This cross-sectional nationwide survey in Japan was conducted targeting executive physicians at hospitals providing cancer treatments as respondents. The survey form was distributed in November 2017, and non-responding institutions were reminded 3 weeks after the first mailing. Responses to the survey in written format were considered consent to participate. Based on the national ethical guideline of epidemiological studies in Japan, this study was exempted from review by the Ethics Committee at the Kyoto University Graduate School and Faculty of Medicine, Kyoto University Hospital.
Two target samplings were identified: one was 399 DCHs, where the Ministry of Health, Labor, and Welfare-authorized high-quality cancer treatment was provided, and the other sample was non-DCHs that potentially manage patients with cancer because a considerable number of patients could receive cancer treatment at non-DCHs. Non-DCHs were randomly sampled and then stratified based on regions and inpatient bed numbers.
The sampling strategy is summarized in Figure 1. To identify DCHs, the list of DCHs was obtained from the Ministry website as of April 2017. To identify non-DCHs, all hospital data were acquired from Japan Medical Press, Inc., in October 2017. To identify non-DCHs that offer cancer treatment, we excluded the following hospitals: (i) hospitals with <100 general ward beds (this was because there were no DCHs with <100 inpatients beds), (ii) national sanatorium, (iii) hospitals not delivering cancer treatment based on the hospital name and clinical departments (e.g., rehabilitation hospitals or no cancer treatment departments), and (iv) others (e.g., breast cancer-specified hospitals). To ensure representativeness and comparability between DCHs and non-DCHs at each region, stratified random sampling was performed based on the region and inpatient bed number. Regional strata were divided into nine categories: Hokkaido, Tohoku, Tokyo, Kanto other than Tokyo, Chubu, Kansai, Chugoku, Shikoku, and Kyushu-Okinawa. Inpatient bed number strata were divided into four categories based on the actual distribution of DCHs: <299 beds, 300–499 beds, 500–699 beds, and >700 beds. Considering the possibility that the response rate from non-DCHs may be low, three times more non-DCHs were extracted in each stratum. Moreover, responding hospitals that did not provide three cancer treatment modalities (surgery, systemic chemotherapy, and radiation therapy) at their own hospitals were excluded from the analysis in order to ensure comparability according to predefined exclusion criteria.
Sample size calculation
We did not define a primary outcome owning to the explanatory nature of the survey. However, to compute the confidence interval of the point estimate within 10%, responses were needed from at least 96 hospitals. Therefore, at least 40% of responses were estimated to be obtained, consisting of 240 hospitals from both types of hospital.
The questionnaire was created after a comprehensive literature search. A pilot test was performed on three physicians with positions equivalent to that of executives in cancer hospitals to ensure face and content validity.
i) Current status of IOP
To clarify the current status of IOP, the international consensus was used as the IOP indicator , consisting of five sections: clinical structure, clinical process, clinical outcomes, education, and research, which were categorized as major or minor indicators. Clinical outcome indicators were excluded because the target respondents were experiencing difficulties in answering these questions due to the heterogeneity among specialties within the same hospitals based on the pilot test results. Thus, a total of 25 questions (Table 2-4) were primarily used based on the categorical answer format from no (0%), limited (1%–24%), approximately half (25%–74%), mostly (75%–99%), and all (100%) departments.
ii) Executive physicians’ perception toward IOP
To investigate executive physicians’ perceptions of the oncology department toward IOP, 16 questions were used based on the literature search [6, 15-21]. Each question was rated on a 5-point Likert-type scale, from 1 (strongly disagree) to 5 (strongly agree). In addition, a free text query was prepared by asking opinions regarding IOP.
Descriptive statistics was performed to summarize the data. To adjust the biased distribution of inpatient bed number between the responding DCHs and non-DCHs, each stratum of non-DCHs was weighted according to DCH distributions in inpatient bed number (<299 beds, 300–499 beds, 500–699 beds, and >700 beds). Missing data was not imputed. The t-test or Cochrane–Armitage trend test was used to compute differences in the proportion, as appropriate. A P-value of <0.05 was considered statistically significant. Owning to the exploratory nature of this survey, the adjustment of multiple testing was not performed. The GNU R software (version 3.2.0; R Project for Statistical Computing, Vienna, Austria) was used for all statistical analyses.
Free comment responses were qualitatively analyzed using inductive content analysis method [22, 23]. Two independent investigators (M.N. and S.I.) reviewed and generated the codes. Then, emerging codes were compared and discussed with an expert PC physician (Y.U.) to achieve agreement of the codes labeled from the data. To ensure rigor and trustworthiness, an experienced investigator (T.M) supervised and examined the consistency of results.