Adverse events occur after one in five medical discharges, and contribute to readmission. About half of such events are preventable or ameliorable and usually involve communication issues, often related to medication management.1

Failure to convey accurate, complete and up-to-date information across interfaces in care is a major, avoidable risk to patient safety, yet improving care at hospital discharge has recently been described as an “unmet challenge.”2 The National Service Framework for Older People3 stated that “the current emphasis on providing more care in the community requires better communication than ever between health professionals.” Yet the Royal College of Physicians has recognised “serious problems with the validity of clinical information in interim discharge documents that may affect patient care … resource management (and) performance indicators.”4

Recent changes in UK working practice have contributed to communication problems. Following introduction of the new UK General Practitioner (GP) contract, patients needing out-of-hours care are less likely to see a familiar doctor. In secondary care, increasing specialisation and the European Working Time Directive have led to fragmentation of care and more frequent handovers. These changes generate potential “gaps” in communication about a patient’s present and previous clinical condition.5 While an individual gap may appear insignificant, the cumulative “loss” of information across several gaps can lead to serious adverse events.6 Recognition of such “gaps” is essential to avert predictable adverse events.78 Where handover was once regarded as a poor substitute for continuity of care,9 continuity now depends on meaningful handover.

Our study was set up as part of an Integrated Service Improvement Programme (ISIP) at Nottingham City Hospital, UK, to provide baseline information for the development of an electronic discharge document. Currently, as in other UK centres, a handwritten interim TTO (To Take Out) form incorporating the discharge prescription is either given to the patient or posted to the GP on discharge. A detailed, typed letter often follows later.

Insufficient or inaccurate discharge information can adversely affect patient care,10^–13 and a recent study has shown a relationship between discrepancies in discharge medication and readmission.14 By studying a group of patients likely to have predictable problems following discharge, we aimed to (1) identify communication gaps at hospital discharge for older people who are readmitted within 28 days; (2) assess the contribution of incomplete discharge information to readmission; and (3) identify information that should be included in electronic discharge documents.

METHODS

Selection of cases

The hospital information bureau provided a list of 141 consecutive patients aged 75 and over who were readmitted within 28 days of a previous discharge (1 January to 13 February 2004). Thirty-three (23%) were not emergency readmissions (for example, planned gastroscopy appointment) and were excluded from the review. Medical records were missing for one patient. For the remaining 108 patients, the records were reviewed by EW after the readmission.

Recording of data

On the basis of the Royal College of Physicians draft standards for discharge information,15 we developed a form which EW used to record whether or not certain items of information were recorded in the inpatient notes, the discharge letter or the TTO form. The main elements of this assessment are shown in table 1. Data were entered into a database using an electronic scanning system (Teleform© Form Recognition Ltd), and 20% of data were double entered onto a separate database to enable a manual check of the accuracy of the scanning system. Where drug-related problems were identified, case summaries were made for later review to assess their contribution to readmission.

Analysis of data

SPSS was used for statistical analysis of the information content of discharge documents. The chi-squared test was used to assess the significance of differences in information recorded on the TTO form and discharge letter.

For drug-related cases, two reviewers (EW and OP) independently used explicit criteria for causality,16 preventability17 and contribution of drugs to the readmission, as described in a previous study,18 and outlined in the tables 2–4. Where there was disagreement, a third reviewer acted as moderator (AA), in keeping with accepted practice, and a majority decision was taken for classification. The proportion of patients considered to have a drug-related readmission was calculated as the percentage of patients who were classified as having a definite or probable drug-related morbidity that made a dominant or partial contribution to the readmission. The proportion of patients considered to have a preventable drug-related readmission was assessed using the criteria for preventability shown in the box.

All four criteria must be fulfilled to confirm preventability.

RESULTS

Patient characteristics at the preceding admission

There were 58 male and 50 female patients, and the median age was 80 years (range 75–98). The most frequent primary reasons for admission were respiratory disorder (18), surgical procedure (14), shortness of breath (12), falls/poor mobility (11) and chest pain (11). Inpatient notes recorded a high level of underlying chronic disease including coronary heart disease (29 (27%)), malignant disease (24 (22%)), chronic obstructive pulmonary disease (23 (21%)) and poor mobility (21 (19%)). “End-stage disease” was recorded for seven of 18 patients with metastatic disease and for two patients with non-cancer disease.

The mean number of drugs per patient at discharge was 7 (range 0–18), and chronic cognitive impairment was documented for 36 (33%) patients.

Readmissions

Thirty patients (28%) returned within 3 days of discharge and 48 (44%) within 7 days. The median length of stay for the readmission was 10 days (range 1–83). Most patients (90 (83%)) returned with symptoms related to their previous admission. At readmission, 67 (62%) patients had no discharge letter available (in 25 of these cases, the letter was waiting to be typed). Twenty-eight patients (26%) died during the readmission.

Discharge documents

TTO forms were available for 104 (96%) patients. Although “author” (the person writing the form) was recorded on 100 (96%) forms, only 22 (21%) authors’ names were legible, and contact details were provided for only 30 (28%). Discharge letters were available for 66 (61%) patients.

Table 1 shows the extent to which items of information were available in the inpatient notes, TTO form and discharge letter. It can be seen that for many information items, the discharge letter gave a more complete account of the episode than the TTO form. In contrast, the TTO form was more likely to contain a list of medicines along with dosages and frequencies of administration.

Medication management information

Some prescribed medication was missing from 19 (18%) TTO forms and 14 (21%) discharge letters. Where drugs were missing from the TTO form, the same drugs were usually missing from the discharge letter. Medication changes were made during the preceding admission for 93 (86%) patients, and 67 (62%) had an increase in numbers of medications. Often, it was impossible to tell from the TTO form whether drugs listed on admission, but not on discharge, had been stopped deliberately or unintentionally omitted. This information was missing for 17/26 (65%) of patients for whom only one drug was stopped, and 17/20 (85%) of those for whom more than one drug was stopped.

Although 74 (71%) TTO forms listed medication changes that required monitoring, such as new drugs or dose changes within 24 h of discharge, only 12/74 (16%) gave any monitoring advice. Such advice was often incomplete: one TTO form stated “started amiodarone—please check TFTs in 6 months” but gave no indication that other cardiovascular drugs had been changed shortly before discharge and that blood pressure was falling. Another requested monitoring of urea and electrolytes but did not indicate that the haemoglobin was already falling following introduction of low-dose aspirin.

Preventable readmission

Many patients had several problems after discharge that contributed to readmission, including inadequate support service provision for 40 (37%) patients and/or carers. Detailed assessment was required to determine which of several problems tipped the balance. One patient for example, whose support services were delayed, was readmitted after 8 days with drug-induced diarrhoea. Another patient with mobility problems was readmitted with a urinary-tract infection: the GP had not been informed of the positive MSU result from the previous admission. Medication management emerged, however, as a major preventable problem, whether or not it resulted in readmission. Sometimes, missing medication management information was one of several reasons, but for others it was the dominant reason, for readmission.

Medication-related readmission

Readmission was judged to be medication-related for 41 (38%) patients (see table 5) and preventable for 25 (61%) of these. There were communication gaps, including monitoring advice, for 22 (54%) of the 41. For 23 (21%) patients, medication was the dominant reason for readmission, and 19 (83%) of these were preventable. Where incomplete medication discharge information was the dominant reason, 11 (92%) of the 12 cases were judged to be preventable. This represents 10% of the total sample.

Twenty-four (59%) of the 41 patients had more than one medication-related problem, most commonly drugs missing from the TTO form (six patients) which resulted in the patients not receiving these medications in the community.

Another 43 (40%) patients who were readmitted for other reasons also had medication problems following discharge.

Types of medication-related problems

Medication related problems included deterioration following changes in cardiovascular medication for nine patients, failure to recognise poor compliance for seven and insufficient analgesia for three. The most common problems arose from recently changed NSAIDs, associated with bleeding in 11 cases. For six patients (two of whom had started low-dose aspirin alone), this was the reason for readmission (see table 5). For three patients started on aspirin, two of whom bled, a past history of peptic ulcer was omitted. In another case, aspirin and diclofenac, which were written up on the inpatient drug chart, but had not been administered, were erroneously added to the TTO form by a pharmacist in the absence of any indication to the contrary in the notes. This patient died from gastrointestinal bleeding on readmission.

DISCUSSION

We found that information important for safe continuity of care was often missing at discharge, and medication problems following discharge were common. Accurate information on changes in medication was missing from two-thirds of all discharge documents. At review, 38% of readmissions were considered to be medication-related, and 61% of these were preventable. Where medication management information was the dominant reason for readmission, 92% of cases were considered preventable.

Strengths and limitations of the study

Many patients were at high risk of readmission regardless of any preventable communication-related adverse event, and many problems arose in the community, beyond the influence of secondary care. Many medication-related problems were however both predictable and preventable. Despite a high (33%) level of cognitive impairment, for example, adherence problems were rarely recognised as a high risk requiring closer monitoring.

Although this study demonstrates many gaps in discharge information, we could not establish a causal link between missing information and adverse events or readmission. Judgements about the contribution of medication to readmission were based on information available in case notes. Other relevant details were unknown—for example, the accuracy of admission drug information from primary care, or how discharge drug information was processed in GP surgeries.

Correction of some drug-related problems may only have delayed a readmission that was inevitable for other reasons. Many readmissions are multifactorial, but this study shows that inadequate prescribing information plays an important and remediable part. In some cases, early return, for example with oedema, revealed another unrelated, but more serious problem, such as gastrointestinal bleeding. Both problems could have been ameliorated with better handover at discharge. In other cases, readmission for other reasons such as poor mobility led to a timely correction of adherence problems.

To minimise the risk of misclassifying cases, explicit criteria were used to assess whether the readmission was related to medication and whether it was preventable.1617 We included only cases where there was strong evidence in the case-notes that the readmission was definitely or probably related to medication, based on independent scoring by two observers. As a result, the figures may represent an underestimate of the actual number of medication-related cases.

Implications for clinical practice: filling the gaps

Adverse drug events contribute to about 6.5% of hospital admissions, and around two-thirds are preventable.18 Preventable drug-related admission rates are even higher in readmission studies,19 and two-thirds of potentially hazardous/contraindicated drug combinations in patients in general practice may involve medication initiated by hospital doctors.20

More serious adverse effects, for example gastrointestinal bleeding, are more often preventable, yet less likely to be recognised.21^–23

Although electronic methods expedite communication, safe continuity of care depends on accurate and complete information provision.

This cannot be regarded as an “add-on” task for junior doctors or pharmacists with little or no knowledge of the patient: it was often impossible to discern from the TTO form how ill a patient had been, or which of several altered drugs needed most urgent follow-up. Even when a discharge letter was available, however, prescribing information was often limited. Our study suggests that safe medication management at the interface between secondary and primary care requires considerable improvements in communication. Changes that could reduce risks include systematic recording of changes to medication, and the reasons for these. A predischarge review could detect early adverse effects of altered medication and inform monitoring recommendations. GPs have consistently requested more discharge information,11^–13 but must also ensure that accurate information is available on admission. While pharmacists provide invaluable support, responsibility rests with prescribers on both sides of the interface. A higher index of suspicion is needed for medication-related problems.

We suggest that discharge documents should be regarded no longer as a “record of events” but as a “referral” back to primary care, with adequate information for immediate continuity of care.

Further research

Further research is needed to help develop and evaluate interventions designed to improve communication on hospital discharge and to reduce the risks of preventable readmission.

CONCLUSIONS

Incomplete discharge documentation was common, on both TTO forms and discharge letters. Two-thirds of patients were discharged with incomplete documentation of changes to medication. Three-quarters of patients had medication problems following discharge and over half of 41 medication related readmissions were potentially preventable with better medication management information.