The challenge for general practitioners to keep in touch with vulnerable patients during the COVID-19 lockdown: an observational study in France

This paper builds on selected data collected during the second national observational survey conducted among GPs in France by ACCORD network. This survey was aimed at documenting adaptations of GPs at the end of the first lockdown in France (May 2020) and identifying individual, organizational and territorial factors associated to these adaptations, with a special interest for organizational factors. It was based on a questionnaire that could be filled in on-line using the free LimeSurvey tool. The questionnaire was created by the survey team members, inspired by clinical experience-based hypotheses, available literature, and answers to the previous survey (March 2020) [14], then revised by a panel of 7 primary care experts and finally piloted by the survey team members before the survey being launched. A message with the link to the survey was sent using national mailing lists of GPs (N ~ 25,000) from the study partners (see Additional file 1) and the mailing list of GPs (N = 4436) who participated in the previous survey organized in March 2020 by the same partners [14]. The link to the survey was also disseminated using social networks. Participants were invited to send the survey link to their colleagues. After reading information concerning aims, methods, sources of funding, institutional affiliations of the researchers and the anticipated benefits, subjects gave their informed consent to participate by clicking and answering the questionnaire. Data were collected between 7th and 20th May 2020.

The questionnaire for this survey was composed of 63 questions. A first part explored adaptations to the pandemic situation in relation with 7 domains of interest (activity, prescriptions, occupational health, patients with COVID-19, nursing home residents, vulnerable patients, territorial partnerships). The second part collected GP’s individual characteristics (gender, age, teaching activity, other complementary clinical activities, usual density of activity, recent quantitative activity changes, number of patients hospitalized for/who died due to COVID-19, being at risk of severe COVID-19, fear of SARS-COV2), organizational characteristics (type of practice: alone/monodisciplinary practice/independent multidisciplinary group/care centre; size of practice) and territorial characteristics (location of the practice, relations with the hospital, local partners and networks).

This paper specifically builds on the answers to the following question (thereafter called “question of interest”): “Did you take the initiative to contact by phone some of your patients (e.g. vulnerable, with chronic diseases)?” with three possible answers: “No, I did not call any patient”, “Yes, the ones I thought about”, and “Yes, I made a list using some criteria (e.g. patients with 100% coverage by the French national insurance due to a chronic disease, body mass index)”. A fourth answer choice was also proposed: “Other, please specify”. These latter answers were manually recoded into one of three previously described answer categories when the verbatim was explicit enough (See Table 1). In the other cases, observations were excluded and considered as missing data for the question of interest.

Incomplete questionnaires and duplicates were excluded. The questionnaire was considered complete if including at least the GPs adaptations as well as organizational and territorial factors, age and sex. The duplicates were identified searching a same token (individual number of access to the questionnaire for the responders to the previous study who agreed to be surveyed again), registration number in the national directory of health professionals or email address. The sample was then limited to GPs practicing in Metropolitan France. Questionnaires from GPs who declared no clinical activity in the last 7 days were finally excluded because these GPs were considered not to have a good insight into practice adaptations in a quickly evolving situation, as well as those with missing data for the question of interest.

In the first analyses the variable of interest was considered as a dichotomous variable Yes/No; thus, the two answer categories “Yes, without criteria-based strategy” and “Yes, with criteria-based strategy” were considered together as “Yes”.

For the final model, independent multidisciplinary group and care centre were merged into one modality: “multidisciplinary practice”. Moreover, the practice type and practice size were expected to be highly collinear. As both were considered very relevant for the question of interest, a composite variable was constructed rather than favouring one over the other. Therefore, a new variable “type and size of practice” was created with five modalities: i) alone, ii) monodisciplinary practice with 5 ≤ professionals, iii) monodisciplinary practice with > 5 professionals, iv) multidisciplinary practice with ≤20 professionals, and v) multidisciplinary practice with > 20 professionals.

The GPs’ characteristics were first described. Then, bivariate analyses were carried out using the question of interest considered as a dichotomous variable, followed by bivariate analyses using the question of interest considered in three answer categories, using chi-2 tests for categorical data. Then, an unordered multivariate multinomial logistic regression model was used to assess the relative contribution of the different factors. To assess the potential role of all variables of interest, no selection of variables was undertaken. The reference categories were commonly those with the larger population, except for age and number of patients who died of COVID-19 for which we tested specific hypothesis. For age group, in the demographic context of the current switch of the older generation by a younger workforce, we explored the effect of being among GPs in their first 10 years of practicing and of being in their last 10 years of practicing, versus reference category of 40-55 years. The likelihood ratio chi-square and the Score and Wald tests were used to assess how well the multivariate model fitted the data. SAS software (version 9.4) was used to undertake these analyses and the statistical significance threshold was set at 5%.

A qualitative analysis of the verbatim material for the answer “Other, please specify” was performed to explore key components of the reported contact initiatives. A general inductive approach was used to extract units of meaning and articulate emerging concepts. It consisted of open labelling followed by categorization [15], using NVivo10®. The same verbatim section could be labelled with more than one unit of meaning. The analysis was performed by one of the authors (TB) and the final categorization was generated during two consensus meetings with the whole research group.

Among the 4699 questionnaires that were filled in, 3096 were retained after exclusion of duplicates and incomplete questionnaires. Among the 3068 (99.1%) questionnaires by GPs practicing in Metropolitan France, 38 were excluded because these GPs declared no clinical activity in the last 7 days and 18 because of missing data for the question of interest, resulting in a final sample of 3012 questionnaires (Fig. 1). Among this study sample, 1659 (55.1%) were women, and 1127 (37.5%) participants were younger than 40 years of age. Moreover, 469 (15.6%) participants declared that they worked alone, 1300 (43.4%) in a monodisciplinary practice, and 1228 (41.0%) in a multidisciplinary practice (n = 1099, 89.5%, in an independent multidisciplinary group and 129, 10.5%, in a care centre) (Table 2). The sample corresponds to 5.6% (3012/53,339) of all currently active French GPs [16]. Responders were younger (< 40 years, 37.6% vs 17%), more frequently women (55.1% vs 44%) [16] and worked less frequently alone (15.6% vs 39%) [10] than the whole French GPs’ population (Table 2). All French regions were represented (Additional file 2).

Among the 3012 responding GPs, 123 (4.1%) filled in the “Other, specify” text field. The inductive analysis of the 115 unique answers resulted in 58 labels, classified in four categories (Fig. 2), and articulated in three concepts: “vulnerability” “organization” and “mission”. The concept of vulnerability covered a wide range of patients’ selection criteria indicated by GPs. It gave insights into a composite and empirical definition of vulnerability in the COVID-19 pandemic context: age, ongoing disease or follow-up, administrative criteria about chronic disease coverage, limited mobility, social criteria. Many GPs described a multicriteria approach, sometimes with individual assessments. Some said “the one I thought about”, highlighting the inconscient process of assessment based on multiple “gut” factors. This “human brain” tool seemed important. Indeed, some young GPs said that they felt helpless because “they did not know the patients”. Very few had a pre-existing list of vulnerable patients. Some simply contacted all their patients. The concept of organization covered the resources mobilized by GPs to contact their patients. Technical tools and collaboration were mentioned as resources first to identify vulnerable patients, and then to contact them. Most respondents described the use of their agenda to list the patients they met the past months or were supposed to meet during the lockdown. Fewer used medical records or national health system data to identify vulnerability criteria. Many GPs also described an informal interprofessional network (e.g., nurses, community pharmacists, medical secretaries) to identify patients. To contact patients, GPs used different strategies: collective contacts (email messages, texts, or website pages), but mainly individual contacts (telephone, video consultations). They often reported relying on collaboration with medical secretaries, medical students, nurses and especially public health nurses, associate GPs, or informal caregivers to perform the actual contact. Mission was the emerging concept about the GP-patient relation. Respondents described their initiative as a mission and more than a one-call procedure. They wrote about regular contacts, closer to a guardian role. Their motivation for this remote contact was to avoid any physical contact and the related risk of infection for their patients. They wrote about “keeping in contact”, an almost physical expression, while the questionnaire used the technical word “phoning”. However, some considered that their complete availability was sufficient to ensure that patients would contact them if needed (as usually done) and did not call them.

Finally, the analysis of the collected data indicated a significant variability and a dynamic process. Keeping in touch during the lockdown was an evolving task, in a rapidly changing environment.