A random sample of 350 admissions was selected from 3701 inpatient admissions during 2009 at a 52-bed Orthopedic Department at a university hospital in Stockholm, Sweden. Both elective and acute admissions were included in the study. We reviewed the healthcare process during the care of the patient. The study emanates from a previous study that compared the ability of two RRR methods, the Harvard Medical Practice Study (HMPS) method and the Global Trigger Tool (GTT) to identify AEs in the same sample of admissions . In the present study the HMPS method was evaluated for the detection of no-harm incidents.
The inclusion criteria were that no-harm incidents had to be related to care provided by the Orthopedic Department, and the meeting of any one of the following criteria:
The no-harm incident was to have been caused within 30 days before index admission, and detected during index admission.
The no-harm incident had to be caused and detected during index admission.
The no-harm incident was to have been caused during index admission and detected within 30 days of index discharge from the Orthopedic Department at an inpatient or outpatient visit.
The hospital’s computerized medical record system – referred to as a “record” in this article - was the source of all data since it included documentation from all healthcare professionals. In addition to reviewing index admissions and readmissions, a review of outpatient notes for no-harm incidents connected to the inpatient index admission was carried out.
The review team consisted of one registered nurse (RN) and two physicians. The RN (MU) was highly experienced in using screening criteria, and well informed of the orthopedic context. One physician was a senior orthopedic surgeon (OM) experienced in reviewing records. The other physician had no specific orthopedic competence or RRR experience, but had expertise in patient safety (KPH).
To ensure validity, the review process was standardized in a written manual. Detailed examples were developed, discussed and defined by the team prior to the start of the study. In the familiarizing process, each team member independently reviewed 11 written training records followed by a consensus process allowing all involved to engage in discussions of no-harm incident assessment.
RRR based on the method used in the original HMPS [24, 25], is a two-stage review process to identify the incidence of AEs using 18 screening criteria. Over the years, modifications of the original method have been made in subsequent studies; e.g. by adding preventability to the review process [26–29].
A two-stage RRR was performed according to HMPS methodology [25, 27, 29]. In review Stage 1, all records from the random sample were reviewed by the orthopedic nurse. She screened the index admission plus/minus 30 days for the presence of one or more of 18 predefined screening criteria. For every screening criterion detected, a judgment was made by the RN on whether it reflected the presence of a potential no-harm incident or not. This was performed parallel to screening for AEs . The finding was briefly described in a study database to facilitate the next step of the review process. Only records with potential no-harm incidents were forwarded to the physicians for review. Each physician reviewed half the records forwarded by the RN. In addition, every tenth record was reviewed double-blinded by both physicians, which consequently demonstrated inter-rater reliability.
In review Stage 2, the two physicians each performed an independent review. Initially, an assessment of healthcare causation was performed using a 6-point scale [24, 28]. Healthcare causation implies the actions of individual healthcare workers as well as the systems and care processes used in delivering healthcare rather than the patient’s underlying illness. It includes both acts of omission (for example, failure to diagnose or treat) and acts of commission (for example, incorrect diagnosis or treatment, or poor performance) . Only potential no-harm incidents with a score of four or higher (i.e. greater than 50% likelihood of healthcare causation) were classified as no-harm-incidents. The physicians then judged if the no-harm incident was considered preventable or not. A preventable no-harm incident was defined as an event in healthcare management due to failure to follow accepted practice at the level of the individual or of the system . A similar 6-point scale was used and a score of four or higher (i.e. more than 50% likelihood) meant that the no-harm incident was considered preventable .
The physicians documented the related screening criteria for each no-harm incident. If a record contained more than one no-harm incident, each was reviewed separately, and the associated screening criterion documented separately. In accordance with the HMPS methodology, supplementing assessments were performed for an overview of no-harm incidents. First, no-harm incidents were classified into eight categories describing their nature (e.g. drug related). Also, place of occurrence (e.g. Emergency Department) and contributing factors  (e.g. team factors) were documented. The contributing factors that influenced practice were classified according to the framework by Vincent et al.  and were adapted to the no-harm incidents.
In order to check that no potential no-harm incident had been excluded by the RN in review Stage 1, every tenth admission that had not been forwarded to a physician for review was screened by one of the two physicians (OM). If the physician found any additional no-harm incidents, they were added. Similarly, any no-harm incident found by the physicians during review Stage 2 was added. These two steps constituted the validation of the RN review.
A comparison between RRR and one local and four nationwide systems for incident-reporting regarding the 350 admissions was also carried out. Documentation of no-harm incidents from the five incident-reporting systems was identified by two RNs (KS and MU). In Sweden each person has a unique personal code number. The five systems are not confidential to the involved healthcare clinic, enabling the investigation of no-harm incidents connected to the same patient sample. The five systems were: 1) the obligatory clinical incident-reporting system at the hospital, 2) hospital reports, named Lex Maria, according to Swedish legislation, 3) malpractice claims reported to the Swedish County Council’s Mutual Insurance Company, 4) the Medical Responsibility Board and 5) patient/relative complaints reported to the Patient Advisory Committee (obligatory in every county council according to Swedish legislation).
Ethical approval was obtained from the Regional Ethics Committee of Stockholm (Number 2008/951-31/3).