This investigation was conceived as an ancillary study of the multicenter, randomized SENTIPAT trial [17]. We took advantage of the trial to examine patients’ opinions on the organization of their hospital discharge. SENTIPAT participants were also asked to describe their experience with this process.

This multicenter (5 adult acute care units in a Parisian teaching hospital participated voluntarily: digestive and general surgery, gastroenterology, hepatology, infectious diseases, and internal medicine), randomized trial focused on the evolution of patients’ health on returning home posthospitalization (follow-up duration: 6 weeks). The principal objective was to determine whether the information directly transmitted by the volunteer patients via a dedicated website was comparable with that obtained during a telephone interview. It was a noninferiority trial (the main judgment criterion was the percentage of patients reporting at least one clinically significant adverse event occurring during the 42 days after hospital discharge).

Consecutive patients with Internet access at home were eligible for inclusion. They were enrolled the day of hospital discharge and randomized into 2 groups (stratified by department): Internet or telephone follow-up. Patients not eligible (ie, with the same characteristics as those randomized but without Internet access at home) were also included at a ratio of 1:4 of noneligible to eligible patients.

Lastly, 2550 patients (510 from each unit) were initially planned. Between February 25, 2013 and September 8, 2014, we enrolled 2090 patients who were not cognitively impaired and did not have a behavioral disorder, who spoke and wrote French, and were returning home after an acute care hospitalization, regardless of the type of stay—standard hospitalization (scheduled or not) on weekdays only (maximum Monday to Friday or any combination thereof) or outpatient hospitalization (1 day)—and not opposed to participating in the trial.

A patient satisfaction score concerning the hospital discharge process was calculated for each of the 3 items (see scores accorded in Table 1). Each subscore was calculated with a normalized sum of all the item questions (range 0-1), each question within a given item having the same weight, and the total score was computed (mean of the 3 subscores, range 0-1). Questions not concerning the patient (eg, Q11D did not concern patients who used public transportation) were not included in the calculations.

The global distribution of the responses for each discharge questionnaire question was assessed. Between-group qualitative variable (eg, sex) differences were compared with Fisher’s exact test and quantitative variables (eg, scores, age, hospitalization duration, and level of education that was considered as a 4-level ordinal variable) were compared with either the Mann-Whitney-Wilcoxon test or, for matched-paired data, Wilcoxon’s signed rank test. A P value ≤.05 defined significance of usual comparisons. However, the significance threshold was lower when a Bonferroni correction was applied for multiple comparisons, as indicated where appropriate. Missing data were taken into account as follows: nonresponding patients and incomplete questionnaires were excluded from the analyses. Also, for questions not concerning the patient, the response “nonapplicable” was used and they were not included in the analyses. All statistical computations used the R program.

A total of 755 (66.17%) completed discharge questionnaires were collected from the 1141 patients included after a hospital stay of 2 or more days (Figure 1). The relative contribution of each unit ranged from 13.41% (153/1141) to 35.58% (406/1141) and their response rates did not significantly differ from one unit to another (P=.08) and ranged from 60.3% (132/219) to 70.0% (284/406). Patients’ median age was 55 (IQR 39-66) years and 591 of 1141 (51.80%) were women. Hospitalization lasted a median 6 (IQR 3-10) days (median 7, IQR 4-11 and median 3, IQR 2-3 days for standard and weekday-only hospitalizations, respectively) (Table 2). Responders were significantly older than nonresponders (P<.001) for comparable sex distributions, level of education, and hospitalization durations. Internet, telephone, and noneligible group patients completed the questionnaire within median 6 (IQR 3-16), median 7 (IQR 7-9), and median 7 (IQR 7-8) days postdischarge, respectively, with respective Internet and telephone response rates of 39.1% (168/430) and 87.2% (381/437, P<.001). Noneligible patients were significantly older than telephone patients were (P<.001) and their response rate was significantly lower (75.2%, 206/274 vs 87.2%, 381/437, P<.001).

Box plot comparisons between the Internet and telephone groups for each of the 3 items or their total scores (Figure 2) revealed no important differences as confirmed by the corresponding statistical comparison results; all were associated with nonsignificant P values (Table 3). Moreover, the telephone and noneligible groups did not differ significantly for the total score or its 3 subscores.

This study was designed to investigate patients’ opinions of the hospital discharge process with a questionnaire administered either by self-reporting directly online or by a traditional telephone interview. Response rates to our questionnaire showed that patients are willing to assume an active partnership role—87.2% for the telephone group (with Internet access at home) and 75.2% for noneligible patients (without Internet access at home)—values close to the average response rate for 13 studies that included a telephone inquiry to obtain patients’ impressions of hospital care delivered (ie, mean 70%, range 24%-91%) [21]. However, such interviews are more cumbersome and expensive to organize (eg, interviewer, scheduling of calls) and implement as routine practice over the long term. Responders were significantly older than nonresponders, with a similar observed pattern in the Internet and the telephone groups, and this might reflect individuals’ greater availability or interest in health care questions, which increases globally with age.

Given the rising availability and utilization of the Internet in all populations, this easy, low-cost approach as a means of collecting patients’ opinions might be an attractive alternative to telephone calls. In our study, the Internet patients’ response rate was lower (39.1%, 168/430) than the telephone patients’ response rate (87.2%, 381/437). In many respects, it is not surprising. For example, ignoring an invitation to actively enter personal data on a website is much easier than ignoring a live person who reaches another by phone. Nevertheless, the 39% response rate observed in our study exceeded that usually reached with Internet surveys, according to a meta-analysis of 39 studies (median 27%, range 14.5%-51%) [22]. Nonetheless, the information reported by online patients did not differ significantly from those collected by phone. Obtaining patients’ opinions on the discharge process is in-line with current initiatives to achieve a patient-centered health care system [23-25]. Our observations suggest that long-term implementation of an information system, similar to that developed for this study, would enable patients to directly transmit their hospital discharge experiences. The scores observed for the 3 dimensions of the discharge process explored herein indicate an overall positive patient perception: discharge logistics organization (median 0.86, IQR 0.8-1), preplanned posthospital continuity-of-care organization (median 0.67, IQR 0.58-0.92), and patients’ impressions of the process (median 1, IQR 0.75-1). Discharge logistics organization, in particular, was judged globally satisfactory even though approximately half of patients were not involved in the scheduling of their discharge date and time.

Pertinently, our results also identify several difficulties, notably hospital transmission of information to primary care physicians and the patient, and thereby also indicate how to potentially improve performance. Only 20.3% (153/755) of patients declared having left the hospital with a discharge summary and/or letter for their primary care physician. These observations agree with those previously reported by authors investigating discharge summary availability at the time of discharge for health care professionals responsible for posthospitalization continuity of care [26-33]. For example, in their review, Kripalani et al [28] indicated that only a median 14.5% (range 9%-20%) and median 52% (range 51%-77%) of primary care physicians had received discharge summaries 1 and 4 weeks after discharge, respectively.

In addition, more than a quarter of patients deemed the medical and/or nursing explanations of their discharge documents as poor or unsatisfactory, or had received none. Moreover, one-fifth of patients reported a lack of information at discharge. This absence of information and/or delivery of information not corresponding to patient expectations was also noted previously [13,27,34,35]. Other than the strict enumeration of the hospital discharge instructions provided to the patient by health professionals at discharge, the patient’s understanding of them is not always optimal [36,37], notably concerning medical treatments to be pursued [38-41], and can be underestimated by health professionals [42,43]. The findings of Horwitz et al [37] are particularly interesting because despite the demonstration of a gap between the information given to the patients and their understanding of it, the patients “were uniformly positive in their assessment of discharge care” as in our study. In a 2014 systematic revue [44] of 36 studies targeting patients’ opinions on quality of care, only 2 addressed the quality of the discharge process [45,46] and they reported globally positive impressions. However, 2 other studies [10,14] examined the association between patient satisfaction with the discharge process and the hospital readmission rate within 30 days, an important health outcome measure, and found it to be significant suggesting this impression reflects, at least in part, the quality of hospital care. Nevertheless, the associated performance differences were relatively modest, thereby suggesting that improving patient satisfaction with discharge organization would also have a minor impact on health in terms of solid outcome measures.

Our study has several limitations. Inherent to studies requiring an active participation, responders constitute an intrinsic biased selection sample of patients. An example of a demographic status significantly associated with the responder status is age. Moreover, differences are introduced by the questionnaire administration mode (Internet vs telephone). The response rate observed in the Internet group is somewhat disappointing, but another study on patient satisfaction also reported similar response rates: 34% and 78% in the Internet and telephone groups, respectively [47]. Nevertheless, even if a similar response rate had been observed in the telephone and Internet groups, this would not exclude different selection biases from one group to the other (eg, inherent to the mode of administration). The major result of the study is that despite the biases of this study (potential or not), the estimates issued from the 2 groups are very similar; therefore, an Internet-based survey in the domain investigated should be considered as a useful alternative to a “reference” telephone survey. Nonetheless, collecting patients’ opinions via the Internet, as done in this study, has numerous advantages. First, unlike the telephone interview that inserts a third person and a potential information bias (survey subjectivity), resorting to the Internet allows self-administration of the questionnaire, triggering the patient’s active participation. Finally, this method of data collection is less costly than managing a telephone cohort and yields comparable information. However, the similar scores in the telephone and Internet mode of administration observed in this study are based on a particular newly developed questionnaire deployed in a given patient population; therefore, this limits the generalizability (external validity) of the results.

The results of this study advocate for establishing a permanent information system that would enable volunteering patients to express their opinions concerning hospital discharge. Such an information system could also be used for other management issues related to health care organization. Those planning to design similar surveys via the Internet should anticipate a response rate comparable to that observed in the present study. Nevertheless, the concept of sentinel patient explored in this study could constitute, in the future, an essential tool in a patient-centered approach to the organization of care.