Abstract
Creutzfeldt-Jakob disease (CJD) is a kind of rare, rapidly progressive fatal central nervous system disorders. In China, the surveillance for CJD has started since 2006. As one of the major issues in CJD surveillance, the follow-up process via telephone plays important role in CJD diagnosis and surveillance. Although the follow-up process was conducted by the experiential staffs from CJD surveillance center in China CDC, it is frequently encountered that some interviewed family members do not cooperate well during follow-up. To screen the possible factors influence on the compliances of the interviewees during CJD follow-up, 11 independent variables from patient aspect and 4 variables from interviewee aspect were selected and a questionnaire was prepared. Based on 199 suspected sporadic CJD cases reported to CJD surveillance center in 2013, a telephone-inquiring was conducted and the degree of compliances of the interviewees were given as good, fair or poor. After screened with univariate analysis and evaluated ordinal logistic regression analysis, several indictors, such as the patient gender, CJD diagnosis, numbers of clinical symptoms, continual medical treatment after diagnosis, medical treatment mode, as well as the relationship with the patient and CJD knowledge of the interviewees, showed influence on the compliance in CJD follow-up process significantly. The data here provide for the first time the factors related with the compliances of the interviewed family members of the suspected CJD patients during follow-up process, which supplies useful clue for us to improve CJD follow-up process and increase the capacity of CJD surveillance.
Introduction
Human prion diseases, also termed as human transmissible spongiform encephalopathies (TSEs), are a group of transmissible neurodegenerative diseases including Creutzfeldt-Jakob disease (CJD), Kuru, fatal familial insomnia (FFI) and Gerstmann-Straüssler-Scheinker (GSS) syndrome.Citation1,2 Human prion diseases may occur sporadically, genetically or acquiredly.Citation2 Despite of the presence of various subtypes, human prion diseases usually share a number of common neuropathological characteristics, such as spongiform degeneration, extensive neuronal loss, gliosis, and completely uniquely, the accumulation of PrPSc, a misfolded and protease-resistant form of the normal prion protein (PrPC).Citation1
CJD, a kind of rare disease, was first recognized in the 1920s.Citation3 As the dominant subtype in human prion diseases, CJD occurs world-wide with an incidence around 1 and 2 cases per million of population per annum, among them, the majority (roughly 85%) of these cases are idiopathic and termed as sporadic CJD (sCJD).Citation3 And about 5–15% of CJD cases are familial or genetic CJD (fCJD or gCJD) that are directly associated with a list of mutations (including point mutations, insertions and deletions) in the open reading frame of the prion protein gene, PRNP. Less than 1% CJD cases are acquired, which usually have definite medical histories and termed as iatrogenic CJD. Since the outbreak of bovine spongiform encephalopathy (BSE) in the 1980's, a new subtype of CJD, variant CJD (vCJD) had emerged, which is linkd with consumption of contaminated beef by BSE.
The clinical manifestations of CJD are multifarious. Besides of rapid progressive dementia, myoclonus, visual or cerebella disturbance, pyramidal or extrapyramidal disfunction and akinetic mutism are also included in the diagnostic criteria for CJD issued by WHO. The clinical course of CJD is usually shorter than 2 years, most of who die within one year after onset. Clinically, those symptoms or signs may not appear simultaneously at the moment of visiting neurologist. Meanwhile, those clinical manifestations are not CJD-specific, which can also be observed in many other neurological diseases. Although there are several reliable clinical examinations and laboratory tests, the definite diagnosis of CJD still lies on the relevant tests of brain tissues. Therefore, clinical follow-up is one of the routine issues for the diagnosis and the surveillance of CJD.
China national CJD surveillance has started since 2006.Citation4,5 Thousands of suspected cases have been reported to the surveillance center in China CDC. Except for dozens of various fCJD, FFI and GSS that were diagnosed with genetic evidences, 8 hundred were probably and possibly diagnosed as sCJD roughly. The percentage of the definitely diagnosed as sCJD cases in China was extremely low due to the low rate of brain postmortem in Chinese. Thus, implementation of follow-up process does not only benefit the patients and their members, but also enhance the capacity for CJD diagnosis and surveillance.
In this study, we conducted a telephone-inquiring during the CJD follow-up process, in order to screen the potential factors that may affect the compliances of the interviewed family members. All reported cases in CJD surveillance network from January to September 2013 were enrolled. As the diagnostic criteria of human genetic prion diseases differs from that of sCJD, 2 cases of E200K gCJD and 6 FFI cases were not included. Based on the designed questionnaire covering dozens of independent variables from both aspects of the patient and the interviewee, the degree of compliance of each interviewed family member was given. After evaluated with univariate analysis and ordinal logistic regression analysis, several indictors, such as the patient gender, CJD diagnosis, numbers of clinical symptoms, continual medical treatment after diagnosis, as well as the relationship with the patient and CJD knowledge of the interviewees, were selected as the significant influencing factors for the compliance in CJD follow-up process.
Results
General information of the suspected CJD patients and the interviewed family members
Totally 199 suspected CJD cases reported to China CJD surveillance system in 2013 were enrolled in this study. The general information of those patients was summarized in . The male patients were slightly more than the female (M/F: 107/92). Most patients (124 cases) were at the group of 50–70 year-old. About half of the cases (107 cases) lived in urban areas. About 56 cases accepted the follow-up interview less than 2 months after onset, 77 less than 6 months and 66 more than 6 months. All patients were described to have dementia. Besides of dementia, 6 patients appeared 4 other major clinical symptoms, 42 appeared 3, 66 appeared 2, 32 appeared one and 53 appeared none. And 192 out of 199 cases had EEG examinations, 26 patients showed typical changes. MRI scanning revealed typical abnormalities in 78 cases out of 199 patients. Besides, 195 cases had CSF 14–3–3 tests, 63 were 14–3–3 positive. Most of the patients (192 cases) received the routine treatments of Western medicine, while only 7 cases took traditional Chinese medicine. About 167 cases stayed at hospitals or transferred to other small hospitals for medical service and 32 cases discharged from hospitals and stayed at home after diagnoses. And 158 cases received continually medical treatment after diagnosis and 41 cases almost stopped medical treatment except daily care. For medical reimbursement modes, 25 cases had socialized medicine, 124 cases had various medical insurances, and 50 cases were self-paid.
Table 1 Summarized basic information of the patients reported as suspected CJD cases and the interviewed family members as interviewees based on the diagnosis
The registered family members of 199 suspected CJD patients were interviewed via telephone-inquiring. Most of them (170 cases) were in the group of 30 to 60 years old. Approximately 118 interviewees were children of the reported patients, 58 were spouses, 8 were parents and 15 were others (brothers/sisters, cousins or nephews). About 72 individuals lived in countryside, and the rests worked in the cities with various professions, among them 11 were medical staff working in different medical institutions. The assessment of the knowledge of CJD among the interviewees revealed that 108 interviewees were unknown about CJD, 44 knew the knowledge of CJD limitedly and 47 were familiar with CJD.
The degrees of compliances of the interviewees based on the selected variables
The degree of compliance (good, fair and poor) of each interviewee was given by the staff in China CJD surveillance center after telephone-inquiring. Generally, most interviewees were willing to cooperate with the staffs during follow-up survey, showing good and fair cooperation when communicated with the staffs in telephone. The percentages of good, fair and poor compliances were 33.67% (67/199), 52.26% (104/199) and 14.07% (28/199), respectively. The evaluating data of the interviewees and distributions based on each variable were summarized in . Regarding to the good compliance, more interviewed family members from the female patients (41.3 0% vs 27.10%) and the patients with medical background (62.50% vs 32.46%) cooperated well in follow-up survey. More portions of the interviewees were in the groups of good compliance from the patients with the diagnosis of pending (63.16%) and probable sCJD (45.33%), with more numbers of clinical manifestations, with abnormal results in EEG and MRI and positive CSF 14–3–3. Patients treated with traditional Chinese medicine, reimbursed with socialized medicine and medical insurance, and received continual medical treatment after diagnosis also showed positive influences on their family members to cooperate with the staffs during telephone-inquiring. On the other side, the interviewees who had medical background (90.91%), who were parents (75.00%) of the patients, and who were familiar with CJD (63.83%) showed good compliance with the staffs.
Table 2 Summarized evaluating data of the degree of compliance of the interviewees by telephone-inquiring based on the variables
Significant indicators in univariate analysis
After transferred into SPSS file, the data were subjected into cross-tabulations and the statistical significance of the relationship between individual independent variables and outcome variable was examined by chi-squared test. No significant effect on the outcome was addressed for the following factors from the aspect of the patients, such as sex (p = 0.062), age at onset (p = 0.658), permanent residence (p = 0.227), follow-up time from the onset (p = 0.656), numbers of appearances of major clinical symptoms (p = 0.095), typical changes in EEG (p = 0.098), medical treatment (p = 0.162), medical service after diagnosis (p = 0.930), as well as factors from the aspect of the interviewees, such as age (p = 0.522) and relationship with the patient (p = 0.141). However, profession of the patient (p = 0.047), CJD diagnosis (p = 0.000), typical changes in MRI (p = 0.001), positive CSF 14–3–3 (p = 0.001), medical reimbursement mode (p = 0.033), continue medical treatment after diagnosis (p = 0.000), as well as the profession of the responder (p = 0.001) and the knowledge of CJD (p = 0.000) had significant effect on the degree of compliance of the interviewed family members().
Table 3 Analyses of the potential factors in the compliance degree of the interviewees by chi-square statistics
Significant indicators in ordinal logistic regression analysis
All above variables with P-value less than 0.2 were put into ordinal logistic regression analysis. The first round of assay showed that some variables with significance in the univariate analyses did not reveal any statistical significance in the ordinal regression model, including medical associated profession of the patient, MRI, positive CSF 14–3–3, medical reimbursement mode, as well as professions of interviewees. Furthermore, all variables with P-value larger than 0.1 were excluded from the model and the rest ones were put into the second round of ordinal regression assay. Totally, 7 variables illustrated significant association with the degree of compliance of the interviewed family members, including gender, CJD diagnosis, numbers of appearances of major clinical symptoms besides of dementia, continue medical treatment after diagnosis and medical treatment from the aspect of the patients, as well as relation with the patient and knowledge of CJD from the aspect of the interviewees ().
Table 4 Analyses of the relational influence factors in the compliance degree of the interviewers by ordinal logistic regression model
Among the indicators of the patients, gender appeared to be significant on the degree of compliance, that the interviewed family members of the female patients cooperated better during inquiring than those of male ones (b = −0.823). CJD diagnosis became another significant indicator. Compared with those of the patients of probable sCJD (b = 0.000), the interviewees of the group of pending (b = 1.538) showed better compliance, whereas those of the groups of non CJD (b = −0.256) and possible sCJD (b = −1.744) showed relatively worse compliance. The patients’ clinical manifestations were considered to be associated with the degree of compliance as well. The inquired family members of the patients with more clinical symptoms cooperated more actively with the investigations than those of the patient with less clinical symptoms. Additionally, whether receiving continual medical treatment after diagnosis influenced the interviewees compliances significantly, in which the family members of the patients who gave up treatment after diagnosis (b = −1.007) were more likely to be uncoordinated in the follow-up process.
Among the factors of the interviewees, the relationships of the interviewees with the patients were significance on the degree of compliance. Parents (b = 3.894) and spouse (b = 0.834) showed better compliance than children (b = 0.000) and others (b = −0.111). Besides, CJD knowledge of the interviewees was another significant indicator, that the individuals who received higher scores in the test of CJD knowledge were easier to be cooperated during inquiring than those who got fair (b = −1.664) or poor (b = −1.893) score.
Discussion
Since we conducted the national surveillance for CJD in 12 provinces in mainland of China from 2006, the patient follow-up has become one of main issues in the national CJD surveillance center. Due to relative rapid progression of disease, the family members of the suspected CJD patients are usually the interviewees during follow-up. In this study, we have performed for the first time a telephone-inquiring among the suspected CJD patients reported to China CJD surveillance center in 2013, aiming to address the potential factors affecting the compliance of the interviewed family members in the follow-up process. Eleven independent variables from patient aspect and 4 variables from interviewee aspect are selected. After ordinal logistic regression analysis, several significant indicators have been identified.
We have found that the situations of CJD diagnosis of the patients at the moment of follow-up are significant indicators to influence on the compliances of the interviewees. The interviewed family members of the patients with temporally uncertain diagnosis (pending) seem to be easier to be communicated, reflecting that they want to have more advisements or suggestions from the doctors at this very time. Good compliances are also more frequently observed in the family members of the patients with the diagnosis of probable sCJD than those of possible sCJD and non-CJD. In fact, we have also noticed that the family members of the patients with probable sCJD are more likely to ask the questions about daily patient care initiatively during telephone-inquiring. Possibly linked with the variable of CJD diagnosis, the factor of the numbers of CJD associated clinical manifestations is also a significant indicator for the compliances of the interviewees, as probable sCJD cases usually have more numbers of the 4 major symptoms/signs besides of dementia than possible sCJD.Citation4,5 CJD typical changes in MRI and positive CSF 14–3–3 show statistical significance in the univariate analysis, but not in the following ordinal logistic regression analysis. Since those 2 factors are directly associated with CJD diagnosis, we believe that the contributions of those factors on the compliances of the interviewees may mix with the variable of CJD diagnosis in logistic regression analysis. Despite that typical EEG changes are also important in diagnosis of CJD as MRI and CSF 14–3–3 are, our study here does not figure out statistical difference in the analyses of the potential factors in the compliance degree of the interviewees by chi-square statistics. The exact reason for the difference among EEG, MRI and CSF 14–3–3 in chi-square statistics is unknown. However, the numbers of the patients with typical EEG changes (26/199) in this study is markedly less than those with typical changes in MRI (78/199) and positive CSF 14–3–3 (63/199), which might influence the subsequent statistical assays. Surprisingly, the interviewees from the families of female patients appear to be more cooperative in this telephone-inquiring. The reason for such gender difference is not clear. However, reviewing of the CJD diagnosis of the enrolled cases illustrate more male patients in the group of non-CJD and more female patients in the group of probable sCJD. This observation let us speculate that the gender difference on the compliances of the interviewees may also be closely linked with the variable of CJD diagnosis.
During telephone-inquiring, 41 reported patients are described to almost stop further medical service except for daily care after diagnosis, especially after discharged from central hospitals. The reasons for discontinuing medical treatment are various, but the ratios of the family members who are poorly cooperated during follow-up in this group are remarkably higher than those in the group of continual medical treatment, being a significant factor both in the univariate and logistic regression analysis. This anticipatable result may truly reflect the hopeless moods of the family members at this stage. Although majority (81.40%) of the suspected CJD patients are being hospitalized during telephone-inquiring, the compliances of the family members in this group are quite comparable with those stayed at home. Seven patients have received TCM as the main medical treatment and their family members appear better compliance in the follow-up process. However, 6 of them are probable sCJD cases. Therefore, we believe that the good compliances of the interviewees in the group of TCM are due to the indicator of CJD diagnosis.
Two indicators in the aspect of the family members significantly affect the degree of compliance during follow-up process. One is the relation with the patient. Parents and spouses are usually more cooperative than children and other relatives in our study. The other indicator is CJD knowledge of the interviewed family members. Our study confirms that the individuals who are familiar with CJD tend to cooperate with our staffs better than those who are unknown. Such profile has been observed in the follow-up survey for some other diseases.Citation6 Additionally, the interviewees with medical background themselves or the interviewees from the patients with medical background show higher portion of good compliance in the telephone-inquiring. It indicates that the knowledge of prion disease or relative medicine is positive for the interviewees in further CJD follow-up process. As a rare neurological disease, ordinary Chinese people, even physicians working in small clinics, usually know little about CJD. On the other hand, inappropriate descriptions of BSE from some media increase people's panic and fear. We have frequently noticed that people, including some medical staffs, have believed that CJD is a kind of contagious infectious disease. Such misconception may disturb the communication between family members and medical staff. Our data here highlight again the importance of education of CJD knowledge in increasing the capacity of CJD surveillance.
Like other developing countries worldwide, the economic and living conditions of city and countryside vary largely, which may also be a candidate to influence the compliances of the family members during follow-up. Although the portions of good compliances in the groups of the interviewees of the patients living in countryside and the interviewees who are farmers are lower, ordinal logistic regression analysis does not identify statistical significance, highlighting the contribution of this variable is limited. Despite that the coverage of medical insurance in the mainland of China has improved dramatically in the past 20 years, it still does not reach universal coverage. Currently, there are 3 major medical reimbursement modes in China. One is the so-called socialized medicine that mainly covers the state personnel working in the government or state-owned entities. Another is universal medical insurance that comprises the population who work or live in the city region or in some countryside. The third mode is self-paid or having limited reimbursement source, usually involving in peasants in countryside and city populaces without fixed jobs. The economic status and payment capacity of the population in the group of self-paid are relatively poor. However, no significance of this variable has been addressed on the compliances of the interviewees after analyzed with ordinal logistic regression, even in the analysis when mixing the data of socialized medicine and medical insurance together as non-self paid. It might, at least partially, indicate that the economic situations of the patients and their families do not significantly influence on the compliances of the interviewees during CJD follow-up.
To sum up, in this study we have screened the potential indicators that may affect the compliances of the family members in follow-up process based on the suspected CJD cases reported to China CJD surveillance network in 2013. We have found that the CJD diagnosis of the patients, whether the patients receive continually treatment after diagnosis, the relationship of the interviewees with the patients and the CJD knowledge of the interviewees are markedly significant. Certainly, the main conclusions of this study derive from the surveillance data of less than one year, with relatively limited case numbers. The selection of independent variables, the design of questionnaire and the assessment of the degree of the compliances for the interviewees need to be further improved and perfected in the future practices. Nevertheless, the outcome of this study will help us to improve our CJD follow-up process and increase the capacity for CJD surveillance.
Methods
Ethics statement
This surveillance program has been approved by Ethical Review Committee of National Institute for Viral Disease Prevention and Control, China CDC. All signed informed consents have been collected and stored by the China CJD Surveillance Center.
Study population
All referrals reported to China CJD surveillance center as suspected CJD from January to September 2013 were enrolled into this study. The diagnosis and subtype of the reported cases were conducted according to the CJD diagnostic criteria issued by China CDC, which was constituted based on the diagnostic criteria for CJD issued by WHO.Citation7 After exclusion of 8 cases of genetic prion diseases (2 cases of E200K gCJD and 6 cases of FFI), totally 211 cases were collected. Twelve cases were excluded during follow-up process, because the phone number was unable to connect or the registered contactor (usually being family member) did not answer the phone more than 3 times. Among 199 cases, 88 were diagnosed as CJD cases, including 75 probable sCJD and 13 possible sCJD. Around 92 patients were excluded the possibility of CJD after expert consultation, with or without further diagnosis at the moment of telephone inquiring. Those 92 cases were referred to as non-CJD. The rest 19 cases were grouped as pending, who were likely to be CJD cases from the clinical examinations or laboratory tests, but still did not fulfill the diagnostic criteria from clinical signs at the moment of inquiring.
Determination of independent variables
The independent variables in this study consisted of the information from both patients and interviewees (family members). The information from patient aspect covered the demographic factors (sex, age at onset and profession), the diagnosis of CJD, the time of follow-up from onset, the numbers of appearances of major clinical symptoms besides of dementia, the results of clinical examinations and laboratory tests (electroencephalograph (EEG), Magnetic Resonance Imaging (MRI), cerebrospinal fluid (CSF) 14–3–3), the patterns of medical treatment and medical service after diagnosis, the modes of medical reimbursement, continue medical treatment after diagnosis(yes or not). The information from interviewee aspect composed the demographic factors (age and profession), the relationship with the patient, and the knowledge of CJD. The CJD knowledge of the interviewees was tested by 4 questions, including etiology, transmission and genetic CJD (Supplementary ). Each question was given 1 score. Therefore, the score was between 0 and 4. The degree of CJD knowledge was categorized into 3 groups, familiar (score higher than 2), known limitedly (score of 2) and unknown (score of less than 2).
Design of the follow-up questionnaire
Based on the selected independent variables, we developed the follow-up questionnaire for the interviewees (family members) of the patients reported as suspected CJD in China national CJD surveillance network (Supplemental ). The questionnaire consisted of the information from both patients and interviewees. In the part of patients, sex (male or female), age at onset (<50, 50–70 or >70 y), permanent residence (urban or rural) and medical associated profession (yes or no) were collected as the demographic factors; time from onset (<2, 2–6 or >6 m) and numbers of appearances of major clinical symptoms besides of dementia (4, 3, 2, 1 or 0) as the clinical factors; typical changes in EEG and MRI, CSF-14–3–3 positive (yes, no or not available) as the examination data; non-CJD, possible sCJD, probable sCJD, and pending as the diagnosis information; medical treatment (Western medicine or traditional Chinese medicine), medical service (hospitalization or at home), modes of medical reimbursement (socialized medicine, medical insurance or self-paid) and continue medical treatment after diagnosis (yes or no) as the medical service factors. In the part of interviewees, 4 major information were collected, including age (<30, 30–60 or >60 y), profession (farmer, officer clerk, worker, medical staff or other), relationship with the patient (spouse, parent, child or other), knowledge of CJD (unknown, known limitedly or familiar). The compliances of the interviewees during telephone-inquiring were degreed as good, fair and poor.
Telephone-inquiring and determination of the degree of compliance
A telephone-based follow-up survey was conducted by the staff of China CJD program with the designed questionnaires. Each telephone inquiring was supervised by another senior epidemiologist and final score was given after discussion. The compliance degree of the interviewee was categorized into 3 degrees, poor: (reject to answer the phone more than one time, or reject to answer the question during follow-up, or be indifferent to the outcome of the tests and diagnosis proposed by CJD surveillance center), fair (reject to answer the phone only one time, or answer the questions presented by the staff passively, or accept the diagnosis given by CDC only with no other requirements), good (be willing to answer the phone, or pay attention to the diagnosis and the outcome of the tests proposed by CJD surveillance center, or ask for recommendation on his or her own initiative).
Statistical approaches
Firstly, the association between outcome variable and independent variables was examined by cross-tabulations and the statistical significance of such relationship was examined by chi-square test. A 2-tailed P value of less than 0.05 was considered to be statistically significant. Considering that compliance degree of the interviewee was an ordinal categorical variable, ordinal logistic regression analysis was conducted.Citation8-11 Variables with a P value less than 0.2 in the univariate analysis were included in the ordinal logistic regression analysis, since usage of the variables with P value less than 0.05 may miss some meaningful variables.Citation12-14 After first round of ordinal logistic regression analysis, the variables with P value larger than 0.1 were excluded and the rests were re-entered into the ordinal model in order to identify significant indicators in the presence of the other variables. The variable with P value of less than 0.05 was considered to be statistically significant. Data were inputted by Epidata software. Statistical analyses were performed using SPSS20 (IBM, USA). Overall, the processes of telephone-inquiring and statistical analyses were shown in .
Figure 1. Flow chart of variable screening.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
KPRN_A_983747_Supplemental_Table_1.doc
Download MS Word (51 KB)Acknowledgment
We greatly appreciate all interviewed family members of the suspected CJD patients for their kind cooperation in the telephone-inquiring. We are also indebted by Drs. HY Yao and Q Xiao from the Office of Epidemiology, China CDC, for their helpful discussions in statistical analysis.
Funding
This work was supported by Chinese National Natural Science Foundation Grants (81301429), China Mega-Project for Infectious Disease (2011ZX10004–101, 2012ZX10004215), and SKLID Development Grant (2012SKLID102, 2011SKLID211).
Refernces
- Prusiner SB. Prions. Proc Natl Acad Sci U S A 1998; 95:13363-83; PMID:9811807; http://dx.doi.org/10.1073/pnas.95.23.13363
- Collinge J. Prion diseases of humans and animals: their causes and molecular basis. Annu Rev Neurosci 2001; 24:519-50; PMID:11283320; http://dx.doi.org/10.1146/annurev.neuro.24.1.519
- Aguzzi AaK V, ed. Introduction to Prion Disorders, in Neurodegeneration: The Molecular Pathology of Dementia and Movement Disorders, Second Edition. Oxford, UK: Wiley-Blackwell, 2011
- Shi Q, Gao C, Zhou W, Zhang BY, Chen JM, Tian C, Jiang HY, Han J, Xiang NJ, Wang XF, et al. Surveillance for Creutzfeldt-Jakob disease in China from 2006 to 2007. BMC Public Health 2008; 8:360; PMID:18928564; http://dx.doi.org/10.1186/1471-2458-8-360
- Gao C, Shi Q, Tian C, Chen C, Han J, Zhou W, Zhang BY, Jiang HY, Zhang J, Dong XP. The epidemiological, clinical, and laboratory features of sporadic Creutzfeldt-Jakob disease patients in China: surveillance data from 2006 to 2010. PLoS One 2011; 6:e24231; PMID:21904617; http://dx.doi.org/10.1371/journal.pone.0024231
- Madadi M, Zhang S, Yeary KH, Henderson LM. Analyzing factors associated with women's attitudes and behaviors toward screening mammography using design-based logistic regression. Breast Cancer Res Treat 2014; 144:193-204; PMID:24510010; http://dx.doi.org/10.1007/s10549-014-2850-9
- WHO manual for surveillance of human transmissible spongiform encephalopathies including variant Creutzfeldt-Jakob disease. WHO Communicable Disease Surveillance and Response. Geneva: WHO, 2003:71-2
- Brant R. Assessing proportionality in the proportional odds model for ordinal logistic regression. Biometrics 1990; 46:1171-8; PMID:2085632; http://dx.doi.org/10.2307/2532457
- Bath PM, Gray LJ, Collier T, Pocock S, Carpenter J. Can we improve the statistical analysis of stroke trials? Statistical reanalysis of functional outcomes in stroke trials. Stroke 2007; 38:1911-5; PMID:17463316; http://dx.doi.org/10.1161/STROKEAHA.106.474080
- Bath PM, Geeganage C, Gray LJ, Collier T, Pocock S. Use of ordinal outcomes in vascular prevention trials: comparison with binary outcomes in published trials. Stroke 2008; 39:2817-23; PMID:18669897; http://dx.doi.org/10.1161/STROKEAHA.107.509893
- P. McCullagh. Regression models for ordinal data. J Roy Statist Soc B 1980; 42:109-42
- Bursac Z, Gauss CH, Williams DK, Hosmer DW. Purposeful selection of variables in logistic regression. Source Code Biol Med 2008; 3:17; PMID:19087314; http://dx.doi.org/10.1186/1751-0473-3-17
- Mickey RM, Greenland S. The impact of confounder selection criteria on effect estimation. Am J Epidemiol 1989; 129:125-37; PMID:2910056
- Robert B. Bendela AAA. Comparison of stopping rules in forward “stepwise” regression. J Am Stat Assoc 1977; 72:46-53