Article Text

PDF

Medication incidents in primary care medicine: protocol of a study by the Swiss Federal Sentinel Reporting System
  1. Markus Gnädinger1,
  2. Alessandro Ceschi2,3,
  3. Dieter Conen4,
  4. Lilli Herzig5,
  5. Milo Puhan6,
  6. Alfred Staehelin1,7,
  7. Marco Zoller1
  1. 1Institute of Primary Care, University of Zurich, Zurich, Switzerland
  2. 2Department of Clinical Pharmacology and Toxicology, University Hospital Zurich, Zürich, Switzerland
  3. 3National Poisons Centre, Tox Info Suisse, Associated Institute of the University of Zurich, University Hospital Zurich, Zürich, Switzerland
  4. 4Patientensicherheit Schweiz, Zürich, UK
  5. 5Policlinique Médicale, University of Lausanne, Lausanne, UK
  6. 6Epidemiology, Biostatistics, and Prevention Institute, University of Zurich, Zürich, Switzerland
  7. 7Sentinel Surveillance Network, Swiss Federal Office of Public Health, Bern, Switzerland
  1. Correspondence to Dr Markus Gnädinger; markus.gnaedinger{at}hin.ch

Abstract

Background/rationale Patient safety is a major concern in healthcare systems worldwide. Although most safety research has been conducted in the inpatient setting, evidence indicates that medical errors and adverse events are a threat to patients in the primary care setting as well. Since information about the frequency and outcomes of safety incidents in primary care is required, the goals of this study are to describe the type, frequency, seasonal and regional distribution of medication incidents in primary care in Switzerland and to elucidate possible risk factors for medication incidents.

Methods and analysis Study design and setting: We will conduct a prospective surveillance study to identify cases of medication incidents among primary care patients in Switzerland over the course of the year 2015. Participants: Patients undergoing drug treatment by 167 general practitioners or paediatricians reporting to the Swiss Federal Sentinel Reporting System. Inclusion criteria: Any erroneous event, as defined by the physician, related to the medication process and interfering with normal treatment course. Exclusion criteria: Lack of treatment effect, adverse drug reactions or drug–drug or drug–disease interactions without detectable treatment error. Primary outcome: Medication incidents. Risk factors: Age, gender, polymedication, morbidity, care dependency, hospitalisation. Statistical Analysis: Descriptive statistics to assess type, frequency, seasonal and regional distribution of medication incidents and logistic regression to assess their association with potential risk factors. Estimated sample size: 500 medication incidents. Limitations: We will take into account under-reporting and selective reporting among others as potential sources of bias or imprecision when interpreting the results.

Ethics and dissemination No formal request was necessary because of fully anonymised data. The results will be published in a peer-reviewed journal.

Trial registration number NCT0229537.

  • CLINICAL PHARMACOLOGY
  • EPIDEMIOLOGY
  • PUBLIC HEALTH

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/

Statistics from Altmetric.com

Strengths and limitations of this study

  • First Swiss prospective and systematic collection of incident data in primary care.

  • Covering three linguistic regions and two distribution systems.

  • Bias from selective and under-reporting or non-detection.

Introduction

Patient safety is a major concern in healthcare systems worldwide. Although most safety research has been conducted in the inpatient setting,1 evidence indicates that medical errors and adverse events pose a serious threat for patients in the primary care setting as well, since most patients receive ambulatory care.2–4 Gandhi and Lee5 noted that safety concerns in the outpatient setting differ from those in the hospital setting in obvious and non-obvious ways. Medication-related incidents are also important in primary care, but are perhaps not as well documented as in secondary care. A study by Gandhi and Lee6 in primary care showed that 25% of 661 ambulatory care patients with at least one prescription had such an incident, of which 13% were classified as serious and 39% as preventable or ameliorable. In a study by Pirmohamed et al in the UK, medication-related incidents caused about 6.8% of all hospitalisations7 and in Switzerland, these incidents are responsible for about 7.2% of hospitalisations.8 Diagnostic errors9 and adverse drug events have been identified as frequent safety concerns; furthermore, there is a body of literature about the safety of outpatient procedures and the consequences of coordination as well as continuity-of-care failures.10 Hospital and outpatient care also differ in their infrastructure and in many processes, as well as in their ability to detect, monitor and address safety issues. Information about the frequency and outcomes of safety incidents in primary care is required to identify risks or ‘hot spots’, to prioritise them and to take action as needed.

Definition of terms

For the purpose of our study, we use the terminology of the International Classification for Patient Safety (WHO).11 Here, a patient safety incident is defined as an event or a circumstance which could have resulted or did result in unnecessary harm to a patient. There are several possible causes. (1) An error, defined as: failure to carry out a planned action as intended or an application of an incorrect rule; (2) A violation, defined as: deliberate deviation from an operating procedure, standard or rule; (3) An (external) circumstance, defined as: a situation or factor that may influence an event, agent or person. We want to distinguish incidents from adverse drug reactions (ADR) which are defined as: unexpected harm resulting from a justified action where the correct medication process was followed for the context in which the event occurred. The same applies to drug–drug or drug–disease interactions (DDI). Critical Incident Reporting System (CIRS) refers to a voluntary anonymous database system to which Swiss family physicians or paediatricians may report incidents as those occurred in their practices.

We will restrict the topic of our study to medication-related incidents. Medication or medicine refers to a pharmaceutical drug, officially called a medicinal product, which can be loosely defined as any chemical substance—or product comprising such—intended for use in the medical diagnosis, cure, symptomatic treatment, or prevention of disease (Wikipedia).

Information retrieval

Methods to collect information about adverse incidents are manifold.2 ,11–14 In the literature, different ways to collect adverse incident information are described: voluntary versus mandatory reporting systems, patient questionnaires, or pharmacists reporting to registries. Interviews can be held with physicians or questionnaires can be filled out by them. Charts of deceased patients can be audited meticulously or medicolegal cases may be analysed. Information may be derived prospectively in an actual case by case manner or retrospectively along case vignettes. The two methods most commonly used are incident reporting and chart review. Both methods have the potential to systematically cover information on the entire range of safety events in medical offices in common.

Incident reporting has a long tradition in clinical risk management and is increasingly used in outpatient care.15–18 Indeed, incident reporting has been the dominant method for the study of safety incidents in primary care.19 It is based on voluntary and usually anonymous reports of physicians and nurses, and is used to describe the types and characteristics of patient safety incidents. These reports may vary considerably with respect to the information which is included and the likelihood of ‘true’ incidents being reported is unclear. Studies based on this method describe large variations in the number of reports submitted.20 Moreover, professional groups differ in their frequency of reporting; in-hospital care physicians reported preferentially severe incidents, while nurses cover the whole spectrum of impact levels.20 Selective reporting by physicians may have multifactorial causes such as lack of time, thinking that an ADR which is already known is not worth mentioning, or concerns about data confidentiality. Non-reporting may be as frequent as 94%.21 ,22 O'Beirne et al concluded from a very low report rate (<1 report per person per year) that incident reporting may be “a costly but not very effective way to study safety problems in primary care”.23 A recent Swiss study analysed safety issues in primary care.12 This was a semiquantitative, retrospective investigation involving over 300 nurses and physicians. Seven of 23 issues were related to drug treatment. Frequently named issues were insufficient monitoring of potential side effects, missing prescriptions of required treatment, and errant medication relating to the route of administration, dosage or timing. The low and perhaps, selective reporting of incidents in any system makes it difficult or virtually impossible to collect valuable quantitative information from CIRS.24

In chart review, medical records are analysed by independent experts in order to identify adverse events, and to assess potential harm and preventability in each case.25 Such analysis requires complete and correct patient documentation to provide valid results. In many cases relevant information may be unavailable.25 As chart review is a time-consuming approach, many resources are needed to analyse a large number of patient records at different primary care offices; furthermore, this kind of analysis is usually retrospective.

Sandars and Esmail2 described a frequency of 5–80 medical errors per 100 000 consultations in primary care patients. Incidents, as defined above, may also result from circumstantial factors (without any errors) and therefore, occur somewhat more frequently. In order to not overburden the participating physicians with our study, we decided to limit our focus to medication incidents. These make up between 9% and 42% of all registered incidents2 ,9 ,26–28 and of these, approximately 70% may be prescription errors.29 In contrast to the study by Sandars and Esmail, our pilot study found a rate of approximately one medication incident per 2 months and per physician (see online supplementary appendix 1), while the former reported a rate of about one in 2 years.

Concerning predisposing factors, Avery et al30 found a propensity for becoming a victim of medication errors with the young (<15 years) or elderly (>64 years) ages. The latter was confirmed by Salanitro et al.31 Two studies by Field et al32 ,33 reported morbidity as promoting errors. All studies listed polymedication as a key factor.30–34 Better knowledge of factors associated with medication errors would be helpful to implement preventive measures and therefore, reduce the frequency of avoidable incidents in the future.

The Swiss National Sentinel Reporting Network (http://www.sentinella.ch)

To apply a supplementary method to gaining insight into safety hot spots in primary care, we aim to assess medication errors by using the Swiss National Sentinel Reporting Network. Founded in 1986, it was mainly designed to survey transmissible diseases. Later, it also assessed other health problems of public interest. In 2015, besides our project, the areas covered are: surveillance of mumps, whooping cough, flu or Borrelia, pneumonia and middle ear infections, tick bites, vaccinations against measles or whooping cough, and alcohol use among adolescents. One hundred and forty general internal medicines and 27 paediatric practices report to the system, 109 of them in the German, 44 in the French, and 14 in the Italian part of Switzerland. It generates daily to weekly current data and covers the entire geographic and linguistic regions of our country.

Method

Aims of the project

To describe the type, frequency, seasonal and regional distribution of medication incidents in primary care in Switzerland and to elucidate possible risk factors such as age, gender, polymedication, morbidity, and previous hospitalisation.

Study design

We will conduct a prospective surveillance study to identify cases of medication incidents among primary care patients over the course of 1 year.

Population

Any person undergoing drug treatment in general internal or paediatric practices participating in the Sentinella network. The latter covers a representative sample of patients in primary care for Switzerland (see above). These patients include children, individuals with mental retardation or the elderly—all of whom might be at increased risk for medication errors.

Inclusion criteria

  • Any erroneous event (as defined by the physician) relating to the medication process and interfering with the normal treatment course.

Exclusion criteria

  • Lacking treatment effect, ADR, or drug–drug interaction or DDI without detectable treatment error.

  • Refusal of patients to refer data to the Sentinel system.

Questionnaire development

As we could not identify questionnaires suitable for continuous reporting and adaptable to our local conditions, we had to develop new ones. We tested these in an 8-week pilot study (see online supplementary appendix 1).

In the Sentinella study, the Italian speaking physicians had to decide whether to report in German or French. For the main study, we developed only two language sets of questionnaires, since these had to be filled in only by physicians, not by patients. We deemed assessment of construct validity of the questionnaires not imperative, since we did not measure hidden constructs (like ‘depression’) by our questions and information was mainly needed about influencing our target variable, that is, the type and frequency of incidents.

Data to be collected

The data to be collected in our study are summarised in boxes 1 (physician related) and 2 (patient related/incident related); for the latter, only pre-existing data from medical records will be collected. Furthermore, we collected denominator data as depicted in box 3.

Box 1

Physician-related data

Initial questionnaire

Sentinella identification number

Gender*

Age*

Specialisation (general practitioner or paediatrician)*

Number of physicians in practice, and among them those reporting to Sentinella*

Working hours per week*

Drug distribution system (by pharmacist or by physician)

Drug prescription system (electronic, machine written, hand-written)

Electronic drug–drug interaction (DDI) system

Availability of X-ray, ECG, ultrasound

Medical history (electronic or paper based)

Quality certificate

Team sessions

Physician's participation in quality circle

Localisation (urbanity, language region)*

Special education/interests

Caring for institutions

Final questionnaire

Proportion of non-reporting incidents during the study

The data are collected from each physician reporting to the Sentinella system. The questionnaire and the coding plan are available at http://www.medication-incidents.ch.

*These data is delivered by Sentinella administration. All other data will be collected by use of questionnaires.

Box 2

Patient-related data

Patient-related/incident-related items

Week of reporting

Year of birth

Gender

Physician-to-patient relationship

Dwelling situation

Social problems

Dementia or learning disabilities

Psychiatric problems

Use of psychotropic drugs

Linguistic problems

Smoking or substance abuse

Visual blurring or hearing loss

Gait disturbance

Renal insufficiency

Liver cirrhosis/insufficiency

Previous hospitalisation (12 months)

Care dependency

Number of chronically administered active drugs*

Number of diagnoses for chronic disease

Scale value of ‘Thurgau Morbidity Index’ (TMI)35

Description of incident

Who noticed the incident

What went wrong

Name of drug

Other drugs used possibly related to the incident

Endangering of patient

Amount of damage

Organ system involved

Duration

Recovery

Treatment/surveillance

Causal triggers

Interface problems

Information to the patient about the incident and his reaction

Consequences of the incident

Responsibility

Possibility to anticipate the incident

Whether a similar incident was previously notified within the study

General proposals

Physician-related items

Sentinella identification number

The data are collected for each incident. The basic reporting of a medication incident is collected weekly. It includes the Sentinella identification number, the gender of the patients and their year of birth. The questionnaire and the coding plan are available at http://www.medication-incidents.ch.

*Each regularly prescribed or administered pharmacological specialty counts per active substance contained, according to the latest available medication list. All therapies regularly taken during at least 1 month are considered. Drugs that are prescribed for shorter periods (eg, antibiotics for a week) are not included. Regularly administered eye drops, inhalations, nasal sprays count if a general systemic effect is intentional (eg, calcitonin nasal spray) or must be taken into account (eg, timolol eye drops). Also transcutaneous, subcutaneous or vaginal hormone-releasing systems, and preparations administered by the specialist (eg, gynecologist: birth control pills) should be recorded. Herbal drugs count—regardless of the number of plants—as one medication. Homeopathic drugs, cell salts, etc, are not counted. Multivitamins count if they are taken due to a medical indication (eg, short bowel syndrome), but not when the administration was adopted as ‘roborant’; multivitamins are considered one drug. Also therapies at the hospital administered such as oncology are counted. Whether the patient also applies the drug (compliance) is irrelevant to the study, the important thing is that it is so prescribed. Prescribed medications to relieve on requirement which do not need to be taken daily, or self-medication is not recorded.

Box 3

Denominator data

Fortnight analysis (only once in study, from 7 to 20 March 2015)

Previous hospitalisation (during the preceding year)

Care dependency

Number of active drugs chronically administered

Number of chronic diagnoses

Scale value of ‘Thurgau Morbidity Index’ (TMI)35

Year of birth*

Gender*

Multiple consultation (within the 14-day period)

Sentinella identification number

Daily analysis

Physician-to-patient contacts

Sentinella identification number

The data are collected from each patient consulting the practice during the year 2015, irrespective of the presence of an incident.

*Year of birth and gender will be collected during another 14-day period in fall but without the other items. The questionnaire and the coding plan are available at http://www.medication-incidents.ch. Definition of medication count: see box 2.

Time schedule

The pilot study took place from July until September 2013. The final questionnaires in both languages and their English translation are available at http://www.medication-incidents.ch. The main study takes place from January until December 2015. Evaluation of the study data and writing of the publication will be performed in 2016.

Statistics

Expected number of cases

On the basis of our pilot trial (see online supplementary appendix), we expect at least one drug-related incident to be reported every 2 months per physician. In the Sentinella system, there are currently 167 physicians actively reporting, among them are 27 paediatricians. When assuming that they work 10 months per year, this would result in 660 reported incidents. We, hence, have to decrease this number to some extent because not all physicians work 100%, there will be some non-reporting or non-detection of incidents, and paediatricians report lower rate of incidents (according to the results of our pilot study). So we expect approximately 500 incidents in our study.

Statistical methods

For the analysis of our data, we will use descriptive statistics in order to describe the type, frequency, seasonal and regional distribution of medication incidents. We will use logistic regression to assess the association of medication incidents with potential risk factors. We will use SPSS.

Independent ethical committee

The ethical committee of Canton Zurich decided that our study did not need formal approval, because the data are completely anonymous (KEK-ZH 2014-0400). The study was recorded in http://www.ClinicalTrials.gov: NCT02295371, as well as in our national study registry (http://www.kofam.ch; SNCTP000001207).

STROBE statement: Where applicable, our publication will follow the general STROBE guidelines (http://www.equator-network.org/).

Results from a pilot study

From the existing literature,12 we included questions on the social and clinical state of the patients, on the type of incident, and on possible causative factors, and tested these in a pilot study, from July until September 2013, with a sample of general physicians or paediatricians; these were in two language sets (German and French) for the three different language regions of Switzerland (German 11, French 7, Italian 3) for 8 weeks. Fifty-one cases were recorded, leading to an incident rate of 0.4 (median, IQR 0.4) per family physician, and week or 4 (5) per family physician and 1000 patient contacts. Virtually no incidents were observed by paediatricians (details of this study are presented in online supplementary appendix 1).

Discussion

Strength and limitations

Our study is the first prospective and systematic collection of incident data in Swiss primary care. The well-motivated reporting physicians, the duration of 12 months and the coverage of the three main linguistic regions as well as of two drug distribution systems will deliver new and relevant insights.

Not all cases may be reported; there may be some selective reporting of cases of higher clinical importance, and some cases may not be reported due to lack of time or legal considerations. Some cases may not be detected by the physicians. Some non-prescription treatments may be missed if they are not causally linked to the incident. The sample size may be too small for inferential statistics. The 14-day denominator period may not be representative for the patient collective over the whole year. The definition of inclusion criteria may be interpreted varyingly by the physicians. Qualitative information may be missed in this study and has perhaps to be further addressed in qualitative research. The low and perhaps selective reporting of incidents in any system makes it difficult or virtually impossible to collect valuable quantitative information from CIRS.24

Conclusion

Data on safety issues in ambulatory primary care patients is scarce and this is also true for Switzerland. The retrospective qualitative study by Gehring et al12 and the study by Livio et al8 on hospitalisations makes it reasonably clear that things are in no way better than that in UK7 or USA,6 where 6.5% of all hospitalisations and 4.0% of all hospital care days as well as 1 in 667 hospital deaths are caused by medication incidents. Since up to 72% of these appear to be preventable or ameliorable, it seems worthwhile to focus on these so as to further elucidate risk factors.

The data of our study should allow for describing the type, frequency, seasonal and regional distribution of medication incidents in Swiss primary care practices and for helping to identify risk factors. The definition of ‘hot spots’ could sensitise the physicians to focus on dangerous situations and help them to redefine their standing operational procedures (SOPs).

Acknowledgments

The authors are grateful to Lee Wennerberg for English language corrections, the Federal Health Agency for the translation of their pilot study questionnaires into French, Dr Sven Staender, Männedorf, for the helpful comments. They thank all colleagues participating in the pilot study (Banderet HR, Bigler M, Dafflon M, Ebnöther S, Ferrazzini G, Frey J, Fröhlich F, Haftgoli N, Köhler B, Marty F, Mühlemann N, Merz A, Oppliger S, Pasche O, Pilet F, Quack C, Sébo P, Schwarzenbach HR, Spirig M, Valentin J).

References

View Abstract
  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Contributors MG lead the study, did the pilot study (questionnaire development, data entering and processing), wrote all documents, made all the contacts with the Sentinella administration, ethics committee and others, programmed the electronic questionnaires, will enter the hand-written questionnaires into the database, will do the data processing and will write the publication after data collection. LH is French speaking and will help others to understand the French-printed questionnaires. MP was responsible for sound methodology. AS conceived the idea for the study. All the authors have seen all the study documents and contributed intellectually to this publication. All the authors have contributed to the revision of the draft of this publication, and approved the submitted version of this publication.

  • Funding The study was funded by Bangeter-Rhyner Foundation, Basel (grant number 8472/HEG-DSV).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.