Reasons for acute referrals to hospital from general practitioners and out-of-hours doctors in Norway: a registry-based observational study

The study is registry-based, using data from national health registries covering the whole population in Norway in 2017.

All GP and OOH contacts result in a claim to the Control and Payment of Reimbursement to Health Service Providers database (KUHR), managed by The Norwegian Health Economics Administration (HELFO). A single claim contains the patient’s national identification number, time and date for the contact, and if the contact is a simple contact (e.g. telephone contact), an office consultation, or a home visit. Also, it is mandatory to include one or more diagnostic codes according to International Classification of Primary Care (ICPC-2) [18]. This diagnosis will be routinely transferred to the referral letter and recorded as the reason for referral. In our material, 17% of the KUHR claims contained more than one diagnosis. We used the main (primary) diagnosis in our analyses. The ICPC-2 codes are divided into chapters denoted by a letter that identifies an organ system, followed by a two-digit number referring to either a symptom (code 01–29) or a disease (code 70–99). ICPC-2 was developed to describe the reasons for encounter in primary health care or general practice and reflects the content of primary care, last updated in 2016 [19, 20].

Psychiatric hospitals were not included. We therefore use the term somatic hospital admissions in our study. All hospital stays are recorded in the Norwegian Patient Registry (NPR), which includes information on patient’s national identification number, time and date of the admission, degree of urgency, and one or more discharge diagnoses using the International Statistical Classification of Diseases and Related Health Problems version 10 (ICD-10). For the hospital discharge diagnoses we used the main diagnosis with three characters. An acute hospital admission was defined by the NPR’s data form as an admission required immediately or within 24 h and lasting for more than 24 h. These admissions were then included if they were related to a GP or OOH contact.

In this study we defined a GP or OOH doctor contact as a consultation or a home visit, telephone contacts were excluded. All such contacts with GPs (N = 14,457,247) or OOH doctors (N = 1,361,731) during 2017 were included. For GPs this represent both acute and follow up contacts.

Statistics Norway (SSB) created a pseudo-anonymized identification number which replaced the national identification number in the KUHR and NPR databases. Thereby data from both sources could be combined without revealing the patients’ identities.

A GP or OOH-doctor contact in KUHR was defined as the referral contact for a patient if it occurred within 24 h before an acute admission in NPR. By this definition we identified 265,518 referrals to hospital from a GP or an OOH doctor.

A suitable grouping of ICPC-2 codes into presenting complaints in emergency departments (EDs) has been described by Malmström et al. [21]. We applied this for the symptom groups abdominal pain, chest pain and shortness of breath. The abdominal pain symptom group was defined by the following codes: abdominal pain/cramps general (D01), abdominal pain epigastric (D02) and abdominal pain localized other (D06). The chest pain symptom group was defined by these codes: chest pain NOS (A11), heart pain (K01), pressure/tightness of heart (K02) and cardiovascular pain NOS (K03). Shortness of breath was included pain in respiratory system (R01), shortness of breath/dyspnoea (R02), wheezing (R03), and breathing problem other (R04). This group was named after the most frequent reason, shortness of breath.

We used the referral disease diagnoses pneumonia (R81), appendicitis (D88), acute myocardial infarction (K75), and stroke (K90) to study associations between a specific disease diagnosis given at referral and the discharge diagnoses after hospital stay.

Numbers and frequencies for ICPC-2 chapters and the 30 most common ICPC-2 diagnoses that led to a referral were obtained. Referral rates were calculated by dividing GP and OOH contacts leading to a referral by all contacts with the same diagnosis.

For referrals in the abdominal pain symptom group, we identified the ten most frequent discharge diagnoses. Generalized linear model (GLM) log-binomial regression were used to estimate relative risk (RR) of the patient being discharged with each of these diagnoses compared with all other discharge diagnoses after a referral with abdominal pain symptom. The patients’ ages, genders and if the patient had been referred by a GP or OOH doctor were used as explanatory variables and were adjusted for each other. RR was calculated for females with males as reference, age was divided by 10 and then used as a continuous variable giving RR with a 10-years increase, and OOH doctor referrals were compared with GP referrals as reference. The RRs were presented with 95% confidence intervals (CI). For referrals in both the chest pain symptom group and shortness of breath symptom group the ten most frequent discharge diagnoses were identified, and equivalent GLM were used for each of the ten discharge diagnoses.

For the four disease-specific referral diagnoses pneumonia (R81), appendicitis (D88), AMI (K75) and stroke (K90) we conducted frequency analyses for discharge ICD-10 diagnoses representing > 3% of the patients referred with these diagnoses, respectively. GLM were used to estimate the RR for the discharge diagnoses. For the referral diagnosis pneumonia (R81) the corresponding discharge ICD-10 diagnosis pneumonia was used as comparison. For appendicitis (D88), AMI (K75) and stroke (K90) we used respectively acute appendicitis (K35), AMI (I21) and cerebral infarction (I63) as comparison in the analyses. Explanatory variables were gender, age, and referral from an OOH doctor or a GP. For three of the models the log-binomial regression did not converge, and a Poisson regression with robust variance estimates was applied to estimate RR [22].

The analyses were performed using Stata 16.1. (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.)

The median patient age for the abdominal pain symptom group referrals was 46 years (Table 3), and 60% were women. Every fourth patient referred with abdominal pain was discharged with a similar symptom diagnosis from ICD-10, abdominal and pelvic pain (R10). The second and third most frequent discharge diagnoses were acute appendicitis (K35) (12%) and cholelithiasis (K80) (6%).

For the abdominal pain symptom group, there was a higher relative risk for the discharge diagnosis to be ileus (K56) if the patient was referred from an OOH doctor compared with a referral from a GP (RR = 2.12 [95%CI: 1.84–2.45]). The opposite was found for diverticular disease (K57) with a lower relative risk of being discharged with diverticular disease if referred from OOH compared to GP (RR = 0.51 [95%CI: 0.46–0.56]). The relative risk for the discharge diagnosis of abdominal and pelvic pain (R10) was higher for women compared to men (RR = 1.38 [95%CI: 1.32–1.44]), but lower for acute appendicitis (K35) (RR = 0.59 [95%CI: 0.55–0.62]), acute pancreatitis (K85) (RR = 0.54 [95%CI: 0.47–0.61]) and calculus of kidney and ureter (N20) (RR = 0.40 [95%CI: 0.34–0.47]).

The median patient age in the chest pain symptom group was 62 years, and 44% were women (Table 4). One third of patients referred with chest pain were discharged with the ICD-10 code describing the similar symptom, pain in throat and chest (R07). AMI (I21) accounted for 12%, and angina pectoris (I20) for 10%.

Among patients in the chest pain symptom group, a discharge diagnosis of heart failure (I50) was associated with higher age (RR = 2.13 [95%CI: 1.93–2.36]). Women referred with chest pain had a lower relative risk of being discharged with a diagnosis related to ischemic heart disease: AMI (I21) (RR = 0.54 [95% CI 0.50–0.59]), angina pectoris (I20) (RR = 0.70 [95%CI: 0.64–0.77]) and chronic ischemic heart disease (I25) (RR = 0.47 [95%CI 0.40–0.55]) compared with men.

In the shortness of breath symptom group the median age was 70 years, and 50% were women (Table 5). Heart failure (I50), pneumonia (J12–18) and chronic obstructive pulmonary disease (J44) were the most common discharge diagnoses with 12, 11 and 8%, respectively. The discharge diagnoses after referral for shortness of breath showed a larger variation compared to abdominal pain and chest pain and had fewer discharges with a symptom describing diagnosis. Among patients discharged with the diagnosis acute bronchiolitis (J21) 75% were less than 2 years.

The four referral diagnoses: pneumonia (R81), appendicitis (D88), AMI (K75), and stroke (K90) accounted for 16,811 admissions (6% of all). 59% of the patients referred with the diagnose pneumonia (R81) were discharged with a corresponding ICD-10 pneumonia-diagnosis (J12–18) (Table 6). Only 1% of patients referred with pneumonia were discharged with pulmonary embolism (I26), and 0,5% with AMI (I21).

Of patients referred with appendicitis (D88) 51% were discharged with the corresponding ICD-10 diagnosis acute appendicitis (K35), and 24% with the symptom describing diagnosis abdominal and pelvic pain (R10) (Table 7).

For patients referred with the diagnosis AMI (K75), 43% were discharged with the corresponding ICD-10 diagnosis AMI (I21), 12% with pain in throat and chest (R07), and 7% with angina pectoris (I29) (Table 8). The discharge diagnosis heart failure (I50) was associated with higher age (RR = 1.96 [95%CI: 1.51–2.55]).

After referral with the diagnosis stroke (K90), 30% of the patients were discharged with the ICD-10 diagnosis cerebral infarction (I63), 10% with transient cerebral ischemic attack (TIA) (G45) and 4% with intracerebral haemorrhage (I61), whereas 56% had other diagnoses (Table 9).

Of all GP contacts only 1% resulted in a referral for acute admission to somatic hospital, whereas OOH doctors referred 11%. Referral rates for GP and OOH contacts vary greatly by referral diagnosis. Abdominal pain, chest pain and shortness of breath were the dominant symptom diagnoses in referrals and had a considerable variation in discharge diagnoses. For both abdominal pain and chest pain the corresponding symptom diagnosis was the most frequent. For the referral diagnoses pneumonia, appendicitis, AMI and stroke the corresponding discharge diagnoses were dominant. Some of the discharge diagnoses were associated with high patient age, female gender, and if the referral came either from a GP or an OOH doctor.

The study includes all residents in Norway in 2017, all GP and OOH doctor contacts and all acute referrals to somatic hospitals. Therefore, there is no selection bias. The data sources are registries containing activity data delivered to manage funding of primary and specialist healthcare, and therefore presumably accurate and complete. This ensures the representativity for the situation in Norway.

In our analyses we used only the primary diagnosis. Therefore, we may have missed some information on the reason for referral in cases with more than one diagnosis given.

Diagnosis coding may vary between doctors, both in general practice and in OOH services. In a clinical situation with abdominal pain when appendicitis is suspected, the physician may apply either a symptom diagnosis abdominal pain/cramps general (D01) or the disease diagnose appendicitis (D88). Probably, in some cases a more precise or severe diagnosis than the right one according to diagnosis criteria may be used to prepare the hospital receiving the patient. Distinction between symptom and disease diagnosis in referrals must be interpreted with caution. In a prehospital setting in helicopter medical services in Finland, ICPC-2 coding was tested for inter-rater reliability for six written fictional cases [23]. The researchers found an overall agreement of only 52%. Nevertheless, research in Norwegian general practice concludes that GPs’ use of ICPC-2 codes corresponds well with the patient records in 85% of the consultations [24]. This supports the use of ICPC-2 diagnosis to describe the reason for referral. Our results reveal a coherence between referral ICPC-2 diagnosis and the hospital discharge diagnosis, supporting the design.

The linkage between the GP and OOH doctor contacts in KUHR and hospital admission in NPR has some uncertainties. The primary care contact prior to admission might be random, and not related to the admission. In a previous study we found a distinct accumulation of GP and OOH contacts within 24 h before acute hospital admission [9, 10]. This indicates that the assumption that a contact within 24 h before an admission is the referral contact, is valid. The present study supports this further by demonstrating the relationship between the referral and the discharge diagnoses.

Our overall OOH referral rate (11%) is higher than the rate described in studies from England (7 and 10%) [25, 26], and Denmark (4–8%) [27], but lower than a study from a single OOH primary care centre in Norway (14%) [5].

Diagnoses from the ICPC-2 chapter general and unspecified conditions (A) was most frequent for referrals from OOH doctors, similar to findings from other studies [11, 27]. The variation of referral rates for different ICPC-2 diagnoses reflects the severity of the conditions as well as the nature of the services. For atrial fibrillation, the GPs’ referral rate was only 0.01 whereas the OOH referral rate was 0.45. According to national guidelines, newly discovered atrial fibrillation should be referred for immediate further investigations and treatment, whereas chronic atrial fibrillation requires comprehensive GP follow up, but not admission. There is a similar effect for GP contacts with a diagnosis of appendicitis (D88) where as much as 70% were not referred. This probably represent GP follow up, rather than acute assessment. These figures underline the different nature of GP daytime services and OOH services, and therefore the proportions of contacts for a diagnosis related to admission are not directly comparable.

Abdominal pain is a common symptom in general practice, and the second most frequent reason for contacting OOH services [13, 21, 28]. In a Norwegian study of patients offered acute appointment with a GP because of abdominal pain, the GP referred 26% acutely to hospital [13]. In our study only 4% of GP patients with abdominal pain/cramps or localized abdominal pain were acutely referred to hospital, illustrating that GPs provide care for both acute and chronic complaints. OOH services provide emergency primary healthcare with only acute appointments and correspond better with the patients in the other Norwegian GP study. For OOH patients with abdominal pain/cramps or localized abdominal pain, 26 and 24% were referred in our material, which fits well with the GP study.

We found that the referral rates for chest pain and heart pain from OOH doctors were 0.37 and 0.66, respectively. In a prospective study from Norwegian OOH services 50 out of 100 patients presenting at the casualty clinic with chest pain as the main symptom were referred to hospital [12]. However, we do not know the referral diagnosis for these referrals.

For referrals with abdominal and chest pain symptoms, the most frequent ICD-10 discharge diagnoses were the corresponding symptom-based diagnoses abdominal and pelvic pain (R10) and pain in throat and chest (R07). Our previous study showed that these discharge diagnoses were the overall second and third most frequent discharge diagnoses after acute admissions to hospital, irrespective of referral agent [9, 10]. Such extensive use of symptom diagnoses both in referrals and discharges could serve as a reminder of the diagnostic challenges in both primary care and in hospitals but might as well be an indication of defensive medicine.

Only 1% of patients referred with chest pain were discharged with gastro-oesophageal reflux disease (K01). This could be due to effective gatekeeping treating these patients in primary care and is in line with previous research from Norway where 5 out of 100 patients attending OOH services with chest pain were diagnosed with dyspepsia by OOH doctor, and none were referred to hospital [12].

Chest pain is a core presenting symptom for AMI for both women and men, although atypical symptom presentations in women have received increased attention [15, 29, 30]. We found that women referred with a chest pain diagnosis had a lower relative risk of being discharged with a discharge diagnosis of ischemic heart disease compared with men. This adds to previous knowledge of gender differences in acute ischemic heart disease presentation.

GPs accounted for a somewhat larger share of the disease-specific referral diagnoses pneumonia (R81), appendicitis (D88), AMI (K75), and stroke (K90) than the symptom groups abdominal pain, chest pain, and shortness of breath, suggesting that GPs are more likely than OOH doctors to use specific disease diagnoses. This could be due to better knowledge of the patient’s history or better diagnostic facilities.

For referrals with diagnoses pneumonia, appendicitis and AMI the most frequent discharge diagnoses were the corresponding disease diagnosis. Correspondingly, referrals with the diagnosis stroke were often discharged with a diagnosis of cerebral infarction, TIA, or intracerebral haemorrhage. This indicates that when the referral diagnosis is a disease-specific diagnosis, the primary care doctor is more certain of a specific disease compared with cases where a symptom diagnosis is used. Hospital doctors could take this into consideration when receiving patients referred acutely from primary care with a disease-specific diagnosis.

Patients from OOH services with the referral diagnosis AMI are less likely to be discharged with the diagnosis angina pectoris or chronic ischemic heart disease than patients from general practice. Angina pectoris and chronic ischemic heart disease typically have a less acute presentation, leading to contact with the GP rather than the OOH services, again illustrating that GPs treat cases with a lower degree of urgency.

An adequate gatekeeper role for acute hospital admissions must balance the task to discover all patients with serious conditions with the risk of unnecessary admissions [5, 31]. The aim is to avoid missed diagnosis (false negatives) at an acceptable level of false positives where no or minor disease are revealed in hospital. Low risk taking would lead to defensive medicine, increasing costs without gained health and overdiagnosis. On the other hand, too restrictive referral practice would lead to an increased number of undiagnosed severe conditions. Such underdiagnosis will severely affect the individual patient’s health. Finding the right balance in referral practice is a major challenge for primary care doctors performing a gatekeeper function in prehospital acute care. Continuity of care in general practice may help in this task. Length of patient continuity with a named regular GP is associated with lower use of OOH services, fewer hospital admissions and even lower mortality [32].

Our study reveals that one third of patients referred with chest pain and one quarter of patients referred with abdominal pain were discharged with a symptom diagnosis, hence no severe condition was revealed at hospital. Likewise, 24% of patients referred with appendicitis, and 12% referred with AMI were discharged with a symptom describing discharge diagnosis. We believe this to be a reasonable level of accuracy, but attention to this topic should be high and an objective for further research. The topic of referral practice should be emphasized in medical education, and policy makers should be aware of the issue over- and underdiagnosis.

Referrals to hospital are always a matter of clinical medical assessment. In a gatekeeping system this also applies for emergencies. Both GPs and OOH doctors seem to perform gatekeeping for acute hospital admissions based on their setting and the different patient populations. Our findings suggest that there should be an accept of more symptom-based referrals from OOH services. The GP’s knowledge of the patient’s medical history is valuable also when performing gatekeeping for acute referrals to hospital. The large variation of referral diagnoses implies that a broad medical competence is necessary when assessing emergencies.