Secondary use of hospital information system data for safe bedside radiography in terms of patient factors

Abstract

Objective:

The purpose of this study was to use patient data gathered by a hospital information system (HIS) to improve the safe performance of bedside radiography.

Method:

Hierarchical cluster analysis was used to investigate the factors of hospitalised patients who had undergone radiography in the X-ray room or at the bedside. Logistic regression analysis was then performed to quantify patient factors and calculate the probability of undergoing general radiography or bedside radiography.

Results:

Patients were grouped into six clusters by hierarchical cluster analysis on the basis of their factors. We found a remarkable difference between clusters for the ratio of bedside radiography. Results indicated that “types of transportation” and “level of mobility” related to the ratio of bedside radiography. Logistic regression analysis of the associations between the probability of undergoing bedside radiography and patient factors indicated that type of transportation and level of mobility were highly correlated with bedside radiography or general radiography.

Conclusion:

Our results suggested that the secondary use of HIS data for the quantitative evaluation of patient factors and implementation of those quantitative values in medical records may be useful for the safe performance of bedside radiography as well as providing a method of decision support for doctors to order bedside radiography.

Keywords

hospital information systems patient care management patient safety data analysis radiography

Introduction

In serious cases or for difficult patients, radiography can be performed at the bedside with a portable X-ray machine (Eklund et al., 2012). Radiological technology has been investigated with regard to the reduction of radiation exposure (Ben-Shlomo et al., 2016; Den Harder et al., 2016; Goodman and Amurao, 2012; Iezzi et al., 2012; McCollough, 2012; Sarti et al., 2012; Thabet et al., 2012), the positioning and clarity with which the area under observation is visualised (Hospers et al., 2009; Huang and Schweitzer, 2014; Kijowski et al., 2006; Siddiqui et al., 2014), and the improvement of image quality through advances in devices and systems (Hamer et al., 2005; Inamura and Kim, 2011; Körner et al., 2007; Uffmann et al., 2005; Utsunomiya et al., 2013). Recently, more hospitals have been switching from the use of X-ray film to filmless diagnosis via images displayed on monitors, meaning that images can be rapidly transferred after scanning, and clear images can be accessed on hospital information system (HIS) terminals at any time (Jeong et al., 2014; Utsunomiya et al., 2013). Consequently, even with imaging in the general wards using a portable X-ray machine, it is possible to immediately obtain a digital image of the same quality as one taken in the X-ray room. Therefore, it is possible to prevent the risk of patients falling or of sudden changes in condition during patient transfer from their bed to the X-ray room. Portable X-ray also enables imaging in a stable environment for the patient.

The Japanese population is currently aging at a rate unparalleled elsewhere in the world. Future projections of the age structure of the Japanese population anticipate that the proportion of older people aged 65 years and over will increase from the current (2016) figure of 27.5% to 33.9% in the next 20 years, and in another 40 years, the proportion of the population of non-working-age (young and older) people will almost equal the working-age population (National Institute of Population and Social Security Research, 2016). The time spent on radiography scanning increases for patients with age-related risk factors, and the number of aged patients has increased rapidly. In this aging society, assessments limited to radiography equipment and scanning techniques are insufficient to evaluate the safety of radiography scanning. It is essential that patient factors that may impede safety, such as diminished activities of daily living (ADLs) and comprehension, be taken into account in the evaluation of radiological technology. However, few reports are available on the investigation of this problem (Igarashi et al., 2010; Sasaki et al., 2014).

Aim of the current study

Secondary use of medical information for evaluating and improving the quality and safety of healthcare is also now an active subject of development (Yamamoto et al., 2008). Our aim in this study was to investigate the possibility of using data gathered by an HIS to provide optimal support of doctor’s orders for safe bedside radiography, taking into account patient factors.

Method

Participants and study period

Of the patients hospitalised once at Kagoshima University Hospital between April 1, 2013, and March 31, 2014, 5609 individuals aged ≥15 years old underwent either general or bedside radiography. They were stratified by age into three groups: the working-age population (age 15–64), the young-elderly population (age 65–74), and the older-elderly population (age ≥75). Patients aged 0–14 were excluded from the analysis owing to marked biases in their type of transportation around the hospital and level of mobility.

Definitions of terms

Kagoshima University Hospital began developing an HIS in 1984, and its electronic medical records system came into operation in 2006. One of its subsystems is the nursing care system, in which the ability of each patient to engage in ADL is recorded over time from admission to discharge. Other input parameters related to ADL include type of transportation, level of mobility, and direct nursing care, with data input almost in real time in accordance with changes in patient conditions. Type of transportation was used as an indicator of patient mobility, and level of mobility was used to indicate the ability of patients for ADL. These indicators are shown in Boxes 1 and 2. In addition to basic data on age and sex, the patient factors analysed in this study included type of transportation, level of mobility, whether or not they had undergone surgery, and mean nursing care time per day, all of which might have a major effect on bedside radiography. These factors were defined as follows:

Age groups: Three groups, working-age population (15–64 years), young-elderly population (age 65–74), and the older-elderly population (age ≥75).

The World Health Organization defines young people as those aged 0–14, working-age people as those aged 15–64, and older people as those aged ≥65 years. In Japan, older people are classified into two groups: young-elderly aged 65–74 and older-elderly aged ≥75 (Regulation of Act on Assurance of Medical Care for Elderly People, 2015).

Type of radiography: classified as either general or bedside. General radiography refers to radiography performed in an X-ray room in the Department of Radiology for patients admitted to general wards who visit the X-ray room. Bedside radiography refers to radiography performed at the bedside with a portable X-ray machine for patients admitted to general wards, emergency wards, or the intensive care unit.

Direct nursing care and time spent on direct nursing care (see Box 3): Uto and Kumamoto (2005) carried out a survey of the time spent on nursing care at Kagoshima University Hospital, allowing quantification of the amount of nursing care provided to patients. “Direct nursing care” refers to tasks that directly affect the patient, such as giving bed baths and treating wounds. Median values based on actual measured times were designated for the time required for blood pressure measurements and bed baths, which constituted individual subcategories of direct nursing care.

Box 1.

Types of transportation.

Term	Definition
Stretcher	Transport by stretcher
Escort	Transport by wheelchair
Independent	Walking independently

Box 2.

Level of mobility.

Term	Definition
I	Bedridden
II	Some movement in bed
III	Ability to walk within the room
IV	Ability to walk with complete independence

Box 3.

Nursing care activities.

Measurement	Time (minutes)	Medical examination assistance	Time (minutes)	Assistance in daily living	Time (minutes)
Measurement of body weight	3	Collection of venous blood sample	3	Application of cold compress	2
Measurement of pulse rate	3	Intravenous drip management	1	Bed bath	10
Measurement of blood pressure	3	Oxygen inhalation therapy	1	Foot bath	10
Measurement of respiratory rate	5	Catheter management	3	Stoma care	5
Measurement of oxygen	5	Sputum suction	5	Tube feeding	2
Measurement of urine volume	2	Wound treatment	5	Turning or positioning	2

Data analysis

Patient clustering by hierarchical cluster analysis

Hierarchical cluster analysis was carried out to investigate the factors of the participants, who were clustered into groups. The variables used, taken from data recorded in patients’ electronic medical records, were age, sex, type of transportation around the hospital, level of mobility, whether or not surgery was performed, and mean nursing care time per day, all of which were considered to be related to the amount of time and effort required for general and bedside radiography. The amount of the nursing care each patient received changed daily. Therefore, the total time spent on nursing care during the entire hospitalisation period was summed, and the mean nursing care time per day was calculated as an index of the level of time and effort required for nursing each patient. The distance between clusters was measured using the two quantitative values of age and mean nursing care time per day using Ward’s method of analysis. If an individual patient underwent radiography several times in the course of a single hospitalisation, the type of scan, level of mobility, and type of transportation on each day of scanning were used for analysis. In addition, for all patients who underwent surgery and patients in each obtained cluster group, we classified the image type as corresponding to before surgery, day of surgery, and within 3 days of surgery.

Probability of bedside radiography and quantification of patient factors by logistic regression analysis

Logistic regression analysis was used to predict whether patients should undergo bedside radiography or were capable of undergoing general radiography on the basis of patient data. The logistic model incorporated 10 different variables. The estimated weight coefficients of these 10 explanatory variables were calculated using the maximum-likelihood method to create a probability model for undergoing bedside radiography. For each of the 12 combinations of the four variables for level of mobility and the three variables for type of transportation, the mean value and standard deviation of the probability of undergoing bedside radiography in each of the groups were calculated and compared with the actual proportion of patients who underwent this type of radiography.

Software used for analysis

Data analyses were carried out with the statistical software SPSS Statistics version 22 and Open Office statistical software, R version 2.12.2.

Results

Patient clustering by hierarchical cluster analysis

Figure 1 shows the dendrogram constructed from the patient data. The patients were grouped into six clusters, with cluster 1 containing 244 patients; cluster 2, 2162 patients; cluster 3, 1856 patients; cluster 4, 1012 patients; cluster 5, 311 patients; and cluster 6, 24 patients. The total number of radiographic procedures undergone by patients in cluster 1 was 1603 (67.5% of which were bedside radiography), in cluster 2, 5383 (27.8% bedside radiography); in cluster 3, 3152 (8.4% bedside radiography); in cluster 4, 3172 (44.9% bedside radiography); in cluster 5, 1212 (46.7% bedside radiography); and in cluster 6, 165 (94.5% bedside radiography). Figure 2 shows each cluster in terms of age groups, types of transportation, level of mobility, types of radiography, and whether or not patients had undergone surgery. Figure 3 shows the mean nursing care time per day in each cluster. The patients in cluster 1 included a high proportion of nonindependent patients, and most of the patients in this group had undergone surgery (Figure 2(a)). The patients in cluster 2 were non-elderly, and compared with cluster 1, more patients required escorting rather than a stretcher. This cluster also included fewer surgical patients compared with cluster 1, and 40% of them underwent bedside radiography (Figure 2(b)). The patients in cluster 3 had the highest ADL scores of any cluster and mainly comprised non-elderly, independent patients; few had undergone surgery, and only a low proportion underwent bedside radiography (Figure 2(c)). The patients in cluster 4 had low ADL scores, and although most were non-elderly, many required assistance, indicating that this group included a high proportion of surgical patients (Figure 2(d)). The patients in cluster 5 were similar to those in cluster 4 in that most had low mobility scores, but this group included a higher proportion of elderly patients, many of whom were not independent (Figure 2(e)). Cluster 6 contained only 24 patients, far fewer than any of the other clusters. The patients in this group had extremely low mobility scores; most were older-elderly patients who were bedridden or required assistance, and a low percentage had undergone surgery (Figure 2(f)). In addition, Figure 4 shows the classification of image type by day before surgery (a), day of surgery (b), and within 3 days of surgery (c) for all patients who underwent surgery and patients in each obtained cluster group. In both groups, the rate of bedside radiography was higher after surgery compared to before surgery, and bedside radiography accounted for 92.7% of the images taken on the day of surgery.

Figure 1.

Dendrogram constructed for the patient data.

Figure 2.

Breakdown of (a) cluster 1, (b) cluster 2, (c) cluster 3, (d) cluster 4, (e) cluster 5, and (f) cluster (6) in terms of age groups, types of transportation, level of mobility, types of radiography, and whether or not patients had undergone surgery.

Figure 3.

Breakdown of the mean nursing care time per day in each cluster.

Figure 4.

Type of radiography (a) before surgery, (b) on the day of surgery, and (c) within 3 days of surgery in each cluster.

Probability of bedside radiography and quantification of patient factors by logistic regression analysis

Notably, this analysis excluded data from 4 days after surgery. Further, when imaging was performed on the same patient more than once within one hospital stay, data on each imaging date were used. This brought the total number of cases to 22,838. Categorical variables were reduced to dichotomous variables based on the explanatory variables outlined in Table 1.

Table 1.

Target variable (Y) and explanatory variables (x) incorporated into the logistic model.^a

Target variable	Bedside radiography or general radiography
Y = 1 or 0	Bedside radiography or general radiography
Explanatory variable	Coefficient parameter	Estimated value	p-Value
Intercept	β ₀	−3.233 ± 0.113	<2e–16
Type of transportation (based on independent)
x ₁₁: Set at 1 for stretcher and 0 for other means	β ₁₁	1.336 ± 0.110	<2e–16
x ₁₂: Set at 1 for escort and 0 for other means	β ₁₂	0.378 ± 0.094	3.42E–05
Level of mobility (based on level IV)
x ₂₁: Set at 1 for level I and 0 otherwise	β ₂₁	3.079 ± 0.117	<2e–16
x ₂₂: Set at 1 for level II and 0 otherwise	β ₂₂	1.873 ± 0.105	<2e–16
x ₂₃: Set at 1 for level III and 0 otherwise	β ₂₃	0.700 ± 0.097	1.26E–09
Relationship between day of radiography and surgery (based on no surgery)
x ₃₁: Set at 1 for before surgery and 0 otherwise	β ₃₁	1.445 ± 0.066	<2e–16
x ₃₂: Set at 1 for the day of surgery and 0 otherwise	β ₃₂	3.544 ± 0.089	<2e–16
x ₃₃: Set at 1 for within 3 days of surgery and 0 otherwise	β ₃₃	1.217 ± 0.070	<2e–16
Sex (based on women)
x ₄	β ₄	−0.103 ± 0.041	4.08E–04
Age
x ₅	β ₅	−0.012 ± 0.001	7.24E–15

^aThe five explanatory variables are expressed as 10 items by converting qualitative to quantitative variables. β were coefficients estimated by the maximum-likelihood method.

Table 1 shows the results of coefficient estimation by the maximum-likelihood method. When independent walking was set as the standard (0) for type of transportation, the weighting coefficient for use of a stretcher was 1.336, larger than that for requiring an escort, at 0.378. When level IV was set as the standard (0) for level of mobility, the weighting coefficient was greatest when the level of mobility was lowest, at 3.079 for level I, 1.873 for level II, and 0.700 for level III. With regard to the day of imaging before and after surgery, if no surgery was set as the standard (0), the weighting coefficient before surgery was 1.445, the weighting coefficient for imaging on the day of surgery was 3.544, and the weighting coefficient for within 3 days of surgery was 1.217; thus, the weighing coefficient was greatest on the day of surgery. When female sex was set as the standard (0), the weighting coefficient for male sex was lower, at −0.103. Taking all the explanatory variables together, the weighting coefficient was highest for level of mobility I, followed by type of transportation (stretcher). The estimated weighting coefficient for age was low, at −0.012, and the intercept was −3.233. The estimated values of the weighting coefficients for these 10 explanatory variables were consistent with formula 1 in the probability model for undergoing bedside radiography. Nagelkerke’s R ² was 0.659.

The classification table is shown in Table 2. The percentage of cases the model classified correctly was 85.6% overall. This model formula can be used to calculate the mean probability of undergoing bedside radiography for all 12 individual combinations of type of transportation and level of mobility (Table 3). The probability of bedside radiography was highest (0.879 ± 0.090) for transport by stretcher with a mobility level of I, followed by transport by stretcher with a mobility level of II (0.622 ± 0.153) and escorted transport with a mobility level of I (0.601 ± 0.145). Independent walking with a mobility level of IV was associated with the lowest probability of bedside radiography (0.061 ± 0.076). Moreover, the actual rate of bedside radiography for the group with the highest value (transport by stretcher with mobility level of I) was 0.881, which was similar to the probability obtained from the model (0.879 ± 0.090), with a difference in probability of less than 10%. Four of these combinations were considered unlikely to occur in actual clinical practice: transport by stretcher with mobility level of III or IV and independent walking with mobility level of I or II. The differences between the actual rate of bedside radiography in these groups and the probabilities of undergoing bedside radiography predicted by the model were 10% or more. The difference in probability between the actual and predicted values for the other eight groups was less than 10%.

P (Y = 1 | x_{11}, x_{12}, ..., x_{5}) = \frac{e x p [β_{0} + β_{11} x_{11} + ... + β_{5} x_{5}]}{1 + \exp [β_{0} + β_{11} x_{11} + ... + β_{5} x_{5}]}

Formula 1. Probability model formula for undergoing bedside radiography.

Table 2.

The classification table of model formula.

Observed		Predicted
		Bedside radiography		Percentage correct
		No	Yes	Percentage correct
Bedside radiography	No	10724	1658	0.866
Bedside radiography	Yes	1618	8838	0.845
Overall percentage				0.856

Table 3.

Mean probabilities of undergoing bedside radiography for all 12 individual combinations of transportation type and mobility level and the actual proportions of patients undergoing each type of radiography.

	Number	Type of transportation	Level of mobility	Proportion of bedside radiography	Proportion of general radiography	Mean probability of bedside radiography (Y)
1	8867	Stretcher	I	0.881	0.119	0.879 ± 0.090
2	1134	Stretcher	II	0.601	0.399	0.622 ± 0.153
3	73	Stretcher	III	0.548	0.452	0.539 ± 0.299
4	39	Stretcher	IV	0.667	0.333	0.498 ± 0.285
5	186	Escort	I	0.527	0.473	0.601 ± 0.145
6	2693	Escort	II	0.315	0.685	0.305 ± 0.144
7	3520	Escort	III	0.145	0.855	0.147 ± 0.126
8	650	Escort	IV	0.063	0.937	0.077 ± 0.086
9	13	Independent	I	0.308	0.692	0.517 ± 0.210
10	16	Independent	II	0	1	0.311 ± 0.249
11	1015	Independent	III	0.110	0.890	0.105 ± 0.111
12	4632	Independent	IV	0.061	0.939	0.061 ± 0.076

Discussion

Patient trends revealed by hierarchical cluster analysis

Matsumoto et al. (2013) showed that the amount of nursing care provided to patients depends on factors such as their age and ability to engage in ADL and reported that the lower a patient’s ability to engage in ADL, the longer the time spent on nursing care. Therefore, in this study, we used medical data from electronic medical records to carry out cluster analysis in patients who underwent either general or bedside radiography during a single hospitalisation and identified the factors of patients in each cluster. Six clusters were categorised on the basis of the resulting dendrogram.

The highest rate of bedside radiography was seen in cluster 6, but this cluster only included 24 patients, constituting a tiny proportion (0.42%) of the patient data as a whole. This group contained only bedridden patients, albeit few who had undergone surgery. Among the other clusters, the highest rate of bedside radiography was in cluster 1, in which around 70% of radiographic procedures were performed at the bedside. In this cluster, 64.3% of patients were elderly, approximately 68% required a stretcher, and 84.1% had a mobility level of I or II. Surgery had been performed on 46.7%, and the mean nursing care time per day was 456 min. The rates of bedside radiography for the day before surgery, the day of surgery, and within 3 days of surgery were 61%, 94.4%, and 88%, respectively. Thus, cluster 1, which had a high rate of bedside radiography, included patients who were older and less capable of engaging in ADL than those in clusters 2–5.

In the results of this cluster analysis, “types of transportation,” “level of mobility,” “types of radiography,” and “surgery” were the main factors in classifying patients. In particular, differences between clusters were observed in the factors of “types of transportation” and “level of mobility.” Moreover, we found that there is a remarkable difference between clusters as for the ratio of bedside radiography. Thus, it is indicated that “types of transportation” and “level of mobility” relate to the ratio of bedside radiography. This result appears reasonable, because doctors are already fairly proficient at selecting immobile patients for bedside radiography.

Bedside radiography and quantification of patient factors

In the logistic regression analysis, coefficient estimation by the maximum-likelihood method identified six factors for which the estimated weighting coefficient exceeded 1: mobility level of I (3.079), mobility level of II (1.873), transport by stretcher (1.336), radiography on the day before surgery (1.445), radiography on the day of surgery (3.544), and radiography within 3 days of surgery (1.217). In surgical patients, the weighting coefficients became greater, especially for imaging performed immediately after surgery, suggesting that nursing care temporarily increases after surgery. Furthermore, among the 12 individual combinations of transportation type and mobility level, the probability of undergoing bedside radiography was highest for the combination of transport by stretcher and mobility level of I, followed by transport by stretcher and mobility level of II and then by escorted transport and mobility level of I. These results demonstrated that type of transportation and level of mobility, both of which are directly related to patients’ ability to engage in ADL, had a heavy weighting toward increasing the probability of undergoing bedside radiography.

The difference between the results predicted by the model with the actual proportions of patients undergoing each type of radiography exceeded 10% for four groups: transport by stretcher with level of mobility III or IV and independent walking with level of mobility I or II. These combinations were seen in very few patients, less than 1% of the total number, and are inconsistent with typical clinical patient profiles. These patients may have been anesthetised in order to undergo therapy and treatment and undergone radiography while still recovering from those therapies.

Secondary use of HIS data for safe bedside radiography and application to the allocation of radiological technologists

Both general and bedside radiography are ordered by a requesting doctor or attending doctor within 1 day before radiography is performed. However, radiography may also be ordered on the same day for emergency patients or those whose condition has changed rapidly. Whether radiography is performed in the X-ray room or at the bedside is based on the experience of the ordering doctor. Radiological technologists perform radiography in accordance with these orders. Although doctors normally take the patient’s condition into account when issuing an order, it is possible that they might not accurately understand the patient’s condition in real time. For example, there are cases where, with orders (general radiography or bedside radiography) from doctors with little experience, the actual images taken by the technician do not match the patient’s condition. Furthermore, there is a possibility that the doctor may issue an order without confirming the patient’s condition.

On the other hand, nurses at the bedside input data on patients’ conditions into their electronic medical records in real time, meaning that a patient’s condition when the doctor orders radiological scans will have been updated on the day of the scan. The use of up-to-date patient data enables quantitative, objective information to be added to the doctor’s order, enabling the technologists performing the scans to use the appropriate method to maximise safety.

In particular, because the evaluation parameters of the patients analysed in this study are input into the patient summary section of the electronic medical records, the use of the probability model formula for undergoing bedside radiography developed in this study could enable the generation of weighting coefficients for individual patients. Therefore, it is considered a useful decision support method for doctors to order bedside radiography based on quantitative data. According to the results of this study, in the patient group carrying stretcher with a mobility level of I, the actual rate of bedside radiography was 88.1%, and the probability of bedside radiography as determined by logistic regression analysis was 87.9%. Thus, for these patient groups, bedside radiography can be performed with 87.9% probability, and the weighting coefficient obtained in this study can contribute to safe bedside radiography. This method could provide support for the optimisation of bedside radiography on the basis of quantitative data. It would also not only enable provision of the assistance required by individual patients during scanning but also allow the demand for bedside radiography to be predicted each day based on the number of orders from the preceding day and on the day itself and the weighting coefficients for each patient. In addition, this method could be useful for managing the schedule of radiological technologists and enabling staff to be allocated appropriately, potentially improving management in terms of medical safety as well as streamlining scanning tasks. It could thus enable safe radiological technology with a low burden on patients.

Our study has several limitations. First, the data were limited to Kagoshima University Hospital. Differences in patient data from other hospitals compared to our hospital, or changes in patient data due to the progression of aging, development of new medical technology, or typical patient populations in hospitals are assumed to lead to differences in patient weighting. Second, if a patient undergoes a sudden change in condition and their electronic medical record is not updated, the radiography order may be erroneous in some cases. However, using the logistic model of this study, it might be possible to identify such a patient immediately. Specifically, an explanatory variable on the day of imaging is substituted into the model formula, and the probability of bedside radiography is calculated. If the probability is close to 0 even though the imaging order of the day was for bedside radiography, there is a possibility that their data have not been updated. Finally, in this study, validation on the obtained model formula is yet insufficient, and this method has limitations on the accuracy of prediction. In further work, we will verify the model formula using the following two approaches: (i) the method for evaluation of the model accuracy using the cross-validation and (ii) mounting model formula on electronic clinical record system and verifying the model using future data.

Conclusion

Our results suggest that the secondary use of HIS data for the quantitative evaluation of patient factors and implementation of those quantitative values in medical records may be useful for the safe performance of bedside radiography as well as for decision support method for doctors to order bedside radiography.

Footnotes

Acknowledgment

The authors would like to thank the staff of the Department of Radiology and the Joint Research Laboratory, Kagoshima University Graduate School of Medical and Dental Sciences, for the use of their facilities.

Human subject protections

The study was performed in compliance with the World Medical Association Declaration of Helsinki on Ethical Principles for Medical Research Involving Human Subjects and was reviewed by the Institutional Review Board of Kagoshima University Hospital.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Ben-Shlomo

Bartal

Mosseri

. (2016) Effective dose reduction in spine radiographic imaging by choosing the less radiation-sensitive side of the body. The Spine Journal 16(4): 558–563.

Den Harder

De Heer

Maurovich-Horvat

. (2016) Ultra low-dose chest CT with iterative reconstructions as an alternative to conventional chest X-ray prior to heart surgery (CRICKET study): rationale and design of a multicenter randomized trial. Journal of Cardiovascular Computed Tomography 10(3): 242–245.

Eklund

Klefsgård

Ivarsson

. (2012) Positive experience of a mobile radiography service in nursing homes. Gerontology 58(2): 107–111.

Goodman

Amurao

(2012) Medical imaging radiation safety for the female patient: rationale and implementation. Radiographics 32(6): 1829–1837.

Hamer

Sirlin

Strotzer

. (2005) Chest radiography with a flat-panel detector: image quality with dose reduction after copper filtration. Radiology 237(2): 691–700.

Hospers

van der Laan

Zeebregts

. (2009) Vertebral fracture assessment in supine position: comparison by using conventional semiquantitative radiography and visual radiography. Radiology 251(3): 822–828.

Huang

Schweitzer

(2014) The role of radiology in the evolution of the understanding of articular disease. Radiology 273(2 Suppl): S1–22.

Iezzi

Santoro

Marano

. (2012) Low-dose multidetector CT angiography in the evaluation of infrarenal aorta and peripheral arterial occlusive disease. Radiology 263(1): 287–298.

Igarashi

Fukushi

Shinoda

. (2010) Investigation of errors by radiological technologists and evaluation of preventive measures: general and mobile X-ray examinations. Radiological Physics and Technology 3(2): 136–143.

10.

Inamura

Kim

(2011) History of PACS in Asia. European Journal of Radiology 78(2): 184–189.

11.

Jeong

Joo

Ryu

. (2014) Development of a mini-mobile digital radiography system by using wireless smart devices. Journal of Digital Imaging 27(4): 443–448.

12.

Kijowski

Blankenbaker

Stanton

. (2006) Radiographic findings of osteoarthritis versus arthroscopic findings of articular cartilage degeneration in the tibiofemoral joint. Radiology 239(3): 818–824.

13.

Körner

Weber

Wirth

. (2007) Advances in digital radiography: physical principles and system overview. Radiographics 27(3): 675–686.

14.

Matsumoto

Uto

Muranaga

. (2013) DPC in acute-phase inpatient hospital care. Visualization of amount of nursing care provided and accessibility to nursing care. Methods of Information in Medicine 52(6): 522–535.

15.

McCollough

(2012) Automated data mining of exposure information for dose management and patient safety initiatives in medical imaging. Radiology 264(2): 322–324.

16.

National Institute of Population and Social Security Research (2016) Available at: http://www.ipss.go.jp/syoushika/tohkei/newest04/sh2401top.html (accessed 10 February 2016).

17.

Regulation of Act on Assurance of Medical Care for Elderly People (2015) Available at: http://law.e-gov.go.jp/htmldata/S57/S57HO080.html (accessed 20 August 2015).

18.

Sarti

Brehmer

Gay

(2012) Low-dose techniques in CT-guided interventions. Radiographics 32(4): 1109–1119.

19.

Sasaki

Uto

Kawahara

. (2014) Demand forecast for portable radiography as a result of changes in hospital management policy: with the aim of improving services for safe imaging. The 34th Japan Association for Medical Informatics 34(34): 434–437.

20.

Siddiqui

Galizia

Almusa

. (2014) Evaluation of the tarsometatarsal joint using conventional radiography, CT, and MR imaging. Radiographics 34(2): 514–531.

21.

Thabet

Kalva

Liu

. (2012) Interventional radiology in pregnancy complications: indications, technique, and methods for minimizing radiation exposure. Radiographics 32(1): 255–274.

22.

Uffmann

Neitzel

Prokop

. (2005) Flat-panel-detector chest radiography: effect of tube voltage on image quality. Radiology 235(2): 642–650.

23.

Uto

Kumamoto

(2005) Study on weighting of amount of nursing care using data on index of patient’s need for nursing and system approach. Journal of Medical Systems 29(2): 165–177.

24.

Utsunomiya

Monzen

Akimoto

. (2013) A feasibility study on reduction of the entrance-surface dose to neonates by use of a new digital mobile X-ray system. Radiological Physics and Technology 6(1): 157–161.

25.

Yamamoto

Matsumoto

Tada

. (2008) A data capture system for outcomes studies that integrates with electronic health records: development and potential uses. Journal of Medical Systems 32(5): 423–427.