higher risk of getting a late diagnosis, compared with patients with no family history of disease (OR ¼ 1.63; p ¼ 0.016). We found a similar relationship when comparing patients with AD family history involving other relatives with patients with no AD history in the family (OR ¼ 1.67; p ¼ 0.013). However, we found that having a parent with AD was not associated with differences in the odds of receiving a late diagnosis (p ¼ 0.119).
Table 1 provides descriptive statistics of our sample. On average, patients experiencing first warning signs spent 2.2 years (SD: 2.1) before the AD diagnosis. In our sample, 31.1% of the patients had AD history running in their family (SD: 46.3), involving brothers/sisters (7.5%), parents (14.4%) or other relatives (9.1%). Less than half of the patients had an informal caregiver involved between the date of first signs and the diagnosis date (43.1%): 18% were spouses, 5.7% were sons, 14.9% were daughters and 4.5% were other relatives (friends, cousins or uncles). The mean informal caregivers' age at the patients' diagnosis date was 61. The mean patients' age at diagnosis was 78, and 30.8% of the patients were males. Only 15.1% of patients had a diploma higher than the baccalaureate. When they were active, most of patients were non-executive employees (36.2%), workers (14.4%) or traders (14.2%). Executive employees represented 7.1% of our sample. A quarter of our sample received surgery between the dates of first signs and diagnosis, and 15.2% received a diagnosis for depression within that time span.
4.4. Correlation between diagnosis delay and informal care Table 3 also provides the results of the association between diagnosis delay and the presence of an informal caregiver at the
Table 2 Wilcoxon tests (n ¼ 995).
4.2. Wilcoxon tests Table 2 provides results of the Wilcoxon test for equality of means. The mean diagnosis time for the patients with no AD history in the family (2.19 years) was lower than the mean diagnosis time of the group defined by patients with AD history in the family (2.51 years). Results confirmed the presence of statistical differences
Mean Presence of AD history No Yes Presence of informal caregiver No Yes
2.19 2.51 2.85 1.55
p-value 0.0108
<0.01
T. Rapp / Social Science & Medicine 118 (2014) 9e16
13
Table 3 Determinants of a delayed diagnosis for Alzheimer's disease (AD).a Dependent variable: 0,1,2,3,4þ years of delay
Model 1
Model 2
Odds ratio
P-value
% Change
1.502
<0.01
50.2
Odds ratio
P-value
% Change
1.638 1.327 1.670
0.016 0.119 0.013
63.8 32.7 67.0
0.249 0.211 0.211 0.202 Reference 0.985 1.104 1.170 1.002 Reference 1.286 1.204 0.978 1.380 1.437 1.408 1.290 1.044 0.982 1.763 1.761
<0.01 <0.01 <0.01 <0.01
75.1 78.9 78.9 79.8
0.932 0.640 0.535 0.993
1.5 4.2 5.1 0.2
0.338 0.372 0.933 0.060 0.120 0.102 0.078 <0.01 <0.01 <0.01 <0.01
28.6 20.4 2.2 38.0 43.7 40.8 29.0 4.4 1.8 76.3 76.1
Presence of AD history Involving brother/sister Involving parent Involving another relative Presence of informal caregiver Spouse Son Daughter Other Income <760 euros/month 760 < Income < 1525 euros/month 1525 < Income < 2300 euros/month 2300 < Income < 3000 euros/month 3000 < Income Executive Farmer Traders Other Non executive employee Worker Baccalaureate Patient is male Patient's age at diagnosis Caregiver's age at diagnosis Surgery after warning signs Depression diagnosis after warning signs
Reference 0.978 1.110 1.176 0.997 Reference 1.292 1.212 0.980 1.376 1.448 1.386 1.316 1.043 0.983 1.776 1.761
t1 t2 t3 t4
0.166 1.977 3.225 4.083
0.107 1.918 3.168 4.028
Pseudo R2 Brant test Number of observations
0.0684 0.254 995
0.0689 0.137 995
0.226
<0.01
77.4
0.902 0.623 0.521 0.990
2.2 11.0 17.6 0.3
0.322 0.353 0.937 0.061 0.109 0.113 0.055 <0.01 <0.01 <0.01 <0.01
29.2 21.2 2.0 37.6 44.8 38.6 31.6 4.3 1.7 77.6 76.1
a This table presents the results of the ordered logit models where the dependent variable measured 5 time span categories: 0,1,2,3, 4þ years of delay between the date of first warning signs and the diagnosis date.
time of the first warning signs. In Model 1, the presence of primary informal caregivers reduced the risk of a late diagnosis (OR ¼ 0.22; p < 0.01). In Model 2, we see that this effect was found when the caregiver was either the patient's spouse, son or daughter, or any other family relative (p < 0.01 for all variables). 4.5. Additional determinants of AD diagnosis delays Table 3 provides results of additional determinants of diagnosis delays. We found that income categories were not associated with differences in the odds of receiving a late diagnosis. Compared with executives, the non-executive job category was associated with higher odds of receiving a delayed diagnosis (OR ¼ 1.37; p ¼ 0.061). Education was not associated with diagnosis delay (p ¼ 0.113). An increase in patients' age (atdiagnosis) was associated with an increase in the odds of receiving a late diagnosis (OR ¼ 1.04; p < 0.01). Compared to females, male patients had higher odds of being diagnosed with delay (Model 1: OR ¼ 1.31; p ¼ 0.055; Model 2: OR ¼ 0.29; p ¼ 0.078). The presence of a depression diagnosis after the date of first warning signs was a strong predictor for late diagnosis receipt (OR ¼ 1.76; p < 0.01). Patients receiving surgery after the date of warning signs had a higher risk of having their diagnosis date postponed, controlling or not for the presence of confusion events (OR ¼ 1.76; p < 0.01). Finally, an increase in informal caregivers' age was associated with a decrease in the risks of a late diagnosis (OR ¼ 0.98; p < 0.01 in Models 1 and 2).
4.6. Association with QoL differences In Table 4, we found that a greater QoL difference was associated with higher late diagnosis risks (Model 3: OR ¼ 1.02; p ¼ 0.036). In these analyses, the positive role of informal caregivers was also confirmed (OR ¼ 0.23; p < 0.01). Models 3, 4, and 5 also confirmed that the presence of AD history in the family increased the risks of a late diagnosis. When running these analyses among patients with informal care, we found that the effect of the QoL difference was no longer significant (p ¼ 0.372). When running these analyses among patients with no informal care, we found that QoL difference increased the odds of a late diagnosis (OR ¼ 1.02; p ¼ 0.028). All other variables had similar effects and significance than in Models 1 and 2. 5. Discussion 5.1. Key findings Our principle findings were that family history slows diagnosis, but informal care speeds it up. These results must be carefully discussed, as they were not intuitive. We first found that the presence of AD history in the family increased the risks of receiving a delayed diagnosis. This result can be explained by the fact that patients with AD history in the family already have some knowledge about the disease and their future states of health, and that the diagnosis news are not going to move
14
T. Rapp / Social Science & Medicine 118 (2014) 9e16
Table 4 Differences in quality of life estimations and odds of a delayed diagnosis.a Dependent variable: 0,1,2,3,4þ years of delay
Model 3 Odds ratio
Presence of AD history Presence of informal caregiver Patient is male Patient's age at diagnosis Caregiver's age at diagnosis Surgery after warning signs Depression diagnosis after warning signs Quality of life difference
t1 t2 t3 t4 Pseudo R2 Brant test Number of observations
1.555 0.233 1.362 1.039 0.983 1.744 1.788 1.022
Model 4
Model 5
P-value
% Change
Odds ratio
P-value
% Change
Odds ratio
<0.01 <0.01 0.042 <0.01 <0.01 <0.01 <0.01 0.036
55.8 76.2 44.5 3.8 1.6 72.5 81.6 2.2
1.713
0.015
65.3
1.421
0.026
48.3
1.167 1.038 0.986 1.482 1.949 1.015
0.534 0.018 0.099 0.146 <0.01 0.372
25.1 3.8 1.5 55.2 92.9 1.1
1.465 1.039 0.980 1.949 1.629 1.029
0.013 0.018 0.013 <0.01 0.032 0.028
60.1 3.8 1.6 80.5 67.4 2.9
0.078 1.787 3.038 3.881
1.141 2.785 4.076 4.881
0.433 2.019 3.277 4.145
0.0695 0.307 883
0.0398 0.413 372
0.0344 0.585 511
P-value
% Change
a This table presents the results of the ordered logit models where the dependent variable measured 5 time span categories: 0,1,2,3, 4þ years of delay between the date of first warning signs and the diagnosis date. Model 4 was run on a subsample of 372 patients who had an informal caregiver between the date of first warning signs and the diagnosis date. Model 5 was run on a subsample of 511 patients who did not have informal caregivers.
their beliefs by much. Indeed, previous evidence suggests that disease knowledge is more likely to be related to the family experiences than “outside” information: while there is evidence that dementia risk factors provided by scientific evidence has little impact the general public knowledge (Low and Anstey, 2007), there has been empirical evidence that families' past experience and knowledge of the disease must be considered when exploring diagnosis-seeking in AD (Carpentier et al., 2010). Intuitively, patients with family history faced the dramatic consequences of the disease, and are aware of the large uncertainty surrounding the diagnosis, because of the absence of an effective medical treatment. Therefore, they prefer to remain undiagnosed rather than diagnosed with a medical treatment that has a reduced and uncertain impact on their future well-being. This intuition was strengthened by the fact that the effect was higher when the AD history involved a brother or sister, and not significant when AD history involved a parent. In the case of brothers or sisters, it can be assumed that the patient has observed the low efficacy of current AD treatments. We then found that the presence of an informal caregiver reduced the risks of receiving a late diagnosis, regardless of the informal caregiver concerned (spouse, son, daughter, cousin or friend). Previous research underlined the impact of health risk perception on diagnosis decisions (Cohen et al., 2008; Etner and Jeleva, 2013), and provided evidence of the central role of informal caregivers on the perception of the disease risks in dementia (Hansen et al., 2008). Following these theoretical and empirical evidences, our interpretation is that siblings involved in informal caregiving increase the chances of an early diagnosis by influencing patients' risk perception. This result confirms previous empirical evidence showing the importance of caregivers networks in diagnosis decisions (Carpentier et al., 2010). Indeed, a recent French survey showing that more than 85% of people declare that they would encourage their relative (spouse or parent) getting diagnosed (TNS-SOFRES, 2013). This result underlines the importance of informal care networks at the early phases of AD management. 5.2. Additional factors influencing diagnosis seeking We found four additional results that must be interpreted. First, we provided evidence that differences in subjective QoL estimations were associated with greater risks of a delayed diagnosis,
underlining the need to inform patients and informal caregivers about AD symptoms. In France, a priority was given to inform and educate the public to AD warning signs to improve early diagnosis. Specifically, the 9th objective of the 2008e2012 Alzheimer Plan (Plan-Alzheimer, 2012) was focused on that matter trough the implementation of a dedicated hotline, a dedicated website, the implementation of regional conferences, and research program focusing on the public's perception of the disease. The AD diagnosis announcement is strictly regulated and monitored (HAS, 2011). It has to be provided by a specialist in a dedicated place allowing intimacy and confidentiality. Moreover, the diagnosis can be progressively provided to the patient, following a set of several meetings with the specialist in charge of the diagnosis. Our findings might be used to enhance diagnostic practice. Following previous research (Hansen et al., 2008), we recommend the development of education programs for general practitioners dealing with the benefits of a formal diagnosis of dementia and the need to improve patients' awareness. Second, demographic variables were strongly associated with diagnosis delays, confirming previous evidence (Andersson and Lundborg, 2007; Slovic, 2000). The difference between men and women confirmed a result that has been observed in other diseases prevention campaigns. For instance, previous research has documented that gender was a determinant of late diagnosis in HIV (Kivela et al., 2010). This could be explained by overconfidence rates within that population: previous results obtained in behavioral finance showed that for some decisions men are more subject to overconfidence than women (Barber and Odean, 2001; Bernasek and Shwiff, 2001). According to these articles, gender would explain specific attitudes towards risks and specifically that men are more likely to adopt risky financial investment behaviors. Our interpretation is that similar effects may also be relevant for diagnosis decisions. The effect of patients' age confirmed the presence of a potential bias against the elderly people's value of life, which was described in previous AD research (Fillit et al., 2010). It can also be explained by the fact that in France, after 70 years old, only 9% of people are willing to get an AD diagnosis (TNS-SOFRES, 2013). Third, the economic variables had a reduced impact, while previous evidence showed that diagnosis seeking decisions are likely to be associated with economic variables such as coverage (David et al., 2012). The absence of effect related to income and education and the small statistical effect of job categories can be
T. Rapp / Social Science & Medicine 118 (2014) 9e16
explained by the French health expenditures coverage system. In France, the decision to postpone AD diagnosis should not be explained by budgetary constraint reasons, as AD medical expenditures are fully reimbursed by the National Health Insurance System. Moreover, public financial support is provided to patients with functional and cognitive declines to finance professional and informal care use. Fourth, our results underlined that the presence of surgery and depression diagnosis were associated with diagnosis delays. The effect of these two variables shows that care producers were unable to determine the patients' propensity of having AD, raising the key role of the physician in the diagnosis decision. Further research will have to explore the behavior of care producers facing patients at risks of AD, and specifically their propensity to provide a diagnosis and to treat patients. From a theoretical perspective, this case study is interesting because in AD the medical decision-makers simultaneously face both diagnostic ambiguity and therapeutic ambiguity. The overall effect of diagnostic ambiguity and therapeutic ambiguity on treatment rates is difficult to predict. In the presence of ambiguity aversion, ambiguity regarding the diagnosis of a patient will increase the physician's propensity to provide a treatment for to the patient, while treatment ambiguity will reduce the probability of treatment (Berger et al., 2013). In our sample, 80% of the patients started to use anti-Alzheimer therapies the same year as the diagnosis date, which was high given the small clinical effect of these drugs. These large treatment rates at the time of diagnosis could be explained by the presence of diagnosis ambiguity and ambiguity aversion among physicians in our sample, even if using the year of treatment and the year of diagnosis did not allow for accurate explorations. 5.3. Limitations Our study was faced with some limitations. Cancer data analyses provide evidence that the presence of private insurance has a significant role to explain differences in information seeking behaviors (Walsh et al., 2012). We were not able to control for that parameter in our analyses. However, the French national insurance system reduces the likelihood of such a mechanism in our sample. The generalizability of our results also is of concern. The PLASA study was designed to test the impact of an innovative intervention in AD. It is well known that clinical trials participants usually are more likely to be care seekers. In consequence, the patients that participated in the PLASA study could have some specific (unobserved) characteristics that could impact their history of disease and the outcome that was explored in our paper. The absence of precise records (day/month/year) for the date of first signs and diagnosis date did not allow us running parametric survival analyses, which could have increased the accuracy of our results. However, the use of ordered logistic regression was relevant given the distribution of our outcome, and it was confirmed by the non-significant Brant tests. Finally, our sample excluded 136 patients with missing information for one or more variables. We found that these excluded patients had lower late diagnosis rates and had higher rates of informal caregiving use. The exclusion of this subsample may have leaded us to underestimate the effect of the informal care variable. However, our sensitivity analyses (analyses not reported in the paper but available upon request) confirmed the robustness of our results. First, we used a sample selection model to explore whether there was a correlation between the unobserved factors associated with the probability of having non-missing values and the unobserved factors associated with the probability of receiving an early diagnosis. We controlled in the first stage for a dichotomous variable indicating the presence of missing values for the job variable.
15
We did not find the presence of a significant correlation (Chi2 ¼ 0.32, p ¼ 0.482) and therefore concluded that the potential selection bias was reduced. Second, we found similar results when running a more restricted Model that did not control for income but used more observations of 1005 patients. 5.4. Health policy implications Our results have central health policy implications. Being able to improve early detection in AD is expected to have important economic benefits. There is evidence that the cost of AD dramatically increases with the severity of cognitive, behavioral and functional symptoms, and that being able to manage the disease when the patient has moderate symptoms can lead to large savings (Rapp et al., 2012). Moreover, there is evidence that ambiguity about health status has an impact on quality of life, which can be stronger than physical problems (Graham et al., 2011). Knowing the most important determinants of late diagnoses in AD is therefore crucial for early detection campaigns. We identify several opportunities for early detection campaigns. Families with history of disease should be targeted, as we confirmed the presence of avoidance behaviors among patients with disease awareness. Campaigns should also target isolated patients, who do not benefit from informal care. We recommend that public policies encourage the organization of informal care networks when the patients experience the first warning signs. Our results underlined the importance of improving diagnosis for old patients and for men. We found that early detection can be blurred by clinical events such as surgery or depression diagnosis, revealing that care suppliers' behaviors can also explain delayed diagnoses. Specific campaigns should also be designed to help care producers identifying AD symptoms. Acknowledgments s d’e te This paper was presented at Sciences Po Lille (Universite Alzheimer), and at the Ethics, Economics and Health Seminar of APHP. I wish to thank Bruno Vellas, Nicolas Sirven and Lise Rochaix, and the three anonymous reviewers for providing helpful comments on the manuscript. All errors remain mine. References Andersson, H., Lundborg, P., 2007. Perception of own death risk. An analysis of roadtraffic and overall mortality risks. J. Risk Uncertain. 34, 67e84. Barber, B., Odean, T., 2001. Boys will be boys: gender, overconfidence, and common stock investment. Q. J. Econ. 116, 261e292. Berger, L., Bleichrodt, H., Eeckhoudt, L., 2013. Treatment decisions under ambiguity. J. Health Econ. 32, 559e569. Bernasek, A., Shwiff, S., 2001. Gender, risk, and retirement. J. Econ. Issues 35, 345e356. Brant, R., 1990. Assessing proportionality in the proportional odds model for ordered logistic regression. Biometrics 46, 1171e1178. Breslow, N.E., 1970. A generalized Kruskal-Wallis test for comparing K samples subject to unequal patterns of censorship. Biometrika 57, 579e594. Carpentier, N., Bernard, P., Grenier, A., Guberman, N., 2010. Using the life course perspective to study the entry into the illness trajectory: the perspective of caregivers of people with Alzheimer's disease. Soc. Sci. Med. 70, 1501e1508. Clare, L., 2003. Managing threats to self: awareness in early stage Alzheimer's disease. Soc. Sci. Med. 57, 1017e1029. Clare, L., Rowlands, J., Bruce, E., Surr, C., Downs, M., 2008. ‘I don't do like I used to do’: a grounded theory approach to conceptualising awareness in people with moderate to severe dementia living in long-term care. Soc. Sci. Med. 66, 2366e2377. Cohen, M., Etner, J., Jeleva, M., 2008. Dynamic decision making when risk perception depends on past experience. Theory Decis. 64, 173e192. Dartigues, J.F., 2011. Alzheimer's disease: early diagnosis. Rev. Prat. 61, 926e930. David, G., Saynisch, P., Acevedo-Perez, V., Neuman, M.D., 2012. Affording to wait: medicare initiation and the use of health care. Health Econ. 21, 1030e1036. Eeckhoudt, L.R., Lebrun, T.C., Sailly, J.C., 1984. The informative content of diagnostic tests: an economic analysis. Soc. Sci. Med. 18, 873e880. Ellsberg, D., 1961. Risk, ambiguity, and the Savage axioms. Q. J. Econ. 75, 643e669.
16
T. Rapp / Social Science & Medicine 118 (2014) 9e16
Etner, J., Jeleva, M., 2013. Risk perception, prevention and diagnostic tests. Health Econ. 22, 144e156. Fillit, H., Cummings, J., Neumann, P., McLaughlin, T., Salavtore, P., Leibman, C., 2010. Novel approaches to incorporating pharmacoeconomic studies into phase III clinical trials for Alzheimer's disease. J. Nutr. Health Aging, 640e647. Ganz, M.L., 2001. Family health effects: complements or substitutes. Health Econ. 10, 699e714. Gehan, E.A., 1965. A generalized Wilcoxon test for comparing arbitrarily singly censored samples. Biometrika 52, 203e223. Graham, C., Higuera, L., Lora, E., 2011. Which health conditions cause the most unhappiness? Health Econ. 20, 1431e1447. Hansen, E.C., Hughes, C., Routley, G., Robinson, A.L., 2008. General practitioners' experiences and understandings of diagnosing dementia: factors impacting on early diagnosis. Soc. Sci. Med. 67, 1776e1783. HAS, 2011. Diagnostic et prise en charge de la maladie d'Alzheimer et des maladies es: recommandation 71. Haute Autorite de Sante , Saint-Denis. apparente Hutchinson, S.A., Leger-Krall, S., Wilson, H.S., 1997. Early probable Alzheimer's disease and awareness context theory. Soc. Sci. Med. 45, 1399e1409. Kaplan, E.L., Meier, P., 1958. Nonparametric estimation for incomplete observations. J. Am. Stat. Assoc. 53, 457e481. Kivela, P.S., Krol, A., Salminen, M.O., Ristola, M.A., 2010. Determinants of late HIV diagnosis among different transmission groups in Finland from 1985 to 2005. HIV Med. 11, 360e367. Koszegi, B., 2003. Health anxiety and patient behavior. J. Health Econ. 22, 1073e1084. Lin, P.J., Cangelosi, M.J., Lee, D.W., Neumann, P.J., 2013. Willingness to pay for diagnostic technologies: a review of the contingent valuation literature. Value Health 16, 797e805. Logsdon, R.G., Gibbons, L.E., McCurry, S.M., Teri, L., 2002. Assessing quality of life in older adults with cognitive impairment. Psychosom. Med. 64, 510e519. Low, L.F., Anstey, K.J., 2007. The public's perception of the plausibility of dementia risk factors is not influenced by scientific evidence. Dement. Geriatr. Cogn. Disord. 23, 202e206. McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., Stadlan, E.M., 1984. Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work
Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology 34, 939e944. Meechan, G., Collins, J., Petrie, K., 2002. Delay in seeking medical care for selfdetected breast symptoms in New Zealand women. N. Z. Med. J. 115, U257. Neumann, P.J., Cohen, J.T., Hammitt, J.K., Concannon, T.W., Auerbach, H.R., Fang, C., et al., 2012. Willingness-to-pay for predictive tests with no immediate treatment implications: a survey of US residents. Health Econ. 21, 238e251. Neumann, P.J., Hammitt, J.K., Mueller, C., Fillit, H.M., Hill, J., Tetteh, N.A., et al., 2001. Public attitudes about genetic testing for Alzheimer's disease. Health Aff. (Millwood) 20, 252e264. Nourhashemi, F., Andrieu, S., Gillette-Guyonnet, S., Giraudeau, B., Cantet, C., Coley, N., et al., 2010. Effectiveness of a specific care plan in patients with Alzheimer's disease: cluster randomised trial (PLASA study). BMJ 340, c2466. Nourhashemi, F., Gillette-Guyonnet, S., Andrieu, S., Rolland, Y., Ousset, P.J., Vellas, B., 2008. A randomized trial of the impact of a specific care plan in 1120 Alzheimer's patients (PLASA Study) over a two-year period: design and baseline data. J. Nutr. Health Aging 12, 263e271. Plan-Alzheimer, 2012. Plan Alzheimer 2008e2012. Rapp, T., Andrieu, S., Molinier, L., Grand, A., Cantet, C., Mullins, C.D., et al., 2012. Exploring the relationship between Alzheimer's disease severity and longitudinal costs. Value Health 15, 412e419. Slovic, P., 2000. The Perception of Risk. Earthscan Publications Ltd, London and Sterling. l'anticipation de la maladie d'Alzheimer. In: T. TNS-SOFRES, 2013. Les français face a SOFRES (Ed.). TNS SOFRES, Paris. Viscusi, W., Magat, W., Huber, J., 1991. Communication of ambiguous risk information. Theory Decis. 31, 159e173. Walsh, B., Silles, M., O'Neill, C., 2012. The role of private medical insurance in socioeconomic inequalities in cancer screening uptake in Ireland. Health Econ. 21, 1250e1256. Wimo, A., von Strauss, E., Nordberg, G., Sassi, F., Johansson, L., 2002. Time spent on informal and formal care giving for persons with dementia in Sweden. Health Policy 61, 255e268.