We found 64/1237 (5.2%) of admissions to be caused by MRH. In general, this MRH was fairly well documented in all sources examined within the complete medical record. However, failure to transfer information from inpatients’ medical records to discharge summaries was identified, and there was no one place within the paper medical record where complete MRH information was consistently found. Finally, less than one in ten patients with a confirmed MRH-related diagnosis had this reflected in their ICD-10 codes.
Interpretation and implications for practice
The A&E triage data source contained a symptom statement in 93% cases. A substantially higher percentage of A&E clerking documents included statements of diagnosis, identification of the causative agent, and action or plan than A&E triage.
The one-page PTWR form should capture concise, specific information about a patient’s diagnosis and management plan. Documentation of all potential MRH statements was found in at least 65% of the PTWR data sources. The percentage of implicit statements within this source was relatively high (23% of all cases). However, the PTWR doctors’ time–pressure [
21] may be reflected in the use of the single-word or very short statements identified, perhaps assuming the next reader will link together the information on the PTWR form with that mentioned elsewhere within the patient’s medical record [
22].
Of interest was the contribution the contents of the drug chart made to increasing overall documentation in the 3 pm data source. This suggests that the drug chart is not only used as a record of prescribed and administered drugs, but is sometimes used to communicate clinical information which may not be found elsewhere. As electronic prescribing systems become more widespread, it will be important to consider whether these systems can also be used to document this additional contextual information.
In total, information regarding 14 statements was not transferred from the paper medical record to the EDC; conversely, seven additional statements were found on the EDC and not in the paper medical record. The presence of additional statements on the EDC may reflect new information discovered during the patient’s stay. The data collection method for this study focussed primarily on the first two or three days after admission, and at discharge. Information regarding a patient’s condition which was established after first ward-transfer and documented in the paper medical notes before discharge was not collected. However, most of the information presented at discharge was identified during the first two or three days of patients’ hospital stays.
One aspect of MRH where a degradation of information seemed to occur was illustrated by the decrease in percentage availability of symptom and diagnosis statements between paper medical record and the EDC: there were six instances of symptom information not being transferred to EDCs, and four instances of diagnosis information not being transferred. This failure to transfer information may be a reflection of the way the information is spread throughout the inpatient record making comprehensive transfer of information difficult. Patients’ medical records are often large and comprise many pages. To provide an accurate discharge summary, the discharging doctor must condense and communicate information dispersed throughout the entire paper medical record: they must be able to identify the most recent and accurate data entries on which to base their discharge summary. It may be unreasonable to expect a discharging doctor to review every patient’s entire medical record and extract all relevant information. Electronic records may allow storage of clinically important information in easier-to-navigate patient chronologies, which would allow ready access to the most up-to-date information. However, given current variation in extent and approach to electronic health record system use within the UK, more work is likely to be needed to realise these potential benefits [
23].
Four or five different data sources were examined in the paper medical record compared with one at discharge; therefore poor documentation within individual sources in the paper medical record does not necessarily translate into poor cumulative availability when patients’ paper medical records are considered as a whole.
Overall, our findings suggest that while key information is generally transferred onto the EDC, important information is sometimes lost. As failure to pass on information has been identified as a patient care problem [
3,
11‐
14] there may be potential to avoid injury at minimal cost to the NHS by encouraging and facilitating better communication of the information which already exists in patients’ medical records.
Regarding the structure of the EDC specifically, there were instances when the statement relating the causative agent to the symptom or diagnosis was in an inappropriate section of the EDC. The person who made the entry was not always stated, but some entries were specifically attributable to pharmacists. For instance, relevant information was sometimes found in the “details of information booklets and other information given with medicines” section or the “additional notes” section where pharmacists can add additional comments. The pharmacist may add valuable medicine-related information to the EDC but the next reader may not be alerted to this information because it is printed in an unexpected section of the EDC. Future EDC systems must allow both doctors and pharmacists to make entries in a section of the EDC dedicated to medication changes, and to enable easy documentation of MRH.
The role of patient adherence as a contributing factor to admission was sometimes poorly communicated on the EDC; in one instance there was no indication that a patient admitted following a collapse had mistakenly taken their anti-hypertensive medicines twice. Adherence problems commonly fall to primary care providers to manage; it is therefore crucial that this be communicated to GPs and to community pharmacists wherever possible. However, as the patient is discharged with a copy of the EDC, the discharging doctor may be reluctant to apportion them blame and possibly cause offence, or wary of making statements based on potentially inaccurate assumptions in the medical records. Discharging doctors are unlikely to have admitted the patient, and may not have built a sufficient relationship with the patient to document culpability.
ICDs related to MRH were rarely used. Although we studied a small sample of patients from one organisation, the lack of MRH-related coding was notable. Both health care planning and research are underpinned by good quality evidence based on accurate data. Since coders use patients’ complete medical records as the primary source of information for coding, the accuracy of coding is likely to reflect the clarity of that medical record. Although MRH information was present in most cases, it was often in a format which needed interpretation, and was spread throughout the entire record. The EDC often provides a more interpretable (81% explicit statements) and accessible summary, and use of unambiguous statements (e.g. drug induced neutropenia) may facilitate assignment of MRH-related ICD codes.
Implications for research
Investigation into the reasons for the ambiguousness of some information on the PTWR form data source may enable improvement of the PTWR form, or contribute to the restructuring of the PTWR. Future research should also explore how GPs, community pharmacists, patients and other recipients interpret MRH-related information on patients’ discharge communications and how this information is acted upon. The reasons for failing to document adherence problems should also be explored, as should the reasons why doctors use implicit or short statements within the medical record, and whether use of explicit or implicit statements is related to the probability that the admission was MRH-related.
Any work reporting a rate of MRH-related admissions calculated only using routinely collected ICD-10 codes is likely to underestimate the true occurrence. Exploring the reasons why ICD-10 codes are not used for MRH-related admissions should be a next step.
Comparison with existing literature
The 5.2% (64 of 1,237) of patients admitted and diagnosed with MRH is in line with previous UK studies [
4‐
6] using similar data collection methods, but focussing only on admissions caused by ADRs and poor adherence, which report 5.3-8.0%, and with international figures [
9,
24]. We also included admissions due to errors but these were responsible for only five admissions. We also found the causative agents to be similar to those reported elsewhere [
25].
The ICD codes attributed to each hospital admission have previously been used to identify ADR-related admissions [
26,
27], while others suggest that using ICD codes underestimates their true occurrence [
28,
29].
In our study a total of 6 (0.5%; 95% confidence interval 0.1-0.9%) of 1237 admissions were found to have one or more MRH-related ICD-10 codes. Wu et al. [
26] found 557,978 of 59,718,694 (0.9%) of admissions over a ten year period to have an ADR-related ICD-10 code. When Wu et al.’s inclusion criteria are applied to the present sample, only 3 (0.2%; 95% confidence interval 0.0-0.5%) of our 1237 admissions had an ICD-10 code from the selection of ICD codes used by Wu et al. It is not clear why the proportion of admissions coded as being due to ADRs was lower in our study, but may be due to our sample being much smaller and from a more specific admission pathway in just one hospital.
Strengths and limitations
As well as being the first study to explore these issues, a particular strength of our study is that MRH-related admissions were detected during the PTWR, when each patient is examined by a multi-disciplinary team with great experience in detecting causes of admission, thus the diagnosis of MRH reflects the opinion of a team, rather than a single individual.
Limitations include the fact that the contents of the patient’s handwritten case notes were only examined up to the time of first ward transfer (typically two or three days after admission). As patients in our sample stayed for up to 65 days, it is unsurprising that extra information was found on the discharge summary when compared with those data sources examined in the first few days of admission. Furthermore, we were not able to capture information from every data source for every patient, reducing the completeness of the data set. Data collection took place between October and June and so any seasonal variation may not be fully represented.
The inter-rater reliability testing performed for categorising the presence of MRH-related information showed only ‘fair’ agreement, and our sample was relatively small. We did not explore whether agreement was affected by the profession of the assessor. Finally, while we recorded whether or not MRH was predominantly caused by error we did not assess the preventability of other types of MRH detected.