Impact of patient-accessible electronic medical records in rheumatology: use, satisfaction and effects on empowerment among patients

In total, 415 patients diagnosed with RA, conform the 1987 American College of Rheumatology classification criteria, were selected from the patient database of the Arthritis Centre Twente in Enschede, the Netherlands. Treating rheumatologists (n = 6) were asked to exclude those patients from the selection who were deceased (n = 10), or had had their last consultation longer than one year ago (n = 24). Other reasons for rheumatologists to exclude patients were: not diagnosed with RA (n = 3), severe co-morbidity (n = 4) or change of hospital (n = 1). In total, 42 patients were excluded. The 373 remaining patients were sent a personal invitation letter and a paper-and-pencil questionnaire on T0. A reminder was sent to those patients who did not respond within two weeks. One invitation was returned as undeliverable and 6 patients called or e-mailed to report that they were not interested in participation. After the pretest, all (but these 7) patients received an invitation to visit the portal and to log in with a personal account, which could be obtained in person in the clinic. Five months later, the same sample was approached with a personal letter and posttest questionnaire, with a reminder after 2 weeks, excluding the 7 patients that had withdrawn (n = 366). At T1, four questionnaires were returned as undeliverable and two patients were deceased, resulting in 360 patients that received questionnaires on both T0 and T1. This study did not need approval of the ethical review board.

To investigate predictors of patient web portal use, the questionnaire on T0 comprised of socio-demographics (age, gender, education level, marital status, and employment), health literacy [23, 24], Internet use (access, quantity of use and self-perceived skills), and disease characteristics (moment of diagnosis, number of clinic visits in the past 6 months, self-perceived general health).

To investigate the impact of the patient web portal, five validated instruments were included on which change could be expected due to the use of the portal. Satisfaction with care was measured by items based on the QUOTE rheumatic patients [25]. Six items were used measuring satisfaction with the rheumatologist and the nurse practitioner, respectively. Response options ranged from “not at all” (1) to “definitely” (4). The possible range of scores for both scales was 6 to 24, with 24 representing a high satisfaction with care. Cronbach’s alphas were .83 and .87 respectively. Trust in the rheumatologist was measured with the Trust In Physicians short form (TRIP_sf), which is based on the Cologne-Patient-Questionnaire scale “trust in physicians” and which measures different aspects of a trusting physician-patient-interaction [26]. Items could be answered using a 5-point Likert scale ranging from “do not agree at all” (1) to “completely agree” (5), with a possible score ranging from 5 to 20, with 20 representing a high trust in the rheumatologist. Cronbach’s alpha was .93. Self-efficacy in patient-provider communication was assessed with the 5-item version of the Perceived Efficacy in Patient-Physician Interactions questionnaire (PEPPI-5), which assesses the subjective sense of patients’ confidence when interacting with their physicians [27]. The instrument has been translated and validated for the Dutch situation [28]. Participants respond to each question on a 5-point Likert scale ranging from “not at all confident” (1) to “very confident” (5). The range of possible scores is 5 to 25, with 25 representing the highest patient self-efficacy. Cronbach’s alpha of this instrument was .96 in our data. Illness perception was assessed with the Revised Illness Perception Questionnaire (IPQ-R). The subscales ‘Personal control’ (6 items), ‘Treatment control’ (5 items), and ‘Illness coherence’ (5 items) were used, which assess personal control and self-efficacy beliefs, belief in the treatment, and understanding of the illness, respectively [29]. All items can be answered on a 5-point Likert scale ranging from “strongly disagree” (1) to “strongly agree” (5), which makes the possible range of scores 6 – 30 and 5 – 25, with 30 and 25 representing the highest perceived control. Cronbach’s alphas of the subscales were .52, .72, and .76, respectively. Medication adherence was assessed with the Morisky Medication Adherence scale (MMA), which measures medication-taking behavior using eight items. Response categories are yes/no for seven dichotomous items and a 5-point Likert response ranging from “always” (1) to “never” (5) for the last item, which was dichotomized [30, 31]. Scores were recoded, so that higher scores represent a better medication adherence, with eight representing perfect adherence. Cronbach’s alpha was .66.

The questionnaire on T1 comprised of the same measures as the pre-test, except for health literacy. To assess use and satisfaction with the patient web portal, several questions were added to the post-test, including: (1) use and moment of use of the website section and login section of the portal, (2) sharing of the personal information from the portal with others, (3) perceived ease of use, clarity, usefulness and completeness of the portal, (4) problems encountered on the portal, and what was done to solve them, (5) difficulties with understanding information in the login section, and (6) occurrence of wrong information in the login section. Additionally, questions were asked on the self-perceived impact of the portal. These questions covered the same constructs as the aforementioned instruments, but asked patients directly if they felt that the portal caused an increase, decrease, or did not change anything concerning these outcomes.

Analyses were performed using the Statistical Package for the Social Sciences (IBM SPSS Statistics 20). To analyze differences in age and gender at baseline between patients who did return the questionnaire at T0 and/or T1 and those who did not, the Mann–Whitney test was applied for age and the chi-square for gender. Descriptive statistics were used to summarize socio-demographics, Internet-related and disease characteristics, portal use and satisfaction, and perceived impact of the portal. To explore relationships between patient characteristics and portal usage (non-use, website only use, and login use), the Kruskal-Wallis test was applied for continuous variables and the chi-square for discrete variables. Whenever a significant difference was found, pairwise comparisons were performed to further analyze differences between groups, using Mann–Whitney or chi-square tests. Additionally, multi-nominal logistic regression was used to analyze which variables uniquely predicted patient web portal use. To analyze the effect of portal use, Analysis of Covariance was used, in which scores on T1 of non-users, website users and login users were compared, including their T0 scores as covariate. For all analyses, p-values <0.01 (two-tailed) were used as criterion for statistical significance, as multiple comparisons were executed.

Of the 372 patients who received the questionnaire at T0, 259 (70%) sent it back completed. At T1, 360 patients were sent the questionnaire, of which 214 (59%) completed it. A total of 194 (54%) patients completed both questionnaires. There were no differences in age or gender between responders and non-responders on T0. At T1 and in the paired samples, there were no differences between responders and non-responders in gender, but the mean age of responders was 4.2 years higher (P = .01) and 3.6 years higher (P = .02), respectively.

Of all respondents on T1, more than half (54%) reported to have used the portal, and 86 respondents (40%) reported to have logged in to view their personal information (Table 1). Of all respondents with Internet access, 70% had used the portal and 54% had logged in. Lack of Internet access was the most frequent reason for not using the portal (n = 56). Other reasons not to have used the portal were: “I planned to but didn’t have time yet” (n = 30), “I’m not interested” (n = 19), “I tried, but something went wrong” (n = 7), and “I don’t know how to visit the portal” (n = 5).

Of the respondents who logged in on the portal, 60 (70%) reported to do this in the week before a consultation with their rheumatologist or nurse practitioner. Sixteen respondents (19%) reported doing this after their consultation (data not shown in Table). Of the respondents who logged in on the portal, 29 (34%) shared their personal information with a family member.

Table 2 shows the personal and Internet-related characteristics of the respondents at T1. The overall mean age was 62 (SD = 13.3), ranging from 20 to 86 years old. Two thirds of the respondents were female, which is representative for our population. Overall, respondents reported using the Internet regularly, but only a minority (31%) rated their own Internet skills as “good” to “very good”. Univariate analyses showed that age, marital status, education level, employment, health literacy and all Internet-related characteristics were significantly related to portal usage. Non users were more often older, single, lower educated and unemployed. Respondents with a higher level of health literacy were more inclined to log in on the portal, as well as respondents who used the Internet more often, had more years of experience, and perceived their own skills as better.

Table 3 shows an overview of health-related characteristics of the respondents on T1. Most patients had been diagnosed with RA for more than a year, and visited the rheumatology clinic regularly. The majority of respondents perceived their general health as good or excellent. None of these characteristics were significantly related to portal use.

Further analyses with multi-nominal logistic regression, showed that all variables together explained 59% of the variance (R² = .59 (Nagelkerke), model χ ²(22) = 94.04, P < .001). Patient web portal use was significantly predicted by age (b = .09, Wald χ ²(1) = 4.72, P = .03), with younger respondents being more inclined to use the portal. Logging in at the portal was significantly predicted by self-perceived Internet skills (b = −.96, Wald χ ²(1) = 4.74, P = .03) and amount of Internet use (b = −.70, Wald χ ²(1) = 6.07, P = .01).

The portal was positively appraised and most login users found their personal information “fairly easy” to “very easy” to understand (Table 4). When logging in to the portal, 15 respondents experienced a single problem. As a result, three respondents left the portal, three asked for help, and nine kept trying until they succeeded. Two respondents reported reoccurring problems when logging in to the portal. One of them requested a new account, and one called the web host. Nine respondents who had logged in to the portal reported finding incorrect information. In all cases, this concerned medication or blood test information that was outdated. Three respondents mentioned this during a consultation with their doctor and one participant called the hospital. Five respondents reported not taking any action (yet) because: “I thought it wasn’t important”, “I didn’t know who to contact”, and “I was too insecure to contact anyone”.

Several positive changes were perceived by patients who had logged on to the portal (Table 5). A large part of the respondents felt that they were more involved in their treatment and that they understood their treatment better due to the patient web portal. One third of all login users felt that the quality of care was higher as a result of the portal. Also, according to a large part of the respondents, knowledge about the disease, understanding of what care providers explain, communication with the care providers, and trust in the care providers was increased. Additionally, some patients reported to search less for health information by themselves, as a result of the information provided by the hospital-based portal. Only one participant perceived a negative change, he/she felt less involved in the treatment due to the patient web portal. No further adverse effects were reported by the participants.

Analyses of Covariance revealed that the T1 scores on the impact outcome measures did not differ among the three groups (non-users, website only users and login users) (Table 6). Website users and login users did not show a larger improvement in the patient-provider relationship and on empowerment than non-users did, as we would have expected. It should be noted, however, that ceiling effects at T0 were found for four out of eight outcome measures. A ceiling effect is present when at least 15% of respondents scored the highest possible score on the scale. Our results showed that 45% of the respondents scored the highest possible score on satisfaction with the rheumatologist, 52% on satisfaction with the nurse, 30% on trust in the rheumatologist and 29% on self-efficacy in the patient-provider communication. This shows that room for improvement on these measures was limited in our sample. Additionally, 56% of the respondents had a score of 7 on medication adherence, with a highest possible score of 8; while this cannot be defined as a ceiling effect, room for improvement was limited on this outcome as well.