Article Text
PDF
Abstract
Objectives Because of fundamental differences in healthcare systems, US readmission data cannot be extrapolated to the European setting: To investigate the opinions of readmitted patients, their carers, nurses and physicians on predictability and preventability of readmissions and using majority consensus to determine contributing factors that could potentially foresee (preventable) readmissions.
Design Prospective observational study. Readmitted patients, their carers, and treating professionals were surveyed during readmission to assess the discharge process and the predictability and preventability of the readmission. Cohen’s Kappa measured pairwise agreement of considering readmission as predictable/preventable by patients, carers and professionals. Subsequently, multivariable logistic regressionidentified factors associated with predictability/preventability.
Setting 15 hospitals in four European countries
Participants 1398 medical patients readmitted unscheduled within 30 days
Main Outcome(s) and Measure(s) (1) Agreement between the interviewed groups on considering readmissions likely predictable or preventable;(2) Factors distinguishing predictable from non-predictable and preventable from non-preventable readmissions.
Results The majority deemed 27.8% readmissions potentially predictable and 14.4% potentially preventable. The consensus on predictability and preventability was poor, especially between patients and professionals (kappas ranged from 0.105 to 0.173). The interviewed selected different factors as potentially associated with predictability and preventability. When a patient reported that he was ready for discharge during index admission, the readmission was deemed less likely by the majority (predictability: OR 0.55; 95% CI 0.40 to 0.75; preventability: OR 0.35; 95% CI 0.24 to 0.49).
Conclusions There is no consensus between readmitted patients, their carers and treating professionals about predictability and preventability of readmissions, nor associated risk factors. A readmitted patient reporting not feeling ready for discharge at index admission was strongly associated with preventability/predictability. Therefore, healthcare workers should question patients’ readiness to go home timely before discharge.
- Communication
- Healthcare quality improvement
- Hospital medicine
- Human factors
- Patient-centred care
Statistics from Altmetric.com
Introduction
Unplanned readmissions within 30 days are perceived as a major adverse event after hospital discharge. Current readmission rates vary from 10% to 30% internationally and the related costs are increasing.1 Internationally, readmission rates are often used to audit and reimburse units and as a quality of care monitoring tool.2 3 In the Netherlands, readmission rates were introduced as an official quality indicator in 2016.4 Due to an ageing population and healthcare policy changes, emergency department (ED) admissions have risen in the last few years.5 6 As a result, hospital beds are becoming scarce and physicians are under constant pressure to discharge patients rapidly.
The problem with the use of readmissions as a quality indicator is that not all readmissions are preventable and their causes might find their origin in natural progression or unavoidable recurrence of underlying diseases.7 8 In addition, despite many initiatives to reduce readmissions predischarge and postdischarge, such as financial penalties levied on hospitals with higher rates and telephonic follow-up interventions, it is still questionable whether these interventions truly result in lower readmission rates.5–9 Moreover, since preventability has not been defined uniformly, it remains uncertain whether this quality indicator, in its current form, is reliable.10 11 Several studies have used the opinion of physicians as the gold standard to determine whether readmissions are preventable, and have defined factors that would predict preventability from these findings.12–14 However, recent work has shown that even among doctors there is poor consensus about the preventability of a readmission.15
Although many attempts have been made to create readmission prediction models, there are still no widely used and internationally validated tools predicting the chance of readmission.16–18 Most risk models were developed in the USA and Canada, and due to differences in healthcare systems and case mix they may not be suitable to be used in European populations.19–21 In comparison to US patients, European patients tend to be healthier and have easier access to care because of obligatory healthcare insurance.22 In addition, European primary care is much better streamlined with the rest of the healthcare chain.23 Another problem with these risk models is that they do not measure preventability. A recent study demonstrated that causes of potentially preventable readmissions are mostly human-related coordination and communication failures.24 A few studies have been performed investigating opinions of patients and healthcare workers on the preventability of readmissions and discharge planning.25 26 Yet most of these studies did not include all stakeholders and questioned patients retrospectively after discharge from a readmission.
Therefore, we performed the first prospective observational study of 1398 unscheduled medical readmissions to 15 centres in four European countries aiming to investigate the following:
the opinions of readmitted patients, their carers, nurses and physicians on the predictability and preventability of the readmissions;
the contributing factors that could potentially predict (preventable) readmissions using the consensus opinion of the interviewed.
Methods
Sites and participants
This research project ‘CURIOS@’ (CaptUring Readmissions InternatiOnally to prevent readmissions by Safer@home consortium) took place in academic and non-academic centres using the network of the Safer@home (Scoring Acute admissions For Estimating Readmission) consortium. The study was an initiative of a research collaboration (Safer@home consortium), focusing on readmissions and safer discharge processes, formed in 2014 consisting of 12 acute physicians, emergency physicians and epidemiologists from Europe. Fifteen centres participated (nine centres in the Netherlands, three in the UK, one in Ireland and two in Denmark). The data collected were derived from two clinical episodes: readmission (RA) and index admission (IA) between 1 January 2016 and 1 November 2016. Data collection took place on the unit of readmission at any time within 72 hours after the readmission. Primary approval was obtained from the medical ethical committee of the VU University Medical Center. Subsequently, local approval was obtained in other participating centres.
Eligible patients were those who had unscheduled admissions through ED, acute medical unit or other clinical areas, were aged above 18 years, and readmitted to the hospital following a previous hospital admission to any clinical specialty for a minimum of one night in the previous 30 days. If a patient was readmitted more than once, only the first readmission was included. To be included, readmissions had to be to a medical ward (cardiology, geriatrics, gastroenterology, haematology, internal medicine, nephrology, neurology, oncology, pulmonary medicine, rheumatology). Excluded were patients readmitted electively for procedures, surgery or chemotherapy; with an IA for psychiatry or gynaecology; who stayed shorter than one night during IA or RA; transferred to the ward from an initial admission to the intensive care unit (ICU); and admitted to another institution in their IA. All patients gave written informed consent.
Data collection
After instructions and training from the coordinating investigator (LG), centres were allowed to participate, aiming to include at least 50 patients consecutively. Investigating site researchers were deemed competent once they followed adequate training, and enough time for them was allocated to perform interviews, data collection and root cause analysis. They were all medically trained but not involved in patient care during the inclusion period. After obtaining written informed consent, site researchers surveyed readmitted patients: the questionnaire consisted of seven questions about their readiness for discharge during IA and predictability and preventability of their readmission (see online supplementary 1). This questionnaire was constructed using available literature and after reaching consensus among Safer@home group members.27–31 In addition, it consisted of questions that could easily be asked by physicians and answered by patients in daily practice, and were reproducible to be used in other settings. The questionnaires were tested in a pilot patient group before agreeing on a final amended version. Subsequently, after obtaining patient’s permission, a carer (defined as a person providing unpaid intensive and long-term care because of a personal relationship) if available was approached in person or by telephone to answer two questions about predictability and preventability of the readmission. Lastly, a doctor and a nurse responsible for the patient during readmission were interviewed. To reduce bias, all interviews were performed separately by site researcher. This investigating researcher also answered the same two questions after having assessed answers given by the other interviewees. The researcher also had access to patients’ charts. In addition, data items preidentified through literature as being potentially predictive of a (preventable) readmission were collected.16 17 These variables providing information of the IA (ie, Charlson Comorbidity Index32 and Clinical Frailty Scale33) were extracted during readmission, using information collected directly from patient and using clinical notes of IA and discharge communication. The data set contained no patient identifiable variables.
PRISMA analysis
To assess predictability and preventability, we asked all interviewed about the reasons for readmission. Subsequently, site researcher qualified these into one or more root causes categorised into disease, patient, healthcare worker, organisational or other causes, originally identified by prevention and recovery information system for monitoring and analysis (PRISMA) supplying us with more information about probable root causes for the readmission. PRISMA analysis has previously shown to provide objective and structured insight into causes for adverse events, by composing root causal trees for adverse events.24 34 This method has been accepted by the World Alliance for Patient Safety (see online supplementary 2 for more detailed description).35
Measurement of predictability and preventability
Since a gold standard defining predictability and preventability is not available, after reaching consensus in our half-yearly consortium meetings a new variable was composed. It was decided that if a majority (50% or more) of interviewed groups (patients, carers, doctors, nurses, researchers) assessed the readmission as predictable or preventable (options yes, no, unknown), the readmission was decided as predictable or preventable. For example, if a carer was not available, if two out of four remaining interviewed assessed readmission as preventable, it was deemed preventable. If all five interviewees answered the questions, three ‘yes’ answers were needed for this conclusion. In a separate analysis, we used the answers given by the members of the five interviewed groups separately and regarded option ‘unknown’ as missing.
Statistical analysis
Statistical analysis was performed in SPSS V.22.0. Categorical variables are summarised as frequencies and percentages. Continuous variables are summarised by mean and SD in case of a normal distribution or median and ranges otherwise.
Cohen’s kappa (κ) was used to measure agreement of predictability and preventability assessments (yes, no, unknown) separately for each pair of five interviewed groups. McNemar-Bowker test was performed to assess whether different pairs of interviewed groups varied in proportions of ‘yes’, ‘no’ and ‘unknown’ answers.
Logistic regression analyses were used to find variables that were associated with assessments of preventability and predictability of readmissions. For all categorical variables where ‘unknown’ or ‘don’t know’ was an option, these were considered as ‘missing’. However, these constituted less than 10% of the total answers. For predictability and preventability assessed by the individual groups, the option ‘don’t know’ was considered as missing in multivariable models, regardless of its percentage (tables 1 and 2). Separate models for predictability and preventability were built for each interviewed group and for the composed variable of consensus judgement. Only variables with a two-sided p≤0.10 in univariable logistic regression analyses were included in a multivariable logistic regression analyses where backward elimination was used to find a minimum set of variables that were independently associated with predictability and preventability. To account for differences between countries, country was included as a predictor in all models. The magnitude of the association between predictors and outcome was quantified using ORs together with their 95% CI. The discriminative ability was quantified by means of the area under the receiver operating characteristic curve (AUROC).
Results
Patient and hospital and readmission characteristics
During the study period 1961 patients were eligible, of whom 1398 patients participated, resulting in an inclusion rate of 71.3%. The reasons for exclusion were the following: patient was already sent home on day of intended inclusion, 41.0% (231 of 563); unwilling to participate, 20.4% (115 of 563); being too ill, 19.2% (108 of 563); language barrier, 8.6% (48 of 563); patient deceased at RA, 2.1% (12 of 563); and other reasons, such as being in quarantine, 8.7% (49 of 563). The median age was 70 (range 18–96), and the median number of included patients per hospital was 71 (range 48–226); other characteristics are given in table 1 (for additional characteristics: see online supplementary 3).
Patient characteristics*†
The majority of interviewed deemed 27.8% (390 of 1398) of the readmissions as potentially predictable and 14.4% (202 of 1398) potentially preventable. The assessment per interviewed is listed in table 2.
Predictability and preventability according to the interviewed*†
Consensus on readmission (κ)
Table 3 shows Cohen’s kappa for the two questions put forward to all individuals interviewed: (1) ‘Do you feel the current readmission was expected?’ and (2) ‘Do you feel the current readmission was preventable?’ (yes, no, don’t know). For predictability, none of the kappas were satisfactory; they were all below κ=0.7. The poorest consensus was found between patient and physician, and patient and nurse, with a rate of κ=0.173 and κ=0.153, respectively. The highest kappa was found for physician and researcher (κ=0.607). The consensus on preventability of readmission was also unsatisfactory; the highest score in this analysis was κ=0.473, measuring agreement between researcher and physician. The poorest scores in this group, κ=0.105 and κ=0.135, were found comparing patient with physician and nurse, respectively. McNemar-Bowker test was p<0.05 for all, indicating that proportions of respondents answering ‘yes’, ‘no’ and ‘don’t know’ differed between all pairs of interviewed groups.
Consensus on readmission
Contributing factors to assessing the predictability and preventability of readmission
Using multivariate models, factors potentially contributing to predictability and preventability of readmission were identified according to the opinion of the majority. Subsequently, models were composed per interviewed group. The variables used in the models (table 1) associated with assessing readmission as potentially predictable and preventable were different for every group interviewed (tables 4 and 5). Using the majority consensus on predictability and preventability of the interviewed (table 4), factors significantly associated with a higher predictability of readmission included the following: having a non-elective IA (OR 2.55; 95% CI 1.59 to 4.08), having more than five admissions in the year before IA (OR 2.69; 95% CI 1.72 to 4.20), lower age (OR 0.98; 95% CI 0.97 to 0.99), higher clinical frailty scale (OR 1.29; 95% CI 1.18 to 1.42) and a higher charlson comorbidity score (OR 1.08; 95% CI 1.02 to 1.15). When a patient reported having felt ready at discharge during IA, the readmission was deemed less likely (OR 0.55; 95% CI 0.40 to 0.75). Having a follow-up planned (OR 0.52; 95% CI 0.35 to 0.78) and feeling ready for discharge (OR 0.35; 95% CI 0.24 to 0.49) were significantly associated with readmissions being deemed less preventable. As illustrated in table 5, with the exception of the physician for preventability, all interviewed considered readmissions more predictable and preventable when a patient reported not feeling ready for discharge. Tables 4 and 5 demonstrate the discriminative ability of the 12 models by means of the area under the curve varied from moderate to good (0.65–0.74).
Factors positively associated with predictability and preventability of readmissions assessed by the majority*
Factors positively associated with readmission being assessed as predictable and preventable per interviewed group*
Root causes
All interviewed were asked to qualify reasons for readmission into one or more of five available root causes. Each readmission could have more than one root cause per interviewed and the maximum cause per patient was 25. In total 6895 root causes were identified, with a mean of 5.3 per patient (SD 1.9). Most root causes were disease-related (67.9%; 4686 of 6895), followed by healthcare worker-related (17.5%; 1208 of 6895), patient-related causes (10.9%; 749 of 6895), organisational (2.2%; 153 of 6895) and non-classifiable (1.4%; 99 of 6895). In univariate analysis, when the composed variable for majority opinion was used, disease-related root causes were negatively associated with a readmission being considered predictable (83.7% vs 71.8%) (OR 0.5; 95% CI 0.37 to 0.66) and preventable (87.6% vs 38.0%) (OR 0.09; 95% CI 0.06 to 0.12); healthcare worker-related root causes were positively associated with readmissions being more predictable (11.9% vs 20.6%) (OR 1.92; 95% CI 1.39 to 2.65) and preventable (5.6% vs 66.3%) (OR 33.43; 95% CI 22.48 to 49.727).
Discussion
This prospective European multicentre study of 1398 unscheduled medical readmissions revealed that there was poor consensus on predictability and preventability among readmitted patients, their carers, nurses and physicians. Especially, there was little consensus between patients and their physicians. In addition, factors that could potentially contribute to (preventable) readmission were notably different according to every group interviewed. Not feeling ready for discharge was strongly associated with predictability and preventability when opinion of the majority interviewed was used. This was also underlined when opinions of the interviewees were taken separately. In addition, according to the opinion of the majority, 27.8% of the readmissions were deemed predictable and 14.4% preventable. Readmissions thought to have been caused by healthcare worker failures were more often deemed preventable.
Our study is the largest multicentre study performed to date and the first investigating European readmissions prospectively. Previous large-scale studies were mostly performed in the USA, and due to the differences in healthcare systems, the results of these studies are probably not applicable to a European setting.21 This is because broader issues other than the underlying, medical condition, such as social and environmental factors, are increasingly recognised as having a contributory role in the cause of readmission.19 21 In addition, our study is one of the first studies in Europe where opinions of the most important stakeholders in the care chain were taken into account. Most previous studies were smaller retrospective studies, or did not involve all stakeholders.7 24 36
The percentages of readmissions that were deemed preventable are in line with previous research, but their numbers do distinctly differ per interviewed group.12 An important finding is the discrepancy in assessment between all interviewed, especially between patient and their treating physician, the latter assessing less readmissions as likely preventable. This implies that these healthcare professionals do not agree with th17 eir patients about the predictability and preventability and associated factors.
Although multiple risk models have been composed trying to create readmission prediction models, most models do not perform satisfactorily in a European population.21 This was also demonstrated in our cohort; 53.8% had a low risk of readmission according to the HOSPITAL score (see online supplementary 1).17 In addition, many variables in these models are not modifiable, and therefore not suitable as interventions to improve the healthcare chain.19–21 It must be acknowledged that the area under the receiver operating characteristic curves (AUROC) of our regression models were not higher than other prediction models. However, it is doubtful whether the many prediction models already created really add value to the day-to-day practice.
According to our results, if a patient reported not feeling ready for discharge, this was a risk factor associated with a higher chance of readmission. Previous work has already demonstrated the importance of readiness for discharge by using comparable questions in their study designs. Auerbach et al 26 reported this by stating that a proportion of US readmissions may be prevented with better attention to patients’ readiness for discharge. They used patients’ perspective to identify potentially changeable factors such as discharge and advance care planning, which could be helpful to prevent readmission. We used another perspective by enquiring all interviewed for root causes that could better qualify the lack of readiness. Readmissions were more often deemed preventable if attributed to healthcare worker-related causes, which are probably modifiable and therefore a potential focus for improvement. One of the added values of our study is collecting the perspective of the patient’s caregiver per readmission, which has not been performed on such a large scale in the past.37 38 Caregivers also assessed readmission more preventable when patients did not feel ready for discharge. Literature shows that caregiver engagement can improve continuity of care by providing a consistent relationship for patients during their journey through the multiple and complex transitions in care.39 Caregivers play an important role in the transfer and additional care at home, which could potentially prevent readmission. This study emphasised their opinion should be highly valued and taken into account during the discharge process.
Commencing improvement by simply asking patient at the bedside whether he feels ready before discharge may be the first step in understanding each other’s perspectives and could make other prediction models less relevant. A recent study underlined this by showing that early preparation for discharge resulted in significant reductions in patients reporting they were unaware of the problems to watch out for postdischarge, and patients who did not understand their recovery plan.40 In addition, Greysen et al 37 reported the most readmitted patients understood their postdischarge plan but were not explicitly asked about anticipated difficulties carrying out the aspects of this plan.
The interviewed groups did not agree on predictability, let alone on preventability. A high readmission rate should be a signal for a hospital to further look into the causes, but solely looking at the rate and penalising the hospitals, without correcting for case mix, and looking more closely into the type of care provided on the work floor are questionable. Our work demonstrates that defining a gold standard integrating preventability in a quality indicator is difficult. In addition, disunity exists whether 7-day readmissions are not a better indicator of quality of care since the timeframe of 30 days may not be homogeneous and readmissions that occur later may often be related to new problems.41 42 Our work does indicate so; in a subanalysis dividing the cohort in a 0–7 days (612 readmissions, 43.8%) versus an 8–30 days group (786 readmissions, 56.2%), univariate regression analysis using consensus opinion on preventability showed a significantly higher percentage of preventable readmissions in the early readmissions, when compared with the late readmissions (54.0% vs 46.0%) (p=0.002, OR 0.619; CI 0.459 to 0.836). However, in further analysis no distinct risk factors were found between these groups.
The limitations of this study must be acknowledged. We collected data from four different countries in 15 hospitals all using different healthcare systems, which may make results less generalisable. However, European healthcare systems are more alike when compared with the US system, and we corrected for country in our analysis. We did not correct for medical staff capacity or trainee exposure (less staff and more trainee exposure might affect patients’ readiness for discharge) as collecting data on these aspects was difficult. Another limitation is asking opinions during readmission about the IA, which could potentially lead to (recall) bias and subjectivity. Ideally, this study should be conducted at discharge of IA. However, sending patients home after their confirmation of not feeling ready for discharge brought up ethical and feasibility issues, marking the fact that physicians are hesitant to ask this question since it might lead to not being able to discharge the patient. Majority consensus was used to decide on preventability and predictability, which may be questionable. Therefore, we also reported on perspectives per interviewed group separately.8 A final caveat is the composition of our own questionnaire; since this is not a validated measure, subsequent use in different settings should be tested more rigorously.
Conclusion
The European acute healthcare chain is under increasing pressure, potentially resulting in more unscheduled readmissions. This international multicentre study performed in 2016 is the first to prospectively assess this problem in Europe. The 1398 readmitted patients, their carers, treating nurses and their physicians do not agree on the predictability and preventability of readmissions, let alone associated risk factors. This raises the question of the validity of readmissions as a quality monitor. Extensive research has been performed on risk models, but healthcare professionals simply asking the patient whether they are feeling ready for discharge at IA may be one of the key questions to target in preventing unnecessary readmissions. Future work is warranted to assess if asking these questions consistently leads to more patient-directed discharge plans and if they could potentially decrease the percentage of (preventable) readmission.
References
Footnotes
Competing interests None declared.
Patient consent All patients included in this study signed informed consent form before inclusion. This informed consent form was composed by our group and approved by the ethical committee.
Ethics approval Primary approval was obtained from the Medical Ethical Committee of the VU Medical Center (ID: 2015.293). The Medical Ethics Review Committee of VU University Medical Center is registered with the US Office for Human Research Protections (OHRP) as IRB00002991. The FWA number assigned to VU University Medical Center is FWA00017598. Using this declaration other participating centres received local approval. In the UK ethics approval was obtained from the Wales Research Ethics Committee 5, Bangor (ID: 15/WA/0424). In Denmark ethical approval was obtained from The Secretariat of the Regional Committees on Health Research Ethics for Southern Denmark (ID: S-20152000-115 CSF). In Ireland ethical approval was obtained from the SJH/AMNHCH Research Ethics Committee (ID: 2015-03 Chairman’s Action (17)).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Anonymised full data set and statistical codes are available to readers on request.
Collaborators Safer@home collaborator group (alphabetically ordered): L Ackermans, A B Arntzenius, D G Barten, M H van derBie, T Boeije, D A Chamles, E Chaudhuri, M Diepenbroek-Meekes, D P Cooper, E M Durinck,J A van Erven, T A Graaff-de Kooter, F Holleman, J Huussen, W Jansen, J JJensen, J Kellet, T Knol, I Lee, T S van Lieshout, A J Meinders, N EMullaart-Jansen, S C van Nassau, J L Pedersen, R O Jensen, A Pronk, M R Kristensen,A M Ridge, T C Roeleveld, M C Schipper, D Vedder, J P van der Vorst, J Wachelder.
Correction notice This article has been corrected since it first published Online First. The collaborator list has been included,This paper has been amended since it was published Online First. Owing to a scripting error, some of the publisher names in the references were replaced with ’BMJ Publishing Group'. This only affected the full text version, not the PDF. We have since corrected these errors and the correct publishers have been inserted into the references.