In the years 2010 – 2015, nearly six million diagnoses were made in the 17 public and two private pathology institutions in Norway (4). In light of this figure, the number of compensation claims is relatively low. Injury compensation from the NPE is a voluntary option, and applications are submitted by individual patients or their guardians. The number of cases in the NPE records therefore reveals nothing about the actual number of injuries caused by errors made in pathology departments or laboratories. Because the NPE does not register pathology-related cases as a separate category, we cannot exclude the possibility that some cases have been omitted from our searches.
The observation that false-negative diagnoses were the most frequent cause of claims for compensation is natural, for two reasons. First, inadequate or no treatment of cancer is potentially fatal, and as a rule more serious than overtreatment. Second, there is no system in place to capture false-negative diagnoses.
False-positive diagnoses are prevented by having two or more pathologists confirm a diagnosis of malignancy. Benign lesions, on the other hand, which predominate in daily diagnostics, are not controlled in a similarly systematic way. Diagnostics of large amounts of material expected to be benign may also increase the risk of errors if the diagnostics assume the character of routine screening (5).
Diagnostics of pigmented skin lesions are among the pathologists’ most common tasks. Most of these lesions are benign moles or keratoses and easily diagnosed. However, the group of moles is also where pathologists face some of their most difficult diagnostic considerations when distinguishing between benign, irregular or atypical, and malignant, and where a misjudgement at worst may lead to a fatal outcome. Underdiagnosis of malignant melanomas is also reported as the most frequent cause of compensation claims internationally (6, 7).
The cause of false-negative or false-positive diagnoses is difficult to identify. Diagnosis of pathological changes largely depends on human factors, and even specialists with long-standing experience may commit errors of judgement. Some years ago in Sweden, underdiagnosis of malignant melanomas was revealed in a large group of patients (8, 9). The diagnoses were made by pathologists who had many years of experience (10).
There are no good studies of whether large workloads over time give rise to a higher number of errors in pathology diagnostics. Renshaw and Gould undertook studies of diagnostic safety with the aid of a blinded review of recently diagnosed biopsies and concluded that the workload had no significance for the number of errors (11).
However, this study was undertaken among a small number of doctors over a limited period of time, and the participants were aware of the fact that a study was in progress. The doctors included in the study had previously had a limited workload, varying between 23 to 32 cases/diagnoses per day. Measuring workload in terms of diagnoses is also questionable as long as the type of diagnostics remains unspecified. The workload per case depends on the type of sample, the number of sections and further analyses, and will thus vary significantly, within as well as between different departments and laboratories.
The reasons why the diagnoses were overlooked in the Norwegian cases could not be identified by this study. Most likely, however, a double review of all samples by two pathologists, as practised when a malignant diagnosis is expected, might have prevented some of these errors.
A study of 359 pathology laboratories in the USA showed that departments that have systems for repeated review of a defined proportion of the cases prior to communication of a final result yielded a significantly lower proportion of revised diagnoses (12). However, no information is provided regarding the types of cases that were selected for review. Repeated review of all samples would require considerable resources and is regarded as unrealistic at a time when the shortage of pathologists has given rise to long waiting periods (5, 13).
Screeners are specially trained bioengineers who are schooled in assessing normal cell samples and in recognising samples with aberrant changes. False-negative cervical cytology diagnoses may have arisen due to erroneous assessments made by this group of professionals, although methodological problems may also have been a significant factor (14, 15).
The classical smear test is fraught with artefacts resulting from air drying, for example, which complicates the assessment. Mass screening of cervical cytology is therefore currently undertaken with the aid of fluid-based techniques, which eliminate sample collection errors such as poor smear technique, contamination by blood and air drying of the material. Although fluid-based testing is a relatively new technique, there is reason to assume that the improved morphology that can be achieved with this technique will considerably reduce the number of false-negative diagnoses in the future (16).
The number of pre-diagnostic and post-diagnostic errors accounted for 17 per cent of compensation claims. In the majority of the cases of mix-up (six of eight), the cause was registered as uncertain. Mix-ups can occur at many levels and involve patient identities or requisitioners upon registration, sample containers upon reception, blocks or sections upon processing or diagnoses upon dictation of results.
Outdated ICT solutions and insufficient use of automated solutions are particular to pathology. The chain of safe patient identification is broken when samples arrive in the department. Today, only three of 17 public departments in Norway have systems for electronic requisitioning and possibilities for scanning patient IDs from test phials into their computer systems, but none of them use electronic requisitioning for all samples (according to a survey of public Norwegian pathology departments in March 2016 undertaken by G.C. Alfsen, unpublished data). Most departments transfer data on patients and requisitioners manually to their own computer systems. Hand-written clinical information is scanned from paper request forms, which are manually labelled along with the sample containers using stickers with bar codes internal to the department.
The handling of the material is also largely undertaken manually. Small tissue samples such as from curettage or pipelle, endoscopic biopsies or needle biopsies are moved manually several times – from the containers to transport cassettes and subsequently to casting moulds. Each handling procedure involves a risk of loss of tissue, contamination of one sample by another and mix-up. Many types of small tissue samples can be put directly into the cassette by the clinician when collecting the sample, but this requires electronic requisitioning and/or pre-marked cassettes and is used only in a single department in Norway for a limited proportion of samples. Automatic casting machines reduce the amount of manual handling of small samples, but have so far been procured by only a couple of departments and are used only for a selected proportion of the smallest samples.
Systematic reviews have detected contamination from various sources in the laboratories in 0.6 per cent of all sections (17, 18). We know from our own experience that contamination or mix-up of samples regularly cause near-misses which are prevented through control routines internal to the departments. There are no reliable figures for such near-misses.
The risk of erroneous diagnoses caused by mix-up of IDs or contamination has spurred the development of a system for routine comparison of the DNA of the patient with that of sample material in paraffin blocks (19). Mix-up of IDs may equally well occur when the sample is collected, i.e. before the sample arrives in the department. The US study described only prostate cancer diagnoses and detected no differences in risk of ID mix-up between the doctor who collected the sample (urologist) and the pathology department, with a total risk of mix-up in 0.26 per cent of the cancer samples. Contamination of the test blocks was detected in 0.67 per cent of the cases and was mainly caused by handling in the pathology departments. On the basis of this study, the number of false-positive prostate cancer diagnoses was estimated at approximately 4 500 per year and the annual costs were estimated at USD 880 million (20).
The incidence of overlooked diagnoses has not been systematically investigated in the same way, and thus the costs of the consequences ensuing from a delayed cancer diagnosis have not been estimated. Although there are only few compensation claims resulting from mix-up or contamination in our study, there is little reason to believe that the real incidence of such errors is lower in Norway than in other countries. It is therefore likely that an upgrade of ICT systems and introduction of electronic requisitioning and ID tracing of samples throughout the process would entail a considerable improvement in patient safety. Upgrading of the equipment base with a focus on preventing contamination is another measure that might reduce the number of pre- and post-diagnostic errors (21).