To our knowledge, this is the first study addressing the epidemiology of shaft fractures of femur and humerus, studying both birth-related and later fractures during the first year of life, with account taken of association to perinatal characteristics. As far as we know, this is also the largest population-based report on such fractures, and the first report on the incidence of humerus shaft fractures diagnosed during infancy but after birth.
The overall incidence of birth-related shaft fractures of femur and humerus in the present study was 0.024‰ and 0.101‰, respectively. These results differ substantially from previously reported hospital case studies with incidence rates of 0.13–0.14‰ [1, 2] and 0.04–0.06‰ [2, 3]. Shaw et al.  reported that 20% (n = 5) of humerus shaft fractures in children under the age of one year were birth-related, compared with 60% in our study. It is unlikely that previous studies have had hidden cases, while hidden cases may exist with our register design. However, both the small sample sizes of previous reports [1–3, 11] and the differing settings might have contributed to the discrepancies compared with our population-based findings. Another reason could be our definition of birth-related fractures, which also included fractures identified in the early neonatal period without fall accidents, but not coded as birth-related. The existence of such cases indicates a possible underestimation of shaft fractures that might be associated with birth.
One of the issues emerging from this study is the challenges related to epidemiologic surveillance of birth-related femur and humerus shaft fractures, especially birth-related long bone fractures. The current P codes of birth injury to the skeleton in ICD-10 encompass the categories femur fractures and other long bone fractures. The ICD-11 released in 2018 has only one category: Birth injury to long bones (KA45.6). Our findings support the use of double coding, i.e., adding S codes, for improved precision of shaft diagnoses. Diagnosies of birth-related fractures might be delayed for femur fractures, [1, 14] especially among the preterm-born,  and for humerus shaft fractures [3, 8, 11]. We have previously reported that some birth-related fractures of femur or other long bones are diagnosed after the first week of life and in some cases at even higher age . Our findings show an overlap between birth-related femur fractures and other long bone fractures. Whether these double diagnoses are due to coding errors, the identification of a cause, or delayed diagnosis cannot be ascertained, due to the study design with no access to patient records. In a clinical perspective, assessment of whether a fracture is birth-related or not has less importance for treatment; however, if the diagnosis is delayed and abuse is a potential cause, this can be of greater importance [8, 16].
The incidence of femur shaft fracture not related to birth found in this study (0.154‰) was in line with previous findings in Sweden . This is higher than what was reported from the UK by Talbot et al.; however, the authors stated that there might be an underestimation of the incidence due to missed cases in their dataset obtained from the Trauma and Audit Research Network and exclusion of children with open fractures and associated injuries . In these previous studies, the risk of a femur shaft fracture was reported as higher among boys than among girls after the age of one year, but neither this study nor these past studies found any statistically significant sex differences during infancy [4, 5]. We cannot explain why boys in our cohort had a higher risk of birth-related diaphyseal humerus fractures but not of femur shaft fractures. Sex is not always reported but – interestingly – birth-related humerus shaft fractures previously reported by Basha et al.  and Sherr-Lurie et al.  were all in boys.
The current study found an incidence of humerus shaft fracture later during infancy of 0.073‰. A previous hospital study reported that among 101 children younger than 12 months treated at an emergency clinic, 2 patients were diagnosed with a humerus shaft fracture . Another study reported an incidence of 0.17‰ (n = 3) in children aged < 24 months .
Our cases shared perinatal risk factors confirming the state of knowledge. For femur shaft fracture, these factors were breech presentation [1, 2, 6], multiple birth, and prematurity [1, 6], and for humerus shaft fractures, they were vacuum-assisted delivery, shoulder dystocia, and heavy birth weight, as reported for other birth-related long bone fractures . A high proportion of cesarean delivery in connection with shaft fractures [1–3] was not evident from our findings. Among those with a femur shaft fracture during the first to fifth month, we found a preponderance of the perinatal risk factors breech, prematurity, and SGA, and among those with humerus shaft fracture, shoulder dystocia, and heavy birth weight. Few perinatal risk factors were seen among those with femur or humerus shaft fractures diagnosed from the sixth month of life. For the early femur shaft fractures, this might be interpreted as a sign of bone fragility; however, this cannot be ascertained, due to the study design. We have previously shown that long bone fractures, not specifically shaft fractures, among infants of less than 6 months of age, especially those without accidents reported, are associated with risk factors for a metabolic bone disease, such as being born preterm . Case reports have described rickets as being associated with humerus shaft fracture [19, 20] and femur shaft fractures . Case series have reported long bone shaft fractures associated with metabolic bone disease of infancy . We found only few shaft fractures with a diagnosis of rickets or vitamin D deficiency, as previously reported for fractures of any type .
Femur and humerus shaft fractures are closely associated with abuse diagnosis among infants. Our proportion of femur shaft fractures attributed to abuse, 9%, is higher than in the Swedish nationwide study, which reported a figure of 4.2%  probably because we included those having an entry in the RCWM. These numbers can be compared with those in previous studies reporting that 21–31% of femur fractures (including fractures of the entire femur) in children under the age of 1 year were caused by child abuse [12, 23]. Shaft fractures of the humerus are also common in connection with abuse . Our outcome of 10% of shaft humerus fractures having been assessed as abuse is lower than in a U.S. hospital study and a UK multicenter hospital study, which both reported that one third of humerus shaft fracture in children under the age of one year were associated with probable or indeterminate abuse [11, 12]. Whether this discrepancy is due to different criteria for determination of abuse or a true difference cannot be ascertained.
In the absence of other abuse-suspected findings, e.g., bruises, the conclusion that a femoral fracture or humerus fracture in an infant has arisen as a result of abuse cannot be based solely on exclusion of accidental trauma. The importance of making a thorough risk factor analysis, including bone fragility disorders, is underlined by the fact that routinely used radiological methods have low sensitivity for detecting low bone mineralization . Tests for metabolic bone disease such as measuring levels of calcium, phosphorus, alkaline phosphatase (ALP), parathyroid hormone (PTH), vitamin D, copper, and ceruloplasmin should be done routinely , and testing for genetic disease, e.g., osteogenesis imperfecta, should be considered.
Eight out of ten children with a femur shaft fracture occurring later in infancy had a reported accident; the vast majority had a reported fall accident as described earlier [4, 5], usually occurring in children aged 6–12 months. Our finding of 58% of children with a humerus shaft fracture having an accident reported is higher than the 29% reported by a hospital study (7/24) . However, much like in the study on femur shaft fractures by Talbot et al. , 29% of the children in our study had no reported accident.
Strengths and limitations
A major strength of this study was the population design derived from diagnoses in a national patient registry considered to have high validity . Although shaft fractures in infants have not been specifically validated in the patient registry, this fracture type can be considered to have high precision, although there is still the potential for misclassification bias or missing data. A strength was also the linkage to the RCWM to detect cases associated with abuse; we found cases that had no diagnosis of maltreatment, but had an entry in the out-of-home care register. A further strength was the linkage to a birth registry . The rate of birth fractures was estimated by combining femur and humerus shaft fractures diagnosed in days 1–7 and fractures diagnosed as birth-related, in the absence of accidents in the first week of life. The major weakness of the study was that we did not have information from patient records regarding treatment and cause of injury. This also limited our analysis of the clinical circumstances of shaft fractures registered as both birth-related (P coded) and other during infancy (S coded), and of shaft fractures in the early neonatal period not diagnosed as birth-related. Furthermore, the extent of testing for biochemical markers or genetic disease could not be determined, meaning that there might have been cases with an abuse diagnosis disorder in which proper differential diagnostics had not been undertaken.