Gaining information about home visits in primary care: methodological issues from a feasibility study

To reasons for non-response, a short (5 to 10 minutes) semi-structured guideline-based telephone interview was conducted. GPs who did not participate in the study were asked to participate in the non-responder analysis exploring barriers to participate. Interviews were recorded and analysed in the same manner as the study design interviews, with pseudonymous recordation and analysis of data. If GPs opted not to participate in the telephone interview, their reasons for unwillingness were asked and documented.

To compare socio-demographic data and organisational features of the different subgroups (participating documenting GPs, non-participating GPs, and GPs who dropped-out (Table 1)), we used exact tests (chi² test, Fisher test, median test) due to the small sample sizes of the subgroups.

After postal invitation, 29 of 92 GPs (32%) confirmed their interest in participation per fax. Twenty-one of 92 GPs (23%) documented their home visits during the allocated week and sent the completed survey instruments back. One participating GP interrupted the documentation after 2 home visits. The response rate of 23% is in line with response or participation rates (ranged between 19 and 36%) of other studies that recruited German teaching general practices [31, 37, 38].

Four GPs did not start the documentation because of personal reasons (2) and/or disagreement with obtaining patients’ consent (3). Two GPs indicated that they documented home visits and assessed study design and instruments in the telephone interview, but failed to send back documentation materials. Two GPs were could not be reached for follow up for unknown reasons. There were a total of eight dropouts with regards to documentation. Documenting GPs and dropouts were asked to assess study design and instruments. 22 telephone interviews were conducted (19 documenting GPs, 3 dropouts) as outlined in the first study arm of Figure 1. The response rate of responders was 76%. This higher response rate is almost the same as those of follow-up surveys [27, 39].

All non-participants were contacted by phone for invitation for telephone interviews to analyse reasons for non-response. 20 of the 63 non-responders (30%) participated (see Figure 1, second study arm). The response rate was in line with another German study assessing reasons for non-response or non-participation in a follow-up [40].

Comparing the descriptive data of the participating documenting (Table 1, column 3) and non-participating GPs (Table 1, column 5/6) there were no significant and visible differences regarding type of practice and estimated number of home visits. Median age was similar in the groups of documenting and interviewed non-participating GPs. Some differences were nevertheless visible but not significant. Based on the small sample sizes only strong effects could be measured. In comparison to non-participating GPs, documenting GPs tended to care for an older patient population (≥65 years) and less for nursing home residents. Documenting GPs delegated their home visits to other medical professionals less frequently than interviewed non-participating GPs. There were on average ten percent less GPs from urban regions in the subgroup of documenting GPs compared to non-participating GPs as well as to the total sample. This was also observed in a former study, where significantly more GPs from rural regions in Saxony participated [9]. This association is not consistent with international data that showed no impact of urban milieu and response rates among physicians of 17 different specialty areas [41]. However, differences in organisation of medical care based on urban rural differences in the context of differing health care systems limit this international comparison.

Drop outs were characterized by younger age, more single practices and smaller number of home visits as well as they were more often from rural regions than the other subgroups. But based on the small dropout sample size of three, the analysis is quite limited.

Most frequently reported reasons for participation by the GPs were: interest in the topic of the study (5), to support research in primary care (4), to influence health policy in a way that will improve their working conditions (4) and to support family medicine in general (5).

Majority of the GPs (13 of 21) stated that the incentive of 3 Euro we gave in the study was not a motivation to participate and would have equally been likely to participate without monetary incentive. Regarding the question whether the incentive could motivate GPs that are indecisive to participate, several GPs assessed a monetary incentive of 3 Euro per documentation sheet as ridiculous and not attractive for GPs. One GP assessed it as a small incentive, but argued that GPs who do not want to support a study would also not do it for a greater incentive. Five GPs stated that the interest in the study/topic is crucial and not the financial incentive. This is not in accordance with most of the studies analysing factors that influence response rates of physicians. Independent on study designs, the majority of these studies showed an association between even modest monetary incentives (between 1 and 5 US Dollars) and higher response rates when compared with no incentive [19, 25, 26]. One possible explanation is that teaching practices (our study sample) are more motivated to take part in research studies than physicians that are not connected to the academic medicine.

In summary, reasons to participate beyond work-related and personal factors include personal or professional interest in a topic and the perceived relevance in research in primary care [19, 27, 40, 42].

The primary reasons given for non-participation were lack of time or the heavy burden by documentation (15 von 21).These are well documented main reasons for GPs non-responding in one-time surveys as well as in longer lasting clinical trials [19, 22, 29‐32, 38, 40].

Three GPs stated that the study design did not fit with their home visit schedules. One of them conducted planned home visits every six weeks. The two other GPs assessed that they did not have enough home visits for the study. Two GPs who were willing to participate forgot to confirm their interest because they misplaced the request in a pile of paperwork. These reported reasons show potential for improving the response rate by presenting more detailed study information for example while a personal delivery of the study instruments in the practice [27]. Delivering an information sheet to the practice staff, including workflow instructions combined with a checklist, could support the implementation of the study while decreasing drop-out rate [43, 44].

Reported reasons for physician drop-out in our study were 1) personal reasons such as illness and 2) unwillingness to obtain patients consent. GPs argued that it could disrupt the doctor-patient relationship and to explain the information of the content sheet and to get the consent (sign) was too complicated and took too much time. The recruitment problem based on obtaining patient consent is not new and inherent in clinical research [28, 29].

Two of 21 documenting GPs did not document at allocated week because they forgot to bring the study instrument or the consent sheet with them to the home visits. In these cases we arranged another documentation week with both GPs and they did the documentation at this later date.

The majority of participating GPs (15 of 21) confirmed in principle the feasibility of documentation of all home visits of a whole week. Some GPs stated that it was feasible and reasonable if they got a good preparation by their practice nurses. This confirms the non-negligible impact of medical staff on feasibility of research with GPs. Some of the GPs also pointed to the time-consuming process of documentation home visits over a whole week period. Most frequently reported reason was the effort to obtain the consent by the patients.

Comparing the estimated number of home visits (mean 20 ± 11) with the documented number of home visits (mean 12 ± 9), there seems to be a gap between estimation and documented reality. Majority of GPs documented significant less home visits than they conduct in average per week (see Figure 2, the points under the red line that mention the optimised scenario describing that number of documented correspondent to estimated number of home visits).

In the interviews, eleven GPs confirmed that they documented all home visits of a week whereas seven GPs pointed out problems in documenting all home visits of a week. Three GPs did only document home visits of four days because of a free day for the team in the allocated week. One GP forgot the documentation in cases of urgency home visits. Three GPs reported that they did not document all home visits because of the missing patients’ consents. In summary, the answers in the interviews validated and gave explanation to the observed under-representation of documented home visits in our study.

Documentation of home visits was primarily done by GPs conducting the home visits (n = 20). In one case the health care assistant did the documentation although the GP conducted the home visits. In another case health care assistant and GP shared the documentation depending on the person conducting the home visit.

Problems in obtaining patients consent were predominant in home nurse residents where a visible number of patients are unable to provide informed consent due to cognitive deficits. Some GPs documented home visits of these patients without consent. Based on the explained problem with the consent, a GP constrained the feasibility of the study only on home visits that took place in private homes. Consequently, he did not document his home visits at home nurse residents. GPs 244 patients had a survey instrument completed during a home visit. 45 of these survey instruments were excluded because of missing consent form (see Figure 3). Besides these “visible” missing patients there were an unknown number of non-documented patients that were unable to provide informed consent due to cognitive deficits. These missing values are an important source of bias when describing characteristics of home visits.

The one-side questionnaire about demographic characteristics of GPs and practices was send back by 20 of 21 GPs. An oral (at the end of the telephone interview) and a postal reminder (including the questionnaire) were mentioned to the one missing GP, but he did not send it back. One question except, all questions were fully answered and answers seems to be plausible. Two GPs gave no answers regarding the question of number of patients who live in nursing homes.

All GPs assessed the two-sided semi-structured documentation sheets as comprehensive and easy to handle. Only one GP expressed the wish concerning a shorter questionnaire. Another GP asked for more detailed information in the instructions regarding documenting travelling time in case of home visits in nurse home residents. Several GPs pointed to doubling of contents in the questions concerning results for encounter and decision on therapies. All of these arguments highlight the importance of using a short questionnaire. The impact of shorter questionnaires on response rates in primary care setting is well assessed [25, 26, 41, 45]. But it is to consider, that a short questionnaire automatically includes a limitation of number of obtaining information.

Missing rate in answers was below 5 percent in the majority of the questions (concerning date and duration of the home visit, travelling time, sex and age of the patient, housing situation, kind of home visit, assessment of the need for a home visit). Somewhat higher missing rates were observed concerning care level (6%), assessment of social support (9%) and time of home visit (11%). Analysis of the pattern of incompleteness for time of home visits showed significant accumulation of missing data in two GPs (16 of 23 missing data). The remaining missing data were distributed over four GPs.

Analysing the GPs with accumulated missing data, there were no unusual characteristics compared to the other GPs. In the interviews only one of both GPs reported problems in filling the documentation sheets that concerned problems with remembering traveling time. But there was no explanation concerning problems with documenting the time of the home visit.

Several interviewed GPs also reported difficulties with recording and documenting time variables because of recall problems. In conclusion, instructions for GPs participation should point out more clearly how GPs should document these variables to avoid recall problems. GPs reported no problems in documenting patients’ morbidity, reasons and results for encounter as narrative answers.

The questions concerning assessment of the need for the documented home visit showed strong tendencies (floor effect described by a positively skewed distribution) in answers (see Figure 4). This allows several interpretations. 1) The data reflect from GPs point of view the need for home visits that seems to be very high. 2) Answers could be influenced by social desirability in this way that GPs who decided to conduct a home visit would not question their own behaviour. Asked for the general comprehensibility of the questions in the documentation sheets, one GP focused on these questions. This GP pointed out that his decision of conducting a home visit automatically includes the need for a home visit. Furthermore this GP explained that there are only two alternatives in need for home visits: yes or no and nothing in between.

Comparable, but inverse patterns of distribution were observed regarding the questions 1) whether the patient could also visit GPs’ practice and 2) whether a health care assistant or practice nurse could also conduct this home visit. In summary, the inversed patterns of response confirm the answers in the question of the need for GPs’ home visit.

The majority of the GPs did not miss basic topics regarding home visits. One GP who stopped her documentation after two home visits commented that the documentation sheet does not collect information on the narrative counselling of patients, which is the most important and time-consuming part of home visits. Additional recommendations included: focusing more on content and on counselling chronic and palliative patients, dividing acute and urgent home visits, and analysing contents of home visits after a hospital stay.

To our knowledge, this is the first study focusing on the methodological feasibility of obtaining data on GPs’ home visits. From a methodological point of view, the pilot study sample was useful to assess whether the study instruments were feasible. Because the population of the main study will also include non-teaching GPs, the results regarding home visits might differ between this study pilot and our future planned study group. We found that the role of monetary incentives did not play a major role in participation. This may be because the monetary incentive was too low and thus not an actual factor in GPs decision to participate. Other factors involved the GPs decisions to participate are most likely a reflection of the intangible factors stemming from the longitudinal relationship of the GP teaching practice network affiliated to the University Halle.