Skip to content

Advertisement

BMC Medical Informatics and Decision Making

What do you think about BMC? Take part in

Open Access
Open Peer Review

This article has Open Peer Review reports available.

How does Open Peer Review work?

Adverse events recording in electronic health record systems in primary care

  • Sabine E. M. de Hoon1Email author,
  • Karin Hek1,
  • Liset van Dijk1 and
  • Robert A. Verheij1
BMC Medical Informatics and Decision MakingBMC series – open, inclusive and trusted201717:163

https://doi.org/10.1186/s12911-017-0565-7

Received: 31 March 2017

Accepted: 22 November 2017

Published: 6 December 2017

Abstract

Background

Adequate record keeping of medication adverse events in electronic health records systems is important for patient safety. Events that remain unrecorded cannot be communicated from one health professional to another. In the absence of a gold standard, we investigate the variation between Dutch general practices in the extent to which they record medication adverse events.

Methods

Data were derived from electronic health records (EHR) of Dutch general practices participating in NIVEL Primary Care Database (NIVEL-PCD) in 2014, including 308 general practices with a total practice population of 1,256,049 listed patients. Medication adverse events were defined as recorded ICPC-code A85 (adverse effect medical agent). Between practice variation was studied using multilevel logistic regression analysis corrected for age, gender, number of different medicines prescriptions and number of chronic diseases.

Results

In 2014 there were 8330 patients with at least one medication adverse event recorded. This corresponds to 6.9 medication adverse events per 1000 patients and is higher for women, elderly, patients with polypharmacy and for patients with comorbidity. Corrected for these patient characteristics the median odds ratio (MOR = 1.92) suggests an almost twofold difference between general practices in recorded medication adverse events.

Conclusion

Our results suggest that improvement in terms of uniformity in recording medication adverse events is possible, preventing potential damage for patients. We suggest that creating a learning health system by individual practice feedback on the number of recordings of adverse events would help practitioners to improve their recording habits.

Keywords

Recorded medication adverse eventsBetween practice variationGeneral practicePatient safety

Background

Medication adverse events such as opioid-induced constipation and myalgia caused by statins, are important patient safety indicators and a priority topic according to the World Health Organization [1, 2]. In this paper a medication adverse event is defined as a response to a drug which is noxious and unintended, and which occurs at doses normally used in man for the prophylaxis, diagnosis, or therapy of disease, or for the modification of physiological function [3]. A recent review demonstrates a bidirectional causal link between medication adverse events and non-adherence to medication therapy [4]. Adverse events can therefore negatively affect the pharmacological treatment effect which can eventually lead to hospital admissions.

Patient awareness and notification of relevant medication adverse events as well as adequate patient-communication when medicines are prescribed can improve patient safety. A report of the Institute of Medicine states that poor exchange of medical information and communication between healthcare professionals are responsible for medication adverse events and harm patient safety [5]. Safety research in other industries, such as aviation, also show that lack of communication and teamwork, and miscommunication are factors that contribute to adverse events [6]. Adequate recording of medication adverse events in healthcare can help overcome this since recorded events can be communicated. Furthermore, it is a prerequisite for creating a learning health system, both at the level of society as well as at the level of health care practices.

Most patient safety research in healthcare focuses on hospitalized patients. However, in many countries primary care is the first point of contact between patients and the healthcare system. As such, most of the care is provided there and the general practitioner has a large share in the medication that is prescribed. Because of the gatekeeper role, general practitioners (GPs) play an important part in patient safety by adequately signaling and recognizing medication adverse events [2, 7, 8]. Subsequently, uniform recording these events in patient’s electronic health record (EHR) is important to monitor progress and to ensure that all responsible parties are aware of this possible safety hazard [9]. Routine EHRs can play an important role in achieving a learning health system [10], improving healthcare. At patient level, EHRs play a key role in communication since events that remain unrecorded will not be communicated. This communication is crucial when multiple health care providers are involved in patient care, especially since patient care becomes more complex [8, 11].

Other countries with a comparable primary care structure, such as the UK, already showed that EHR data can be used to identify medication adverse events [12] alongside for example national incident reporting systems. In this paper we investigate the extent to which medication adverse events are routinely recorded in Dutch general practice and the variation between practices. We control for patient characteristics that are assumed to be strongly associated with the actual occurrence of adverse events. We control for age as it is known that older patients experience more medication adverse events [12, 13]. Other patient characteristics such as gender, polypharmacy, and comorbidity are also included, as they have found to be risk factors for medication adverse events as well [13, 14]. Underlying assumption in this paper is that the larger part of the remaining between practice variation does not represent differences in the actual occurrence of adverse events, but differences in the recording of these events.

Methods

Study population

Data were derived from EHRs of Dutch general practices that participated in NIVEL Primary Care Database (NIVEL-PCD) in 2014 [15]. Like in several other European countries (e.g. UK, Denmark), GPs in the Netherlands have a fixed population list that can change every three months. Only general practices with complete data (at least 46 weeks of recording) were included. Patients with missing data on gender (n = 12) and age (n = 13) were excluded from the analyses and we could use data from 308 general practices with a total listed population of 1,256,024 individuals.

Measures

Symptoms and diagnoses are recorded routinely in general practice by practice nurses as well as GPs using the International Classification of Primary Care (ICPC) [16]. Medication adverse events should be recorded as patient contacts with ICPC-code A85 (adverse effect medical agent) in the EHR. Medication adverse events were calculated as the number of patients with at least one contact or episode of medication adverse event per 1000 listed patients. Medicines prescribed are recorded according to the Anatomical Therapeutic Chemical (ATC) classification system. The number of different medicines (unique ATC-codes) prescribed to an individual patient in 2014 was used as an indicator for polypharmacy and categorized in four categories (0, 1–4, 5–9 and ≥10 different medicines per patient). The number of chronic diseases in each individual patient was counted and categorized in four categories (0, 1, 2, and >2 chronic diseases per patient).

Statistical analysis

To study between practice variation in recorded medication adverse events, multilevel logistic regression was performed adjusting for patient characteristics. In this multilevel logistic model patients (first level) were nested within general practices (second level). We first fitted an empty model (model 0) which allowed us to detect a possible general practice effect. We then estimated a model including patient age (centered to the mean) and gender (model 1) followed by a model including the number of different medicines prescribed to a patient (model 2) and number of chronic diseases in this patient (model 3). Differences between software packages in completeness of recording have been shown in other electronic health record based studies [17]. There may be differences between software brands in the extent to which they help the general practitioner to recognize, acknowledge and record adverse events. Analyses of differences between software package brands showed considerable variation in recorded adverse events, ranging from 5.1 to 10.4 adverse events per 1000 listed patients, dependent on the software brand. Therefore, all models were adjusted for software package used by the general practice. Odds ratio’s (OR), 95% confidence intervals (95%-CI), and p-values were calculated. To describe between practice variation in recorded medication adverse events, we calculated practice variance components, median odds ratio’s (MOR) and intraclass coefficients (ICC) [18]. The practice variance component indicates the variance between practices. The ICC indicates the proportion of variance that is attributable to differences between general practices. In this study, MOR refers to the increased probability of a recorded medication adverse event between two randomly chosen practices. Furthermore, we reported the probability of medication adverse event recorded and the 95%-CI around this mean per general practice. In order to make these results more interpretable, all determinants were centered to their mean with the most common software package as reference. All analyses were conducted using STATA version 14.0.

Privacy

The study was carried out in accordance with Dutch legislation on privacy. According to Dutch legislation, neither obtaining informed consent nor approval by a medical ethics committee is obligatory for observational studies [19]. This study has been approved by the applicable governance bodies of NIVEL Primary Care Database under number NZR00315.055.

Results

Medication adverse events

Of the 1,256,024 patients aged between 0 and 112 years the average age was 40.7 years. The age and gender distribution of the patients in this study corresponds with the general Dutch population. As can be found in Table 1, 68.0% of the patients received different medicines and 38.7% had at least one chronic disease. This is comparable to the general Dutch population. There were 8330 patients with a least one recorded medication adverse event. This corresponds to 6.9 medication adverse events per 1000 listed patients. Recorded medication adverse events are twofold higher in women compared to men (9.4 and 4.3 per 1000 listed patients respectively) and gradually increases with age, except for 5–17 year olds having lower rates than younger children. Similarly, the number of recorded medication adverse events increases with the number of medicines prescribed and with the number of chronic diseases (Table 2).
Table 1

Characteristics of 1,256,049 patients in 308 general practices in the study

 

Study population n (%)

Gender

 Male

620,000 (49.4)

 Female

636,024 (50.6)

Age

 0–4

64,080 (5.1)

 5–17

188,189 (15.0)

 18–44

429,686 (34.2)

 45–64

353,843 (28.2)

 65–74

125,886 (10.0)

 75–84

68,175 (5.4)

 85+

26,165 (2.1)

Number of different medicines prescribed

 0

401,783 (32.0)

 1–4

559,774 (44.6)

 5–9

201,738 (16.1)

  ≥ 10

92,731 (7.4)

Number of chronic diseases

 0

769,604 (61.3)

 1

280,156 (22.3)

 2

111,195 (8.9)

  > 2

95,069 (7.6)

Table 2

Number of patients with medication adverse events recorded, by patient group

 

N

Per 1000 patients

Gender

 Male

2584

4.3

 Female

5746

9.4

Age

 0–4

132

2.3

 5–17

269

1.5

 18–44

2085

5.1

 45–64

2321

6.7

 65–74

1671

13.5

 75–84

1315

19.9

 85+

537

21.9

Number of different medicines prescribed

 0

148

0.4

 1–4

2174

4.0

 5–9

2888

14.5

  ≥ 10

3120

34.4

Number of chronic diseases

 0

2289

3.1

 1

2039

7.5

 2

1564

14.4

  > 2

2438

26.5

Between practice variation

Table 3 presents the results of the models analyzing the probability of recording a medication adverse event using ICPC-code A85. The empty model shows the probability that a general practice records a medication adverse event as A85 is 0.93% (95%-CI: 0.24–3.55%). The between practice variance was 0.48 (standard errors (SE) = 0.05) which corresponds to an ICC of 0.129. This means that 12.9% of the variability in recorded medication adverse events can be attributed to differences between practices (and not individuals). When taking into account patient characteristics, similar practice variances with corresponding ICCs are observed. Practice variance is 0.47 (SE = 0.05) when adjusted for age and gender (model 1) and 0.46 (SE = 0.05) when number of different medicines was added to this model (model 2). This suggests that variability in recorded medication adverse events can still be attributed to between practices differences. Due to the high correlation between number of different medicines prescribed and chronic diseases (r = 0.58), adding the latter to this model did not improve the model (practice variance of 0.46, ICC = 12.2%, and MOR = 1.91). The number of medicines prescribed was most strongly associated with medication adverse events. The median odds ratio calculated from the practice variances is fairly high in the empty model (MOR = 1.94) and remains equally high after adjustment (MOR = 1.92). This means that after accounting for patient characteristics, the probability of a recorded medication adverse event is still almost two times higher in one practice compared to another. This variability is also presented in Fig. 1.
Table 3

OR and 95%-CI for a medication adverse event recorded for an patient within a practice

 

Empty model

Model 1a

Model 2b

Fixed effects

Patient characteristics

Age

 

1.03 (1.03–1.03)c

1.00 (1.00–1.01)

Gender

 Male

 

1

1

 Female

 

2.07 (1.97–2.17)

1.64 (1.57–1.72)

Number of different medicines prescribed

 0

  

1

 1–4

  

9.93 (8.40–11.73)

 5–9

  

34.74 (29.37–41.09)

 ≥10

  

81.22 (68.49–96.33)

General effects at practice level

Practice variance (SE)

0.48 (0.05)

0.47 (0.05)

0.46 (0.05)

MOR

1.94

1.92

1.92

ICC

0.129

0.124

0.124

SE standard error, MOR Median Odds Ratio, ICC intraclass correlation coefficient

amodel 1 adjusted for age and gender, bmodel 2 adjusted for age, gender, number of different medicines prescribed. All models were adjusted for software package used by the general practice

crepresented as Odds Ratio (95% confidence interval)

Fig. 1

Adjusted probability of a medication adverse event recorded in a patients’ EHR, per practice. Adjusted for age, gender, number of different medicines and software package (model 2). Variables were centered around their mean and the most common software package served as reference. Each dot represents a general practice. The error bars represent the 95% confidence interval around the estimate of that practice

Discussion

We investigated the variation in frequency of recording of medication adverse events between general practices in the Netherlands and to what extent this variation can be attributed to differences between practices or patients within the practices. We found that on average 6.9 medication adverse events are recorded per 1000 patients. General practices differ considerably in recording medication adverse events, with a median odds ratio of 1.94. As expected, more medication adverse events were found in elderly, patients with polypharmacy and chronic diseases. Other studies confirm these findings [14, 20, 21]. After accounting for these patient characteristics, a considerable amount of between practice variation in recorded medication adverse events remains. The median odds ratio still shows almost a twofold difference when comparing two random general practices, with 12.4% of the variability attributed to the level of practices.

Other studies examining medication adverse events in primary care focus on person-years or number of consultations as the unit of analysis. In a supplementary analysis we found that 2.3 consultations (n = 11,004) per 1000 consultations with the general practice concerned a recorded medication adverse event. Number of contacts were calculated on the basis of insurance claims codes representing (telephone) consultations and home visits. In an observational study using an English general practice research database Tsang et al. showed a slightly lower overall incidence of 6.0 medication adverse events per 1000 person-years, and 8.0 adverse events per 10,000 consultations [14]. Another study in English general practices based on routinely recorded data found 1.26 adverse events per 1000 consultations [12]. Compared with these studies, the overall incidence of recordings of medication adverse events seems to be relatively low. Whether this is due to differences in recording or real incidence remains to be investigated.

This study is one of the first studies to investigate between practice variation in medication adverse event recordings. It could be that this process can be facilitated by the software package used in the general practice. Differences between software packages in completeness of recording have been shown in other electronic health record based studies [17], as the software package also seems to affect the quality of prescribing [22]. Optimizing the electronic health record system can help the general practitioner to detect and record adverse events. We also found considerable variation in recorded adverse events between six different software packages. All models were subsequently adjusted for software package. However, between practice variation remained suggesting that other practice characteristics such as age or years of experience of the general practitioner can explain this variation. Also social and socio-economic difference between general practices and migration background would be relevant additional demographic characteristics to explain remaining variation. Within the scope of this paper it was not possible to explore these issues further.

One possible explanation for the between practice variation found in this study may be that it represents the actual ‘true’ variation between practice populations. However, given the fact that we controlled for the most important individual patient level factors of medication adverse events (i.e. age, gender, polypharmacy, and chronic diseases), we believe other explanations may be more important. Another possible explanation is that doctors differ in the extent to which they ‘recognize’ adverse events. Another possibility is that medication adverse events are assigned to other ICPC codes by some GPs. Rather than the diagnosis code ‘adverse effect medical agent’ (ICPC code A85), general practitioners may also use ICPC code A13 which indicates the symptom ‘concern of medical agent’. However, sensitivity analysis showed that ICPC code A13 and A85 are only modestly correlated at practice level (correlation coefficient = 0.4) suggesting that A85 represents medication adverse events as diagnosed by the general practitioner. Yet another possibility is that general practitioners may also record medication adverse events as a symptom presented by patients. For example myalgia may have been recorded as an adverse event (ICPC code A85) related to the use of statins and as myalgia (ICPC code L18) by another. Yet another explanation may be that some general practitioners use free text to record medication adverse events instead of the required fields. A qualitative study exploring above mentioned possible explanations can help to better understand the variation in recorded events, but this was not possible within the scope of this paper. Moreover, because of the important role pharmacists play in medication management of patients, it is possible that part of the unexplained between practice variation is due to factors related to the (relation between GP practice and) pharmacy. It could be that patients discuss medication adverse events with their pharmacist (for example when they collect their medicines). If this information is not subsequently communicated with the GP, it will not be recorded in the GPs electronic health records data. Our analysis on the role of software packages showed that GP software packages that have strong links with the information systems used by pharmacies, recorded most medication adverse events. It include the pharmacist-related factors when exploring between practice variation concerning medication.

Whatever the explanation, the relatively large amount of between practice variation found in this study strongly suggests that adverse event recording in electronic health records systems is far from uniform. Uniformity is essential for adequate exchange of information between health professionals and is important with respect to patient safety [6]. Studies looking at incident reporting systems in other industries refer to meaningful feedback information as an important aspect to improve safety. Other industries also tell us that rather than simply providing feedback a learning culture it is also required to attain a successful safe organization [2325]. Also in primary healthcare research it has been found that providing data quality feedback and benchmark information to general practices reduces between practice variation by making practice personnel aware of the variation and of their recording habits and stimulating them to adhere to recording guidelines [17]. This learning health system approach would not only enhance uniformity of recording but also uniformity in identifying or recognizing adverse events.

Conclusions

This study shows that on average 6.9 medication adverse event are recorded per 1000 patients in general practice. This figure is higher for women, elderly and patients with polypharmacy and comorbidity. However, after accounting for these patient characteristics, variation in recorded medication adverse events still shows large differences between general practices. This suggests that improvement in terms of uniformity of recording medication adverse events is possible. To improve this situation, we suggest that practices should be made aware of these differences, using practice feedback and benchmarking tools, as part of a larger scheme to create a learning culture.

Abbreviations

ATC: 

Anatomical Therapeutic Chemical

CI: 

Confidence Interval

EHR: 

Electronic Health Record

GP: 

General practitioner

ICC: 

Intraclass Coefficient

ICPC: 

International Classification of Primary Care

MOR: 

Median Odds Ratio

OR: 

Odds ratio

SE: 

Standard Error

Declarations

Acknowledgements

Not applicable

Funding

This study was financially supported by the Dutch Ministry of Health. The Ministry of Health had no role in the analysis and interpretation of data.

Availability of data and materials

This study has been approved by the applicable governance bodies of NIVEL Primary Care Database under number NZR00315.055.

Authors’ contributions

SH, KH, LD, and RV were involved in the conception of the research question. SH and KH were involved in analyzing the data. All authors had full access to all the data and contributed to the interpretation of the data. SH drafted the manuscript, which was reviewed by all authors. All authors read and approved the final manuscript.

Authors’ information

Not applicable

Ethics approval and consent to participate

The study was carried out in accordance with Dutch legislation on privacy. According to Dutch legislation, neither obtaining informed consent nor approval by a medical ethics committee is obligatory for observational studies.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
NIVEL, Netherlands Institute for Health Services Research

References

  1. Bates DW, Larizgoitia I, Prasopa-Plaizier N, Jha AK. On behalf of the Reseach Priority Setting Working Group of the WHO World Alliance of Patient Safety. Global priorities for patient safety research. BMJ. 2009:338 b1775.Google Scholar
  2. Jha AK, Prasopa-Plaizier N, Larizgoitia I, Bates DW. Patient safety research: an overview of the global evidence. BMJ Qual Saf. 1 februari 2010;19(1):42–47.Google Scholar
  3. World Health Organization. The importance of pharmacovigilance: safety monitoring of medicinal products. Geneva: WHO; 2002.Google Scholar
  4. Leporini C, De Sarro G, Russo E. Adherence to therapy and adverse drug reactions: is there a link? Expert Opin Drug Saf. 2014;13(sup1):41–55.View ArticleGoogle Scholar
  5. Aspden, P, Wolcott, J, Bootman, JL, Cronenwett, LR, eds. Committee on identifying and preventing medication errors, Institute of Medicine. Preventing medication errors. Washington: National Academy Press; 2007.Google Scholar
  6. Helmreich RL. On error management: lessons from aviation. BMJ. 18 maart 2000;320(7237):781–785.Google Scholar
  7. Dovey S, Makeham M, Runciman W, Larizgoitia I. Methods and measures used in primary care patient safety research. Report for Methods and Measures Working Group of WHO Patient Safety. Geneva: WHO; 2009.Google Scholar
  8. Tsang C, Majeed A, Aylin P. Consultations with general practitioners on patient safety measures based on routinely collected data in primary care. JRSM Short Rep. 2012;3:5.View ArticlePubMedPubMed CentralGoogle Scholar
  9. Leape LL. Reporting of adverse events. N Engl J Med. 2002;347:1633.View ArticlePubMedGoogle Scholar
  10. Friedman C, Rubin J, Brown J, Buntin M, Corn M, Etheredge L, et al. Toward a science of learning systems: a research agenda for the high-functioning learning health system. J Am Med Inform Assoc. 2015;22(1):43–50.PubMedGoogle Scholar
  11. Kaelber DC, Bates DW. Health information exchange and patient safety. J Biomed Inform. 2007;40:40–5.View ArticleGoogle Scholar
  12. Tsang C, Majeed A, Banarsee R, Gnani S, Aylin P. Recording of adverse events in English general practice: analysis of data from electronic patient records. J Innov Health Inform. 2010;18(2):117–24.View ArticleGoogle Scholar
  13. Guthrie B, McCowan C, Davey P, Simpson CR, Dreischulte T, Barnett K. High risk prescribing in primary care patients particularly vulnerable to adverse drug events: cross sectional population database analysis in Scottish general practice. BMJ. 2011;342:3514.View ArticleGoogle Scholar
  14. Tsang C, Bottle A, Majeed A, Aylin P. Adverse events recorded in English primary care: observational study using the general practice research database. Br J Gen Pr. 2013;63:534–42.View ArticleGoogle Scholar
  15. NIVEL: NIVEL Zorgregistraties eerste lijn. https://www.nivel.nl/nl/nzr/zorgregistraties-eerstelijn (2017). Accessed 01 feb 2017.
  16. Wood M, Lamberts H. International Classification of primary care: prepared for the World Organisation of national colleges, academies and academic associations of general practitioners/family physicians (WONCA) by the ICPC Working Party. Oxford, New York, Tokio: Oxford University Press; 1987.Google Scholar
  17. van der Bij S, Khan N, Ten Veen P, de Bakker DH, Verheij RA. Improving the quality of EHR recording in primary care: a data quality feedback tool. J Am Med Inform Assoc. 2016;24(1):81–7.Google Scholar
  18. Merlo J, Chaix B, Ohlsson H, Beckman A, Johnell K, Hjerpe P, et al. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health. 2006;60:290–7.View ArticlePubMedPubMed CentralGoogle Scholar
  19. Kuchinke W, Ohmann C, Verheij RA, Van Veen E-B, Delaney BC. Development towards a learning health system – experiences with the privacy protection model of the TRANSFoRm project. In: Gutwirth S, Leenes R, De Hert P (eds) Data protection on the move: current developments in ICT and privacy/data protection. Dordrecht: Springer; 2016. pp 101–134.Google Scholar
  20. Miller GC, Britt HC, Valenti L. Adverse drug events in general practice patients in Australia. Med J Aust. 2006;184:321.PubMedGoogle Scholar
  21. Tsang C, Majeed A, Aylin P. Routinely recorded patient safety events in primary care: a literature review. Fam Pract. 2011;29:8–15.View ArticlePubMedGoogle Scholar
  22. Opondo D, Visscher S, Eslami S, Verheij RA, Korevaar JC, Abu-Hanna A. Quality of co-prescribing NSAID and gastroprotective medications for elders in the Netherlands and its association with the electronic medical record. PLoS One. 2015;10(6):e0129515.View ArticlePubMedPubMed CentralGoogle Scholar
  23. Barach P, Small SD. Reporting and preventing medical mishaps: lessons from non-medical near miss reporting systems. BMJ. 2000;320:759.View ArticlePubMedPubMed CentralGoogle Scholar
  24. Macrae C. The problem with incident reporting. BMJ Qual Saf. 2016;25:71–5.View ArticlePubMedGoogle Scholar
  25. Speirs F, Johnson CW. Safety Culture in the face of industrial change: a case study from the UK Rail Industry. Research report, University of Glasgow, Scotland; 2002. p. 560–568.Google Scholar

Copyright

© The Author(s). 2017

Advertisement