1,153 comments have been coded, of which 504 comments relate to contracted physicians (43.7%) and 649 comments relate to elective physicians (56.3%). Only comments for which a variable of patient empowerment is observed, are considered relevant, resulting in 912 comments for the final sample of this study (79.1%). On average, contracted physicians have 25 comments an elective physicians 32 comments. Two elective physicians are outliers with over 100 comments whereas only one physician with over 100 comments is in the contracted physician group. For contracted physicians, an average of 18.6 and for elective physicians, an average of 27 relevant comments have been written over the past six years. The median shows that 14.5 relevant comments are written for elective physicians and 13 for contracted physicians, which means that there is only a minor difference. For the further calculations, only the relevant comments (n = 912) and thus 372 comments from contracted physicians (40.8%) and 540 comments from elective physicians (59.2%) are used.
Comment length: The number of words assess the length of the comments. Comments of contracted physicians (mean: 59, median: 50) are only minimally longer than the comments of elective physicians (mean: 57.74, median: 49) and therefore not significantly different (F(1, 911) = .214, p = .644). In both groups outliers are identified, for elective physicians with 252 words and for contracted physicians 198 words.
Longitudinal timing of comments
Since the timestamp on DocFinder.at is only given as "from a DocFinder user from x years ago", there cannot be a detailed year be assigned, only the period based on the time of the comment analysis is possible. The time of writing of all comments was on average 3.24 years ago. Comments from contracted physicians have been written on average 3.71 years ago and comments from elective physicians written 2.91 years ago. Looking into the relative distribution most comments have been written 5.0 years ago and thus from 04/2016 to 04/2017. A Mann-Whitney U Test reveals a significant difference (U = 77806.00, p < .001).
Quantitative rating of physicians
The average quantitative rating for contracted physicians is 4.626 stars (out of a maximum of 5.0) and 4.963 stars for elective physicians which are statistically different (F(1, 911) = 48.496, p < .001).
Sentiment of comments
In total 94.4% of the comments are positive and 4.8% are negative. Only 0.8% of all relevant comments are neutral. There is a slightly different relative distribution of the sentiment between the elective and contracted physicians. In the case of elective physicians, 98.7% of the comments are positive, in comparison to 88.2% contracted physicians.
Reason for consultation and health status
Only in a small proportion of comments the patients describe the reason for consultation and/or their health status. More than 80% of comments do not include information on the reason for consultation or health status at all.
Time for patients
Comments have been coded with two different variables regarding time spent in consultation. One is time in general and the other is time specifically related to questions. Elective physicians spent significantly more time on patients in general (U = 10790.00, p < .001) as well as significantly more time on patients' questions (U = 93.00, p = .046) compared to contracted physicians. The calculated effect size is rs = .253 for the variable time in general and rs = .343 for the variable time for questions and corresponds to a medium effect in both cases.
Health literacy
For the presentation of the results of the health literacy dimension, we report the following sub-areas
Relationship-relevant aspects: Drawing on the patient-doctor relationship, we observe differences in the patients' health literacy in the doctor-patient relationship (responding to patients, listening, asking questions, and doctor showing interest). A significant difference (U = 7265.00, p < .001) is found with a medium effect size (r = .204). In addition, the Spearman Rank Correlation is used to test whether the relationship is related to the factor time. A highly significant medium positive correlation of these two variables was found (r = .373, p < .001).
Language aspects: We examined whether we find a correlation between elective and contracted physicians and the expressions physicians use when talking to patients, which yield no significance (p = .377), whether in the physician's use of technical terms or simple language.
Communicative aspects: Finally, the communication by the physicians is investigated. We test whether elective and contracted physicians were able to convey enough information by verbal or also visual and in written information. The main purpose of this section was to determine whether written or visual aids had been used and whether patients felt they received sufficient information. Only a low number of codes are found in the written and the visual communication. Also, the highlighting of key points by physicians was mentioned only five times, which means that these variables have to be excluded from the presentation of results. In the area of communication, we further investigate whether patients subjectively felt that they received enough information. Significance was determined based on the cell frequency expected in two cells below 5 by Fischer's Exact Test, and effect size is determined by the Phi Coefficient. A medium and significant correlation was found between the subjective feeling of having received enough information and the type of physician (φ = .248, p = .009).
Shared decision making
We first investigate whether elective and contracted physicians differ regarding the presentation of treatment options. With p = .251 (Fischer's Exact Test), no significant correlation is found between the presentation of treatment options and the type of physician. If treatment options had been discussed in principle in the conversation, it is relevant how precisely the treatment options were discussed. This aspect is rarely described in the comments (n = 24) and subsequently the Mann-Whitney U Test yield no significant difference between the two variables (p = .410, U = 51.00, halving of p due to the directional hypothesis). However, when discussing treatment options is associated with the time spent by physicians, a strong positive association of the two variables is observed (rs = .509, p = .037).
We observed discussing risks and answering questions about treatment options only three times, so these aspects of shared decision making cannot be included in any further analysis of results. Similarly, joint decision making between patients and physicians has been coded in only five comments.
Self-management: Self-management addresses two areas: First, knowledge transfer through explanations on the part of physicians and, second, on the relationship between physicians and patients.
We examine whether there is a difference in knowledge transfer, in terms of explanations including specific explanations of test results, diagnoses, therapies, or general tips, between the two groups. With a p-value of p = .019 (U = 10378.00, halving p due to the directed hypothesis) a significant difference can be found. Accordingly, elective physicians explain more accurately and understandably in conversation than contracted physicians do.
For the patient-physician relationship, specifically related to the responsiveness to the individual situation of patients, no significant correlation with the type of physician can be detected (p = .083). Due to the expected cell frequency below 5 in two cells of the Chi-Square Test, the significance value of Fischer's Exact Test is calculated yielding in a p-value less than .10, implying a trend for a correlation. On the relationship level in general, a significant difference (U = 6332.00, p = .011) in the doctor-patient relationship is found (trust, partnership and consideration of the personal situation).
We further examined the correlation of self-management with time. Both the area of explanations (p < .001) and the conditions for self-management at the relationship level (p = .038) are significantly related to the factor time. The effect size between time and explanations is medium with rs = .306, and the effect size between time and relationship level is medium with rs = .211.
Communication
In the communication dimension of patient empowerment, we focus on how the physician behaved toward the patient. For example, whether the physician took the patient seriously and showed respect and honesty toward the patient (factual level). On the other hand, it is considered whether the physician has responded to the individual fears on the interpersonal level, listened and shown understanding. A Mann-Whitney U Test, reveals that elective and contracted physicians differed highly (U = 418.00, p < .001) in conversation at the factual level. The effect size was also r = .410, corresponding to a moderately strong effect. A significant difference (U = 3498.00, p = .008) is also observed at the interpersonal level (halving of the p value, due to the directed hypothesis). The effect size here is r = .175 and corresponds to a weak effect.
As in the other dimensions, the factor time is associated with the variables. For both the factual (p = .015) and the interpersonal level (p < .001), a significant correlation to the variable time is found. Regarding effect strength, the factor time has a moderately strong to strong effect (rs = .441) on the factual level and a strong effect (rs = .509) on the interpersonal level.
In addition to the variables on the relationship, in dimension communication, the possibility of asking questions in the conversation and the degree of answering the questions are investigated. There is no significant correlation (p = .381) between the type of physician and the ability to ask questions. For the determination of significance, due to the expected cell frequency below 5 in two cells, Fischer's Exact Test was used. If the requirements for the Chi-Square Test are met, no significant difference (p = .200) is found here either. Also, regarding answering questions, no significant difference (U = 1646.00, p = .202) between the two groups can be determined using the Mann-Whitney U Test. The significance was again halved due to the directional hypothesis.