Diagnostic error in primary care: where it happens

A 67-year-old man comes to the health unit with fatigue and shortness of breath for a few weeks. He has already been seen twice in the last month, with the same report, and left both times with advice to rest and come back if he got worse. At the third visit, a simple test shows significant anemia, and the following investigation finds bowel cancer. Reviewing the chart, the doctor realizes that the blood count had been ordered at the first visit, but no one ever saw the result.

All three visits were done carefully. At the first, the hypothesis of fatigue from overwork was plausible and the right test was ordered. At the second, another professional saw him, did not know about the pending test and repeated the advice. Only at the third did someone put the pieces together. The error was not in one person, but in the path between them.

Stories like this are common and rarely involve rare diseases or careless doctors. They show that diagnostic error in primary care usually arises from the sum of small failures spread through the process. This text discusses how often the error happens, which diseases are most involved, at which steps of the process it arises, who is most exposed and what helps to reduce it.

Key points

• Diagnostic errors in outpatient care affect about 1 in 20 adults per year (Singh, 2014).
• Most involve common diseases, such as pneumonia, heart failure, cancer and urinary infection.
• Most failures happen within the consultation: history taking, physical examination and ordering tests.
• Results not seen, referrals lost and follow-up that does not close complete the picture.
• People with several diseases and symptoms attributed to already known conditions are more exposed.
• Lists of diagnoses that must not be missed, alarm signs and follow-up systems help to reduce the error.

Summary

• How common diagnostic error is
• Why the error is hard to see
• The diseases involved
• Within the consultation
• Tests, referrals and results
• The follow-up that does not close
• Who is most exposed
• What helps to reduce the error
• Frequently asked questions
• Final checklist

How common diagnostic error is

Estimating the frequency of diagnostic error is difficult, because it is rarely recorded as such. An analysis that combined three large studies of outpatient care estimated that about 1 in 20 adults seen suffers a diagnostic error per year, which corresponds to nearly 12 million people in the country studied (Singh, 2014). The authors calculated that about half of these errors had the potential to cause serious harm (Singh, 2014).

The frequency also appears in lawsuits. A systematic review of 34 studies on claims in primary care found missed or delayed diagnosis as the most common reason, accounting for a fraction ranging from about a quarter to almost two thirds of the actions (Wallace, 2013). Medication error came second (Wallace, 2013).

On the patient side, the perception of error is also measurable. In a study with 1474 outpatients seen in primary care clinics, 57 reported having suffered a diagnostic error (Aoki, 2020). The measure has limits, because it depends on each person's memory and interpretation, but it shows that the problem is perceived by those who are seen.

These numbers do not mean that primary care errs more than other levels of care. They reflect the enormous volume of consultations, the variety of complaints and the early phase in which diseases appear, when the signs are vague. The text on how the Improving Diagnosis report defines diagnostic error discusses the concept in more detail.

It is also worth remembering that not every late diagnosis is an error. Some diseases only reveal themselves with time, and no consultation, however good, would identify them at the first presentation. The error, in the studies cited here, is the missed opportunity: a piece of data available that was not used, a test indicated that was not ordered or a result that was not seen.

Why the error is hard to see

Diagnostic error in primary care has a feature that makes it hard to recognize: it rarely shows up on the day it happens. The person leaves the appointment, the symptom sometimes improves on its own for a while, and the correct diagnosis comes weeks or months later, often at another service. The one who made the error does not always find out, and the one who discovers it does not always come back to tell.

This lack of feedback has consequences. Without knowing about their own errors, doctors tend to trust their own judgment more than they should, because they only see the hits. Reviews of lawsuits and medical records are one of the few windows into what went wrong, and they show that most errors involved conditions the doctor knew well (Wallace, 2013).

Another reason is that the boundary between error and natural progression is blurred. A disease that starts with vague signs may be impossible to recognize at the first visit and obvious at the third. What defines the error, in this case, is not the initial hypothesis being wrong, but the missed opportunity to revise the hypothesis when the signs changed. That was the logic used in studies on missed opportunities, which look at what was available in the record and was not used (Singh, 2009).

Finally, the error is usually collective. When the failure is in a result nobody saw or in a referral that never arrived, there is no single person responsible, and so it is easy for no one to feel ownership of the problem. Recognizing this distributed nature is the first step to designing solutions that do not depend only on the doctor's individual attention.

This does not remove the doctor's responsibility, but it changes the question. Instead of "who made the error?", the useful question becomes "at what point along the way was the information lost, and what would keep it from being lost again?". Teams that ask this question regularly tend to find repeated failures, which can be corrected with simple changes to routine.

The diseases involved

A common idea is that diagnostic error mainly involves rare diseases or exotic presentations. The data show the opposite. A study that analyzed 190 diagnostic errors in primary care found 68 different diagnoses, most of them common conditions (Singh, 2013).

The most frequent missed diagnoses were these (Singh, 2013):

• pneumonia;
• decompensated heart failure;
• acute kidney injury;
• cancer, especially primary cancer;
• urinary tract infection or pyelonephritis.

In lawsuits, the pattern is similar. The most cited conditions were cancer and heart attack in adults and meningitis in children (Wallace, 2013). These are diseases every family doctor knows well, which reinforces that the problem is rarely a lack of knowledge about the disease.

This has an important practical implication. Training the recognition of rare diseases does little to reduce error in primary care. What helps most is recognizing the atypical presentations of common diseases: the older adult's pneumonia that arrives only with confusion, the heart failure that shows up as a night cough, the urinary tract infection that manifests as a fall. Most errors live in these less obvious versions of what is frequent.

Cancer deserves special attention, because the delay is usually long. In a study of 513 people with colorectal cancer, 161 had at least one missed opportunity for diagnosis, with an average of 4.2 missed opportunities and 5.3 clues not followed per person (Singh, 2009). Anemia was the clue associated with the longest median delay, 393 days, and age over 75 was associated with a higher chance of a missed opportunity (OR 2.3) (Singh, 2009).

A large collaboration project between institutions gathered 836 reports of diagnostic error and identified 661 disease-specific pitfalls, grouped into 21 general categories (Schiff, 2022). Among the most common were confusing one disease with another, not recognizing the limitations of a test and not considering an atypical presentation (Schiff, 2022).

Inside the appointment

Most errors arise at the meeting between the doctor and the person. In the study with 190 errors, about 8 in 10 involved failures in the consultation process itself (Singh, 2013). Among these failures, problems in the history appeared in more than half, in the physical examination in almost half and in ordering tests in more than half (Singh, 2013).

The detail is revealing. It is not sophisticated tests that were missing, but questions that were not asked, an incomplete physical examination and a simple test that was not ordered. In practice, these failures appear when the consultation is short, the main complaint takes up all the time and the doctor closes the hypothesis too early.

Think of the person who arrives with back pain. If the history does not ask about fever, weight loss, night pain or a history of cancer, and the examination does not include palpation and a basic neurological assessment, the hypothesis of mechanical pain becomes certainty without having been tested. Most of the time it will be right. The problem is the minority in which it will not be, and for which the consultation left no door open.

Leaving the door open is simple and fits into short consultations. It is enough to tell the person what to expect from the course and which signs should bring them back before the scheduled time, such as fever, loss of strength, pain that worsens at night or unexplained weight loss. This sentence, which takes less than a minute, works as a safety net: if the initial hypothesis is wrong, the person knows how to recognize the moment to seek help again, and the doctor who sees them next finds in the record what was agreed and what needs to be reviewed. In primary care, where the uncertainty of the first consultation is the rule, this net is part of the diagnosis, not a detail.

A study with 56 patients with breast and colon cancer whose diagnosis was delayed helps to understand the type of failure (Poon, 2012). Cognitive errors were present in 53 patients and logistical failures in 31 (Poon, 2012). The choice of diagnostic strategy was the most common cognitive error, both in the initial consultation, in 25 of 56 patients, and in the interpretation of results, in 22 of 50 (Poon, 2012).

Early closure is one of the most studied traps. The person arrives with a complaint that resembles a common condition, the doctor fits it into this hypothesis and stops looking. The text on why we misjudge probability discusses how this shortcut affects the interpretation of tests.

Another source of failure is the test interpreted out of context. A normal result can reassure too much when the probability of disease was high, and a slightly altered result can be ignored when it is part of a pattern. In the study on breast and colon cancer, the interpretation of results was a moment of error as frequent as the initial consultation (Poon, 2012).

The text on premature closure explains why reasoning tends to stop at the first plausible hypothesis and how a short pause helps to reopen the investigation.

Tests, referrals and results

Outside the consultation, the error appears in the handoffs. In the study with 190 errors, failures in referral were present in about 1 in 5, and failures in follow-up and in tracking of tests in about 1 in 7 each (Singh, 2013). In just under half of the errors, more than one process failed at the same time (Singh, 2013).

Tests ordered and not done are a concrete and frequent failure. A clinical trial followed 725 patients of 16 primary care doctors with 1016 delayed imaging tests (Weingart, 2020). In a sample, the doctors were unaware of the delay in 23 of 30 tests reviewed (Weingart, 2020).

The study tested an electronic tracking system that alerted about delayed tests. The conclusion at 90 days was about 1 in 5 in the intervention group versus about 1 in 4 in the control group, with no effect of the system in the multivariate analysis (Weingart, 2020). The result shows that alerting is not enough: someone needs to have the responsibility and the time to act on the alert.

In the study on colorectal cancer, the failures concentrated on the doctor-patient encounter and on following up the clues that were found (Singh, 2009). An anemia seen and not investigated, a bleeding attributed to hemorrhoids without examination and a referral that was never scheduled are examples of clues that were left behind.

These failures share one characteristic: each one, on its own, seems small. The blood count not looked at, the colonoscopy request that stayed in the queue and the follow-up visit that was never booked do not draw attention in daily practice. Only when the diagnosis finally appears, months later, is it possible to see the whole sequence. That is why safety at this stage depends less on one person's attention and more on routines that work even when everyone is busy.

The follow-up that never closes

Primary care has an advantage that other levels do not have: continuity. The same team can see the person several times, and each return visit is a chance to review the hypothesis. This advantage only works when the follow-up closes, that is, when someone checks whether the symptom improved, whether the test was done and whether the result was seen.

A study with 287 people with colorectal cancer and 828 controls showed that, in the year before diagnosis, people with cancer had more consultations (Brandenbarg, 2018). Almost half of them had two or more contacts for digestive complaints, against about 1 in 8 among the controls (Brandenbarg, 2018).

The authors identified traps that repeated in these consultations (Brandenbarg, 2018):

• attributing the symptoms to an already known disease or to a medication in use;
• keeping the first diagnosis even when the course did not match it;
• not reacting to alarm signs recorded in the chart;
• failing to order tests when the symptoms persisted;
• not agreeing with the person when and why to come back.

The person who returns several times with the same complaint is giving a signal. When the symptom does not improve as expected, the useful question is whether the initial hypothesis still makes sense. The text on strategies for deciding without certainty shows how to use the return visit as part of the reasoning.

One simple practice helps: when reviewing the chart before seeing someone who returns with the same complaint, look at how many times it has already been recorded and what was done each time. If the symptom appears for the third time without a clear diagnosis, the consultation stops being an ordinary return visit and becomes an opportunity to start the investigation again, without inheriting the conclusion of previous consultations.

Patients who return many times also stir emotions in the team, such as tiredness or irritation, and these emotions affect reasoning. The text on difficult patients and diagnostic accuracy shows what the experiments found.

Who is most exposed

Some patients are at higher risk. In the study with 1474 outpatient patients, multimorbidity was associated with a higher chance of reporting a diagnostic error (aOR 1.83) (Aoki, 2020). People with several diseases have more symptoms, more medications and more possible explanations for each complaint, which makes it easier to attribute something new to something old.

Age also weighs. In the study on colorectal cancer, people over 75 had a higher chance of a missed diagnostic opportunity (OR 2.3) (Singh, 2009). In older adults, symptoms such as tiredness, weight loss and anemia are easily attributed to age or to other diseases.

Factors belonging to the patient do appear, but in a smaller proportion. In the study with 190 errors, they were present in about 1 in 6 (Singh, 2013). Missing the return visit, not doing a test or not reporting a symptom contribute to the error, but rarely explain it alone.

Even when the patient misses the return visit, it is worth asking whether the system made the return easier or harder. Times that do not fit around work, difficulty booking an appointment and lack of explanation about the importance of the return visit are failures of the service, even if they appear as the patient's absence.

The table summarizes the stages in which the error usually arises and what helps in each one.

Stage	Typical failure	What helps
History taking	A question that was not asked	Review the alarm symptoms of the main complaint
Physical examination	Incomplete or omitted examination	Examine the system linked to the complaint, even with little time
Ordering tests	A simple test not ordered	Ask which serious diagnosis must be ruled out
Results	Result not seen	A defined responsible person and a checking routine
Referral	A request that does not arrive or is not scheduled	Confirm the scheduling and the return of the information
Follow-up	A symptom that persists without a new assessment	Agree on when to return and revisit the hypothesis at the follow-up

The report on diagnostic safety also highlights the outpatient setting as a priority scenario; the text on primary care and global burden summarizes what it says about the topic.

What helps reduce error

Primary care operates within a system that is often chaotic, with short consultations, many demands and scattered information (Olson, 2021). A proposal to reduce errors in this setting highlighted three practices that can be measured: lists of diagnoses that must not be missed, alarm signs and known pitfalls for each complaint (Olson, 2021).

Pitfalls by disease also guide practice. The project with 836 reports proposes that knowing the most common pitfalls of each condition helps the physician recognize when they are falling into one of them (Schiff, 2022). In the study on breast and colon cancer, following the guidelines would have prevented at least one error in about 2 out of every 3 patients (Poon, 2012).

The system matters as much as the reasoning. Having someone responsible for checking results, a routine for delayed tests and a path to confirm that the referral arrived reduces the failures that happen outside the consultation. The trial with electronic alerts shows that technology alone does not solve it, but it can help when there is someone to act on it (Weingart, 2020).

In the routine of a unit, this can mean a short, periodic meeting to review pending tests, a team member responsible for calling those who did not take an important test and a clear record of which professional ordered each test. These are unsophisticated measures, but they target exactly the passages where error appears most.

The person can also take part. Explaining what to expect from the course of the condition, when to return and what to do if a test result does not arrive makes the patient one more safety point. In the Poon study, a patient advocate, someone who follows up and pushes for the follow-up, could have helped in about half of the patients (Poon, 2012).

When the error happens, the way of dealing with it also matters. The text on how to structure the conversation about the error shows how to conduct this conversation, and the text on blame, anonymity and the climate of the room discusses how the team can learn from the cases.

The error also affects the one who made it. Physicians involved in diagnostic errors can suffer guilt, anxiety and fear, which the text on what the second victim is describes in detail. Caring for the team is part of learning from the error.

Frequently asked questions

How often does diagnostic error happen in outpatient care?

One estimate found about 1 in every 20 adults seen per year, and about half of these errors had the potential for serious harm (Singh, 2014).

Do the errors involve rare diseases?

Most involve common conditions, such as pneumonia, heart failure, acute kidney failure, cancer and urinary infection (Singh, 2013).

In which part of the process does the error usually originate?

Most of the time, within the consultation, with failures in the history, the physical examination and the ordering of tests (Singh, 2013).

Who is at greater risk?

People with several diseases reported more diagnostic errors (aOR 1.83) (Aoki, 2020). In colorectal cancer, people over 75 had more missed opportunities (Singh, 2009).

Do electronic alerts solve the problem of delayed tests?

On their own, they are not enough. In a clinical trial, a tracking system did not increase the completion of delayed tests (Weingart, 2020).

What can the physician do during the consultation?

Use lists of diagnoses that must not be missed, alarm signs and known pitfalls for each complaint (Olson, 2021).

Final checklist

• For each complaint, ask which serious diagnosis must not be missed.
• Review the alarm signs before settling on the hypothesis.
• Be suspicious when the person returns with the same complaint without improvement.
• Define who checks each result and each delayed test.
• Confirm that the referral arrived and was scheduled.
• Agree with the person on when to come back and what to do if something does not arrive.

References

1. Aoki T, Watanuki S. Multimorbidity and patient-reported diagnostic errors in the primary care setting: multicentre cross-sectional study in Japan. BMJ Open. 2020. PMID 32819954. DOI 10.1136/bmjopen-2020-039040.
2. Brandenbarg D, Groenhof F, Siewers IM, et al. Possible missed opportunities for diagnosing colorectal cancer in Dutch primary care: a multimethods approach. Br J Gen Pract. 2018. PMID 29203683. DOI 10.3399/bjgp17X693905.
3. Olson APJ, Linzer M, Schiff GD. Measuring and Improving Diagnostic Safety in Primary Care: Addressing the "Twin" Pandemics of Diagnostic Error and Clinician Burnout. J Gen Intern Med. 2021. PMID 33575908. DOI 10.1007/s11606-021-06611-0.
4. Poon EG, Kachalia A, Puopolo AL, et al. Cognitive errors and logistical breakdowns contributing to missed and delayed diagnoses of breast and colorectal cancers: a process analysis of closed malpractice claims. J Gen Intern Med. 2012. PMID 22610909. DOI 10.1007/s11606-012-2107-4.
5. Schiff GD, Volodarskaya M, Ruan E, et al. Characteristics of Disease-Specific and Generic Diagnostic Pitfalls: A Qualitative Study. JAMA Netw Open. 2022. PMID 35061037. DOI 10.1001/jamanetworkopen.2021.44531.
6. Singh H, Daci K, Petersen LA, et al. Missed opportunities to initiate endoscopic evaluation for colorectal cancer diagnosis. Am J Gastroenterol. 2009. PMID 19550418. DOI 10.1038/ajg.2009.324.
7. Singh H, Giardina TD, Meyer AN, et al. Types and origins of diagnostic errors in primary care settings. JAMA Intern Med. 2013. PMID 23440149. DOI 10.1001/jamainternmed.2013.2777.
8. Singh H, Meyer AN, Thomas EJ. The frequency of diagnostic errors in outpatient care: estimations from three large observational studies involving US adult populations. BMJ Qual Saf. 2014. PMID 24742777. DOI 10.1136/bmjqs-2013-002627.
9. Wallace E, Lowry J, Smith SM, et al. The epidemiology of malpractice claims in primary care: a systematic review. BMJ Open. 2013. PMID 23869100. DOI 10.1136/bmjopen-2013-002929.
10. Weingart SN, Yaghi O, Barnhart L, et al. Preventing Diagnostic Errors in Ambulatory Care: An Electronic Notification Tool for Incomplete Radiology Tests. Appl Clin Inform. 2020. PMID 32294771. DOI 10.1055/s-0040-1708530.