Background
Transitions between different healthcare settings put patients’ safety at risk. Transitions between primary and secondary care include referral, discharge after a hospital admission and outpatient clinic visits. The number of such patient transitions is likely to increase worldwide, caused by an aging patient population with increasing prevalence of multimorbidity receiving care at home, with an increasing number of professionals involved in transitional care pathways [
1‐
3]. Indeed, there is a shift from secondary to primary care, triggered by a need to minimize healthcare costs without having to compromise on quality of care [
4,
5].
Every care transition is a potential threat to continuity and safety, with potential risks, near misses and adverse events [
6,
7]. Haggerty et al. divided the concept of continuity of care into three domains, namely managerial continuity, relational continuity and informational continuity [
7]. Several studies found inadequate documentation and lack of communication between different care providers (thus informational (dis)continuity) to be common during care transitions, whilst timely transfer of the medical information after discharge is considered an essential condition for maintaining patient safety during health care transfers [
8,
9].
Given the increase in polypharmacy in the last decade, one of the most challenging parts of exchanging patient information is to safeguard an accurate record of medication use of the patient at referral or discharge [
10,
11]. Healthcare providers are often not fully updated with their patient’s actual medication use [
12,
13]. For example, 27% of drugs that were stopped in the hospital because of adverse drug events were found to be re-prescribed after discharge by the primary care provider (PCP) [
14]. Burnett et al. found a reliability of 81–87% for the several clinical systems studied (such as available outpatient information), of which 20% of reliability failures were associated with a potential risk of harm. [
15] A gap in information is likely to cause medication errors, with the potential to cause patient harm and to result in additional healthcare costs [
16]. Yet, knowledge on the frequency of breaches in the continuity of information on changes in patients’ medication prescriptions between primary and secondary care is scarce [
17]. One reason for this is that until now in The Netherlands the electronic medical records (EMRs) of both settings mostly are not linked.
In the Netherlands, every citizen is listed with a PCP who serves as gatekeeper to the next levels of healthcare. The PCP, a trained family physician working a private general practice in the community, refers patients who need diagnostics or therapy that can solely be provided by medical specialists, to hospitals. Medical specialists see these patients in outpatient specialty care settings or after (acute) hospital admission. When a patient is discharged, or after visiting the outpatient specialty care, changes in treatment plans need to be adequately synchronized with the next care provider, −in the Netherlands always the PCP-, who provides concurrent care or follow-up and is generally the coordinator of care [
18]. The in- and outpatient settings of the hospital work in same EMR, however, PCPs work independently from the hospital, cannot access this record and use different EMR systems. Transitional communication occurs mostly digitally, but also through paper mail and telephone.
So although in-hospital changes (during outpatient clinic visit, hospital admission or visit to the emergency department) should be communicated, letters are not always written or sent timely by the medical specialist, nor correctly received and processed by the PCP. Furthermore, patients may collect their new or changed prescriptions as prescribed by the medical specialist at the community pharmacist, but also at the hospital pharmacist. Only the community pharmacist is connected to the PCP’s record, and therefore medication prescriptions handled by the hospital pharmacist will not be transferred to the PCP record. These sub optimally connected local processes impede tracking the full medication journey from the in-hospital prescription change to actual registration and continuation in the primary care setting and thus impede assessing current performance of settings, as well as the possibility to measure effects of interventions.
Therefore, we aimed for determining the frequency of no or incorrect documentation of in-hospital patients’ prescription changes in their PCP’s medical record, by reviewing individual patient data in a research database with linked primary and secondary care EMRs.
Methods
Study design
A retrospective observational study was performed, using linked EMRs from one academic hospital and 40 of their primary care practices. This was a sub study of the Transitional Incident Prevention Program (TIPP) study. The TIPP study was designed to develop and test a program for enhancing patient safety during healthcare transitions between hospital and primary care. [
19]
Setting and patient selection
For this study we created a research database in which we linked patients’ primary and secondary care EMRs.
We conducted the current medical record study using data from 2013 at the cardiology and gastroenterology department of the University Medical Center Utrecht, a 1042-bed academic hospital, situated in Utrecht, a city in the Netherlands, and referring primary care practices affiliated with the Julius Huisartsen Network (Dutch network of PCPs participating in research). This linked record included the content of the EMR of both the PCPs (Promedico, Medicom) and hospital (Chipsoft); we used the contacts with the healthcare provider (i.e. free text fields), correspondence of the medical specialist to the PCP (e.g. discharge letters or outpatient clinic follow-up letters), and the prescription overviews. Inclusion criteria were adult patients who had transitioned from the hospital to their PCP or vice versa in 2013, e.g. who had been discharged from the two hospital departments (cardiology or gastroenterology) or had had contact with the hospital outpatient clinic or emergency room; and they were registered as a patient at one of our affiliated PCPs. Records were excluded when patients did not experience a care transition upon second look, or the medical record was empty. We randomly sampled 600 (43%) of the 1399 eligible patients for analyses. To secure data transfer in terms of privacy, selection, linking and pseudononymizing of the medical record data was all conducted by a trusted third party (ZorgTTP Houten, the Netherlands; a company specialized in secure linkage and pseudonimisation of data).
Data collection
Transitions
Data was collected regarding age, sex and department that patients visited and the number of transitions. Transitions were categorized into a ‘hospital admissions’ group (at least 1 night in the hospital) and a ‘short hospital contacts’ group. The ‘short hospital contacts’ group consisted of three categories: (1) (telephone) contacts with the medical specialist of the outpatient clinic, (2) emergency room visits, or (3) short-stay hospital admissions (<1 day) in which medication was self-managed during the stay. Intercollegiate consultations were excluded as these did not involve patient transitions. If an outpatient clinic visit or emergency room visit was followed by a hospital admission, this visit was not counted separately, as then prescription changes would only be handed over to the PCP after discharge.
Prescription changes
Subsequently, every transition was separately assessed for changes in prescriptions. We defined changes as any alteration in medication during hospital admission or outpatient clinic visit, and divided these changes into the following 4 types also used by Uitvlugt et al.: prescription of new medication (‘Start’), termination of medication (‘Stop’), change in dose or frequency of a medication was changed (‘Change in frequency/dose’) or a change from one medication to another medication in the same medication group (‘Switch’) [
17]. Changes in dermatological preparations, drugs prescribed for a short period (e.g. antibiotics) and over-the-counter medication were excluded, because these changes typically are more likely not to be communicated to the PCP, noted in the prescription overview of the PCP; knowing this makes interpretation difficult when these prescriptions are not noted. The type of changes, generic name, dosage and corresponding medication group was labelled according to the most commonly used medication database in the Netherlands [
20].
Documentation of prescription change in record of PCP
Subsequently, we evaluated whether the in-hospital prescription changes were documented in the medical record of the PCP, either in the prescription overview or in the free text fields of the medical record. We had data available of every patient for the whole of 2013. Prescription changes should ideally be documented in the record of the PCP the next day after discharge, however, we checked up to three months after the change made in the record of the medical specialist. For each prescription change we calculated the number of days between discharge or outpatient clinic visit and entry in the medical record of the PCP. If no (correct) documentation of the prescription change was found, possible explanations (such as a temporarily change) were explored by further studying the linked medical records.
Outcomes
The primary outcome was the percentage of in-hospital prescription changes that was not or incorrectly documented in the medical record of the PCP. We distinguished three different types of documentation regarding prescription changes. First, when a prescription change (in terms of name, dosage and frequency) was present and correctly noted in the prescription overview of the PCP and also correctly noted in the free text fields, this was classified as “Correct documentation”. Second, an absent or incorrect notation in the prescription overview, but with a correct notation in the free text field was classified as “Inadequate documentation”, as well as correct notation in the prescription overview but incorrect notation in the free text fields: it is better than no documentation at all, but still concerns a discontinuity in information or even incorrect information. Since in daily practice the prescription overview and free text fields both are used by the practice nurse or primary care provider, this presents a potential risk for patient safety. Last, an absent or incorrect notation in both the prescription overview and free text fields was classified as “No documentation”; the latter type corresponded to our primary outcome. An example of this primary outcome was when a prescription was stopped by a cardiologist, but still prescribed later on by the PCP.
Secondary outcome was the time lag between prescription change and documentation in the medical record of the PCP, which for safety reasons ideally would be implemented in the prescription overview of the PCP within one day. We calculated the median number of days and benchmarked the time lag against the cut-off point of one day, two weeks (the duration for which new medications normally are prescribed at the pharmacy in The Netherlands) and three months (the maximum duration for which medications are prescribed). Subsequently, the number and characteristics of transitions, prescription changes and medication groups involved in prescription changes were assessed, as well as the age and sex of patients.
Statistical analysis
Continuous variables are presented as means (± standard deviation, SD) or medians (interquartile ranges, IQR) when data was not normally distributed, while categorical variables are presented as numbers (percentage). All data were analyzed using IBM SPSS Statistics version 21.
Discussion
We reviewed the transitional documentation of prescription changes in patients’ medical records, and found that one third of all in-hospital prescription changes were not documented in patient’s medical record of the PCP. Of these, in 80% of cases we could not trace a satisfying underlying reason (e.g. temporarily given medication) in the medical record of the medical specialist. Changes mostly concerned the start of a prescription in 53.9% (total changes) and 47.6% (not documented changes). When prescription changes were documented correctly, they were found after a median of three days (IQR 0–18.3) in the PCP’s patient records. The types start and switch were more often documented timely than the type dose or frequency change.
Our current findings are grossly in line with other studies. Another Dutch study showed a breakdown in information in each step of the discharge communication process, with only 50% of all prescription changes correctly documented in PCPs’ overviews two weeks after discharge. [
17] This higher rate of incorrect documentation might reflect the shorter follow-up of two weeks, the use of medication reconciliation with the patient (including over-the-counter medication), and the fact that only discharge was assessed (and not also the outpatient clinic transition).
Despite Dutch multidisciplinary guidelines on information transfer from hospital to PCP and on transfer of medication, the criterion of informing the next healthcare provider within 24 h was not fulfilled [
21,
22]. Two small surveys showed that 25% of the PCPs did not receive any communication from the hospital within 4 weeks after visit to the outpatient clinic [
23,
24]. Bell et al. found that for 58% of all discharge summaries the goal of sending the letters within two weeks was not reached [
25]. Kripalani et al. reported that 51–77% of the discharge summaries is not received by the PCP after four weeks, affecting the quality of care in 25% of follow-up visits [
26]. However, this research concerns the deliverance of information from the hospital, and not the subsequent registration of this information in the PCP’s medical record, as assessed by us.
Problems with (timely) deliverance of sufficient information on the part of the hospital are partially responsible for the found high level of undocumented prescription changes. Undoubtedly it is not the only weak link in the process. The ability of the PCP to correctly process prescription changes to the records may be influenced by the amount of time available for administrative tasks, the information processing system and the level of collaboration with the community pharmacist [
27,
28]. We found a relatively high number of documentation of prescription changes within one day. This is most likely due to automated documentation by the pharmacist’s system that currently in the Netherlands often is connected with the PCP’s. A comprehensive and reliable system implementing all prescription changes directly into the medical record of the PCP and/or pharmacist would be a definite solution.
Notably, we found unexpected differences between prescription changes made by the gastroenterologist and the cardiologist, as well as the fact that younger patients seem to be more at risk for absent documentation. A possible explanation may be that many older patients have automated dispensing systems (Baxter), providing the patient with the correct medications on each moment of drug intake during the day, and which are closely monitored by ‘own’ pharmacists. In contrast, younger patients could pick up their medication anywhere without communication to the PCP or their regular pharmacist. However, this finding should be interpreted cautiously, as statistically testing and correcting for interactions was not feasible in our sample.
Our study has several strengths. In contrast to other record studies, we studied the documentation of prescription changes in linked primary-secondary care medical records. Combining patients’ primary and secondary care EMRs provided us with a more complete picture of the implementation of prescription changes on the receiving side. Furthermore, we looked at all care transitions; not only hospital discharge but also outpatient clinic visits, emergency room visits and short stay hospital admissions. Third, we looked at the prescription overview in the medical record as well as free text fields of the medical record of the PCP, providing us optimal information of the medication. Finally, when documentation was absent or incorrect, we tried to find an explanation in the medical record, improving clinical accuracy of interpretation of our results.
Yet, our study also has certain limitations. First, since this concerns a retrospective study, we could often not establish the intention of a prescription change by the medical specialist in the hospital’s medical record. However, this does still not diminish the found results of undocumented prescription changes. Second, adverse drug events such as allergies and serious side effects were not included in our outcome, since data was not available to assess this. This would have made our outcome even more clinically relevant. Third, certain patient characteristics (e.g. educational level or presence of comorbidity) were unavailable, which limited our options to explore which patients were more vulnerable for insufficient transfer of information. Last, this concerned a single center study, decreasing generalizability.
When classifying documentation as in this study, it is important to keep in mind that insufficient quality of documentation is not necessarily insufficient quality of care. An individual case in which a prescription change was made due to an important clinical reason (e.g. a drug allergy) is handled equally as a prescription change that has a lower level of clinical importance (e.g. start of a laxative). Also, the pharmacist or the PCP often intercepts deficits in documentation before and corrects the missing information ‘outside the registration’ before any harm can be done to the patient. Results should therefore not be interpreted as a direct reflection of medical errors or poor quality of care, but as a reflection of documentation quality.
Various literature reviews demonstrate that discharge information is often insufficient or incomplete. A recent review including 19 studies (2007–2014) showed that 21% of the discharge summaries did not include post-discharge medication [
29]. This might partly explain why these changes do not end up in the PCP’s medical record. Moreover, doctors learn to deal with poor documentation on a daily basis and are well capable of preventing patient harm caused by those documentation flaws in most cases. This level of resilience probably is the main explanation as to why outcomes on documentation differ from outcomes on medical errors and adverse events [
30]. However, we do hypothesize that this informational discontinuity will have its effect on patient safety. The more flaws in the healthcare process occur, the higher the risk for fatal slips [
31]. Undocumented prescription changes unnecessarily increase the risk in the patient journey. Unique in this study is the use of a linked EMR, providing a more complete view on this patient journey between different healthcare settings, which was used to measure the extent of the problem, as well as trying to find reasons for undocumented prescription changes. We did provide insight in the extent of the problem, even though causes for omission were mostly not found. Both outcomes are a start for improvement strategies, in order to decrease the risk for harm for patients. The ultimate solution, which would be a single shared EMR through all levels of care, is up until now not within sight in the Netherlands.