Study design
We carried out a randomized, open-label, crossover study, with two study periods (sequences AB/BA) and four assessment points. A wash-out period of 30 days was chosen to avoid remembrance of previous answers.
The study was conducted in the same context of the MEMORA cohort that aims to study the relationship between patient characteristics and functional autonomy change over time among patients attending a MC [10].
Participants and setting
Eligibility criteria for participants were patients attending a memory consultation for the first time at the MC for a diagnostic work-up, aged 50 years or older, living at home, accompanied by an informal caregiver, and who agreed to participate in the study. Exclusion criteria were patients whose caregiver did not wish to participate in the study, patients whose caregiver did not provide a telephone contact, and patients for whom the health status would require institutionalization during the period of the study. The study was conducted at a MC of the Memory Research Center of Lyon (France), between November 2014 and April 2016.
IADL questionnaire and modes of administration
The questionnaire used to assess the level of functionnal independence (or dependence) was the French version of the Lawton IADL including the 8 items: ability to use the telephone, to go shopping, food preparation, to do housekeeping, to do personal laundry, to use transportation, to be responsible for taking medications, to handle finances [2]. The questionnaire answers have been considered as a total score ranging from 0 (dependent) to 8 (independent), as well as 2 sub-scores of 4 based on previous research [11 ,12]. The first sub-score includes the items concerning telephone, transportation, medications and finances. The second sub-score includes the other items. In addition, each item of the IADL questionnaire has been scored as binary variables (1: ability to conduct the activity, 0: no ability).
The IADL questionnaire was collected for the same patients using two modes of administration: by telephone and face-to-face interview. Both interviews consisted of a collection of answers given by the caregiver to the nurses trained for this procedure. The questionnaire was identical in both modes of administration, the questions were asked following the same order, and the nurses had to read the questionnaire exactly as it was written to ensure similar conditions for data collection.
Potentially eligible patients were selected from the list of scheduled appointments in the MC. A letter to inform both the patient and caregiver of the possibility to participate in the study was sent along with the appointment confirmation letter for the memory consultation. The nurse contacted the patients and caregivers depending on the telephone number available, presented the study, checked the eligibility criteria and asked whether they agreed to participate. If they did so, they were assigned randomly to one of the two branches of the study. For the patients in the first branch, the telephone appointment was planned with the caregiver one month after the memory consultation. For the patients in the second branch for whom the telephone appointment was the first administration of the IADL questionnaire, a telephone appointment was planned with the caregiver one month before the memory consultation. In case the call did not succeed, the nurse was to try again up to 4 times.
Patients and study characteristics
Additionaly to the IADL questionnaire, we considered the following patients’ characteristics collected using the electronic case report form (eCRF) of the MEMORA study: age, sex, marital status, relationship between the caregiver and the patient education level, cognitive status, etiology, and the Mini-Mental State Examination (MMSE) score ranging from 0 to 30 and evaluating the overall cognitive performance. Details of the collection of the data in the MEMORA study are available elsewhere [10]. These data were collected at different times i.e. Age, sex, marital status and the relationship between the caregiver and the patient were collected at inclusion of the study (at randomization), while the others data were collected during the face-to-face interview at the MC. Indeed, for organizational reasons, it was not possible to schedule an additional visit for patients whose telephone interview was schedules first, since face-to-face interview would take place 1 month later.
Additional information was collected at the second interview using a paper CRF by the nurses, either by face-to-face interview or telephone in order to detect possible changes between the two measures: change of caregiver respondent between the 2 assessments, change of nurse who administered the questionnaire, and the following events: admission to the emergency department, hospitalization, occurrence of a fall, change of living place, occurence of a death in the family. The number of calls needed to reach the participants, the duration of the telephone interview, and the reason why the questionnaire could not be administered after randomization were also collected.
Randomization
After obtaining oral consent for participation, randomization was performed using a computer-generated list (Microsoft Excel 2010). It was a centralized and restricted randomization with an allocation ratio of 1:1 in a fixed block of 4 individuals. The nurses, who enrolled the patients, assigned each of them to one of the branches according to the random allocation and successively in the order of inclusion.
Sample size
The sample size was calculated using the STATA software version 13 (StataCorp. College Station, TX) (SSQDL function). For Cohen’s Kappa coefficient of 0.8, a proportion of patients dependent for at least 2 items of the IADL questionnaire at 50%, with a power of 80% and a risk alpha of 0.05, the sample size required was 138 patients per branch. With an expected loss to follow-up or missing value proportion of 30%, the total sampe size was estimated to be 197 patients per branch.
Statistical analysis
A flow-chart has been made to describe the recruitment of the population. The characteristics of patients who had completed both the first and second assessments were compared to those of patients who had completed only the first assessment, using the Pearson χ2 test or Fisher’s exact test to compare proportions or independent Student’s t-test to compare means. The characteristics of the final study population were compared between the branches. Charasteristics of the patients were summarized using mean ± standard deviation (SD) or number of patients (percentage), as appropriate. The duration of the telephone interview was presented as mean± SD in minutes.
In the main analysis, the extent of agreement between the two modes of administration was measured using the following statistics: the proportion of participants who gave the same response for both modes of administration (The proportion of patients according to the different cases was compared using the McNemar Chi-squared test); the linearly weighted Cohen’s Kappa coefficient [13] for the total IADL score and for the 2 sub-scores of IADL; the unweighted Cohen’s Kappa coefficient [14] for each item of the IADL score (coded as binary variables) and each level of independence according to the total IADL score e.g. autonomy for 8 abilities vs. 0, autonomy for ≥7 abilities vs. less; the intraclass correlation coefficient (ICC), (“two-way mixed effects, absolute agreement, multiple raters/measurement” form) [15] for the total IADL score; the Bland and Altman analysis [16]. These analyses were conducted in all the study population, and separately in both branches.
The level of agreement according to Kappa coefficients was interpretated as 0-0.2: none, 0.21-0.39: minimal, 0.4-0.59: weak, 0.6-0.79: moderate, 0.8-0.9: strong, ≥0.9: almost perfect [17]. The level of agreement based on the ICC was interpretated as <0.5: poor, 0.5-0.75: moderate, 0.75-0.9: good, ≥0.9: excellent [18].
In an additional analysis, the associations between the patients included in the 95% limits agreement vs. those outside, and the patient characteristics were assessed using logistic regression models. The results were presented as odds ratios and 95% confidence intervals (OR[95% CI]).
Missing data was not replaced. P values less than 0.05 were considered statistically significant. Analyses were performed using SPSS (Statistical Package for the Social Sciences) version 19.0 for Windows (SPSS Inc., Chicago, IL, USA).