Methods

The analysis was performed in the emergency department and acute medical and surgical wards of a French teaching hospital over 2 years (2004–2006). We focused on the process of care provided for outlying inpatients from the admission decision in the emergency department until the day of discharge from hospital. Outlying inpatients were patients hospitalised in a ward different from the appropriate ward according to the provisional diagnosis made in the emergency department.

A multidisciplinary team was formed consisting of three doctors and four charge nurses from medical and surgical wards of the hospital. Doctors, nurses and managers from the quality management team were co-opted to help perform the analysis.

Failure modes, effects and criticality analysis consisted of five stages.

Definition of the process of care

Elements and boundaries of the process were defined. Care provided for outlying patients was split into six chronological steps: emergency department care, transfer from emergency department to outlying ward, first day of care in outlying ward, care in outlying ward from the second day until the day before discharge, day of discharge and post-hospitalisation arrangements. Ishikawa diagrams were used to broadly describe each step with its usual components (workforce, materials, machines, methods and environment) and to enable the team to identify the main items that could be linked to vulnerabilities.

A step-by-step analysis of the process was performed in order to describe these items in detail and to compare current policies and procedures with practices. Between November 2004 and September 2005, we interviewed 40 charge nurses from six medical or surgical departments and 26 doctors from 11 wards included in these six departments using standardised questionnaires. Two questionnaires were developed, one for outlying wards and one for appropriate specialty wards, using WWWWHW methodology (Who What Where When How and Why) and matching the broad structure of the process first described. Members of the team administered questionnaires during the course of 45–120-min interviews.

Failures identification

The team identified potential or observed vulnerabilities from Ishikawa diagrams and the step-by-step analysis of the process (interviewed doctors and nurses were free to report additional organisational dysfunction or specific processes set up to improve the management of outlying patients) and from three charge nurse semistructured interviews (with open-ended questions) in the wards most involved in caring for outlying patients.

Qualitative and quantitative analyses

As several vulnerabilities could be in a single chain of “causes–mode–effects” of failure, we classified them by mode, causes and possible effects of failure.

Six members of the team (doctors or charge nurses from different medical departments and from the quality management team) individually and independently classified from 1 to 10 the likelihood of occurrence (O), the severity (S) and the likelihood of detection (D) for each potential failure mode. Explicit criteria adapted from published assessment scales were used to reduce the subjectivity of individual assessments (table 1).8 9 The criticality index (CI) of each failure mode was calculated by multiplying occurrence, severity and detection scores (CI = O×S×D). A priority ranking was then attributed to each failure mode from CI values (the highest ranking was attributed to the highest CI and the lowest ranking to the lowest).

All failure modes were finally sorted according to the median of individual rankings. The team approved the final classification.

We drew up the final classification using ranking values rather than criticality indices in order to reduce potential classification errors due to variation between individual evaluations.

Candidate process improvements

We suggested measures for improvement to counteract the most critical failure modes of the final classification. These measures had to be feasible, easy to apply with low additional cost in all acute wards of the hospital and subject to the approval of the hospital medical committee.

Follow-up indicators

We developed indicators to regularly assess the implementation of improvement measures. We looked preferentially for indicators that were easily accessible and reproducible annually. Indicators were approved by the hospital medical committee.

Results

Definition of the process, failure modes identification and qualitative analysis

The process of care was systematically described for each step (the description of a substep by an Ishikawa diagram is shown in fig 1 as an example). Through doctor and nurse interviews and according to the different steps, we identified 61 vulnerabilities that were integrated into 41 failure modes (table 2) detailed with their causes and potential effects (Appendices 1 and 2, available online).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Ishikawa diagram of the substep “contact with appropriate specialty and outlying wards” during the “Emergency Department care” step.

Quantitative analysis

The final classification of failure modes with the median of individual priority rankings is reported in table 2. The highest rankings of the classification related to the top priority failure modes, which appeared to be essentially a lack of skilled staff to provide appropriate care and the difficulty in transmitting information to patients, families, nursing and medical staff.

Suggested process improvements

From the classification of the failure modes, the team suggested three major media (specialist doctors, nurse coordinators and standardised medical records) that could effect improvement measures (matches between failure modes and these media are indicated in table 2). Improvement measures were mostly drawn from measures already usefully applied in some departments of the hospital.

Identification of specialist doctors

Those specialty wards most likely to have outlying patients because of lack of vacant beds in their own wards should identify each day a specialist doctor. Specialist doctors should be easily contactable by phone and would be in charge of the care of patients outlying in other departments but falling within their sphere of competence: they would be responsible for prescribing and ordering diagnostic tests, as well as for communicating with patients, their families and doctors in outlying wards. They would follow up outliers on a daily basis, systematically get results of diagnostic tests and plan post-hospitalisation care.

Identification of nurse coordinators

The emergency department, appropriate specialty wards and outlying wards should improve communication between each other by the identification of a nurse coordinator in each department. Nurse coordinators would be in charge of the location of outliers, and they would transmit appropriate information to various healthcare personnel and ensure coordination of the medical services required by the patient.

Use of standardised medical records

Standardised medical records for the whole hospital should facilitate the adaptation of health personnel from one department to another and ensure traceability of information and access to archived data.

Follow-up indicators

The team defined four indicators to follow up the implementation of improvement measures.

1. We planned to enumerate outlying and non-outlying inpatients in outlying and appropriate specialty wards on an annual basis.

2. We planned to compare the length of stay between outlying patients versus non-outlying patients, stratified by Diagnosis Related Group tracers within the three specialty wards most likely to have outlying patients. As cases in each Diagnosis Related Group are quite homogenous regarding diagnosis and length of stay, a longer length of stay for outliers might result from delayed care provision for outlying patients.

3. We intended to assess the satisfaction of outlying patients and hospital staff. Outliers’ satisfaction would be evaluated by means of the patient satisfaction survey that occurs annually in our teaching hospital. The patient satisfaction survey administered in February 2007 could be used as a baseline measure for future comparison (table 3). For medical and nursing staff, we set up a short annual email survey to measure satisfaction and acceptability regarding newly implemented measures.

4. Each year, implementation of improvement measures could be evaluated using the care pathway method, by a prospective study of 20–30 hospitalisations of outlying patients.

Discussion

Using FMECA, we were able to comprehensively identify and rank potential modes of failure of a complex process of care. Main improvement measures suggested were: the daily identification of a specialist doctor in appropriate specialty wards to be responsible for the care provided for patients outlying in other wards but falling within their sphere of competence, the identification of a coordinator in each department likely to have outlying patients and the development of standardised medical records in all wards.

Results of all the follow-up indicators were not available when this paper was written, as improvement measures had just been implemented. The next step will be to use these indicators to assess the efficacy and acceptability of the suggested measures.

The major interest of FMECA was to provide a systematic prospective risk analysis method that was able to achieve an in-depth description of the process of care provided to outlying patients. It allowed us to gather information from many doctors and nurses and to identify various improvement measures already implemented in some departments that could benefit the whole hospital. Involvement of many professionals facilitated acceptance of the suggested measures.

The originality of our study lay in the analysis of an extensive care process: the practice of outlying patients in inappropriate wards includes many dimensions, can cover a potentially long hospitalisation period and involves all acute wards of our hospital. Usually, FMECA is used for more limited and technical processes like blood transfusion,10 11 prescription, preparation or distribution of medication,9 12 13 14 or use of medical equipment.15

The extent of the process was also the first limitation because of the difficulty in clearly defining the boundaries of the medical process. Moreover, the practice of outlying patients could be a fairly subjective notion, particularly for patients with multiple pathologies or in polyvalent units.

The second limitation of FMECA was the subjectivity in ranking the criticality of modes of failure. We observed many differences between individual values of critical indices attributed by members of the team and all of them reported that ranking modes of failure was difficult, partly because failure in care processes could rapidly result in serious consequences. Nevertheless, we tried to limit that potential misclassification by an evaluation based on explicit criteria and by using the median of priority rankings in the final classification. Ranking by a multidisciplinary team was another way to reduce misclassification. Our method supposed that a single member was able to maintain consistent rankings for all failure modes.

FMECA requires substantial staff time.10 11 13 This was a serious barrier to performing a second identification and classification of modes of failure as a follow-up indicator. We preferred to define simpler indicators, easily available or already implemented in the hospital.

As hospital overcrowding often results in outlying hospitalisation, similar improvement measures were suggested for studies on overcrowding management. For example, case managers, placement coordinators or bed managers can achieve better coordination.16 17 18 Development of systems providing reliable and real time information on bed status or diagnostic test results has been suggested to improve coordination.16 18

In conclusion, using the prospective method of FMECA, we were able to identify most critical modes of failure in care provided for outlying patients and to suggest subsequent priority improvement measures. Implementation of those measures will have to be assessed regularly with the follow-up indicators defined.