Proposal for a Global Adherence Scale for Acute Conditions (GASAC): A prospective cohort study in two emergency departments

Background Adherence in the context of patients with acute conditions is a major public health issue. It is neglected by the research community and no clinically validated generic scale exists to measure it. Objective To construct and validate a Global Adherence Scale usable in the context of Acute Conditions (GASAC) that takes into account adherence both to advice and to all types of prescriptions that the doctor may give. To measure adherence and to study its determinants. Materials and method We based the construction of the GASAC questionnaire on a theoretical model and a literature search. Then, between 2013 and 2014, we validated it in a prospective observational study in two hospital emergency departments. Patients were contacted by phone about one week after their consultation to answer several questionnaires, including GASAC and the Girerd self-administered questionnaire about medication adherence as a control. Results GASAC consists of four adherence subscales: drug prescriptions; blood tests/ radiography prescriptions; lifestyle advice and follow-up instructions. An analysis of the 154 sets of answers from patients showed that the GASAC drug subscale had satisfactory internal coherence (Cronbach’s alpha = 0.78) and was correlated with the Girerd score, as was GASAC as a whole (p<0.01). The median score was 0.93 IQR [0.78–1] for a maximum value of 1 (n = 154). In multivariaable analysis, infection was more conducive of good adherence (cut off at ≥ 0.8; n = 115/154; 74.7% [67.0–81.3]) than trauma (OR 3.69; CI [1.60–8.52]). The Doctor-Patient Communication score (OR 1.06 by score point, CI [1.02–1.10]) also influenced adherence. Conclusions GASAC is a generic score to measure all dimensions of patient adherence following emergency departments visits, for use in clinical research and the evaluation of clinical practice. The level of adherence was high for acute conditions and Doctor-Patient Communication was a major determinant of adherence.

1 The notion of adherence uniquely to prescriptions for medication is insufficient [27,28]. Although 2 some authors defined adherence "as the extent to which a patient's behaviour (in terms of taking 3 medication, following a diet, modifying habits, or attending clinics) coincides with medical or health 4 advice" [4], there is still no method to measure adherence that includes all aspects of a patient's 5 behaviour after a consultation. There lacks a standardized tool that is well adapted to clinical research 6 [29] and assessment of health-care quality [1,17], and none suitable for the context of AC [30,31]. The 7 question of how best to measure adherence is still open [2]. 8 A generic scale would be useful to analyse the relationship of between adherence to and other 9 outcomes such as Doctor-Patient Communication (DPC) or satisfaction, and to quantify and compare 10 the impact of measures introduced to improve adherence, such as Patient information leaflets [30,31]. 11 12 Our objective was to create a Global Adherence Scale usable in the context of an Acute Condition 13 (GASAC) based on a theoretical model describing the various dimensions of patient behaviour 14 following a consultation [15] and the results of a literature search. Then, to validate it in two hospital 15 emergency departments and analyse its determinants. 16 17 Materials and methods 18 Literature search 19 We searched the Medline database using the following Mesh terms: patient compliance, adherence 20 AND scale, tool, assessment, measures or questionnaires, in various combinations. We also consulted 21 the Embase and PsycInfo databases, and the Cochrane library in English. 22 Our search filter covered the period from 1985 to 2014. Only meta-analyses, randomized controlled 23 trials, and reviews of the literature were retained. In addition we searched English (NHS) and US 24 (Agency for Healthcare Research and Quality, AHRQ) institutional databases on quality of care 25 assessment and books on the field. Two doctors independently screened titles and if necessary abstracts 26 for all types of articles pertinent to adherence in the context of acute conditions. A manual search was 1 also conducted from the bibliographies of promising articles. Since our literature search did not find 2 specific articles for AC, we based the elaboration of our scale on: 1/ a previously constructed 3 theoretical model, itself based on the literature, 2/ commonly used definitions of adherence [2,4] [15] using a multifactorial approach, as recommended by studies in psychology 12 and sociology [5,32]. It also helped us to avoid the pitfalls of vague terminology [31], poor 13 construction of the scoring system and redundancy between outcomes [15]. This model describes the 14 four aspects of a patient's behaviour following a consultation: taking medications as instructed, 15 following prescriptions for evaluations and tests (radiography, blood tests, appointments with 16 specialists), making appropriate lifestyle changes (i.e. diet, stopping smoking, physical activity, alcohol 17 consumption) and when to engage the healthcare system for worsening or reoccurring symptoms and 18 follow-up. Respecting the model's categories, we constructed a rigorous scale with pertinent items. 19 Moreover the model assisted us in studying the determinants of adherence such as DPC and 20 satisfaction, also defined in the model, with a solid theoretical foundation.

22
Requirements of the new scale 23 The scale needed to respect the following criteria: usable in routine practice, independent of any 24 particular clinical situation, self-reported by the patient, easy to understand, easily evaluable, brief, 25 respectful of the patient's privacy, possible to be completed by or together with carers where necessary, 26 validated and reliable [20]. Self-assessment by the patient was considered the method of choice as it is 27 fast, inexpensive, non-invasive and can potentially help detect the underlying reasons for non-28 adherence [21,37]. In practice certain authors consider it to be the most suitable method for assessing 1 health improvement [20] although it might sometimes suffer from self-deception or dishonest answers.
2 An objective measure such as pill counts, sometimes used in assessing drug adherence, was not 3 feasible in the broad context of assessing other types of behaviour. 4 5 Validation of the questionnaire 6 The pilot study 7 The first version was tested in a pilot study on 30 patients whatever the pathology diagnosed. 8 Immediately after the consultation the patient was given the questionnaire to complete, followed by an 9 additional page about their understanding of the questionnaire and open remarks. 22 Allowing for 20% patients potentially being lost to follow-up we required 180 patients in total. We 23 stopped inclusions when this number was reached. 24 25 Design 1 A two-centre prospective observational study was conducted from November 2013 to May 2014 in the 2 emergency departments of two hospitals. The study was approved by the regional ethics committee 3 (IRB n° 5891 on 31-Oct-2013). 4 5 Physicians who regularly worked in the ED of the two establishments were contacted and voluntarily 6 participated. The physician briefly presented the study (orally and in a patient information letter) and 7 proposed participation to all consecutive adults and children (>15 and accompanied by an adult) 8 diagnosed with a common traumatic or infectious acute condition (ankle sprain or infectious colitis, 9 pyelonephritis, diverticulitis, prostatitis or pneumonia). These acute conditions were chosen from those 10 most commonly seen in primary care [22] and which usually require medication, prescriptions for 11 specialist evaluation or tests, and/or advice and follow-up instructions. We excluded patients whose 12 care led to a hospital stay of more than 48 hours. 13 14 The patient information letter explained the broad aims of the study: to help develop tools to measure 15 the quality of care. Details were not given so as to reduce any self-selection bias. If the patient agreed 16 to participate, they signed a written informed consent. If they declined to participate, this was recorded.
17 Physicians included patients in the study by completing a short inclusion-case report form, describing 18 the patient's baseline and socio-demographic characteristics.
19 Patients were contacted by telephone between 7 and 10 days after the consultation by a study 20 investigator who did not participate in patient recruitment. They were asked the series of questions 21 from the Girerd scale (yes/no) and then the GASAC questions (scored on a Likert scale of 1 to 4). Next 28 percentiles]. The internal consistency of the GASAC score items was assessed by Cronbach's alpha 1 [34]. For quantitative variables, we used the Mann-Whitney test to compare two groups, or the 2 Kruskal-Wallis test to compare more than two groups (non-parametric tests). For qualitative variables, 3 we used the Chi-squared test. Finally, we conducted a multivariate logistic regression to identify 4 factors associated with a "high" adherence according to the distribution of the histogram. To study the 5 determinants of adherence we had to dichotomise the variables, so patients with a GASAC score ≥ 0.8 6 were classed as "highly-adherent", whereas those with a score < 0.8 were classed as "poorly-adherent". 7 All patient characteristics, the level of information, satisfaction and DPC score with p<0.2 in univariate 8 analysis were included into the full model. The final model was obtained by a manual step-wise 9 logistic regression. The correlation between GASAC and the DPC, and correlation between subscales 10 of GASAC scores was explored by calculating Spearman's rho.

Results
12 Literature search 13 Our search extracted 845 records, including 80 reviews. Among these, neither of the two doctors found 14 any reviews or original articles that dealt with an acute condition, nor with global adherence.
15 Concordance was 100%. Among the reviews, four dealt with adherence to exercises (e.g. for 16 musculoskeletal disease); one with showing up at a mammography appointment and one with keeping 17 to a diet. The rest concerned drug adherence in chronic diseases (HIV, psychiatric disease, diabetes).
8 Figure 1. Flow-diagram for the validation of the GASAC scale. Patients lost to follow-up were those 9 who could not be contacted by telephone after 3 attempts.
10 Table 3 shows the baseline and socio-demographic characteristics of the patients. 11 12 8 We calculated the Spearman's coefficient between the different subscales of the GASAC. There was a 9 correlation between the drug subscale and the use of the health care system (Spearman coefficient = 10 0.29 with p = 0.01). In contrast, there was no correlation between the drug subscale and the subscale of 11 recommendations and advice, or between the drug subscale and the test and examination subscale. 12 13 External validity of the GASAC questionnaire 14 We compared the GASAC and Girerd scores (Figure 2). The Girerd analysis included 149 patients with 15 scores distributed as follows: no patient had a score of less than 3/6; 5 (3.3%) had a score of 3/6; 19 16 (12.8%) had a score of 4/6; 35 (23.5%) had a score of 5/6 and 90 (60.4%) had a score of 6/6. Using a 17 non-parametric test there was a statistically significant link between the GASAC score and the Girerd 18 score (p<0.01) and between the drug sub-section of the GASAC score and the Girerd score (p<0.01). 19 We also found a statistically significant correlation between the GASAC and satisfaction scores (n= The determinants of adherence 9 Using an univariate analysis, we determined that the variables associated (p<0.05) with high adherence 10 were the age-band over 40 years, an infectious pathology (as compared to trauma), high patient 11 satisfaction and a high DPC score. Characteristics of the GASAC and strengths 7 The GASAC is a short, patient self-reported questionnaire evaluating four types of patient behaviour 8 following a consultation for an acute condition. The absence of correlations between adherence to 9 drug-prescriptions and adherence to prescription orders for evaluations, tests or specialized 10 consultations, and also with adherence to advice given by the physician, shows that each of the 11 different subscales provides complementary information and are not redundant.
12 The validity of the intrinsic properties of the drug sub-section of the GASAC score was confirmed by a 13 satisfactory Cronbach's alpha coefficient 34