Professional Documents
Culture Documents
10 1093@intqhc@mzy233
10 1093@intqhc@mzy233
doi: 10.1093/intqhc/mzy233
Article
Address reprint requests to: Takumi Nishi, Department of Research Planning and Information Management, Fukuoka
Institute of Health and Environmental Sciences, 39 Mukaizano, Dazaifu-shi, Fukuoka 818-0135, Japan. Tel: +81-92-921-
9941; Fax: +81-92-928-1203; E-mail: nishi@fihes.pref.fukuoka.jp
Editorial Decision 16 October 2018; Accepted 29 October 2018
Abstract
Objective: The aim of this study was to examine whether anesthetic technique is associated with
30- or 90-day mortality and perioperative length of stay (LOS).
Design: We used a retrospective cohort design using a healthcare insurance claims database.
Setting: The Fukuoka Prefecture’s claims database of older patients who underwent hip fracture
surgery under general or regional (spinal or epidural) anesthesia from April 2012 to March 2016
was used for analyses.
Participants: The database under analyses contained 16 125 participants of hip fracture surgery
under general or regional anesthesia.
Main Outcome Measure: We measured 30- and 90-day mortalities and perioperative LOS.
Results: In a propensity score-matched cohort, we found no significant differences in 30- and 90-
day mortalities after adjusting for confounding factors. The reconverted perioperative LOS for the
general and regional anesthesia groups was, respectively, 29.7 (29.1–30.4) and 28.0 (27.4–28.6) days
in the matched cohort. Therefore, the perioperative LOS in the regional anesthesia group was sig-
nificantly shorter by 1.7 days than in the general anesthesia group (P < 0.001).
Conclusions: This study demonstrated that the use of regional anesthesia was not associated with
30- or 90-day mortality, but it was associated with slightly shorter perioperative LOS. Since Japan
has much longer LOS than other countries, our findings have implications for more efficient
healthcare resource utilization and quality assurance in geriatric care.
Key words: anesthetic technique, mortality, length of stay, claims data, propensity score matching
© The Author(s) 2018. Published by Oxford University Press in association with the International Society for Quality in Health Care. All rights reserved.
For permissions, please e-mail: journals.permissions@oup.com 1
2 Nishi et al.
Methods
Participant selection Propensity score matching
We used claims data from the Fukuoka Prefecture Regional We conducted 1:1 propensity score matching in order to reduce the
Association for Late-Stage Healthcare for Older People for fiscal effects of confounding factors including gender, age, type of claim,
2012–15. We obtained claims data on insurance beneficiaries hospi- procedure, timing of procedure, admission day, fracture types,
talized between 1 April 2012 and 31 March 2016. We extracted the admission fiscal year and comorbidities. And we compared mortality
data from 34 344 inpatient records involving a primary diagnosis, and perioperative LOS between regional and general anesthesia
or dominant care resource use, of hip fracture according to the groups after matching. In accordance with previous studies on vari-
International Classification of Diseases, 10th revision (ICD10; able selection with propensity score matching [19–21], we calculated
codes: S72.0-2). Claims data contain information about all proce- propensity scores using a logistic regression model. We aimed to
dures, drugs and devices indexed in the national fee schedule; stand- identify the relationship between regional anesthesia and history
ard codes are assigned to each [12]. From those records, we within 1 year before admission of the following: hip fracture sur-
extracted data for 22 324 patients who had undergone hip fracture gery; chronic obstructive pulmonary disease and treatment with psy-
surgery under general or regional (spinal or epidural) anesthesia. choactive agents, antiplatelets or anticoagulants. As practice style
After excluding duplicate or inappropriate admission records, differed by regions or hospitals, we introduced dummy variables for
we eliminated the following insurance beneficiaries: (i) those who 62 residential municipalities in Fukuoka Prefecture (i.e. 61 variables)
underwent multiple anesthesia; (ii) those admitted to wards other and the interaction terms of residential municipalities and types of
than general or acute care wards; (iii) those aged 65–74 years at first claims in propensity score estimation model, in addition to age, sex
admission (only individuals aged ≥75 or 65–74 years with a specific and the other covariates listed earlier. We employed the Hosmer
disability, such as hemodialysis recipients are eligible for Japan’s −Lemeshow test and C statistic as indicators of how well the logistic
late-stage older people’s health care); (iv) those with multiple frac- regression model fitted the data. The C statistic was 0.753, and the
tures and (v) those who underwent mixed anesthesia. Hosmer−Lemeshow test did not reject the null hypothesis
Ethical approval for this study (No. 27-15) was obtained from the (P = 0.263). Finally, each patient of the regional anesthesia group
Ethics Committee of Life Science, Fukuoka Institute of Health and was matched with a unique control in the general anesthesia group
Environmental Sciences, Fukuoka, Japan, on 5 October 2015. The within a caliper width of 0.02 [22]. As shown in Fig. 1, the number
study used anonymized claims data, so the need to obtain informed of subjects was 9520 (59.0%) in the general anesthesia group and
consent was waived in line with the Ethical Guidelines for Medical 6605 (41.0%) in the regional anesthesia group; we obtained 4708
and Health Research Involving Human Subjects in Japan [13]. pairs from them.
Anesthesia for hip surgery 3
Initial participants
34,344 admission records
Study subjects
16,125 beneficiaries
Statistical analyses subjects admitted in fiscal 2014–15 were higher in the general anes-
We summarized the characteristics of the study subjects using fre- thesia group. The mean CCI was higher in the general anesthesia
quency and proportions for categorical variables and mean and than in the regional anesthesia group: mean (SD) for general anes-
standard deviation (SD) for continuous variables. We employed thesia, 1.1 (1.3), and for regional anesthesia, 1.0 (1.2) (P < 0.001).
standardized differences to check the balance between the regional Other than sex, there were significant imbalances between the two
and general anesthesia groups. We set 0.10 as the threshold value groups.
for acceptable standardized differences. We compared mortality The patient characteristics after propensity score matching
between the regional and general anesthesia groups using Pearson’s appear in Table 2. All imbalances that existed before propensity
Chi-square test. We made comparisons of perioperative LOS using score matching disappeared and the standardized differences of cov-
the unpaired t-test and assumed unequal variance. ariates were at an acceptable level.
Multiple logistic regression analyses were used to estimate the
adjusted effects on mortality. We employed a generalized linear Comparison of outcomes
model (GLM) with a log-link function and gamma distribution to Table 3 shows the comparison of outcomes in the unmatched and
estimate the effects of anesthetic techniques on perioperative LOS matched cohorts. In the unmatched cohort, 98 (1.0%) subjects died
and to control for other confounding factors. We used exponential within 30 days after hip fracture surgery in the general anesthesia
transformed coefficients of those models, the odds ratio in logistic group and 77 (1.2%) subjects did so in the regional anesthesia group.
models and a multiplicative effect in GLM to evaluate the relative Within 90 days after hip fracture surgery, 364 (3.8%) subjects died in
effects of covariates. Furthermore, we reconverted coefficients to the the general anesthesia group and 265 (4.0%) subjects did so in the
original units of mortality (%) and perioperative LOS (days) by esti- regional anesthesia group. There were no significant differences in 30-
mating marginal effects and 95% confidence intervals (CIs). and 90-day mortalities. However, the regional anesthesia group had a
All statistical analyses used STATA for Windows, version 15.1 significantly shorter perioperative LOS (2.1 days) than the general
(StataCorp, College Station, TX, USA). The level of statistical sig- anesthesia group: mean (95% CI), 30.4 (29.9–30.8) days in the gen-
nificance was set at 0.05. eral anesthesia group and 28.3 (27.8–28.7) days in the regional anes-
thesia group (P < 0.001). After adjusting for confounding factors,
there were no significant differences in 30- and 90-day mortalities.
Results The reconverted perioperative LOS for the general and regional anes-
Descriptive statistics thesia groups was 29.9 (29.5–30.4) and 28.9 (28.4–29.4) days,
As evident in Table 1, the mean age was significantly higher in the respectively. Thus, the regional anesthesia group had a significantly
regional anesthesia than in the general anesthesia group: mean (SD) shorter perioperative LOS (1.0 days) than the general anesthesia
for general anesthesia, 86.1 (5.7), and for regional anesthesia, 86.7 group (P = 0.002).
(5.8) (P < 0.001). The proportions for DPC/PDPS claims, osteo- Similarly, even after adjusting for confounding factors, there were
synthesis and intertrochanteric fracture were higher in the regional no significant differences in 30- and 90-day mortalities in the matched
anesthesia group; the proportions for femoral neck fracture, hip cohort. The unadjusted perioperative LOS was significantly shorter
fracture surgery performed ≥2 days after admission, CCI ≥2 and by 1.7 days in the regional anesthesia than in the general anesthesia
4 Nishi et al.
Table 1 Baseline characteristics of subjects by anesthetic technique for hip fracture surgery
Age
Mean (SD) 86.1 (5.7) 86.7 (5.8) 0.104 <0.001
Sex
Male 1642 (17.2%) 1125 (17.0%) 0.006 0.721
Female 7878 (82.8%) 5480 (83.0%)
Type of claim
FFS 2648 (27.8%) 1264 (19.1%) 0.206 <0.001
DPC 6872 (72.2%) 5341 (80.9%)
Procedure
Osteosynthesis 6066 (63.7%) 4873 (73.8%) 0.014 <0.001
Total hip arthroplasty 59 (0.6%) 12 (0.2%) 0.153
Bipolar hip arthroplasty 3395 (35.7%) 1720 (26.0%) 0.150
Timing of procedure
Admission day or day after 2869 (30.1%) 2645 (40.0%) 0.209 <0.001
≥2 days after admission 6651 (69.9%) 3960 (60.0%)
Type of fracture
Subtrochanteric 191 (2.0%) 120 (1.8%) 0.014 <0.001
Intertrochanteric 4879 (51.3%) 3887 (58.8%) 0.153
Femoral neck 4450 (46.7%) 2598 (39.3%) 0.150
Charlson comorbidity index
Mean (SD) 1.1 (1.3) 1.0 (1.2) 0.066 <0.001
0 4008 (42.1%) 2922 (44.2%) 0.043 0.001
1 2851 (29.9%) 2015 (30.5%) 0.012
2 1529 (16.1%) 993 (15.0%) 0.028
≥3 1132 (11.9%) 675 (10.2%) 0.053
Admission fiscal year
2012–13 4379 (46.0%) 3260 (49.4%) 0.067 <0.001
2014–15 5141 (54.0%) 3345 (50.6%)
group: mean (95% CI), 29.7 (29.0–30.4) days in the general anesthe- survival of patients who had undergone surgical repair of femoral
sia group and 28.0 (27.5–28.6) days in the regional anesthesia group neck fracture with respect to anesthetic technique: only the incidence
(P < 0.001). Furthermore, the reconverted perioperative LOS for the of deep vein thrombosis was clearly greater in patients receiving gen-
general and regional anesthesia groups was 29.7 (29.1–30.4) and eral anesthesia, but there were no significant differences for blood
28.0 (27.4–28.6) days, respectively. Therefore, the perioperative LOS loss or mortality [24]. Recent large observational studies for anes-
of the regional anesthesia group was significantly shorter by 1.7 days thetic techniques during hip fracture surgery or joint replacement
than in the general anesthesia group (P < 0.001) surgery for the lower extremities have produced conflicting results.
Some investigations reported that regional anesthesia was associated
with lower risk of postoperative complications or mortality [25–28];
Discussion others found no significant difference in mortality [29–34]. Basque
In this study, we examined the impact of anesthetic techniques on et al. reported that a general anesthesia group had shorter LOS [35].
mortality and perioperative LOS in older patients following hip frac- However, our findings are similar to those reported in USA, UK
ture surgery in Fukuoka Prefecture, Japan. We found that the use of and Taiwan with respect to mortality among anesthetic techniques
regional anesthesia was not associated with 30- or 90-day mortality, [29, 31, 32, 34]. In particular, our assessment of the relative effect of
but it was associated with relatively short perioperative LOS. regional anesthesia on perioperative LOS is similar to that found in
To the best of our knowledge, this is the first study to compare previous studies, though we observed considerably longer LOS
mortality and LOS between anesthetic techniques in Japan. The [29, 32]. Given that LOS in Japan is much longer than in USA or
healthcare system and resource utilization are different in Japan UK [5–7], reducing the perioperative LOS in Japan could have a major
from previously reported countries, so this study makes a notable economic impact. Therefore, if either type of anesthesia could be used
important contribution to the literature. Based on the evidence from on a patient for hip fracture surgery, regional anesthesia would be
two meta-analyses, Japanese clinical practice guideline states that attractive in terms of effective healthcare resource utilization.
either general or regional anesthesia is appropriate [11]. One meta- However, it is necessary to be cautious when interpreting results
analysis integrated the results from randomized or quasi- that are not statistically significant but show a positive association
randomized control trials. It reported that regional anesthesia was with 30- and 90-day mortalities. The mortality we found was lower
associated with decreased mortality at 1 month after surgery; how- than that determined in previous studies. With inpatients aged 65
ever, the results from six trials for 3-month mortality were not stat- years or older after hip fracture surgery, Kondo et al. observed sig-
istically significant [23]. Another meta-analysis compared the nificantly longer LOS and better survival in three Japanese hospitals
Anesthesia for hip surgery 5
Table 2 Subject characteristics after propensity score matching by anesthetic technique for hip fracture surgery
Age
Mean (SD) 86.6 (5.7) 86.5 (5.8) 0.019 0.369
Sex
Male 771 (16.4%) 806 (17.1%) 0.020 0.334
Female 3937 (83.6%) 3902 (82.9%)
Type of claim
FFS 1054 (22.4%) 1042 (22.1%) 0.006 0.766
DPC 3654 (77.6%) 3666 (77.9%)
Procedure
Osteosynthesis 3316 (70.4%) 3330 (70.7%) 0.007 0.756
Total hip arthroplasty 9 (0.2%) 12 (0.3%) 0.014
Bipolar hip arthroplasty 1383 (29.4%) 1366 (29.0%) 0.008
Timing of procedure
Admission day or day after 1711 (36.3%) 1726 (36.7%) 0.007 0.748
≥2 days after admission 2997 (63.7%) 2982 (63.3%)
Type of fracture
Subtrochanteric 91 (1.9%) 92 (2.0%) 0.002 0.997
Intertrochanteric 2662 (56.5%) 2660 (56.5%) 0.001
Femoral neck 1955 (41.5%) 1956 (41.5%) 0.000
Charlson comorbidity index
Mean (SD) 1.0 (1.2) 1.0 (1.2) 0.012 0.575
0 2057 (43.7%) 2003 (42.5%) 0.023 0.687
1 1447 (30.7%) 1463 (31.1%) 0.007
2 695 (14.8%) 724 (15.4%) 0.017
≥3 509 (10.8%) 518 (11.0%) 0.006
Admission fiscal year
2012–13 2313 (49.1%) 2262 (48.0%) 0.022 0.293
2014–15 2395 (50.9%) 2446 (52.0%)
Unadjusted
Number of subjects 9,520 6,605 4,708 4,708
30-Day mortality, n (%) 98 (1.0%) 77 (1.2%) 0.411* 48 (1.0%) 58 (1.2%) 0.488*
Risk ratio (95% CI) 1.13 (0.84 to 1.52) 1.15 (0.78 to 1.68)
Risk difference (95% CI), % 0.1 (−0.2 to 0.5) 0.1 (−0.3 to 0.6)
90-Day mortality, n (%) 364 (3.8%) 265 (4.0%) 0.543* 178 (3.8%) 188 (4.0%) 0.594*
Risk ratio (95% CI) 1.05 (0.90 to 1.23) 1.06 (0.86 to 1.29)
Risk difference (95% CI), % 0.2 (−0.4 to 0.8) 0.2 (−0.6 to 1.0)
Perioperative LOS, mean (95% CI), days 30.4 (29.9 to 30.8) 28.3 (27.8 to 28.7) <0.001** 29.7 (29.0 to 30.4) 28.0 (27.5 to 28.6) <0.001**
Difference in days (95% CI) −2.1 (−2.8 to −1.5) −1.7 (−2.5 to −0.8)
Adjusted
30-Day mortality, estimate (95% CI), % 1.0 (0.8 to 1.2) 1.2 (0.9 to 1.4) 1.0 (0.7 to 1.3) 1.2 (0.9 to 1.5)
Odds ratio (95% CI) 1.14 (0.84 to 1.54) 0.408*** 1.13 (0.76 to 1.67) 0.540***
Marginal effect (95% CI) 0.1 (−0.2 to 0.5) 0.1 (−0.3 to 0.5)
90-Day mortality, estimate (95% CI), % 3.9 (3.5 to 4.2) 4.0 (3.5 to 4.4) 3.8 (3.3 to 4.3) 4.0 (3.4 to 4.5)
Odds ratio (95% CI) 1.03 (0.87 to 1.22) 0.713*** 1.04 (0.84 to 1.29) 0.689***
Marginal effect (95% CI) 0.1 (−0.5 to 0.7) 0.2 (−0.6 to 0.9)
Perioperative LOS, estimate (95% CI), days 29.9 (29.5 to 30.4) 28.9 (28.4 to 29.4) 29.7 (29.1 to 30.4) 28.0 (27.4 to 28.6)
Multiplicative effect (95% CI) 0.97 (0.94 to 0.99) 0.002*** 0.94 (0.91 to 0.97) <0.001***
Marginal effect (95% CI) −1.0 (−1.7 to −0.4) −1.8 (−2.6 to −0.9)
than in two US hospitals [36]. After adjusting for patient and References
national characteristics, Kondo et al. also found that longer LOS fol-
1. Cooper C, Campion G, Melton LJ. Hip fractures in the elderly: a world-
lowing surgery was associated with lower risk of mortality after dis- wide projection. Osteoporos Int 1992;2:285–9.
25. Radcliff TA, Henderson WG, Stoner TJ et al. Patient risk factors, opera- 34. Tung YC, Hsu YH, Chang GM. The effect of anesthetic type on outcomes
tive care, and outcomes among older community-dwelling male veterans of hip fracture surgery: a nationwide population-based study. Medicine
with hip fracture. J Bone Joint Surg Am 2008;90:34–42. 2016;95:e3296.
26. Neuman MD, Silber JH, Elkassabany NM et al. Comparative effective- 35. Basques BA, Bohl DD, Golinvaux NS et al. General versus spinal anaes-